I need a 3 page research paper on 

Title:  Addictive being young and older on Social Media, why activities outdoors can prevent addiction 

In the attached zip file, I have provided 10 journals that you need to use for this research paper.

In the word doc, I have shared the topic and sub-topics that you have to use. And it also has guidelines from the teacher for this paper.

Due on Saturday, 13th March 4PM PST

REQUIREMENT: Write a RESEARCH PAPER

Title:

Addictive being young and older on Social Media, why activities outdoors can prevent addiction.

Hypothesis:

I predict younger people use social media more than older people.

3 pages total.

1-2 paragraphs intro

And then elaborate on 3 topics, using the bullet points and corresponding journals.

What’s  your feeling about the title and why. Use 10 journals, but please don’t copy authors, only (last names and dates) and  While reading they might mention a author you get click that author to see what they’re talking about talk about that to, this is a  research paper. 

I have sub topics I need you to focus on.

Please find things to quote on topics and on bullet points. please stay on the topics and my title

Take the information out that you would like to write about on the bullet points I added below and particularly look at the abstract, method, discussion and results in the journal so you focus on the title and the bullet point. Its a research survey paper

3 Topics (and corresponding sub-topics in each section) to be used in the writing:

1.  Being on Social Media and addiction to on social media

         Interaction online and or activities  

         Addiction with social media 

         Using social media more and more

2. Social Media and a Mental use

         Being unhappy online while often scrolling (McNally et al)

         Having to be liked on social media ( Liang et al)

         Mental unstable and using social media ( Aschbrenner et al)

         Being on social media, being sad, happy, or angry ( McNally et al)

3. Addictive and Social Media isolation or addiction

· Addictive to being on social media (Connolly et al)

· Social media more than interaction (Ulusoy et al)

· Being alone and coping with social media (Bungert et al)

GUIDELINES from Teacher

Introduce the reader to the general problem and provide any necessary definitions.  And clearly state the purpose of the paper (i.e. I want to see a sentence that looks something like this: “The purpose of this paper is…”)

Provide enough background.  You want the reader to understand the subject and some of the general background and past research on the topic.  You need at least 10 sources for the whole paper, and most of them will be in the literature review (so that’s about 3-4 per sub-section).  Focus on peer-reviewed journals. 

Citations in APA style.  (Only last names and the year, and it differs by the number of authors on the paper).  Use online resources for help.

Then build support for the hypothesis.  Take everything you’ve already written (definitions, background, problem), and tie it all together to explain why and how your hypothesis answers the research question, or addresses the problem or the purpose of the study.

Clearly state the hypothesis at the end (in statement form, with clear directional relationships between variables).  This should not be a question, and you have to propose a direction (not something generic or going in both directions for example)

Introduction/Literature Review 

Format of the introduction and literature review:

1.       First 1-2 paragraphs

         The first sentence should engage the reader (basically this is why they should read your paper)

         Then define your variables and important terms

         Finally, clearly state the purpose of your paper

2.       The next 3-5 pages

         This is where you go into the past research (this is the literature review)

         Subheadings (level 2 in APA format) are helpful to help guide your lit review.  Think of 3-4 sub-topics that are related to your variables.

         Example: Let’s say that I think perceptions of sexual assault victims will be different depending on how the research is presented (from the victim’s perspective vs from the perpetrator’s perspective) and I think this difference will be further influenced by gender.  So I can turn these three things into sub-topics:

    Perceptions of sexual assault victims (but doesn’t have to be specifically about perspective taking or based on gender)

    Perspective taking and feelings of empathy (doesn’t have to relate specifically to sexual assault or gender)

    Differences of perception by gender (doesn’t have to be specifically about sexual assault)

         The point of the literature review is to present the past research on the different topic, then you tie them all together at the end of your intro/lit review

3.       Last 1-2 paragraphs

         This is where you tie together all the research and state your hypothesis

   Example: I presented on perceptions, perspective taking, and gender, and here is how they are all connected…

         State your hypothesis as your expected findings in how the variables will relate

   Example: “I believe that perceptions of victims of sexual assault will be more negative when the information is presented from the perpetrator’s perspective, but that this difference will be only be true for men”

Other things that should be included:

         Title page in APA style with running head different first page and page numbers in the top right

         You can begin an abstract (you will add a sentence or two with each section that you write, Intro, Methods, Results, Discussion) and I won’t actually grade it until the final submission. 

Journals/Joural 1 x
A Closer Look at Appearance and Social Media: Measuring Activity, Self-Presentation, and Social Comparison and Their Associations With Emotional Adjustment

Keywords:

depression, body image, social media, body dysmorphic symptoms, social comparison

Abstract (English):

Social media use links with 2 major concerns of adolescents, namely, appearance and comparing favorably with others. Founded on theory, our purpose was to develop a reliable and valid measure of appearance preoccupation online, the Social Media Appearance Preoccupation Scale (SMAPS). In Study 1 (N = 283 Grade 9–12 students), Australian adolescents completed surveys containing 21 SMAPS items. After psychometric analyses, 18 retained items loaded highly on factors tapping (a) online self-presentation, (b) appearance-related activity online, or (c) appearance comparison. The items loading on each factor had high interitem correlations, and girls had higher SMAPS scores than boys. In Study 2 (N = 327 Australian university students <26 years), the SMAPS was confirmed and validated with a range of measures of social media use, emotional adjustment, appearance concerns, and social behaviors. Factor loadings were invariant by gender, SMAPS subscale scores had very small correlations with age, and incremental validity was tested and supported. Additionally, SMAPS subscale scores interacted with general social media use, adding to the explanation of appearance anxiety. The SMAPS will be a useful resource for the study of appearance-related social media use and interactions with friends and others online. (PsycInfo Database Record (c) 2021 APA, all rights reserved) Population: Human Male Female Location: Australia Age Group: Adolescence (13-17 yrs) Adulthood (18 yrs & older) Young Adulthood (18-29 yrs) Tests & Measures: Social Media Appearance Preoccupation Scale   DOI: 10.1037/t76315-000 Method Participants and procedure The participants were 283 high school students ages 13 to 18 years (M = 16.62, SD = .95; 47% male and 53% female) drawn from three schools in an urban area of Australia. The schools were moderate in size and contained Grades 7 to 12. The schools reported that their students were generally from low-middle to high-middle socioeconomic backgrounds. To measure sociocultural background, participants were asked to endorse as many options as applied, with most (84%) of the participants endorsing white Australian, 18% instead or in addition endorsing Asian, two endorsing Australian first peoples/Torres Strait Islander/Pacific Islander, and 11 also describing a diverse range of other backgrounds. Two participants did not respond to any of the social media items and were removed from the dataset, leaving a final sample size of 281. Approval for this study was obtained from the Griffith University Human Research Ethics Committee. Attempts were made to recontact 335 students from three schools that had participated in an earlier study (not focused on social media use) to obtain consent from parents and adolescents to participate in this study. In the initial consent process, students who returned parent and personal consent forms (regardless of participation) were included in a draw to win 5 $100 gift vouchers to a store of their choice. Overall, 309 families were able to be contacted and 26 parents or students declined to participate, resulting in a 92% response rate. Students from two schools completed the 45-min survey either by mail or online. One school opted to have surveys completed during school time under research assistant supervision. Each participant in this study received a $20 gift card when the survey was returned. Measures To investigate appearance-related preoccupation in adolescents’ use of social media, participants responded to 21 items that were developed to tap into the five aforementioned themes identified in the literature including image activities (e.g., “I prefer to upload photos of myself to social media where I look fit and healthy”), investment and self-presentation (e.g., “When I upload photos of myself I usually use filters or alter/change them to make myself look better”), social comparison (e.g., “I feel like I want to change my diet after viewing other people’s pictures online”), active versus passive (e.g., “I approve photos of myself before anyone can tag them” vs. “I am often dissatisfied with my weight or looks in my social media pictures”), and negative responses (e.g., “Seeing pictures of others makes me feel down on myself”). To the creation and wording of such items, both literature on social media use and existing measures related to sociocultural and body image theories of social grooming (Dunbar, 1996; Tufekci, 2008), social comparison (Festinger, 1954; Wheeler & Miyake, 1992), self-presentation (Goffman, 1959; Leary, 1996; Manning, 1992), and self-objectification (Frederickson & Roberts, 1997; Noll & Fredrickson, 1998) were considered. Additionally, attention was also given to represent a mix of body image concerns such as weight and diet, fitness, health and muscularity, as well as general appearance in the wording of the 21 items. Furthermore, 10 additional items were included as filler items that asked about general social media use (seven items), personal trolling behavior (one item), and interactions with friends online (two items). Response options ranged from 1 (strongly disagree) to 7 (strongly agree). Higher scores indicated the tendency to participate in these online activities and to experience stronger feelings associated with these online interactions. Results Item reduction and factor analyses The 21 items were evaluated to assess whether they met two assumptions of exploratory factor analysis (Hair, Black, Babin, Anderson, & Tatham, 2006). First, Bartlett’s test of sphericity was significant, χ2(210) = 4,594.3, p < .001, indicating an acceptable number of significant correlations among variables. Second, the Kaiser Meyer Olkin Measure of Sampling Adequacy for the overall sample was good (.93). The exploratory factor analysis was conducted using principle axis factoring (PAF) with oblique rotation. PAF was used given that this method is recommended for psychological data because it allows for measurement error (Xia, Green, Xu, & Thompson, 2019). In line with best practices, the number of factors to extract was based on Velicer’s minimum average parcel (MAP) test and parallel analysis (Hayton, Allen, & Scarpello, 2004; O’Connor, 2000). When we conducted a PAF, three factors were extracted with eigenvalues over 1 (10.63, 2.02, and 1.46) and Velicer’s MAP test and parallel analysis supported the extraction of three factors, with all three eigenvalues greater than the first three eigenvalues calculated using parallel analysis (O’Connor, 2000). The variance accounted for in the items was 67.20%. As shown in Table 1, the pattern matrix showed loadings above .5 for all but two items and no high cross-loadings for all items, with the exception of one item. Thus, we removed this cross-loading item and removed the two items with loadings below .50 and the PAF was repeated. appearance teasing experienced and witnessed, and general social media use. Method Participants and procedure The participants were 327 university students aged 17 to 25 years (M = 20.1, SD = 1.1; 68% female; note: Two participants were just under age 17). To measure sociocultural background, participants were allowed to endorse as many options as applied, with most (84%) of the participants endorsing white Australian, 12% instead or in addition endorsing Asian, 3% endorsing Australian first peoples/Torres Strait Islander/Pacific Islander, and 9% describing a diverse range of other backgrounds. Participants also reported on their mother and father educational levels: Mother’s education: 21% did not graduate high school, 15% graduated high school, 22% completed vocational training, 39% undertook some university study. Father’s education: 16% did not graduate high school, 24% graduated high school, 22% completed vocational training, 38% undertook some university study. Overall 344 students attempted the questionnaire, but 15 participants were removed for missing data on all measures (i.e., they only completed one or two pages of the questionnaire) and two were removed because they were over age 25. The upper age limit was 25, given that most university students are below this age and because social media use and patterns can change as individuals get older (Olson, O’Brien, Rogers, & Charness, 2012). Measures Social media appearance preoccupation The final 18-item SMAPS from Study 1 was used to measure social appearance online self-presentation (seven items), appearance-related activity (five items), and appearance comparison (six items). Responses options ranged from 1 (strongly disagree) to 7 (strongly agree). See the results section for further details. Depressive symptoms Depressive symptoms were assessed using the Mood and Feelings Questionnaire—Short Version (Angold & Costello, 1987), consisting of a series of 13 descriptive phrases about how the adolescent has been feeling or behaving recently (e.g., “I felt miserable or unhappy”). Items are rated from 1 (not true) to 5 (very true), and the total score was calculated by averaging all items. Higher scores indicate more depressive symptoms, Cronbach’s α = .95. Social anxiety The Social Anxiety Scale for Adolescents (La Greca & Lopez, 1998) assessed symptoms of social anxiety. Eighteen descriptive items (e.g., “I worry about doing something new in front of others”) were rated on a 5-point scale from 1 (not true) to 5 (very true). A total score was calculated by averaging all items, with higher scores indicating higher social anxiety symptoms, Cronbach’s α = .95. Appearance anxiety symptoms The Appearance Anxiety Inventory (Roberts et al., 2018; Veale et al., 2014) was used to measure symptoms of body image anxiety. The Appearance Anxiety Inventory is a 10-item scale (e.g., “I try to camouflage or alter aspects of my appearance”). Participants indicated on a 5-point scale the frequency with which they experienced symptoms 0 (never) to 4 (always or almost always). The total score was formed by summing all items, where higher scores reflected greater appearance symptoms, Cronbach’s α = .93. Disordered eating and related behaviors Six items from the Eating Attitudes Test-26 (Garner, Olmsted, Bohr, & Garfinkel, 1982) were used to measure disordered eating and related behaviors in adolescents. Questions related to eating binges, vomiting, laxative and diet pills, excessive exercise, use of pills or powders to control muscle mass and skipping meals (e.g., “In the past 6 months have you. . . . Gone on eating binges where you feel that you may not be able to stop eating?”). Participants responded on a scale measuring the frequency of these behaviors that ranged from 1 (never) to 6 (once a day or more). A composite score was formed by averaging the item responses. Higher scores indicate more disordered eating, Cronbach’s α = .74. General interpersonal stress Using items from the Responses to Stress Questionnaire (Connor-Smith, Compas, Wadsworth, Thomsen, & Saltzman, 2000), participants reported their personal experience with 10 interpersonal stressors in the past 6 months (e.g., “Being around people who are rude or mean; Having problems with a friend”). Responses ranged from 1 (not at all) to 4 (very much). The items were averaged to form a composite interpersonal stress score, Cronbach’s α was. .84. Coping flexibility One six-item scale from Self-Perceived Flexible Coping with Stress (Zimmer-Gembeck et al., 2018) was used to measure the perceived capacity to use multiple coping strategies when facing stressful events (e.g., “I can come up with lots of ways to make myself feel better if I am stressed”). Response options ranged from 1 (not at all true) to 7 (totally true). A composite was formed by averaging the items, Cronbach’s α was. .84. Sexual harassment Sexual harassment was measured using five items similar to the 14-item American Association of University Women (AAUW) Education Foundation questionnaire (AAUW, 2001). To produce five items, AAUW items were condensed so that much of the content of the 14 items was contained in the five items. However, one item deemed too sensitive (e.g., sexual assault) was removed. Four items gathered reports of verbal and/or nonphysical sexual harassment experiences (“made sexual comments, jokes, gestures or looks at you”; “showed, given, or left you sexual pictures, photos, notes or messages”; “written sexual graffiti about you on a wall or other place”; “spread sexual rumours about you”) and one item related to physical sexual harassment (“flashed you some sexual part of their body”; “touched, grabbed, or pinched you in a sexual way”; or “grabbed you or pulled at your clothing in a sexual way”). Response options ranged from 1 (never) to 5 (very often). Items were averaged so that a higher score indicated greater variety and frequency of sexual harassment, Cronbach’s α was .81. Social media appearance teasing experienced and witnessed One item, with two parts, based on the Perceptions of Teasing Scale (Thompson, Cattarin, Fowler, & Fisher, 1995), assessed frequency of appearance-related teasing from the same- and other-sex (“In the past year, how often have you been teased about the way you look on social media?”). The two parts involved reporting about teasing by the same-sex separately from other-sex peers. A second item with two parts asked about witnessing online appearance-related teasing by the same- and the other-sex (“In the past year, how often have you witnessed but not taken part in teasing on social media?”) Responses ranged from 1 (never) to 5 (very often). Items were averaged to form a composite to indicate online appearance-related teasing separate from a composite of witnessing online appearance-related teasing, Cronbach’s α was .85 for online appearance-related teasing and .93 for witnessing online appearance-related victimization. Appearance-related support from others Support about appearance from important others was measured with a four-item scale adapted from the Important Other Climate Questionnaire (Williams et al., 2006) and the Body Acceptance by Others Scale (Avalos & Tylka, 2006) and was used to assess participants’ perceptions of the support they receive from important others in their lives when they feel bad about their appearance (e.g., “Important people in my life make me feel important regardless of how I look”). Participants responded on a scale ranging from 1 (never or very rarely true) to 5 (very often or always true), where the total score was formed using the average of the items. Higher scores indicate greater perceived support from significant others, Cronbach’s α = .92. Social media use Four items from the Facebook Intensity Scale (Ellison, Steinfield, & Lampe, 2007) were modified (to focus on social media rather than Facebook specifically) to measure emotional connectedness and integration of social media use in daily life (e.g., “Using social media is part of my everyday activity”). Response options ranged from 1 (strongly disagree) to 5 (strongly agree), and items were averaged to produce a total use score, Cronbach’s α = .87. Results Confirmatory factor analyses The confirmatory factor analyses of the 18-item SMAPS was conducted using MPlus with robust maximum likelihood estimation. The three latent factors of Online Self-Presentation, Appearance-Related Activity, and Appearance Comparison were free to covary with each other. The results supported the three-factor structure. As shown in Table 2, all loadings were .60 or above, with one exception. After freeing four covariances between errors, the model had an adequate fit to the data, χ2(128) = 323.60, p < .001, comparative fit index (CFI) = .96, root mean square error of approximation (RMSEA) = .068 (90% confidence interval [CI; .059, .078], p = .001). The factor structure was invariant by gender. More specifically, when the loadings of all items on the three factors were freed to differ (as well as the covariances between factors) for young women and men, the model had an adequate fit to the data, χ2(260) = 511.51, p < .001, CFI = .94, RMSEA = .055 (90% CI [.048, .062], p = .139). However, when all loadings were fixed to gender equality and the fit of these two models were compared, there was no significant difference in the model fits, χdiff2(132) = 187.91, p > .05. Thus, there was no indication that the loadings or the covariances between factors differed significantly for young women and men. For example, the correlation between online self-presentation and appearance comparison was r = .78 for young women and r = .76 for young men. The correlation between online self-presentation and appearance-related activity was r = .41 for young women and r = .45 for young men, and the correlation between appearance comparison and appearance-related activity was r = .39 for young women and r = .43 for young men.
Finally, given the high correlation between online self-presentation and appearance comparison, we also fit a two-factor model with all online self-presentation and appearance comparison items freed to load only on one factor, whereas other items were freed to load only on a second factor of appearance-related online activity. The fit of this model was poor, χ2(130) = 624.23, p < .001, CFI = .89, RMSEA = .108 (90% CI [.100, .117], p < .001). Top of Form Discussion Many adolescents and young adults report a great deal of concern about their appearance, including being judged by others or not conforming to societal ideals (Fardouly & Vartanian, 2015; Holland & Tiggemann, 2016; Ricciardelli & Yager, 2016; Webb & Zimmer-Gembeck, 2014). Such concerns can result in excessive appearance anxiety, body dysmorphia, body dissatisfaction, or disordered eating, all of which can interfere with tasks of daily living, health, and happiness (Roberts et al., 2018; Veale et al., 2014; Zimmer-Gembeck, Webb, Farrell, & Waters, 2018). This preoccupation with appearance may be fuelled even more today by the excessive social comparison and image displays that are a basic part of using social media. Despite this apparent appearance preoccupation and the evidence that it is tied to historical change in social media platforms and their use (Saunders & Eaton, 2018; Seabrook et al., 2016), we could locate no standard measure to assess appearance-related social media use and preoccupation, which could be applied across a range of contemporary social media platforms while also being founded in social theories. Such a measure was needed, as it may be appearance-related use and preoccupation with social media self-presentation and comparisons that will help to identify adolescents and young adults at the highest risk of appearance-related clinical disorders and other related problems. Thus, in this study, a new measure, the SMAPS, was developed and tested in high school (Study 1) and university (Study 2) students. Once developed and confirmed, the SMAPS subscales were validated against a range of measures of social media use and appearance sensitivities, and the subscales were also examined as correlates of emotional maladjustment. Attention was also given to gender differences and invariance, age associations, and the possible of interactions of SMAPs subscales with other social media measures when predicting emotional and appearance-related adjustment problems. Three SMAPS subscales were found in exploratory analyses with high school students and were confirmed in a second study with university students. The SMAPS subscales included Online Self-Presentation, Appearance-Related Online Activity, and Appearance Comparison. These subscales, which were assessed with only 18 items, will allow for consideration of general appearance-related social media use separate from preoccupation in the forms of spending time engaged in self-presentational activities and comparing oneself to others online. As would be expected, appearance-related online activity had moderate correlations with self-presentation and appearance comparison, whereas self-presentation and appearance comparison were more highly correlated with each other. Nevertheless, analyses supported three rather than two factors. Moreover, providing evidence of the validity of all three SMAPS subscales, appearance-related activity was associated with appearance-related problems including disordered eating, but had only small associations with general emotional adjustment (social anxiety and depression). The other two SMAPS subscales tapping preoccupation seemed to generate more risk across a range of problems including appearance-related anxiety, disordered eating, engaging in and witnessing social media teasing about appearance, as well as having moderate positive associations with both general social anxiety and depressive symptoms. In addition, multivariate models showed that SMAPS appearance-related activity and appearance comparison had incremental validity when the outcomes were social anxiety, depression, appearance anxiety, and disordered eating; they had unique associations with all of these outcomes even after considering a range of other social media and appearance-related behaviors, as well as stress, coping, and sexual harassment. Our analyses of gender shows that the SMAPS can be successfully used with both young men and women, with invariance found for the three-subscale factor structure in Study 2. In addition, age tended to show nonsignificant associations with the three SMAPS subscales, across the age range of 13 to 18 years in Study 1 and age 17 to 25 years in Study 2. Yet, as is usually found in studies of body image and appearance (Ricciardelli & Yager, 2016; Webb et al., 2017; Webb & Zimmer-Gembeck, 2014), young women (both high school students and university students) did endorse more appearance-related social media activity and more preoccupation with online self-presentation relative to young men. Young women also engaged in more appearance comparison than young men. In fact, young women scored about 0.5 points higher on activity and about 1 point higher (on a 7-point scale) than young men for self-presentation and appearance comparison. However, rarely were the SMAPS subscale scores more strongly associated with emotional adjustment in one gender relative to the other. In fact, we found only two interaction effects, with both supporting the possibility that greater preoccupation with self-presentation is more strongly associated with emotional maladjustment in young men relative to young women. Given that some research has identified that females are more likely than males to receive emotional support online (from their same-sex friends; Joiner et al., 2014, 2016), it may be that preoccupation with self-presentation is more problematic for young men because they do not get the same level of support received by young women online, which could serve to balance or temper the negative effect of preoccupation. Overall, the findings support the use of the three SMAPS subscales as separate aspects of appearance-related social media use and preoccupation to examine their unique correlations with other behaviors, attitudes, or disorders in youth. Yet, we also showed in the current analyses that they may interact with other existing measures, focusing here on interactions with general level of social media use, to better explain variation in adjustment among youth. In particular, we found that the general level of social media use interacted with all three SMAPS subscales to better explain why some youth reported more elevated appearance anxiety symptoms than others. The consideration of other interaction effects (i.e., between existing measures and the SMAPS subscales) will be a useful addition to future research. For example, future research could examine risk factors for appearance-related preoccupation, to identify whether preexisting vulnerabilities might explain both social media preoccupation and symptom development. Also important, developmental research is needed focusing on social media use and appearance-related preoccupation, ideally beginning in late childhood and continuing all the way throughout adolescence and into young adulthood to capture data within the age periods when body image and eating disorders first onset for both women and men (McCabe & Ricciardelli, 2004; Menzel et al., 2010; Pearson et al., 2017; Sharpe et al., 2018). There is one primary limitation of the present research to note here. The participants in the two studies were Australian young people with the majority endorsing White or Asian race/ethnicity. This may limit the generalizability of the findings. It will be important to test the SMAPS with groups of adolescents and young adults from other countries and from a diversity of other background. Despite the possibility of limited generalizability, the findings provide access to a new measure of appearance-related use and preoccupation, the SMAPS. The SMAPS has three subscales relevant to understanding emotional adjustment and social experiences of young men and young women across the ages of 12 to 25 years. Also, scores derived from the SMAPS were found to be both reliable and valid measure. Future research could build on the findings here to focus on development of appearance-related concerns and disordered eating as potentially related to social media use. References American Association of University Women. (2001). Hostile hallways: Bullying, teasing, and sexual harassment in school. American Journal of Health Education, 32, 307–309. 10.1080/19325037.2001.10603488 Angold, A., & Costello, E. J. (1987). Mood and Feelings Questionnaire. Durham, NC: Duke University. Australian Media Communication Authority. (2013). Like, post, share: Young Australians’ experience of social media. Retrieved from [https://www.acma.gov.au/~/media/mediacomms/Report/pdf/Like%20post%20share%20Young%20Australians%20experience%20of%20social%20media%20Quantitative%20research%20report ] Avalos, L. C., & Tylka, T. L. (2006). Exploring a model of intuitive eating with college women. Journal of Counseling Psychology, 53, 486–497. 10.1037/0022-0167.53.4.486 Bandura, A. (1962). Social learning through imitation. In M. R.Jones (Ed.), Nebraska symposium on motivation (pp. 211–274). Lincoln, NE: University of Nebraska Press. Brown, Z., & Tiggemann, M. (2016). Attractive celebrity and peer images on Instagram: Effect on women’s mood and body image. Body Image, 19, 37–43. 10.1016/j.bodyim.2016.08.007 Burke, M., Marlow, C., & Lento, T. (2010). Social network activity and social well-being. Postgraduate Medical Journal, 85, 455–459. Choukas-Bradley, S., Nesi, J., Widman, L., & Higgins, M. K. (2019). Camera-ready: Young women’s appearance-related social media consciousness. Psychology of Popular Media Culture, 8, 473–481. 10.1037/ppm0000196 Cohen, R., Newton-John, T., & Slater, A. (2017). The relationship between Facebook and Instagram appearance-focused activities and body image concerns in young women. Body Image, 23, 183–187. 10.1016/j.bodyim.2017.10.002 Cohen, R., Newton-John, T., & Slater, A. (2018). ‘Selfie’-objectification: The role of selfies in self-objectification and disordered eating in young women. Computers in Human Behavior, 79, 68–74. 10.1016/j.chb.2017.10.027 Connor-Smith, J. K., Compas, B. E., Wadsworth, M. E., Thomsen, A. H., & Saltzman, H. (2000). Responses to stress in adolescence: Measurement of coping and involuntary stress responses. Journal of Consulting and Clinical Psychology, 68, 976–992. 10.1037/0022-006X.68.6.976 Dijkstra, J. K., Cillessen, A. H. N., & Borch, C. (2013). Popularity and adolescent friendship networks: Selection and influence dynamics. Developmental Psychology, 49, 1242–1252. 10.1037/a0030098 Dunbar, R. I. M. (1996). Grooming, gossip, and the evolution of language. Cambridge, MA: Harvard University Press. Ellison, N. B., Steinfield, C., & Lampe, C. (2007). The benefits of Facebook “friends”: social capital and college students’ use of online social network sites. Journal of Computer-Mediated Communication, 12, 1143–1168. 10.1111/j.1083-6101.2007.00367.x Fairburn, C. G., Cooper, Z., & O’Connor, M. (2008). Eating disorder examination. In C. G.Fairburn (Ed.), Cognitive behaviour therapy and eating disorders (pp. 265–308). New York, NY: Guilford Press. Fardouly, J., Magson, N. R., Johnco, C. J., Oar, E. L., & Rapee, R. M. (2018). Parental control of the time preadolescents spend on social media: Links with preadolescents’ social media appearance comparisons and mental health. Journal of Youth and Adolescence, 47, 1456–1468. 10.1007/s10964-018-0870-1 Fardouly, J., & Vartanian, L. R. (2015). Negative comparisons about one’s appearance mediate the relationship between Facebook usage and body image concerns. Body Image, 12, 82–88. 10.1016/j.bodyim.2014.10.004 Fardouly, J., & Vartanian, L. R. (2016). Social media and body image concerns: Current research and current directions. Current Opinion in Psychology, 9, 1–5. 10.1016/j.copsyc.2015.09.005 Feinstein, B. A., Hershenberg, R., Bhatia, V., Latack, J. A., Meuwly, N., & Davila, J. (2013). Negative social comparison on Facebook and depressive symptoms: Rumination as a mechanism. Psychology of Popular Media Culture, 2, 161–170. 10.1037/a0033111 Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7, 117–140. 10.1177/001872675400700202 Frederickson, B. L., & Roberts, T. A. (1997). Objectification theory: Toward understanding women’s lived experiences and mental health risks. Psychology of Women Quarterly, 21, 173–206. 10.1111/j.1471-6402.1997.tb00108.x Garner, D. M. (2004). Eating Disorder Inventory-3. Professional manual. Lutz, FL: Psychological Assessment Resources, Inc. Garner, D. M., Olmsted, M. P., Bohr, Y., & Garfinkel, P. E. (1982). The eating attitudes test: Psychometric features and clinical correlates. Psychological Medicine, 12, 871–878. 10.1017/S0033291700049163 Gerson, J., Plagnol, A. C., & Corr, P. J. (2017). Passive and Active Facebook Use measure (PAUM): Validation and relationship to reinforcement sensitivity theory. Personality and Individual Differences, 117, 81–90. 10.1016/j.paid.2017.05.034 Goffman, E. (1959). The presentation of self in everyday life. Garden City, NY: Doubleday. Haferkamp, N., Eimler, S. C., Papadakis, A. M., & Kruck, J. V. (2012). Men are from Mars, women are from Venus? Examining gender differences in self-presentation on social networking sites. Cyberpsychology, Behavior, and Social Networking, 15, 91–98. 10.1089/cyber.2011.0151 Hair, J., Black, W., Babin, B., Anderson, R., & Tatham, R. (2006). Multivariate data analysis (6th ed.). Upper Saddle River, NJ: Pearson Prentice Hall. Hayton, J. C., Allen, D. G., & Scarpello, V. (2004). Factor retention decisions in exploratory factor analysis: A tutorial on parallel analysis. Organizational Research Methods, 7, 191–205. 10.1177/1094428104263675 Hogue, J. V., & Mills, J. S. (2019). The effects of active social media engagement with peers on body image in young women. Body Image, 28, 1–5. 10.1016/j.bodyim.2018.11.002 Holland, G., & Tiggemann, M. (2016). A systematic review of the impact of the use of social networking sites on body image and disordered eating outcomes. Body Image, 17, 100–110. 10.1016/j.bodyim.2016.02.008 Holland, G., & Tiggemann, M. (2017). “Strong beats skinny every time”: Disordered eating and compulsive exercise in women who post fitspiration on Instagram. International Journal of Eating Disorders, 50, 76–79. 10.1002/eat.22559 Joiner, R., Brosnan, M., Duffield, J., Gavin, J., & Maras, P. (2007). The relationship between Internet identification, Internet anxiety and Internet use. Computers in Human Behavior, 23, 1408–1420. 10.1016/j.chb.2005.03.002 Joiner, R., Cuprinskaite, J., Dapkeviciute, L., Johnson, H., Gavin, J., & Brosnan, M. (2016). Gender differences in response to Facebook status updates from same and opposite gender friends. Computers in Human Behavior, 58, 407–412. 10.1016/j.chb.2016.01.008 Joiner, R., Stewart, C., Beaney, C., Moon, A., Maras, P., Guiller, J., . . .Brosnan, M. (2014). Publically different, privately the same: Gender differences and similarities in response to Facebook status updates. Computers in Human Behavior, 39, 165–169. 10.1016/j.chb.2014.07.004 Kim, J. W., & Chock, T. M. (2015). Body image 2.0: Associations between social grooming on Facebook and body image concerns. Computers in Human Behavior, 48, 331–339. 10.1016/j.chb.2015.01.009 Kim, M., & Park, W. (2016). Who is at risk on Facebook? The effects of Facebook news feed photographs on female college students’ appearance satisfaction. The Social Science Journal, 53, 427–434. 10.1016/j.soscij.2016.08.007 Kleemans, M., Daalmans, S., Carbaat, I., & Anschütz, D. (2018). Picture perfect: The direct effect of manipulated Instagram photos on body image in adolescent girls. Media Psychology, 21, 93–110. 10.1080/15213269.2016.1257392 La Greca, A. M., & Lopez, N. (1998). Social anxiety among adolescents: Linkages with peer relations and friendships. Journal of Abnormal Child Psychology, 26, 83–94. 10.1023/A:1022684520514 Leary, M. R. (1996). Self-presentation: Impression management and interpersonal behavior. Boulder, CO: Westview. Lonergan, A. R., Bussey, K., Mond, J., Brown, O., Griffiths, S., Murray, S. B., & Mitchison, D. (2019). Me, my selfie, and I: The relationship between editing and posting selfies and body dissatisfaction in men and women. Body Image, 28, 39–43. 10.1016/j.bodyim.2018.12.001 Lup, K., Trub, L., & Rosenthal, L. (2015). Instagram #instasad?: Exploring associations among instagram use, depressive symptoms, negative social comparison, and strangers followed. Cyberpsychology, Behavior, and Social Networking, 18, 247–252. 10.1089/cyber.2014.0560 Manning, P. (1992). Erving Goffman and modern sociology. Cambridge, United Kingdom: Polity Press. Marengo, D., Longobardi, C., Fabris, M. A., & Settanni, M. (2018). Highly-visual social media and internalizing symptoms in adolescence: The mediating role of body image concerns. Computers in Human Behavior, 82, 63–69. 10.1016/j.chb.2018.01.003 McCabe, M. P., & Ricciardelli, L. A. (2004). Weight and shape concerns of boys and men. In J. K.Thompson (Ed.), Handbook of eating disorders and obesity (pp. 606–634). Hoboken, NJ: Wiley. McLean, S. A., Paxton, S. J., Wertheim, E. H., & Masters, J. (2015). Photoshopping the selfie: Self photo editing and photo investment are associated with body dissatisfaction in adolescent girls. International Journal of Eating Disorders, 48, 1132–1140. 10.1002/eat.22449 Meier, E. P., & Gray, J. (2014). Facebook photo activity associated with body image disturbance in adolescent girls. Cyberpsychology, Behavior, and Social Networking, 17, 199–206. 10.1089/cyber.2013.0305 Melioli, T., Rodgers, R. F., Rodrigues, M., & Chabrol, H. (2015). The role of body image in the relationship between internet use and bulimic symptoms: Three theoretical frameworks. Cyberpsychology, Behavior, and Social Networking, 18, 682–686. 10.1089/cyber.2015.0154 Menzel, J. E., Schaefer, L. M., Burke, N. L., Mayhew, L. L., Brannick, M. T., & Thompson, J. K. (2010). Appearance-related teasing, body dissatisfaction, and disordered eating: A meta-analysis. Body Image, 7, 261–270. 10.1016/j.bodyim.2010.05.004 Moreno, M. A., Kota, R., Schoohs, S., & Whitehill, J. M. (2013). The Facebook influence model: A concept mapping approach. Cyberpsychology, Behavior, and Social Networking, 16, 504–511. 10.1089/cyber.2013.0025 Noll, S. M., & Fredrickson, B. L. (1998). A mediational model linking self-objectification, body shame, and disordered eating. Psychology of Women Quarterly, 22, 623–636. 10.1111/j.1471-6402.1998.tb00181.x O’Connor, B. P. (2000). SPSS and SAS programs for determining the number of components using parallel analysis and Velicer’s MAP test. Behavior Research Methods, Instruments, and Computers, 32, 396–402. 10.3758/BF03200807 Olson, K. E., O’Brien, M. A., Rogers, W. A., & Charness, N. (2012). Diffusion of technology: Frequency of use for younger and older adults. Ageing International, 36, 13–145. 10.1007/s12126-010-9077-9 Pearson, C. M., Miller, J., Ackard, D. M., Loth, K. A., Wall, M. M., Haynos, A. F., & Neumark-Sztainer, D. (2017). Stability and change in patterns of eating disorder symptoms from adolescence to young adulthood. International Journal of Eating Disorders, 50, 748–757. 10.1002/eat.22692 Pew Research Center. (2016). Social media update 2016. Washington, DC: Author. Retrieved from https://www.pewinternet.org/2016/11/11/social-media-update-2016/ Rajan, B. (2018). A picture is worth a thousand words: A study of Millenial women’s curation of self on Facebook profile pictures. Comunicación y Genéro, 2, 173–191. 10.5209/CGEN.62674 Ricciardelli, L. A., & Yager, Z. (2016). Adolescence and body image: From developmental to preventing dissatisfaction. New York, NY: Routledge. Roberts, C., Zimmer-Gembeck, M. J., Lavell, C., Gregertsen, E. C., Miyamoto, T., & Farrell, L. J. (2018). The Appearance Anxiety Inventory: Factor structure and associations with appearance-based rejection sensitivity and social anxiety. Journal of Obsessive-Compulsive and Related Disorders, 19, 124–130. 10.1016/j.jocrd.2018.10.004 Rodgers, R. F., Faure, K., & Chabrol, H. (2009). Gender differences in parental influences on adolescent body dissatisfaction and disordered eating. Sex Roles, 61, 837–849. 10.1007/s11199-009-9690-9 Rodgers, R. F., & Melioli, T. (2016). The relationship between body image concerns, eating disorders and Internet use, Part I: A review of empirical support. Adolescent Research Review, 1, 95–119. 10.1007/s40894-015-0016-6 Rutledge, C. M., Gillmor, K. L., & Gillen, M. M. (2013). Does this profile picture make me look fat? Facebook and body image in college students. Psychology of Popular Media Culture, 2, 251–258. 10.1037/ppm0000011 Saunders, J. F., & Eaton, A. A. (2018). Snaps, selfies, and shares: How three popular social media platforms contribute to the sociocultural model of disordered eating among young women. Cyberpsychology, Behavior, and Social Networking, 21, 343–354. 10.1089/cyber.2017.0713 Seabrook, E. M., Kern, M. L., & Rickard, N. S. (2016). Social networking sites, depression, and anxiety: A systematic review. JMIR Mental Health, 3, e50. 10.2196/mental.5842 Sensis. (2017). Sensis social media report. Retrieved from https://www.sensis.com.au/socialmediareport Sharpe, H., Griffiths, S., Choo, T. H., Eisenberg, M. E., Mitchison, D., Wall, M., & Neumark-Sztainer, D. (2018). The relative importance of dissatisfaction, overvaluation and preoccupation with weight and shape for predicting onset of disordered eating behaviors and depressive symptoms over 15 years. International Journal of Eating Disorders, 51, 1168–1175. 10.1002/eat.22936 Smith, A. R., Hames, J. L., & Joiner, T. E., Jr. (2013). Status update: Maladaptive Facebook usage predicts increases in body dissatisfaction and bulimic symptoms. Journal of Affective Disorders, 149, 235–240. 10.1016/j.jad.2013.01.032 Steers, M. N., Wickham, R. E., & Acitelli, L. K. (2014). Seeing everyone else’s highlight reels: How Facebook usage is linked to depressive symptoms. Journal of Social and Clinical Psychology, 33, 701–731. 10.1521/jscp.2014.33.8.701 Tajfel, H. (1982). Social psychology of intergroup relations. Annual Review of Psychology, 33, 1–39. 10.1146/annurev.ps.33.020182.000245 Talbot, C. V., Gavin, J., van Steen, T., & Morey, Y. (2017). A content analysis of thinspiration, fitspiration, and bonespiration imagery on social media. Journal of Eating Disorders, 5, 40. 10.1186/s40337-017-0170-2 Thompson, J. K., Cattarin, J., Fowler, B., & Fisher, E. (1995). The Perception of Teasing Scale (POTS): A revision and extension of the physical appearance related teasing scale (PARTS). Journal of Personality Assessment, 65, 146–157. 10.1207/s15327752jpa6501_11 Thompson, J. K., Heinberg, L. J., Altabe, M., & Tantleff-Dunn, S. (1999). The scope of body image disturbance: The big picture. Washington, DC: American Psychological Association. Tiggemann, M., & Miller, J. (2010). The Internet and adolescent girls’ weight satisfaction and drive for thinness. Sex Roles, 63, 79–90. 10.1007/s11199-010-9789-z Tufekci, Z. (2008). Grooming, Gossip, Facebook and MySpace: What can we learn about these sites from those who won’t assimilate?Information Communication and Society, 11, 544–564. 10.1080/13691180801999050 Uhlmann, L. R., Donovan, C. L., Zimmer-Gembeck, M. J., Bell, H. S., & Ramme, R. A. (2018). The fit beauty ideal: A healthy alternative to thinness or a wolf in sheep’s clothing?Body Image, 25, 23–30. 10.1016/j.bodyim.2018.01.005 Veale, D., Eshkevari, E., Kanakam, N., Ellison, N., Costa, A., & Werner, T. (2014). The Appearance Anxiety Inventory: Validation of a process measure in the treatment of body dysmorphic disorder. Behavioural and Cognitive Psychotherapy, 42, 605–616. 10.1017/S1352465813000556 Webb, H., & Zimmer-Gembeck, M. J. (2014). The role of friends and peers in adolescent body dissatisfaction: A review and critique of 15 years of research. Journal of Research on Adolescence, 24, 564–590. 10.1111/jora.12084 Webb, H. J., Zimmer-Gembeck, M. J., Farrell, L., Waters, A., Nesdale, D., & Downey, G. (2017). “Pretty pressure” from peers, parents, and the media: A longitudinal study of appearance-based rejection sensitivity. Journal of Research on Adolescence, 27, 718–735. 10.1111/jora.12310 Webb, H. J., Zimmer-Gembeck, M. J., & Mastro, S. (2016). Stress exposure and generation: A conjoint longitudinal model of body dysmorphic symptoms, peer acceptance, popularity, and victimization. Body Image, 18, 14–18. 10.1016/j.bodyim.2016.04.010 Wheeler, L., & Miyake, K. (1992). Social comparison in everyday life. Journal of Personality and Social Psychology, 62, 760–773. 10.1037/0022-3514.62.5.760 Williams, G. C., Lynch, M. F., McGregor, H. A., Ryan, R. M., Sharp, D., & Deci, E. L. (2006). Validation of the “Important Other” Climate Questionnaire: Assessing autonomy support for health-related change. Families, Systems and Health, 24, 179–194. 10.1037/1091-7527.24.2.179 Xia, Y., Green, S. B., Xu, Y., & Thompson, M. S. (2019). Proportion of indicator common variance due to a factor as an effect size statistic in revised parallel analysis. Educational and Psychological Measurement, 79, 85–107. 10.1177/0013164418754611 Zimmer-Gembeck, M. J., Skinner, E. A., Modecki, K., Webb, H. J., Rapee, R. M., Hawes, T., & Gardner, A. (2018). The self-perception of flexible coping with stress: A new measure and relations with emotional adjustment. Cogent Psychology. Advance online publication. 10.1080/23311908.2018.1537908 Zimmer-Gembeck, M. J., & Webb, H. J. (2017). Body image and peer relationships: Unique associations of adolescents’ social status and competence with peer- and self-reported appearance victimization. Journal of Adolescence, 61, 131–140. 10.1016/j.adolescence.2017.10.002 Zimmer-Gembeck, M. J., Webb, H. J., Farrell, L. J., & Waters, A. M. (2018). Girls’ and boys’ trajectories of appearance anxiety from age 10 to 15 years are associated with earlier maturation and appearance-related teasing. Development and Psychopathology, 30, 337–350. 10.1017/S0954579417000657   Bottom of Form Journals/journal 10 Self-Construal Moderates Age Differences in Social Network Characteristics Dannii Y. Yeung and Helene H. Fung Chinese University of Hong Kong Frieder R. Lang Friedrich-Alexander-University Erlangen-Nuremberg In this study, the authors examined age differences in social network characteristics (SNC) among Hong Kong Chinese. The sample consisted of 596 Chinese adults, ranging from 18 to 91 years old. Age was positively associated with close social partners and negatively associated with peripheral social partners. For individuals who were more likely to define the self as interconnected with others (i.e., interdependent self-construal), increasing age was associated with a greater number of close social partners. The negative association between age and the number of peripheral social partners, well-documented in the Western literature, was found only among Chinese adults with lower interdependence but not among those with higher interdependence. These findings highlight the importance of examining the underlying mechanism rather than a particular pattern of SNC across cultures. Keywords: social network, socioemotional selectivity theory, interdependence, adulthood, culture Age differences in social network characteristics (SNC) have been widely reported in the Western literature (e.g., Ajrouch, Antonucci, & Janevic, 2001; Carstensen, 1992; Fung, Carstensen, & Lang, 2001; Lang & Carstensen, 1994, 2002). Increasing age was associated with fewer peripheral social partners (peripheral partners), whereas the number of emotionally close social partners (close partners) remained relatively stable across age. Socioemo- tional selectivity theory (Carstensen, Issacowitz, & Charles, 1999) explains these age differences in SNC in terms of older people’s greater desire to seek emotionally meaningful social interactions when they perceive future time as being increasingly limited (Fung & Carstensen, 2004; Lang & Carstensen, 2002). In this brief report, we describe a study in which we examined whether indi- vidual differences in interdependent self-construal may make peo- ple hold different perceptions about what is regarded as emotion- ally meaningful social interactions and thus exhibit different patterns of age-related SNC. Age Differences in SNC Age-related differences in SNC have been widely documented in both cross-sectional (Ajrouch et al., 2001; Fung et al., 2001; Lang & Carstensen, 1994; 2002; Lang, Staudinger, & Carstensen, 1998) and longitudinal studies (Carstensen, 1992; Field & Min- kler, 1988; Lang, 2000; van Tilburg, 1998) conducted among Americans and Europeans. Compared with younger people, older people have fewer peripheral partners but maintain a similar num- ber of close partners in their social networks. Older people’s social networks thus consist of a greater proportion of close partners than do younger people’s. Socioemotional selectivity theory (Carstensen et al., 1999) explains these age differences in SNC in motivational terms. It argues that goals for social interaction change as a function of future time per- ception. Younger people perceive their future time as being relatively more expansive and they prefer interacting with social partners of greater diversity to fulfill their future-oriented goals. Their social networks thus comprise largely peripheral partners. When individuals grow older, however, they perceive their future as being increasingly limited. They shift their attention from long-term future-oriented goals to short-term emotional goals. As a result, they tend to interact with social partners who can best provide them with emotionally mean- ingful experiences. Their social networks thus mainly consist of close partners—such as family members and close friends— but fewer peripheral partners (Carstensen, Gross, & Fung, 1997; Lang & Carstensen, 1994). Longitudinal studies conducted among Germans (Lang, 2000) and Americans (Carstensen, 1992) further demonstrated that changes in network size were mainly due to the reduction in the number of peripheral partners. In particular, when older adults per- ceived their life was coming to an end, they deliberately discontinued interactions with peripheral partners while maintaining the interac- tions with close partners such as family members, relatives, and close friends. Self-Construal To the extent that what people seek under future time limitation is emotionally meaningful social relationships (Fung & Carstensen, 2004), then individual differences in what is considered to be emo- tionally meaningful may lead to different patterns of social relation- ships across adulthood. Self-construal, defined as how one perceives oneself in relation to other people, may be a good determinant of what is considered to be emotionally meaningful in social situations. Ac- Dannii Y. Yeung and Helene H. Fung, Department of Psychology, Chinese University of Hong Kong, Hong Kong, China; Frieder R. Lang, Institute of Psychogerontology, Friedrich-Alexander-University Erlangen- Nuremberg, Germany. The study was supported by Hong Kong Research Grants Council Earmarked Research Grant CUHK4256/03H and a Chinese University of Hong Kong direct grant awarded to Helene Fung. Correspondence concerning this article should be addressed to Helene H. Fung, Department of Psychology, Chinese University of Hong Kong, Room 328 Sino Building, Chung Chi College, Shatin, New Territories, Hong Kong. E-mail: hhlfung@psy.cuhk.edu.hk Psychology and Aging Copyright 2008 by the American Psychological Association 2008, Vol. 23, No. 1, 222–226 0882-7974/08/$12.00 DOI: 10.1037/0882-7974.23.1.222 222 T hi s do cu m en t i s co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . T hi s ar tic le is in te nd ed s ol el y fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . cording to the self-construal theory (Markus & Kitayama, 1991), people with high levels of interdependent self-construal define the self as embedded in ingroups and interconnected with others. Such defi- nitions of self can affect social judgments (Kemmelmeier & Oyser- man, 2001) and guides behaviors (Gudykunst & Lee, 2003; Markus & Kitayama, 1991), including social behaviors (Malikiosi-Loizos & Anderson, 1999; Takahashi, Ohara, Antonucci, & Akiyama, 2002). They may also affect age-related SNC. In particular, those with a higher level of interdependent self- construal may be more likely to maintain interactions with social partners of greater diversity even when they grow older. These individuals may attach great importance to taking care of family members and relatives (i.e., greater familism; Szalay, Strohl, Fu, & Lao, 1994; Yang, 1988) and to maintaining reciprocal relation- ships with all social partners, even peripheral ones (i.e., Renqing or relationship orientation; Cheung et al., 2001; Yeung, Fung, & Lang, 2007; Zhang & Yang, 1998). They may thus be more likely to maintain or even increase the number of close social partners and be less likely to reduce the number of peripheral social partners as they grow older. Present Study Most of the studies on age-related SNC reviewed above were conducted in Western cultures such as North America and Germany. People in these Western cultures have reliably been found to be less interdependent (Hofstede, 1980; Markus & Kitayama, 1991; Oyser- man, Coon, & Kemmelmeier, 2002; Triandis, 1995) than are East Asians. Such differences tend to intensify with age (Fung & Ng, 2006). In this study, we examined the moderating role of interdepen- dent self-construal1 in age-related SNC among Hong Kong Chinese, an East Asian culture with great diversity in self-construal. Hong Kong people, under the influence of their Chinese heritage, emphasize interdependence to a greater degree than do North Americans and Europeans (e.g., Markus & Kitayama, 1991; Oyserman et al., 2000). Yet, much individual difference in self-construal exists within the culture (Gudykunst & Lee, 2003), owing to Hong Kong’s status as one of the top international financial centers and its 150-year history as a British colony (Hong, Morris, Chiu, & Benet-Martinez, 2000). This variability in interdependent self-construal allowed us to test whether individual differences in self-construal would moderate age differences in SNC. We predicted that Hong Kong Chinese with lower levels of interdependent self-construal would show a pattern of age-related SNC similar to that found among Westerners in the prior literature, that is, increasing age was associated with fewer peripheral partners but a similar number of close partners. Yet, among those with higher levels of interdependent self-construal, a different pattern of age-related SNC would be observed: Increasing age would be asso- ciated with more close partners and a weaker reduction in the number of peripheral partners. Method Participants The sample included 596 participants who were between 18 and 91 years old (M � 42.26 years, SD � 19.33), including 219 men and 377 women. Twelve percent of participants had no formal education, 15.1% had primary education, 30% attained secondary education, and over 40% attained tertiary education or above. Almost half of the participants were married, and the remaining participants were single (38.6%), divorced (2.4%), or widowed (12.6%). Procedure Participants were recruited by convenience sampling from uni- versities, community centers, and other public places and stratified by age group (young, middle age, and old) and gender. Although such a sampling procedure might have yielded a sample that was not as representative as those in prior studies that used probability sampling (e.g., Ajrouch et al., 2001; Lang & Carstensen, 2002), the procedure is consistent with that used in an American study that sampled across the entire adulthood (e.g., Fung et al., 2001). When we compared the demographic characteristics of this convenience sample with those of the general population (United Nations Statistics Division, 2007a, 2007b), we found that this sample was generally representative of the population except that participants were better educated and more likely to be widowed. Consent for participation was first obtained, followed by the Wechsler Digit Symbol Test (Wechsler, 1983). Those who dem- onstrated cognitive difficulties in completing the test did not continue the study. The rest completed the following measures. Measures SNC. The Social Convey Questionnaire (Kahn & Antonucci, 1980) was used to measure SNC. Participants nominated social partners to one of three circles that surrounded the word I. The inner circle indicated very close social partners that “one cannot imagine life without.” The middle and outer circles indicated rather close social partners and less close social partners, respec- tively. The number of partners nominated to the inner circle indicates the number of close partners, whereas the number nom- inated to the other two circles indicates the number of peripheral partners. Self-construal. Interdependent self-construal was measured by the subscale of Self Construal Scales (Gudykunst et al., 1996). It consists of 14 items assessing the level of interdependence. Sample items include being willing to stay in a group if one’s participation is needed even if one is not happy about it and consulting others before making important decisions. Participants rated these items on a 7-point scale, with higher scores indicating higher levels of interdependent self-construal. The reliability coefficient for this scale was .83. 1 We originally measured both interdependence and independence be- cause the two types of self-construal were believed to be independent constructs (i.e., orthogonal factors; Markus & Kitayama, 1991; Singelis, 1994). But we eventually excluded independence from all the analyses reported in this article because (a) interdependence and independence were found to be moderately correlated (r � .40) in this sample; (b) preliminary analyses revealed that there was no interaction effect between interdepen- dence and independence on the number of close or peripheral social partners; and (c) independence did not moderate age differences in the number of either close or peripheral social partners, and this lack of moderating effect is consistent with findings from prior studies (Kem- melmeier & Oyserman, 2001; Poon & Fung, in press) in which the moderating role of independence in social psychological phenomena was examined. 223BRIEF REPORTS T hi s do cu m en t i s co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . T hi s ar tic le is in te nd ed s ol el y fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . Demographic, health, and cognitive variables. Age, gender, education level, and marital status were recorded. Perceived health was measured by the Wahler Physical Symptoms Inventory (Wahler, 1983). Participants reported whether they had any of 42 physical symptoms on a 6-point scale ranging from 1 (none) to 6 (almost every day). A mean score was computed, with a higher score indicating more perceived health symptoms. The mean num- ber of perceived health symptoms was 1.94 (SD � 0.53). In addition, as mentioned above, all participants completed the Wechsler Digit Symbol Test (Wechsler, 1983), both as a screening tool and as a measure of cognitive ability. In the test, participants matched symbols with digits in a 90-s period. The total number of matches successfully made during the time period was taken as a rough estimate of nonverbal cognitive ability. Results The mean network size was 13.31 (SD � 7.20), the mean number of close partners was 4.83 (SD � 3.06), and the mean number of peripheral partners was 8.49 (SD � 5.82). Statistically controlling for gender, education level, perceived health, and cog- nitive ability did not affect the results we report below. Moderated regression analyses were conducted to examine the moderating effects of interdependent self-construal on age differ- ences in the number of close social partners and the number of peripheral social partners. Age and interdependence were entered (as continuous variables) in Block 1, and the Age � Interdepen- dence interaction was entered in Block 2. The interaction term was computed after standardizing each variable. Results from Block 1 revealed that, similar to prior Western findings (e.g., Germans, Lang & Carstensen, 1994; Americans, Fung et al., 2001), age was negatively associated with the number of peripheral social partners, B � �0.098, SE � 0.012, � � �.327, p � .001. However, although these Western findings did not reveal any age differences in the number of close social partners, our results showed that age was positively associated with the number of close social partners, B � 0.030, SE � 0.006, � � .189, p � .001. Results from Block 2 revealed a significant Age � Interdependence interaction on close partners, B � 0.379, SE � 0.137, � � .110, �R2 � .012, �F(1, 592) � 7.638, p � .01, and on peripheral partners, B � 0.926, SE � 0.256, � � .141, �R2 � .019, �F(1, 592) � 13.110, p � .001 (see Table 1). To illustrate the significant Age � Interdependence interaction effects on SNC, interdependence was divided into the high, me- dium, and low levels by a tercile split.2 Age was positively associated with the number of close partners among participants with medium and high levels of interdependence, B � 0.034, p � .01, and B � 0.051, p � .001, respectively, but not among those with a low level of interdependence, B � �0.008, ns. Moreover, age was negatively associated with the number of peripheral partners among participants with low and medium levels of inter- dependence, Bs � �0.170 and �0.119, respectively, ps � .001, but not among those with a high level of interdependence, B � �0.046, ns (see Figure 1). Discussion In this study, we tested whether interdependent self-construal moderated age differences in SNC among Hong Kong Chinese. We predicted that those with lower interdependent self-construal would show a pattern of age-related SNC similar to that found among Germans (e.g., Lang & Carstensen, 1994; Lang et al., 1998) and Americans (e.g., Fung et al., 2001) in the prior literature, that is, increasing age would be associated with fewer peripheral part- ners but the same number of close partners. But those with higher interdependent self-construal might attach greater importance to emotionally close social partners as these social partners are the ingroup (Szalay et al., 1994; Yang, 1988). They might thus be more likely to maintain or even increase their number of close social partners with age. Moreover, these individuals might also feel more obligated to maintain reciprocal relationships with all 2 The patterns of results were similar regardless of whether interdepen- dent self-construal (a continuous variable) was split up by a tercile split, by a median split, or by the 1 standard deviation above the mean, mean, 1 standard deviation below the mean division as recommended by Aiken and West (1991). We chose to use a tercile split because such a split was less likely to result in misclassification than was a median split (MacCallum, Zhang, Preacher, & Rucker, 2002), and it has been successfully used in a prior study on a similar topic (Lang & Carstensen, 2002). Table 1 Moderated Regression Analyses of the Number of Close and Peripheral Social Partners Number of close social partners Number of peripheral social partners Unstandardized coefficients Standardized coefficients R2 �R2 Unstandardized coefficients Standardized coefficients R2 �R2B SE �s B SE � Block 1 .095 .095*** .122 .122*** Age 0.030 0.006 .189*** �0.098 0.012 �.327*** Interdependence 0.808 0.169 .195*** 1.955 0.317 .248*** Block 2 .106 .012** .141 .019*** Age 0.028 0.006 .178*** �0.103 0.012 �.341*** Interdependence 0.743 0.170 .179*** 1.796 0.317 .228*** Age � Interdependence 0.379 0.137 .110** 0.926 0.256 .141*** Note. ** p � .01. ***p � .001. 224 BRIEF REPORTS T hi s do cu m en t i s co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . T hi s ar tic le is in te nd ed s ol el y fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . social partners, even peripheral ones (Cheung at al., 2001; Fung & Ng, 2006; Yeung et al., 2007; Zhang & Yang, 1998) and thus find it harder to drop peripheral social partners from their social net- works with age. Findings confirmed our predictions regarding the moderating role of interdependent self-construal. The stability of the number of emo- tionally close social partners across age, typically found in Western studies (e.g., Fung et al., 2001; Lang & Carstensen, 1994), was replicated only among Hong Kong Chinese with a low level of interdependent self-construal. In contrast, those with medium and high levels of interdependent self-construal exhibited a positive asso- ciation between age and the number of emotionally close social partners. A consistent pattern was also found for the association between age and the number of peripheral social partners. Although a negative association between age and the number of peripheral social partners was observed for the entire sample, the association was significant only among those with low and medium levels of interdependent self-construal. The association was much weaker and, in fact, no longer significant among those with a high inter- dependent self-construal. Taken together, these findings suggest that the pattern of age differences in SNC typically found in prior literature (e.g., Fung et al., 2001; Lang & Carstensen, 1994; Lang et al., 1998) may not be universal. Increasing age may be associated with stability in the number of emotionally close social partners but a reduction in the number of peripheral social partners only among individuals who are less likely to see themselves as embedded in social groups. For those who do, aging is associated with an increase in the number of emotionally close social partners and a much smaller reduction in the number of peripheral social partners. The basic premise of socioemotional selectivity theory (Carstensen et al., 1999) is that people seek emotionally meaningful social relationships when they perceive future time as being increasingly limited (Fung & Carstensen, 2004). Our findings extend the theory by pointing out that to the extent that cultures or individuals differ in what they consider to be emotionally meaningful, the actual patterns of age differences in social relationships may vary. A few limitations should be considered when interpreting our findings. First, the findings are cross-sectional and the sample included participants from a heterogeneous age range of 18 to 91 years. We could not rule out possible cohort effects on self- construal or SNC, nor could our findings address research ques- tions such as potential developmental transformation from periph- eral relationships into close relationships over time. Second, the convenience sampling limits the generalizability of our findings. In future studies, researchers should investigate the phenomena in longitudinal studies with more representative samples from diverse backgrounds. Such longitudinal studies would also allow closer examinations of developmental changes in perceived closeness over time. Finally, in this study, we only examined age-related patterns of SNC among Hong Kong Chinese, without a direct comparison with Western samples. It is possible that people from different cultures may hold different definitions of close and peripheral social partners, which might have contributed to the observed differences in age-related SNC between Hong Kong Chinese and Western samples. In future studies, researchers should directly compare both the patterns and the definitions of age- related social relationships between East Asian and Western cul- tures to address these issues. Despite such limitations, our findings contribute to the literature by showing that the socioemotional selectivity phenomenon may not be defined by a particular pattern of SNC. Those who place greater emphasis on interdependence may regard all types of social partners as emotionally meaningful and thus are more likely to increase the number of emotionally close social partners and are less likely to reduce the number of peripheral social partners with age. These findings highlight the importance of studying the mechanism rather than a particular pattern of socioemotional aging across cultures. 10080604020 Age 10.00 8.00 6.00 4.00 2.00 0.00 N o . o f E m o ti o n a ll y C lo s e S o c ia l P a rt n e rs High Interdependence Medium Interdependence Low Interdependence B = -.008 B = .034* B = .051*** 10080604020 Age 18.00 15.00 12.00 9.00 6.00 3.00 0.00 N o . o f P e ri p h e ra l S o c ia l P a rt n e rs High Interdependence Medium Interdependence Low Interdependence B = -.170*** B = -.119*** B = -.046 Figure 1. The relationship between age and close social partners and between age and peripheral social partners as a function of interdependence. B � unstandardized regression coefficient. * p � .05. *** p � .001. 225BRIEF REPORTS T hi s do cu m en t i s co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . T hi s ar tic le is in te nd ed s ol el y fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . References Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage. Ajrouch, K. J., Antonucci, T. C., & Janevic, M. R. (2001). Social networks among Blacks and Whites: The interaction between race and age. Jour- nals of Gerontology: Psychological Sciences and Social Sciences, 56B, S112–S118. Carstensen, L. L. (1992). Social and emotional patterns in adulthood: Support for socioemotional selectivity theory. Psychology and Aging, 7, 331–338. Carstensen, L. L., Gross, J. J., & Fung, H. H. (1997). The social context of emotional experience. In K. W. Schaie & M. P. Lawton (Eds.), Annual review of gerontology and geriatrics: Vol. 17. Focus on emotion and adult development (pp. 325–352). New York: Springer. Carstensen, L. L., Isaacowitz, D., & Charles, S. T. (1999). Taking time seriously: A theory of socioemotional selectivity. American Psycholo- gist, 54, 165–181. Cheung, F. M., Leung, K., Zhang, J. X., Sun, H. F., Gan, Y. I., Song, W. Z., & Xie, D. (2001). Indigenous Chinese personality constructs: Is the five-factor model complete? Journal of Cross-Cultural Psychology, 32, 407– 433. Field, D., & Minkler, M. (1988). Continuity and change in social support between young-old, old-old, and very-old adults. Journal of Gerontol- ogy, 43, P100 –P106. Fung, H. H., & Carstensen, L. L. (2004). Motivational changes in response to blocked goals and foreshortened time: Testing alternatives for socio- emotional selectivity theory. Psychology and Aging, 19, 68 –78. Fung, H. H., Carstensen, L. L., & Lang, F. R. (2001). Age-related patterns in social networks among European Americans and African Americans: Implications for socioemotional selectivity across the life span. Interna- tional Journal of Aging and Human Development, 52, 185–206. Fung, H. H., & Ng, S. K. (2006). Age differences in personality among Canadians and Hong Kong Chinese: Personality maturation may differ across cultures. Psychology and Aging, 21, 810 – 814. Gudykunst, W. B., & Lee, C. M. (2003). Assessing the validity of self construal scales: A response to Levine et al. Human Communication Research, 29, 253–274. Gudykunst, W. B., Matsumoto, Y., Ting-Toomey, S., Nishida, T., Kim, K., & Heyman, S. (1996). The influence of cultural individualism– collectivism, self construals, and individual values on communication styles across cultures. Human Communication Research, 22, 510 –543. Hofstede, G. (1980). Culture consequences: International differences in work-related values. Beverly Hills, CA: Sage. Hong, Y. Y., Morris, M. W., Chiu, C.-Y., & Benet-Martinez, V. (2000). Multicultural minds: A dynamic constructivist approach to culture and cognition. American Psychologist, 55, 709 –720. Kahn, R. L., & Antonucci, T. C. (1980). Convoys over the life course: Attachment, roles and social support. In P. B. Baltes & O. G. Brim (Eds.), Life-span development and behavior (pp. 254 –283). San Diego, CA: Academic Press. Kemmelmeier, M., & Oyserman, D. (2001). The ups and downs of thinking about a successful other: Self-construals and the consequences of social comparisons. European Journal of Psychology, 31, 311–320. Lang, F. R. (2000). Endings and continuity of social relationships: Maxi- mizing intrinsic benefits within personal networks when feeling near to death. Journal of Social and Personal Relationships, 17, 155–182. Lang, F. R., & Carstensen, L. L. (1994). Close emotional relationships in late life: Further support for proactive aging in the social domain. Psychology and Aging, 9, 315–324. Lang, F. R., & Carstensen, L. L. (2002). Time counts: Future time per- spective, goals, and social relationships. Psychology and Aging, 17, 125–139. Lang, F. R., Staudinger, U. M., & Carstensen, L. L. (1998). Perspectives on socioemotional selectivity in late life: How personality and social con- text do (and do not) make a difference. Journals of Gerontology: Psychological Sciences and Social Sciences, 53B, P21–P30. MacCallum, R. C., Zhang, S., Preacher, K. J., & Rucker, D. D. (2002). On the practice of dichotomization of quantitative variables. Psychological Methods, 7, 19 – 40. Malikiosi-Loizos, M., & Anderson, L. R. (1999). Accessible friendships, inclusive friendships, reciprocated friendships as related to social and emotional loneliness in Greece and the USA. European Psychologist, 4, 165–178. Markus, H. R., & Kitayama, S. (1991). Culture and the self: Implications for cognition, emotion, and motivation. Psychological Review, 98, 224 –253. Oyserman, D., Coon, H. M., & Kemmelmeier, M. (2002). Rethinking individualism and collectivism: Evaluation of theoretical assumptions and meta-analyses. Psychological Bulletin, 128, 3–72. Poon, C. Y. M., & Fung, H. H. (in press). Physical activity and psycho- logical well-being among Hong Kong Chinese older adults: Exploring the moderating role of self-construals. International Journal of Aging and Human Development. Singelis, T. M. (1994). The measurement of independent and interdepen- dent self-construals. Personality and Social Psychology Bulletin, 20, 580 –591. Szalay, L. B., Strohl, J. B., Fu, L., & Lao, P. S. (1994). American and Chinese perceptions and belief systems: A People’s Republic of China– Taiwanese comparison. New York: Plenum Press. Takahashi, K., Ohara, N., Antonucci, T. C., & Akiyama, H. (2002). Common- alities and differences in close relationships among the Americans and Japanese: A comparison by the individualism/collectivism concept. Inter- national Journal of Behavioral Development, 26, 453– 465. Triandis, H. C. (1995). Individualism and collectivism. Boulder, CO: Westview Press. United Nations Statistics Division. (2007a). Education enrolment by level. Retrieved June 29, 2007, from http://unstats.un.org/unsd/cdb/ cdb_series_xrxx.asp?series_code�25540 United Nations Statistics Division. (2007b). Marital status of population by sex, age group, urban and rural. Retrieved June 29, 2007, from http:// unstats.un.org/unsd/cdb/cdb_series_xrxx.asp?series_code�14880 van Tilburg, T. (1998). Losing and gaining in old age: Changes in personal network size and social support in a four-year longitudinal study. Jour- nals of Gerontology: Psychological Sciences and Social Sciences, 53B, S313–S323. Wahler, H. J. (1983). Wahler Physical Symptoms Inventory manual. Los Angeles: Western Psychological Services. Wechsler, D. (1983). The psychometric tradition: Developing the Wechsler Adult Intelligence Scale. Contemporary Educational Psychology, 6, 82– 85. Yang, C. F. (1988). Familism and development: An examination of the role of family in contemporary China mainland, Hong Kong and Taiwan. In D. Sinha & H. S. R. Kao (Eds.), Social values and development: Asian perspectives (pp. 93–123). Thousand Oaks, CA: Sage. Yeung, D. Y., Fung, H. H., & Lang, F. R. (2007). Gender differences in social network characteristics and psychological well-being among Hong Kong Chinese: The role of future time perspective and adherence to renqing. Aging & Mental Health, 11, 45–56. Zhang, Z., & Yang, C.-F. (1998). Beyond distributive justice: The reason- ableness norm in Chinese reward allocation. Asian Journal of Social Psychology, 1, 253–269. Received June 14, 2007 Revision received October 24, 2007 Accepted November 1, 2007 � 226 BRIEF REPORTS T hi s do cu m en t i s co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . T hi s ar tic le is in te nd ed s ol el y fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . Journals/Journal 2 Journal of Personality and Social Psychology When Every Day Is a High School Reunion: Social Media Comparisons and Self-Esteem Claire Midgley, Sabrina Thai, Penelope Lockwood, Chloe Kovacheff, and Elizabeth Page-Gould Online First Publication, August 13, 2020. http://dx.doi.org/10.1037/pspi0000336 CITATION Midgley, C., Thai, S., Lockwood, P., Kovacheff, C., & Page-Gould, E. (2020, August 13). When Every Day Is a High School Reunion: Social Media Comparisons and Self-Esteem. Journal of Personality and Social Psychology. Advance online publication. http://dx.doi.org/10.1037/pspi0000336 When Every Day Is a High School Reunion: Social Media Comparisons and Self-Esteem Claire Midgley University of Toronto Sabrina Thai Brock University Penelope Lockwood, Chloe Kovacheff, and Elizabeth Page-Gould University of Toronto Although past research has shown that social comparisons made through social media contribute to negative outcomes, little is known about the nature of these comparisons (domains, direction, and extremity), variables that determine comparison outcomes (post valence, perceiver’s self-esteem), and how these comparisons differ from those made in other contexts (e.g., text messages, face-to-face interactions). In 4 studies (N � 798), we provide the first comprehensive analysis of how individuals make and respond to social comparisons on social media, using comparisons made in real-time while browsing news feeds (Study 1), experimenter-generated comparisons (Study 2), and comparisons made on social media versus in other contexts (Studies 3 and 4). More frequent and more extreme upward comparisons resulted in immediate declines in self-evaluations as well as cumulative negative effects on individuals’ state self-esteem, mood, and life satisfaction after a social media browsing session. Moreover, downward and lateral comparisons occurred less frequently and did little to mitigate upward comparisons’ negative effects. Furthermore, low self-esteem individuals were particularly vulnerable to making more frequent and more extreme upward comparisons on social media, which in turn threatened their already-lower self-evaluations. Finally, social media comparisons resulted in greater declines in self-evaluations than those made in other contexts. Together, these studies provide the first insights into the cumulative impact of multiple comparisons, clarify the role of self-esteem in online comparison processes, and demonstrate how the characteristics and impact of comparisons on social media differ from those made in other contexts. Keywords: social comparisons, self-esteem, social media, Facebook, Instagram Supplemental materials: http://dx.doi.org/10.1037/pspi0000336.supp In just over a decade, social media use has skyrocketed. In 2005, only 5% of Americans reported using one or more social media platforms; by 2019, this number had risen to 72% (Pew Research Center, 2019). Furthermore, the majority of Facebook, Instagram, Snapchat, and YouTube users visit these sites at least once per day, contributing to a global average of over 2 hr per day spent on social media per person (Clement, 2020). Although social media can enhance social connection (Ellison, Steinfield, & Lampe, 2007; Liu, Ainsworth, & Baumeister, 2016) and provide opportu- nities for self-disclosure and perceived social support (Davis, 2012; Ko & Kuo, 2009), the preponderance of research indicates that social media use is associated with negative outcomes, such as envy, romantic jealousy, decreased self-esteem and subjective well-being, increased loneliness and social isolation, and depres- sion (Burke, Marlow, & Lento, 2010; Hwang, Cheong, & Feeley, 2009; Kalpidou, Costin, & Morris, 2011; Krasnova, Wenninger, Widjaja, & Buxmann, 2013; Kross et al., 2013; Muise, Christofides, & Desmarais, 2009; Tandoc, Ferrucci, & Duffy, Editor’s Note. David Dunning served as the action editor for this arti- cle.—KK X Claire Midgley, Department of Psychology, University of Toronto; Sabrina Thai, Department of Psychology, Brock University; Penelope Lockwood, Chloe Kovacheff, and Elizabeth Page-Gould, Department of Psychology, University of Toronto. Chloe Kovacheff is now at the Rotman School of Management, Univer- sity of Toronto. This research was supported by a Social Sciences and Humanities Research Council of Canada (SSHRC) Postdoctoral Fellowship to Sabrina Thai and an SSHRC Insight Grant to Penelope Lockwood. Portions of the present research were presented at the 18th Annual Meeting of the Society for Personality and Social Psychology, San Antonio, Texas and the Psychology of Media & Technology Preconference at the 20th Annual Meeting of the Society for Personality and Social Psychology, Portland, Oregon. We thank Laksmiina Balasubramaniam and Ariana Youm for their assistance with data collection. We also thank Joanne Wood, Brett Q. Ford, Geoff MacDonald, and Emily Impett for their helpful comments and suggestions on an earlier version of this article. Data are available by emailing the authors. Study materials, syntax for analyses, and supplementary materials are available on Open Science Frame- work: https://osf.io/kn49t/. Correspondence concerning this article should be addressed to Claire Midgley, Department of Psychology, University of Toronto, 100 George Street, Toronto, ON M5S 3G3, Canada. E-mail: claire.midgley@mail .utoronto.ca T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. Journal of Personality and Social Psychology: Interpersonal Relations and Group Processes © 2020 American Psychological Association 2020, Vol. 2, No. 999, 000 ISSN: 0022-3514 http://dx.doi.org/10.1037/pspi0000336 1 https://orcid.org/0000-0001-6680-2944 http://dx.doi.org/10.1037/pspi0000336.supp https://osf.io/kn49t/ mailto:claire.midgley@mail.utoronto.ca mailto:claire.midgley@mail.utoronto.ca http://dx.doi.org/10.1037/pspi0000336 2015; Verduyn et al., 2015; Valkenburg, Peter, & Schouten, 2006; Vogel, Rose, Roberts, & Eckles, 2014; Woods & Scott, 2016; for reviews, see Best, Manktelow, & Taylor, 2014 and Verduyn, Ybarra, Résibois, Jonides, & Kross, 2017). Despite these negative associations, however, social media use continues to grow (Clem- ent, 2020; Pew Research Center, 2019); thus, it is important to understand when and how social media will result in negative outcomes, and for whom these negative outcomes will be most significant. Social Media Is Associated With Threatening Social Comparisons A growing body of research suggests that social media exerts a negative impact on users through social comparison processes: Individuals see that others on social media appear to be experi- encing more positive outcomes, and consequently feel worse about themselves. Indeed, a number of studies point to associations between Facebook use, upward comparisons, and negative out- comes. Heavy users, in contrast to infrequent users, are more likely to agree that others are happier, have better lives, and are doing better (Chou & Edge, 2012; de Vries & Kühne, 2015). Further- more, making more upward Facebook comparisons has been as- sociated with negative self-perceptions of one’s own social com- petence and attractiveness, increased depressive symptoms, and lower overall well-being (Appel, Crusius, & Gerlach, 2015; de Vries & Kühne, 2015; Fardouly & Vartanian, 2015; Feinstein et al., 2013; Gerson, Plagnol, & Corr, 2016; Liu et al., 2017; Steers, Wickham, & Acitelli, 2014; Tandoc et al., 2015; Vogel et al., 2014; Wang, Wang, Gaskin, & Hawk, 2017). These negative effects, moreover, seem especially pronounced for low self-esteem individuals (Cramer, Song, & Drent, 2016; Jang, Park, & Song, 2016). This past research, however, does not provide clear evi- dence that social comparison is responsible for negative social media outcomes (Appel, Gerlach, & Crusius, 2016). Because these studies relied primarily on retrospective reports, it may be that individuals who are experiencing negative outcomes in these do- mains are simply more likely to recall or report on comparisons with superior others. It is unclear, moreover, which characteristics of the social media context lead to particularly negative social comparison outcomes. Because past studies showing the connection between social media and social comparison have focused exclusively on social media contexts (e.g., Appel et al., 2015), it remains unclear how these processes may differ from other, non-social-media contexts. Pre- sumably, the cognitive mechanisms underlying social comparison will be similar regardless of whether one is exposed to a superior other on social media, another online context, or in real life, with upward comparisons typically leading to threats to self-esteem and diminished mood and life satisfaction (Gerber, Wheeler, & Suls, 2018). One may feel worse if one learns about a friend’s superior academic performance regardless of whether one hears about the friend’s success through social media or face-to-face interaction. Why then, are social media comparisons associated with especially negative outcomes? In the present studies, we identify specific attributes of social media comparisons that are especially damaging to the self. Through examining comparisons in real-time in lab studies, as well as in studies using both experimental and experience-sampling designs, we make three key contributions to the literature: First, we show that social media provides more opportunities for individuals to make comparisons, in particular to superior others; individuals make more frequent upward comparisons when using social media than in other contexts. Second, because social media posts tend to be highly positive, individuals make comparisons that are more extreme in their “upwardness” than in other contexts. This greater frequency and extremity of upward comparisons results in a par- ticularly negative impact of social media use on the self. Third, we advance the literature on social comparison and self-esteem by showing that low self-esteem individuals are especially likely to make more frequent and extreme upward comparisons, which in turn leads to a more negative impact on their self-evaluations; we further examine whether social media amplifies these negative outcomes relative to other contexts. Social Media and Upward Comparison Frequency Social media provides a continuous stream of information about other people’s accomplishments. Past research suggests that social comparisons occur automatically (Chatard, Bocage-Barthélémy, Selimbegović, & Guimond, 2017; Gilbert, Giesler, & Morris, 1995; Mussweiler, Rüter, & Epstude, 2004); when individuals encounter information about another person, their own self- perceptions will be affected. The sheer number of posts in a news feed, each offering a thumbnail sketch about another person, would seem likely to yield numerous comparison opportunities. Furthermore, to the extent that social media posts are positive, they are most likely to yield upward comparisons, resulting in negative outcomes for the self (Gerber et al., 2018). Indeed, evidence suggests that news feed content is predominantly about positive experiences. Although people do not typically post false information about themselves online (Back et al., 2010), they do engage in selectively positive self-presentation (Walther, 2007; Wilson, Gosling, & Graham, 2012) and are more likely to post positive rather than negative content (e.g., Dorethy, Fiebert, & Warren, 2014; Qiu, Lin, Leung, & Tov, 2012; Seidman, 2013). As a result, individuals browsing their news feeds are more likely to see posts about friends’ exciting social activities than dull days at the office, affording numerous opportunities for comparisons to seemingly better-off others. To date, research has not directly tested whether social media exerts a negative impact on the self by eliciting more frequent upward comparisons than in other, non-social-media contexts. A number of studies suggest that individuals who are especially prone to making comparisons experience negative outcomes when using social media (Alfasi, 2019; de Vries, Möller, Wieringa, Eigenraam, & Hamelink, 2018; Hanna et al., 2017; Stapleton, Luiz, & Chatwin, 2017; Vogel, Rose, Okdie, Eckles, & Franz, 2015; Wang et al., 2017). These studies, however, did not explic- itly measure comparison frequency, instead assessing whether an orientation to make comparisons more generally (Gibbons & Buunk, 1999) or comparison tendency on a particular platform (e.g., “Today, when I was on Facebook, I felt less confident about what I have achieved compared with other people”; Steers et al., 2014) might be related to social media use outcomes. Such global reports may not accurately reflect the degree to which individuals were engaging in actual comparison activities (Gerber et al., 2018) and may instead reflect individuals’ lay theories regarding social T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 2 MIDGLEY, THAI, LOCKWOOD, KOVACHEFF, AND PAGE-GOULD comparisons on social media. That is, individuals’ perceptions of how often comparisons are occurring may not match how often they actually make comparisons on social media (Cramer et al., 2016). Moreover, although recent research suggests that individuals with a greater orientation toward making social comparisons ex- perience worse outcomes after using social media (Lee, 2014; Vogel et al., 2015; Wang et al., 2017), research conducted before the advent of social media suggests that these individuals experi- ence similarly negative outcomes, such as increased depressive symptoms and social anxiety, in nonsocial-media contexts (Gib- bons & Buunk, 1999). Consequently, it is unclear whether frequent social media comparisons are associated with more negative out- comes, or whether individuals with a propensity to compare them- selves experience more negative outcomes in any context. One experience-sampling study that did measure comparison frequency demonstrated that physical appearance comparisons made on so- cial media were actually less frequent than comparisons resulting from in-person encounters (Fardouly, Pinkus, & Vartanian, 2017). Given this study’s focus on physical appearance, however, it is unclear how social media comparison frequency predicts compar- ison outcomes more broadly. In the present research, we examined whether upward comparisons would be more frequent on social media than in other contexts, and whether this greater frequency in turn would be associated with more negative social media use outcomes. Social Media and Upward Comparison Extremity We propose that social media will exert a negative impact not only by eliciting a greater number of comparisons to superior others, but also by prompting individuals to make comparisons that are especially “upward.” According to the selective accessibility model (Mussweiler, 2003; Mussweiler & Strack, 1999), individu- als first assess whether they are similar or dissimilar to a superior other at a holistic level; they then go on to test their specific hypothesis either that they are similar or dissimilar to the target. Social media posts are especially likely to highlight large discrep- ancies between the comparer and the poster, leading to a holistic assessment of dissimilarity, and a subsequent test for evidence of dissimilarity. The characteristics of social media will yield partic- ularly compelling evidence that other people are very superior to the self. Posters tend to focus on particularly positive events, showcasing a “newsreel highlight” of their lives (Steers et al., 2014; Zhao, Grasmuck, & Martin, 2008); for example, individuals do not merely post albums of vacation photos, but rather carefully choose, and often digitally enhance, a few select photos that indicate that their vacation was spectacular (Lo & McKercher, 2015). Moreover, such posts are likely to present a strong contrast with the immediate experiences of post viewers, who are, by definition, staring at an electronic device, often while engaged in mundane activities (Tien & Aynsley, 2019). Thus, we argue that comparisons on social media are often to superior others who are not simply better off than the self, but rather who appear to be much better off than the self. These more extreme comparisons may in turn be particularly threatening: To the extent that individ- uals perceive these highly positive outcomes to be unattainable, they will feel worse about themselves (Lockwood & Kunda, 1997). Consistent with this argument, one study on physical ap- pearance comparisons found that women made more extreme upward comparisons with social media targets than to in-person targets, and subsequently evaluated their attractiveness more neg- atively (Fardouly et al., 2017). It is unclear, however, whether this effect extends to domains beyond physical appearance. In sum- mary, the present studies are the first to assess whether individuals make more extreme upward comparisons when using social media than in other contexts, with a negative impact on self-evaluations, mood, and life satisfaction. Self-Esteem and Social Media Comparisons The frequency and extremity of upward comparisons may also explain why social media exerts an especially negative impact on individuals with low self-esteem. To the extent that low self- esteem individuals chronically view themselves as of lower worth than other individuals, it seems likely that they will be especially prone to view successful others as upward comparison targets. Consistent with this possibility, a number of studies indicate that lower self-esteem is indeed associated with more frequent upward comparisons (Locke & Nekich, 2000; Vohs & Heatherton, 2004; Wayment & Taylor, 1995; for a review, see Wood & Lockwood, 1999). Other studies suggest, however, that low self-esteem indi- viduals suffer not because they find many instances in which they are inferior to others but because they find few instances in which they are superior to others (Locke, 2005; Wheeler & Miyake, 1992). Thus, the literature on self-esteem and upward comparison frequency in offline contexts is mixed. Research focused specifi- cally on social media contexts has found that individuals lower in self-esteem report a greater tendency to make Facebook compar- isons in general (Cramer et al., 2016; Jang et al., 2016); however, these studies did not directly examine the role of self-esteem in predicting the actual frequency of upward relative to downward social media comparisons, nor did they compare frequency of social comparisons across different contexts. For low self-esteem individuals, social media offers especially fertile ground for making upward comparisons. Given their nega- tive self-perceptions, they may find it difficult to construe them- selves as being “in the same league” as someone experiencing success (Collins, 1996); as a result, social media posts about even modestly positive achievements may be perceived as upward com- parisons. Thus, although low self-esteem individuals will make more upward comparisons than will higher self-esteem individuals in general, this difference in frequency may be amplified in social media contexts, where opportunities for comparisons abound, re- sulting in especially negative consequences for the self. In addition, to the extent that low self-esteem individuals chron- ically view themselves as having lower self-worth than others, social media posts may yield especially extreme comparisons. Low self-esteem individuals may be habitually likely to make a holistic evaluation that they are dissimilar from successful others. Consis- tent with the selective accessibility model (Mussweiler, 2003), this initial judgment will trigger a search for evidence that they are inferior to successful others. The exaggerated positivity of social media posts (Walther, 2007; Wilson et al., 2012), in combination with the already impoverished self-evaluations of low self-esteem individuals, should lead to perceptions of an especially large discrepancy between the self and the poster—and a more extreme upward comparison. Consistent with this possibility, one study T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 3SOCIAL MEDIA COMPARISONS AND SELF-ESTEEM found that participants with more contingent self-esteem were more likely to make more extreme upward comparisons and sub- sequently experience worse mental health outcomes; however, this research examined only physical appearance comparisons in non- social-media contexts, and focused on contingent rather than low self-esteem (Patrick, Neighbors, & Knee, 2004). We argue that although individuals in general will make more extreme social comparisons when using social media relative to other contexts, this will be especially true of low self-esteem individuals; conse- quently, they will experience the greatest reduction in their self- evaluations, mood, and life satisfaction after social media use. The Present Research In summary, the present research provides the first evidence identifying the characteristics of social comparisons on social media— heightened frequency and extremity—that differ from social comparisons in other contexts. We also provide the first evidence that these particular characteristics in turn contribute to social media’s harmful effects on the self, particularly for individ- uals lower in self-esteem. In Study 1, we examined low and high self-esteem individuals’ actual comparison behavior in real time by assessing their reactions to posts in their own social media news feeds. This enabled us to assess the direction, extremity, and impact of each comparison made, as well as the cumulative effects of these comparisons at the end of the browsing session. In Study 2, we experimentally manipulated post content; this enabled us to assess whether self-esteem would determine the perceived extrem- ity and impact of comparisons, while holding post content con- stant. In Study 3, we manipulated context by asking participants to use social media or engage in other online activities on their smartphone, and then assessed social comparison activity and outcomes. In Study 4, we used an experience sampling methodol- ogy to examine, in a naturalistic setting, whether the frequency, direction, and impact of comparisons on social media would differ from those in other contexts. Across studies, we predicted that participants would make more frequent upward than downward social media comparisons, which in turn would negatively impact state self-esteem, mood, and life satisfaction (Studies 1– 4). We also predicted that these upward comparisons would be to more extremely superior others, and that, for each comparison, extremity would be associated with a more negative impact on self-evaluations (Studies 1– 4). Further, we predicted that the frequency and extremity of upward comparisons would be greater in social media than in other contexts, and so would have more negative outcomes (Studies 3 and 4). In addition, we predicted that low self-esteem individuals would make more frequent and extreme upward social comparisons than high self- esteem individuals, and consequently would experience the great- est comparison threat (Studies 1–3); we predicted that these self- esteem effects would be especially potent among individuals using social media relative to other contexts (Study 3). Study 1 In Study 1, we assessed social comparisons among participants browsing their social media news feeds on either Facebook or Insta- gram. After viewing each post, participants indicated whether they had made a social comparison, and if so, the comparison domain, direction, and impact on their self-evaluations. In addition, to assess the cumulative effects of the posts viewed, we measured participants’ mood, state self-esteem, and life satisfaction after the browsing ses- sion. Whereas past studies have used global retrospective self-reports regarding social comparison activity on social media, participants in the present study reported on the social comparisons they made while browsing their news feed in real time in the lab; this enabled us to more accurately measure the frequency, direction, and extremity of social media comparisons, while reducing potential bias in recall. We predicted that the more extreme the upward comparison, the more negative the immediate impact on self-evaluations. In addition, we predicted that individuals would make more upward comparisons than downward or lateral comparisons, and these upward comparisons would have a cumulative negative impact on their self-evaluations, mood, and life satisfaction. Study 1 also allowed us to examine the role of self-esteem in determining the frequency, extremity, and outcomes of comparisons occurring during an actual social media session. We predicted that lower self-esteem participants would make more extreme upward comparisons than high self-esteem individuals and, thus, report feel- ing worse about themselves after each comparison. In addition, we predicted that low self-esteem individuals would make more frequent upward comparisons, which in turn would result in more negative mood, self-evaluations, and life satisfaction at the end of the session. Participants browsed either their Facebook or Instagram news feed on their smartphones and answered questions about the first 20 posts on a desktop computer. We examined comparisons on two popular social media platforms (Greenwood, Perrin, & Duggan, 2016) to confirm that our results would generalize to more than one site. Method Participants. We recruited 251 introductory psychology stu- dents for a study on social media use. Thirty-eight were excluded from our analyses: Nine participants experienced technical problems with the survey, 18 participants did not complete the survey within the allotted time, four participants indicated at the end of the session that they did not understand the instructions, four participants behaved in a manner indicating that they did not take the study seriously (e.g., stated that they gave random answers to finish the study faster), and three were unable to use their smartphones to browse their social media feeds. Our analyses included 213 introductory psychology students (157 women and 56 men; Mage � 18.98 years, SD � 1.64 years) who participated for course credit. We collected sufficient data (i.e., at least 85 observations; Cohen, 1992) to detect a small effect at both levels of our multilevel models (NL1 � 1796; NL2 � 206). Post hoc power analyses revealed that we had at least 81.06% power for our primary multilevel results. For our other analyses, sensitivity analysis revealed that we had sufficient power to detect a small-to- medium effect (r � .19).1 Procedure. Participants who use both Facebook and Insta- gram were invited to take part in a study on social media use. Upon arrival at the lab, participants first completed the 10-item Rosen- berg Self-Esteem Scale (e.g., “I take a positive attitude toward myself”; � � .87; Rosenberg, 1965). Participants rated themselves on a 7-point scale with endpoints ranging from 1 (strongly dis- 1 Across all studies, we had relatively small samples of males and, thus, limited statistical power to detect significant gender effects. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 4 MIDGLEY, THAI, LOCKWOOD, KOVACHEFF, AND PAGE-GOULD agree) to 7 (strongly agree). The pretest also included questions about frequency of Facebook and Instagram use; all participants indicated using both Facebook and Instagram at least once per week. Finally, to ensure that participants understood what a social comparison was and could complete the task successfully, they each completed a brief training session in which they were pro- vided with a detailed description of what does (and does not) constitute a social comparison, and practiced identifying social comparisons in sample scenarios.2 After the training session, participants were randomly assigned to either the Facebook or Instagram condition and were asked to open the corresponding app on their smartphone. To ensure that participants using both platforms were following a similar proce- dure, we asked them to answer questions about the first 20 posts in their news feed. Participants were asked to complete the question- naire without navigating away from their news feed. For each post, participants indicated the extent to which they had made a social comparison while viewing the post (1 � not at all, 7 � com- pletely). If participants had made a comparison (i.e., answered 2 or above), they answered additional questions about the comparison. Comparison questions. For posts that led to comparisons, participants first indicated in which domain(s) the comparison occurred (from a list based on options provided in an earlier study; Wheeler & Miyake, 1992), with the option of selecting one or multiple domains, or selecting “other” and providing their own domain label (see Table 1 for a list of domain options). Participants then reported the comparison direction, by indicating whether the comparison was to someone worse- or better-off than themselves on a 7-point scale with endpoints ranging from �3 (much worse- off than me) to 3 (much better-off than me). This item enabled us to assess not only the direction of the comparison, but also the extremity. For example, a score of either 1 or 3 would indicate an upward comparison, but the 3 indicates a more extreme upward comparison (Patrick et al., 2004). They then rated themselves on two self-evaluation items (“After making this comparison, I felt better about myself” and “After making this comparison, I felt worse about myself” [reverse-scored]) on a 7-point scale with endpoints ranging from 1 (strongly disagree) to 7 (strongly agree). Responses for the two self-evaluation items, r � �.71, p � .001 were averaged to create a composite score of postcomparison self-evaluations. Post-social media questionnaire. After answering questions about 20 posts, participants were instructed to put down their smartphones. They then completed measures assessing their affect, self-esteem, and life satisfaction. State affect. Participants first completed the 20-item Positive and Negative Affect Schedule (PANAS; Watson, Clark, & Telle- gen, 1988). Ratings were made on a 5-point scale (1 � not at all, 5 � extremely; � � .82). State self-esteem. Participants then completed a state self- esteem measure (Heatherton & Polivy, 1991), indicating how true a series of 20 statements were for them “right now” on a 5-point scale (1 � not at all, 5 � extremely; � � .92). Life satisfaction. Finally, participants completed the five-item Satisfaction with Life Scale (Diener, Emmons, Larsen, & Griffin, 1985), indicating their agreement with each item on a 7-point scale (1 � strongly disagree, 7 � strongly agree; � � .85).3 Results Overview of analyses. We first analyzed our data at the comparison-level, examining the domain, direction, extremity, and impact of individual comparisons. We next examined our data at the session-level, to assess whether frequency of upward compar- isons predicted postsession outcomes. Unless otherwise noted, we analyzed our data using R 3.6.0.4 Individual comparisons. Domains. Both Facebook and Instagram comparisons oc- curred in a variety of domains (see Table 1). For both platforms, the top three domains of comparison were looks/attractiveness, popularity/friendship, and vacations/activities/lifestyle. Direction, extremity, and self-evaluations. Participants made comparisons that were, on average, upward in direction (M � 1.04, 95% confidence interval, CI [0.97, 1.11], SD � 1.54).5 An intercept-only multilevel model indicated this was significantly greater than the scale midpoint of 0, b � 1.04, 95% CI [0.93, 1.15], SE � 0.06, t(193.13) � 17.99, p � .001. Moreover, a multilevel model with direction separated into its between- and within-person components and a random slope of comparison direction revealed that when individuals made comparisons that were more upward in direction than usual, they reported lower self-evaluations after the comparison b � �0.64, 95% CI [�0.72, �0.55], SE � 0.04, t(175.25) � �14.98, p � .001.6 That is, for each participant, their more upward comparisons (relative to that individual’s average comparison direction) resulted in more negative outcomes than their less upward comparisons. Self-esteem and self-evaluations. We then tested whether so- cial media comparisons are especially damaging for individuals lower in self-esteem because they make more extreme compari- sons. To test this hypothesis, we conducted a 2–1–1 multilevel mediation model because the predictor (i.e., self-esteem) varied only at the level of the person, but the mediator (i.e., comparison extremity) and outcome variable (i.e., self-evaluations) varied across all posts, which were nested within individuals. Thus, the person-level average of comparison extremity was included as a covariate in the final model and direction was person-centered (Zhang, Zyphur, & Preacher, 2009). This mediational hypothesis was tested using a variant of the bootstrap procedure (Preacher & Hayes, 2008) amended for 2–1–1 multilevel mediation using the indirect function in R (indirect.mlm; Page-Gould & Sharples, 2016) with 5,000 bootstrapped samples to accurately estimate the indirect effect and its 95% CI. Furthermore, because the relation- ship between comparison extremity and self-evaluations could vary from person-to-person, we modeled this path (i.e., b path) as 2 The full version of the training session, including our comparison scenarios, is available in our study materials on OSF (https://osf.io/acyzu/). 3 Participants in this and subsequent studies answered additional ques- tions not analyzed for this set of studies. The complete questionnaires and datasets for these studies are available by emailing the authors. 4 Syntax for all manuscript analyses are available on OSF (https://osf .io/2k6vd/). 5 All 95% confidence intervals are calculated using the bootstrap per- centile method. 6 We compared a full model that included platform, trait self-esteem, person-mean comparison direction, person-centered comparison direction, and their interactions to a model that included main effects only. Platform did not moderate this effect, �2(2)�2.15, p�.34. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 5SOCIAL MEDIA COMPARISONS AND SELF-ESTEEM https://osf.io/acyzu/ https://osf.io/2k6vd/ https://osf.io/2k6vd/ a random slope (see Figure 1). Our analysis revealed that trait self-esteem was associated with lower self-evaluations as a function of its relationship with comparison extremity, abwithin � 0.01, 95% CI [0.006, 0.02], abbetween � 0.01, 95% CI [0.007, 0.02]: Individuals with lower self-esteem made compari- sons that were more upward, a � �0.03 [�0.04, �0.02], which in turn made them feel worse about themselves, bwithin � �0.42, 95% CI [�0.45, �0.38], bbetween � �0.48, 95% CI [�0.54, �0.39]. Although the effect of trait self-esteem on self- evaluations was significant, c � 0.06, 95% CI [0.05, 0.07], the direct effect of self-esteem on self-evaluations was reduced when the indirect path through comparison extremity was taken into account, c’ � 0.04, 95% CI [0.03, 0.06]. Thus, compared with individuals with higher self-esteem, those with lower self-esteem felt worse after social media comparisons, at least in part because these comparisons were more extreme in their upwardness. Overall session. Comparison frequency and postsession outcomes. Of the 20 posts viewed by participants, an average of 8.42 (Mdn � 8.00, SD � 5.07) resulted in a social comparison. Furthermore, a one- way repeated-measures analysis of variance (ANOVA) corrected using Greenhouse-Geisser estimates (ε � .78) revealed a signifi- cant effect of comparison type on comparison frequency, F(1.56, 329.53) � 140.60, p � .001. Participants made more upward (M � 5.32, SD � 3.85) than downward (M � 1.20, SD � 1.55; t(211) � 14.80, p � .001) or lateral (M � 1.90, SD � 2.46; t(211) � 11.05, p � .001) comparisons. They also made more lateral than down- ward comparisons, t(211) � 3.78, p � .001.7 We then examined whether comparison behavior over all 20 posts influenced subsequent reports of mood, state self-esteem, and life satisfaction. To account for correlations between the outcome measures, we conducted a multivariate regression in which mood, state self-esteem, and life satisfaction were regressed simultaneously on the number of upward, downward, and lateral comparisons made by participants. The total number of upward comparisons predicted outcomes, F(3, 206) � 8.95, p � .001, but number of lateral, F(3, 206) � 1.81, p � .15, or downward comparisons did not, F(3, 206) � 0.87, p � .46. Univariate analyses indicated that making a greater number of upward com- parisons was associated with less positive affect, b � �0.02, 95% CI [�0.04, �0.002], SE � 0.01, t(208) � �2.18, p � .030, r � .15, lower state self-esteem, b � �1.26, 95% CI [�1.84, �0.70], SE � 0.25, t(208) � �4.96, p � .001, r � .33, and lower life satisfaction, b � �0.10, 95% CI [�0.14, �0.05], SE � 0.02, t(208) � �4.22, p � .001, r � .28. There were no effects of number of lateral or downward comparisons, ts � 1.87, ps � .06. Thus, regardless of the number of downward and lateral compar- isons participants made while viewing their news feeds, making more upward comparisons predicted worse mood, lower state self-esteem, and diminished life satisfaction after individuals viewed the 20 posts. 7 Platforms did not differ in terms of number of comparisons reported, b � �0.02, SE � 0.02, z � �0.96, p � .34. However, relative to participants using Facebook, participants using Instagram were especially likely to make upward comparisons relative to any other type of compar- ison, b � �0.54, SE � 0.18, z � �3.03, p � .002, Odds Ratio � 1.71:1. Participants who used Instagram had a 71.39% chance of making an upward comparison, whereas participants who used Facebook had a 59.36% chance of making an upward comparison. Table 1 Comparison Domains Comparison domain Study 1 (N � 1,795) Study 3 (N � 255) Study 4 (N � 1,211) Instagram (n � 941) Facebook (n � 854) Social media (n � 169) Other contexts (n � 86) Social media (n � 124) Other contexts (n � 1,087) Looks/attractiveness 263 (27.9%) 163 (19.1%) 23 (13.6%) 10 (11.6%) 41 (33.1%) 167 (15.4%) Academics/career 54 (5.7%) 89 (10.4%) 14 (8.3%) 5 (5.8%) 11 (8.9%) 254 (23.4%) Dating/relationships 48 (5.1%) 47 (5.5%) 5 (3.0%) 6 (7.0%) 7 (5.6%) 65 (6.0%) Popularity/friendships 98 (10.4%) 119 (13.9%) 15 (8.9%) 5 (5.8%) 3 (2.4%) 46 (4.2%) Vacations/activities/lifestyle 219 (23.3%) 162 (19.0%) 26 (15.4%) 9 (10.5%) 16 (12.9%) 83 (7.6%) Personality/morality 38 (4.0%) 86 (10.1%) 12 (7.1%) 9 (10.5%) 3 (2.4%) 112 (10.3%) Skills/abilities 75 (8.0%) 63 (7.4%) 14 (8.3%) 13 (15.1%) 12 (9.7%) 126 (11.6%) Health/physical fitness 65 (6.9%) 62 (7.3%) 20 (11.8%) 13 (15.1%) 16 (12.9%) 80 (7.4%) Wealth/finances 31 (3.3%) 22 (2.6%) 9 (5.3%) 4 (4.7%) 1 (0.8%) 19 (1.7%) Family 21 (2.2%) 19 (2.2%) 10 (5.9%) 6 (7.0%) 0 (0%) 17 (1.6%) Other 29 (3.1%) 22 (2.6%) 13 (7.7%) 4 (4.7%) 1 (0.8%) 16 (1.5%) Multiple domains — — 8 (4.7%) 2 (2.3%) 13 (10.5%) 102 (9.4%) Note. N is the number of comparisons reported across all participants. In Study 1, participants could select multiple domains; thus, the sum of all domains in Study 1 exceeds the total. Figure 1. The association between self-esteem and self-evaluations is mediated by differences in comparison extremity (Study 1). Unstandard- ized regression coefficients and 95% bootstrapped confidence intervals are reported along the paths they model. Statistics reported within parentheses represent the total effect. Values with the subscript within represent the within-subject effects, and values with the subscript between represent between-subjects effects. The direct and total effects have not been parsed into within- and between-person components. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 6 MIDGLEY, THAI, LOCKWOOD, KOVACHEFF, AND PAGE-GOULD Self-esteem and postsession outcomes. Next, we tested whether individuals lower in trait self-esteem would have lower state self-esteem, life satisfaction, and mood after browsing their news feeds as a result of making more upward comparisons during the session. First, we regressed number of upward comparisons on trait self-esteem using a Poisson regression with a log link function to account for the fact that the number of upward comparisons represented frequency counts and, thus, violated the normality assumption required for traditional regression. Consistent with our hypothesis, low self-esteem individuals made more upward com- parisons, b � �0.04 [�0.06, �0.02], SE � 0.01, z � �6.70, p � .001.8 We then conducted three mediation analyses, one for each outcome, using a bootstrapping procedure (Hayes, 2013) with 5,000 resamples and generating 95% CIs for the indirect effects.9 Number of upward comparisons mediated the positive association between trait self-esteem and state self-esteem, ab � 0.13, 95% CI [0.03, 0.26], SE � 0.06. Although the total effect of trait self- esteem on state self-esteem was significant, c � 2.23, 95% CI [1.90, 2.54], SE � 0.16, the direct effect of trait self-esteem on state self-esteem was reduced when the indirect path through number of upward comparisons was taken into account, c’ � 2.10, 95% CI [1.77, 2.44], SE � 0.16 (see Figure 2 panel A). Number of upward comparisons also mediated the positive association be- tween trait self-esteem and life satisfaction, ab � 0.01, 95% CI [0.001, 0.02], SE � 0.01. Although the total effect of trait self- esteem on life satisfaction, c � 0.16, 95% CI [0.12, 0.19], SE � 0.02, was significant, the direct effect of self-esteem on life satis- faction was reduced when the indirect path through number of upward comparisons was taken into account, c’ � 0.14, 95% CI [0.11, 0.18], SE � 0.02 (see Figure 2 panel B). Finally, we tested whether number of comparisons would mediate the positive asso- ciation between trait self-esteem and affect. Although this indirect effect was in the predicted direction, it was not reliable, ab � 0.001, 95% CI [�0.003, 0.005], SE � 0.002.10 Thus, when using either Facebook or Instagram, participants with lower trait self- esteem reported making a greater number of upward comparisons and, thus, experienced significantly lower state self-esteem and life satisfaction after the social media session. Discussion Study 1 demonstrates a direct link between comparison behav- iors and their immediate consequences when browsing social media. Participants made comparisons that were primarily upward, and many individuals made multiple comparisons in a single session, with a median of 8 comparisons in 20 posts. Across both social media platforms, making more upward comparisons while viewing posts from others was associated with lower state self- esteem and life satisfaction after the social media session, regard- less of the number of downward and lateral comparisons individ- uals had also made. Additionally, we found that, compared with higher self-esteem individuals, those with lower self-esteem reported making more extreme upward comparisons, which predicted lower self- evaluations after each comparison. Furthermore, individuals with lower self-esteem reported a greater number of upward compari- sons, which predicted more negative state self-esteem and life satisfaction after the session. This provides evidence that low self-esteem individuals may be susceptible to making more fre- quent and more extreme upward comparisons, which are both associated with more negative outcomes for the self. Our results indicate that comparisons on social media occurred in a wide range of domains. Whereas past research examining offline comparisons found that individuals tended to make the most comparisons about academics and personality followed by physical appearance and lifestyle (Wheeler & Miyake, 1992), we found that individuals made more comparisons about attractive- ness, popularity, and vacations and leisure activities when using social media. Indeed, only 10% of comparisons made on social media were in the domains of personality and academics (despite the fact that all participants were students). Thus, it appears that, with the rise of social media, the domains in which individuals make comparisons may have shifted, with a greater focus on physical appearance, popularity, and recreation activities. Al- though comparison domain was not a focus of these studies, it is nevertheless useful to consider that social media may prompt comparisons that differ in domain as well as frequency and ex- tremity. We note, however, that we did not directly compare comparisons on social media with those in other contexts; we examine this in Studies 3 and 4. Study 2 Study 1 provides evidence that self-esteem is associated with both more frequent and more extreme upward comparisons while using social media, which in turn are associated with more nega- tive self-evaluations. We note, however, that this study used a correlational design. It is possible that low self-esteem individuals are simply viewing different content than are their higher self- esteem peers. For example, it may be that low self-esteem indi- viduals have negative self-perceptions because they have many superior friends, in which case the posts they view from those friends on social media may be more positive and threatening, resulting in more extreme upward comparisons. In Study 2, we 8 There was a main effect of platform, b � �0.10, SE � 0.03, z � �3.21, p � .001: Participants made more upward comparisons on Instagram than on Facebook; this effect was qualified by a significant trait self-esteem by platform interaction, b � 0.02, SE � 0.006, z � 3.12, p � .002. Although lower self-esteem predicted making more upward compar- isons on both Facebook, b � �0.02, SE � 0.01, z � �2.42, p � .02, and Instagram, b � �0.06, SE � 0.008, z � �7.21, p � .001, this effect was much larger on Instagram than on Facebook. Thus, low self-esteem indi- viduals’ tendency to make upward comparisons is exacerbated when they browse Instagram relative to when they browse Facebook. 9 In the reported models, we treated number of upward comparisons as a continuous variable. However, we tested additional models that treated number of upward comparisons as a count variable (Geldhof et al., 2018). These results were consistent with those reported in the manuscript and are reported in greater detail in online supplemental materials (https://osf.io/ 9vbw6/). 10 We tested whether there was a self-esteem by social media platform effect for any of the postsession outcomes measured. There were no platform effects for affect or life satisfaction, ts � .66, ps � .51; however, there was a significant trait self-esteem by platform interaction for state self-esteem, b � �0.42 [�0.74, �0.12], SE � 0.15, t(208) � �2.70, p � .01, r � .18: This effect was much larger for participants using Instagram, b � 2.67 [2.31, 3.05], SE � 0.22, t(208) � 11.90, p � .001, r � .64, than for those using Facebook, b � 1.84 [1.31, 2.34], SE � 0.21, t(208) � 8.63, p � .001, r � .51. Thus, low self-esteem individuals feel worse about themselves after browsing Instagram than after browsing Facebook. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 7SOCIAL MEDIA COMPARISONS AND SELF-ESTEEM http://dx.doi.org/10.1037/pspi0000336.supp https://osf.io/9vbw6/ https://osf.io/9vbw6/ assessed participants’ responses to a set of Facebook posts we created for the purpose of the study; thus, holding the valence of post content constant for low and high self-esteem individuals. This design is similar to that employed by Vogel et al. (2014) in which participants viewed a post with more or fewer likes, a manipulation of popularity. In our study, we instead manipulated the content of the posts to determine the impact of content valence. Participants were a community sample recruited via Amazon’s Mechanical Turk (MTurk); all viewed social media posts in which individuals described events that varied in valence (positive, neg- ative, and neutral). Participants were told that posts were real examples taken from the social media pages of participants who had taken part in a previous study; in fact, the posts were created by the experimenters for the purposes of the study. After each post, participants indicated whether they made a comparison and how they felt about themselves as a result of the comparison. As in Study 1, we examined whether self-esteem would predict comparison extremity. Here, however, by exposing all participants to the same positive, negative, and neutral posts, we can also rule out the possibility that low self-esteem individuals report making more upward social media comparisons simply because they have friends who post a disproportionate amount of positive content online. We predicted that, after viewing the same posts as people with higher self-esteem, participants with lower self-esteem would report making upward comparisons that were more upward in direction (i.e., more extreme), which in turn would lead to more negative self-evaluations. Method Participants. Through MTurk, we recruited 103 individuals who were paid $1.00 USD. Participants were eligible for the study if they used Facebook at least once per month and passed two standard attention checks (Maniaci & Rogge, 2014). Nine partic- ipants failed one or both attention checks, and three participants indicated they used Facebook less than once per month. Our analyses included 91 participants (63 women, 27 men, and one person of other/undisclosed gender; Mage � 32.95, SD � 10.19 years). As in Study 1, we were interested in within- and between- person effects and, thus, collected sufficient data to detect a small effect (r � .10) at each level (at least 85 observations; Cohen, 1992); post hoc power analyses revealed that for all analyses (except one which had .70 power), our final sample sizes (NL1 � 364; NL2 � 91) had at least .81 power. Procedure. Participants were invited to take part in a study regarding social perceptions on Facebook. First, participants com- pleted the same self-esteem scale (� � .93) used in Study 1. Participants were then presented with four posts, ostensibly written by past participants. Posts were presented one at a time. After each one, participants answered a series of questions. All participants saw two neutral posts that described everyday personal experi- ences (see Table 2); they viewed these as the first and last of the four posts. Participants were then randomly assigned to see either a positive then a negative post or a negative then a positive post. The positive post described either a personal achievement or a Figure 2. The association between self-esteem and postbrowsing outcomes are mediated by number of upward comparisons (Study 1). Panel A depicts the mediational model for state self-esteem, and panel B depicts the mediational model for life satisfaction. Unstandardized regression coefficients and 95% bootstrapped confidence intervals are reported along the paths they model. Statistics reported within parentheses represent the total effect. Table 2 Content of Facebook Posts Viewed by Participants (Study 2) Post ID Post content Neutral 1 Any suggestions for good lunch spots? In the mood to try something new :P Neutral 2 OMGGG to the season finale of Walking Dead! Who else was shocked?!? Negative 1 got some bad news, I’m unexpectedly out of a job . . . but we guess I’ve got to pick myself up and move forward. would really appreciate anyone that knows someone who is hiring! Negative 2 Heartbroken—trust me everyone, never fall in love . . . even when you think you know someone, it’s impossible to actually know what is in their head—it is SUCH a myth that two people can become one . . . Positive 1 Finally got my dream job at John Hopkins Hospital!! Thank you to everyone who helped me get here . . . this is PROOF that hard work, tirelessness, and determination does pay off—you can all reach your dreams :) Positive 2 So thankful to be celebrating 1 year with my babe today . . . you make me smile like no one else can& make my life so much better. We can’t wait for what is next for us. love you xxxooo T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 8 MIDGLEY, THAI, LOCKWOOD, KOVACHEFF, AND PAGE-GOULD pleasant outcome (i.e., getting a good job or a positive relationship experience), and the negative post described a personal negative experience (i.e., a lay-off or a break-up). In summary, all partici- pants viewed four posts—two neutral, one negative and one pos- itive. We used two domains for each condition to ensure that results were not limited to one domain only. In line with the study’s cover story, participants read each post and answered a series of questions about their perception of the posters’ personality. The last question was the comparison direc- tion measure (“To what extent do you feel this person is worse-off or better-off than you?”); participants responded on a 7-point scale with endpoints ranging from �3 (much worse off than me) to 3 (much better off than me) with a midpoint of 0 (neither worse off nor better off than me). Participants then completed a two-item self-evaluation measure (r � .69) similar to the one used in Study 1; they indicated the extent to which they felt worse about them- selves and better about themselves on a 7-point scale ranging from �3 (strongly disagree) to 3 (strongly agree). Results Overview of analyses. We first examined whether partici- pants were more likely to make an upward comparison after viewing a positive (vs. neutral or negative) post and, thus, report lower self-evaluations. We then examined whether lower self- esteem participants were more likely than their higher self-esteem peers to make upward comparisons in response to positive posts. Finally, across all posts, we examined whether, as in Study 1, participants with lower self-esteem made comparisons that were more extremely upward and consequently experienced worse post- comparison outcomes. Unlike Study 1, all analyses in Study 2 were conducted at the level of the individual posts and compari- sons, as participants viewed too few posts to assess frequency of comparisons as an outcome measure. Order effects. There was a significant effect of order on comparison extremity, b � 0.24, 95% CI [0.09, 0.39], SE � 0.08, t(346) � 3.17, p � .001: Participants who saw the positive post before the negative post (i.e., made an upward comparison first) rated comparisons to be more upward than participants who saw the negative post before the positive post (i.e., made a downward comparison first). Thus, although there was no effect of order on self-evaluations, b � �0.03, 95% CI [�0.17, 0.10], SE � 0.07, t(97.88) � �0.50, p � .62, we controlled for the effect of order in all analyses. Comparison direction and extremity. We then examined whether participants made different types of comparisons in re- sponse to the positive, neutral, and negative posts using a one-way repeated measures ANOVA corrected using Greenhouse-Geisser estimates (ε � .76) because post type was nested within person. Comparisons made in response to the positive posts (M � 0.77, SD � 1.46) were more upward than those made to the neutral posts (M � 0.16, SD � 0.77), which were, in turn, more upward than comparisons made to the negative posts (M � �1.46, SD � 1.19; F[1.51, 135.94] � 108.80, p � .001, ts � 3.81, ps � .001). This suggests that our manipulation of post content was indeed effec- tive, with positive content posts leading to upward comparisons, and negative content posts leading to downward comparisons. Self-esteem and comparison extremity. We then examined whether self-esteem influenced comparison extremity for each post type. Comparison direction was modeled as a function of self-esteem (grand-mean centered continuous variable), post va- lence (two dummy-coded variables), and their interaction while controlling for the effect of order. There was a main effect of post valence, �2(2) � 162.41, p � .001: Positive posts resulted in more upward comparisons (M � 0.79, SE � 0.12) than neutral (M � 0.18, SE � 0.09) and negative posts (M � �1.45, SE � 0.12). There was also a main effect of self-esteem, b � �0.20, SE � 0.08, t(93.20) � �2.63, p � .01: Individuals lower in self-esteem made more extreme upward comparisons. The post valence by self-esteem interaction was not significant, �2(2) � 5.27, p � .07. Although the overall interaction did not reach significance, we tested the effect of self-esteem for each post type because of our a priori hypothesis that individuals with lower self-esteem would report comparisons that were more upward after viewing a positive post (Howell, 2013). To test this hypothesis, we recoded the post valence variables so that each post valence was the reference group (Aiken & West, 1991), resulting in three 2-level multilevel models with random intercepts estimated using an unstructured covariance matrix and the Satterthwaite method of estimating degrees of freedom. Self-esteem effect sizes for each type of post was esti- mated using semipartial R2 (Edwards, Muller, Wolfinger, Qaqish, & Schabenberger, 2008). For positive posts, there was a significant effect of self- esteem, b � �0.40, 95% CI [�0.63, �0.16], SE � 0.12, t(315.30) � �3.28, p � .001, semipartial R2 � 0.03 (see Figure 3). In contrast, there was no effect of self-esteem for either neutral, b � �0.15, 95% CI [�0.33, 0.04], SE � 0.09, t(186.14) � �1.57, p � .12, semipartial R2 � 0.01, or negative posts, b � �0.06, 95% CI [�0.29, 0.19], SE � 0.12, t(315.30) � �0.45, p � .65, semipartial R2 � 0.001. Thus, consistent with our hypothesis, when exposed to the same positive posts, individuals lower in self-esteem tended to make more extreme upward comparisons than individuals higher in self-esteem; however, lower self-esteem individuals did not differ from higher self-esteem individuals in comparison extremity after being exposed to neutral or negative posts. Because the overall interaction was not significant, we note that these results must be interpreted with caution. Self-evaluations. To test whether participants felt worse after viewing the positive posts, relative to other posts, as a result of Figure 3. The effect of self-esteem on comparison extremity for each post type (Study 2) while controlling for order. Errors bars represent standard errors. Greater scores on the y-axis indicate a more extreme upward comparison. See the online article for the color version of this figure. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 9SOCIAL MEDIA COMPARISONS AND SELF-ESTEEM making more extreme upward comparisons, we conducted a boot- strapped 1–1–1 multilevel mediation using a similar analytic strat- egy as Study 1. Because we were primarily interested in the occurrence of upward comparisons and, thus, the effect of positive posts, we entered post valence as two dummy-coded variables, one that compared positive to neutral posts (0 � positive, 1 � neutral, 0 � negative), and one that compared positive to negative posts (0 � positive, 0 � neutral, 1 � negative). The predictor (i.e., post valence) varied within participants only. Thus, we did not enter the person-level average of post valence as a covariate in the final model. Furthermore, the relationships between post valence and direction as well as direction and self-evaluations could vary from person-to-person, so we modeled these paths (i.e., a and b paths) as random slopes. Because the function we used allowed us to specify only a single independent variable, we included the other dummy code as a covariate and ran the bootstrap analysis twice, allowing each dummy code to be the independent variable once and the covariate once (Hayes & Preacher, 2014). We also spec- ified the same starting value (i.e., seed) to ensure that the same bootstrap samples were used for both analyses. Finally, we con- trolled for the order in which the posts were presented. Our first analysis revealed that the difference between positive and neutral posts affected self-evaluations as a function of its relationship with comparison direction, abwithin � 0.23, 95% CI [0.09, 0.34], abbetween � 0.20, 95% CI [0.10, 0.35]: When partic- ipants read a positive post, relative to a neutral one, they were more likely to make an upward comparison, awithin � �0.60, 95% CI [�0.85, �0.32], which in turn made them feel worse about themselves, bwithin �0.38, 95% CI [�0.45, �0.19], bbetween � �0.33, 95% CI [�0.50, �0.23]. Although the differ- ence between positive and neutral posts on self-evaluations was significant, c � 0.30 [0.07, 0.53], the direct effect of post valence was not significant when the indirect path through comparison direction was taken into account, c’ � 0.01, 95% CI [�0.21, 0.23] (see Figure 4 panel A). The population covariance for this model was estimated to be �ab � 0, 95% CI [�0.03, 0.07]. This implies that the mediational model was consistent across individuals. Thus, posts appear to exert their impact on self-evaluations through social comparisons. Our second analysis revealed that the difference between posi- tive and negative posts also affected self-evaluations as a function of its relationship with comparison direction, abwithin � 0.93, 95% CI [0.70, 1.32], abbetween � 0.73, 95% CI [0.53, 1.20]: When participants saw a positive post, relative to a negative one, they were more likely to make an upward comparison, awithin � �2.24, 95% CI [�2.53, �1.95], which in turn made them feel worse about themselves, bwithin � �0.42, 95% CI [�0.55, �0.32], bbetween � �0.33, 95% CI [�0.52, �0.24]. Although the differ- ence between positive and negative posts on self-evaluations was significant, c � 0.88, 95% CI [0.63, 1.12], the direct effect of post type was not significant when the indirect path through compari- son direction was taken into account, c’ � �0.09, 95% CI [�0.54, 0.11] (see Figure 4 panel B). As in the first model, the population covariance for this model implies that that the mediational model was consistent across individuals, �ab � 0.002, 95% CI [�0.02, 0.08]. Thus, consistent with our hypothesis, participants felt worse about themselves after being exposed to positive posts, relative to negative or neutral posts, because they made comparisons that were more upward in direction. Self-esteem and self-evaluations. Finally, we tested whether, as in Study 1, self-esteem predicted worse self-evaluations after viewing individual social media posts as a result of making more extreme upward comparisons. We conducted a bootstrapped 2–1–1 multilevel mediation using the same analytic strategy described in Study 1. Furthermore, the relationship between comparison ex- tremity and self-evaluations could vary from person-to-person; thus, we modeled this path (i.e., b path) as a random slope (see Figure 5) as we did in Study 1. We also controlled for order in this model. Our analysis revealed that trait self-esteem was associated with lower self-evaluations as a function of its relationship with comparison extremity, abwithin � 0.11, 95% CI [0.03, 0.20], abbetween � 0.07, 95% CI [0.02, 0.19]: Individuals with lower self-esteem made comparisons that were more upward, a � �0.29, 95% CI [�0.53, �0.08], which in turn made them feel worse about themselves, bwithin � �0.37, 95% CI [�0.43, �0.25], bbetween � �0.24, 95% CI [�0.46, �0.18]. Although the effect of trait self-esteem on self-evaluations was significant, c � 0.45, 95% CI [0.29, 0.62], the direct effect of self-esteem on self-evaluations was reduced when the indirect path through comparison extremity was taken into account, c’ � 0.39, 95% CI [0.20, 0.49]. Thus, consistent with our hypothesis, lower self-esteem participants felt worse about themselves after viewing social media posts at least in part because they made upward comparisons that were more Figure 4. The association between post valence and self-evaluations is mediated by differences in comparison extremity (Study 2). Panel A depicts the mediational model for the difference between positive and neutral posts, and panel B depicts the mediational model for the difference between positive and negative posts. Unstandard- ized regression coefficients and 95% bootstrapped confidence intervals are reported along the paths they model. Statistics reported within parentheses represent the total effect. The direct and total effects have not been parsed into within- and between-person components. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 10 MIDGLEY, THAI, LOCKWOOD, KOVACHEFF, AND PAGE-GOULD extreme in upwardness than those of higher self-esteem partici- pants. Discussion In summary, posts that were positive in valence did lead par- ticipants to make upward comparisons and, thus, feel worse about themselves. The more extreme the upward comparison, the more negative the impact on the self. Furthermore, although all partic- ipants were likely to make an upward comparison in response to the positive posts, lower self-esteem participants interpreted these posts as more upward than did higher self-esteem participants. Indeed, we replicated our first mediation model from Study 1: Compared with participants with higher self-esteem, those with lower self-esteem tended to make more extreme upward compar- isons and, as a result, experienced greater decreases in their self- evaluations after being exposed to the same content. Thus, low self-esteem individuals do not merely have better memory for upward comparisons than their higher self-esteem peers, or view posts with more positive content. Moreover, this effect was limited to positive posts only: Low self-esteem individuals perceived individuals in negative and neutral posts similarly to higher self- esteem individuals; they were not less likely to see the worse-off others as downward comparisons and were not more likely to see neutral posts as upward comparisons. Therefore, these data pro- vide initial evidence suggesting that it is not any comparison behavior in general (Steers et al., 2014), but rather positive posts resulting in more extreme upward comparisons that are a key contributor to low self-esteem individuals’ more negative out- comes after social media use. Because the overall interaction between post valence and self-esteem was not significant, how- ever, we note that the findings related to extremity and self-esteem must be interpreted with caution. In Studies 1 and 2, participants reported on comparisons throughout the time they spent viewing social media posts. Al- though this provided information about participants’ responses to each post they viewed, we note that this procedure may also have created demand characteristics, in that participants may have been especially likely to notice and report on social comparisons. It may be that participants make fewer actual social comparisons when they are not prompted to think about them in this way. Accord- ingly, in Study 3, instead of asking participants to report on each comparison as it occurred, we instead asked them to report on comparisons at the end of the session, without alerting them in advance to the focus on comparison behavior. Study 3 We argue that social media may be especially likely to elicit upward comparisons, and that these comparisons will have a negative impact, particularly among individuals low in self- esteem, who make more frequent (Study 1) and more extreme (Studies 1 and 2) upward comparisons. Up to this point, however, we have not directly compared social media comparisons to those that occur in other contexts. In Study 3, we experimentally ma- nipulated context to examine whether social media would indeed be especially likely to elicit threatening upward comparisons. Specifically, participants were randomly assigned to use their smartphone either to access social media or for any other purposes (e.g., surf the net, text, watch online videos) for 10 min; they then reported their self-evaluations and information about any compar- isons they had made. This manipulation allowed us to examine whether social media comparisons differ from other technology- based comparisons, and to assess whether people feel worse after using social media relative to engaging in other technology-based activities as a result of the type of social comparisons they make. Finally, we evaluated whether low self-esteem individuals would be especially likely to make upward comparisons, and conse- quently be negatively affected, on social media relative to other online contexts. We also used this study to examine two variables found in previous research (Tesser, 1988) to be implicated in comparison outcomes: closeness of the comparison target and domain rele- vance. Past research indicates that individuals are more threatened when outperformed by a close other in a domain that is important to them (Tesser, Millar, & Moore, 1988). Many social media contacts, however, are more distant and may even be past class- mates or celebrities with whom one has no direct contact. More- over, individuals often view posts about leisure activities or activ- ities that may not be directly relevant to the self. This would suggest that, according to the self-evaluation maintenance (SEM) model (Tesser, 1988), social media comparisons might be less threatening than comparisons in other contexts. Accordingly, we measured both closeness and domain importance. Studies 1 and 2 provided initial evidence that comparisons on social media would be threatening to the self. Accordingly, we expected that, despite lower closeness and domain importance, social media comparisons would nevertheless pose a greater threat to the self than those in other contexts, due to their greater frequency and extremity. Method Participants. In total, 482 participants who indicated in a prescreen survey that they owned a smartphone and currently used either Facebook, Instagram, or both completed the study for $1.50 USD. We excluded 67 participants from the analyses: Thirty-five participants did not follow the instructions for the browsing ses- sion (i.e., went on Facebook and/or Instagram in the no social media condition or did not go on Facebook and/or Instagram in the social media condition), four participants failed our comparison training session, and 28 participants submitted the survey more Figure 5. The association between self-esteem and self-evaluations is mediated by differences in comparison extremity (Study 2). Unstandard- ized regression coefficients and 95% bootstrapped confidence intervals are reported along the paths they model. Statistics reported within parentheses represent the total effect. The direct and total effects have not been parsed into within- and between-person components. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 11SOCIAL MEDIA COMPARISONS AND SELF-ESTEEM than once.11 For duplicate responses, we retained the first com- pleted response and deleted subsequent responses. Our analyses included 415 MTurk workers (245 women, 168 men, and two persons of other/undisclosed gender; Mage � 37.55 years, SD � 12.79 years). We collected sufficient data (i.e., at least 85 observations; Cohen, 1992) to detect a small effect at both levels of our multilevel models (NL1 � 255; NL2 � 124). For our other analyses, sensitivity analysis revealed that we had sufficient power to detect a small-to-medium effect (r � .14). Procedure. Prescreen survey. We invited participants to complete a 5-min prescreen eligibility survey that included questions about their technology and social media use. To be eligible for the study, participants had to indicate they owned a smartphone, which they would use during the experimental manipulation, and a second device they could use to complete the actual survey (i.e., laptop or desktop computer); in addition, they had to indicate that they used either Facebook, Instagram, or both platforms. As part of this prescreen survey, participants completed the same self-esteem measure used in Studies 1 and 2 (� � .93). Study questionnaire. We asked participants to use their smart- phones for 10 min, specifying that they either refrain from using social media (no social media condition) or spend the entire 10 min using Facebook and/or Instagram (social media condition). At the end of the browsing session, participants indicated what they did during the browsing session (see Table 3). To encourage honest reporting, we told participants that their compensation would not be affected by whether or not they followed the manipulation instructions. Next, participants in both groups completed the same comparison training session used in Study 1 and then reported whether they had made any comparisons during the 10-min brows- ing session. If they reported making at least one comparison, they were asked to enter the first names of each target to whom they compared themselves. For each target they listed, participants were asked additional questions about the comparison. As in Studies 1 and 2, we asked in what domain the comparison occurred and the extent to which the comparison target was doing better or worse than the self. In Study 3, we also asked participants to indicate how close they felt to the comparison target on a 7-point scale with endpoints ranging from 0 (not at all) to 6 (extremely). Participants also rated how important the domain was to them using a 7-point scale with endpoints ranging from 0 (not at all important) to 6 (extremely important). Finally, participants reported their state self-evaluations using a one-item measure (“Right now, how do you feel about yourself?”) rated on a 7-point scale with endpoints ranging from �3 (much worse about myself than usual) to 3 (much better about myself than usual). Results Overview of analyses. As in Study 1, we first present our analyses at the comparison-level, examining whether comparisons on social media differed from comparisons in other computer- mediated contexts in terms of domain, direction, and impact. We also examine whether, as in Studies 1 and 2, comparisons made by individuals lower in self-esteem were more upward in direction than those made by individuals with higher self-esteem. We then present session-level analyses in which we first examine whether context and self-esteem were associated with frequency of upward comparisons during the 10-min session. In addition, we examine whether spending time on social media (vs. not using social media) predicted greater likelihood of making one or more upward com- parisons and, thus, lower self-evaluations at the end of the session. Finally, we assessed whether this session-level effect was stronger for individuals with lower self-esteem: We assessed whether low self-esteem individuals using social media were especially likely to make one or more upward comparisons and, thus, experience the most negative consequences to the self. Individual comparisons. Domains. Consistent with Study 1, we found that social media comparisons occurred in a variety of domains (see Table 1), and the most common domains of comparison in this context again included looks/attractiveness and vacations/activities/lifestyle. In contrast, the most common comparison domains in other technology-based contexts were health/physical fitness, personal- ity/morality, and skills/abilities. A series of logistic multilevel models, however, indicated that comparisons made on social me- dia (0) compared with those in the no social media condition (1) were not significantly more likely to be about looks/attractiveness, b � �0.42, SE � 0.66, z � �0.63, p � .53, or vacations/activities/ lifestyle, b � �0.43, SE � 0.54, z � �0.79, p � .43, but were less likely to be about personality/morality, b � 1.12, SE � 0.002, z � 598.98, p � .001. Domain importance. We next examined whether domain importance ratings differed by experimental condition using a multilevel model. We modeled domain importance ratings as a function of comparison context (effects-coded: �1 � no social media condition; 1 � social media condition) with a random 11 We considered answering fewer than half of the questions correctly as failing the comparison training session. Table 3 Frequencies of Participants’ Online Activities (Study 3) No social media Social media Condition Condition Activity (n � 213) (n � 202) Made a phone call 20 5 Answered a phone call 12 5 Read a text/message 100 32 Wrote a text/message 67 16 Read an email 96 11 Wrote an email 24 4 Read an article 64 15 Read a book 10 1 Listened to music 26 7 Listened to the radio 7 5 Listened to a podcast 6 3 Watched a video 56 23 Watched TV or a movie 12 6 Used a dating app/website 5 2 Checked Facebook 0 172 Checked Instagram 0 90 Used other social media 36 5 Used Snapchat 7 1 Used Twitter 9 1 Other activity 85 6 Multiple activities 171 109 Note. Frequencies do not total condition sample size because participants were able to report multiple activities. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 12 MIDGLEY, THAI, LOCKWOOD, KOVACHEFF, AND PAGE-GOULD intercept for each person. Experimental conditions did not differ in terms of domain importance, b � �0.01, 95% CI [�0.29, 0.26], SE � 0.14, t(112.79) � �0.10, p � .92. Do- mains of comparisons made on social media (M � 3.88, SE � 0.16) and in other technology-based contexts (M � 3.91, SE � 0.22) were both rated as moderately important. Thus, we found no evidence that social media comparisons involve domains that are more (or less) personally relevant than comparisons in other technology-based contexts. Closeness to target. We then examined whether closeness to the comparison target differed depending on the context in which a comparison was made; we used the same multilevel model as we used for domain importance. Experimental condi- tions did not differ in terms of target closeness, b � �0.04, 95% CI [�0.39, 0.31], SE � 0.18, t(119.37) � �0.24, p � .81. Targets in social media comparisons (M � 2.10, SE � 0.21) and in other technology-based contexts (M � 2.18, SE � 0.29) were rated as relatively low in closeness. In summary, we found no evidence that comparisons on social media were in domains higher (or lower) in importance, or to targets with greater (or lesser) closeness to the self. This suggests that neither closeness nor importance can account for any differences we observed in comparison outcomes between social media and other online contexts.12 Direction, extremity, and self-evaluations. Participants’ self- evaluations after comparisons were modeled as a function of comparison context condition with a random intercept for each person, b � �0.18, 95% CI [�0.41, 0.05], SE � 0.12, t(130.58) � �1.51, p � .13. Although this effect was not significant, it was in the expected direction: Compared with other contexts, comparisons made while using social media were associated with feeling worse about the self after the comparison. Next, extremity of comparison direction was mod- eled as a function of comparison context with a random inter- cept for each person. This analysis revealed that social media comparisons were not more upward in direction than those in other technology-based contexts, b � 0.07, 95% CI [�0.19, 0.33], SE � 0.13, t(127.04) � 0.52, p � .60. Additionally, unlike in Studies 1 and 2, lower self-esteem did not predict making more extreme upward comparisons, b � �0.08, 95% CI [�0.34, 0.18], SE � 0.13, t(132.10) � �0.59, p � .56. We review possible explanations for these findings below. Overall session. Comparison frequency. Next, we conducted a series of Poisson regressions to examine whether conditions differed in terms of number of comparisons. Participants assigned to the social media condition made more comparisons (M � 0.84, SD � 1.50) than those assigned to the no social media condition (M � 0.40, SD � 1.20), b � 0.36, 95% CI [0.14, 0.61], SE � 0.07, z � 5.50, p � .001. Furthermore, participants in the social media condition made a greater number of upward, b � 0.35, 95% CI [0.11, 0.64], SE � 0.08, z � 4.58, p � .001, and lateral comparisons, b � 0.48, 95% CI [0.08, 0.99], SE � 0.19, z � 2.59, p � .01, than those in the no social media condition. There was no difference between the two conditions for number of downward comparisons, b � 0.31, 95% CI [�0.17, 1.04], SE � 0.18, z � 1.69, p � .09. Thus, individuals using social media made more comparisons than those simply browsing the Inter- net. Moreover, social media use was associated with an in- creased number of upward and lateral comparisons, but not downward comparisons. We note that participants reported less frequent comparisons relative to Studies 1 and 2. This is not surprising: Because participants in the social media condition were free to use the platforms in any way they chose (e.g., navigate away from their news feed to read an article, engage in other activities simultaneously such as listening to music or a podcast) rather than viewing only news feed posts (as in Studies 1 and 2), and reported all comparisons at the end of the social media session, rather than after each post (as in Studies 1 and 2), they likely had fewer opportunities to make and/or take note of their individual social comparisons. Although the overall frequency of comparison was lower, we nevertheless observed the predicted difference, with more upward comparisons in the social media condition. See Table 4 for comparison frequency overall and within each condition. Self-esteem and upward comparison frequency. We then tested whether self-esteem was associated with number of up- ward comparisons made during the 10-min session. As predicted, participants with lower self-esteem made a greater number of upward comparisons than participants with higher self-esteem, b � �0.38 [�0.52, �0.24], SE � 0.06, z � �6.63, p � .001. Postsession self-evaluations. Next, we tested whether con- text influenced how participants felt about themselves after the 10-min smartphone session. Participants assigned to the social media condition (M � 0.36, SE � 0.08) reported lower self- evaluations at the end of the study than those in the no social media condition (M � 0.58, SE � 0.08), b � �0.11, 95% CI [�0.22, �0.005], SE � 0.06, t(413) � �2.03, p � .04. We then tested whether making one or more upward comparisons (vs. no upward comparisons) mediated this difference between the two experimental conditions (social media vs. no social media; see Figure 6).13,14 Given the nonnormal distribution of presence of upward comparisons, our mediator, we conducted a nonlinear mediation analysis based on the generalized linear model by specifying a path model with binomial and Gaussian (normal) distributions (Geldhof, Anthony, Selig, & Mendez-Luck, 2018). We used nonlinear mediation analysis (Hayes & Preacher, 2010) to calculate conditional indirect effects (i.e., indirect effects at specific values of X) and used a bootstrapping pro- cedure to calculate confidence intervals with 5,000 resamples 12 We did not have sufficient power to test the full SEM model (i.e. a potential three-way interaction between domain importance, closeness, and context). However, differences in closeness across the two contexts did not mediate the difference in comparison outcomes between contexts. 13 Given the limited variance in number of upward comparisons (1.24), we transformed our count variable to a binary variable to test whether making at least one upward comparison mediated the effect of self-esteem on self-evaluations. Our indirect effects were nonsignificant when we treated number of upward comparisons as a count variable; however, based on the results of Study 1, we would predict that making more upward comparisons would predict worse outcomes. 14 Our results for our simple and moderated mediation models remained consistent when we controlled for number of downward comparisons and number of lateral comparisons. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 13SOCIAL MEDIA COMPARISONS AND SELF-ESTEEM using Mplus 7.4 (Muthén & Muthén, 1998 –2012).15 Consistent with our hypothesis, the indirect effect through presence versus absence of upward comparisons was significant, ab � �0.16, 95% CI [�0.47, �0.04], SE � 0.12, and the direct effect became nonsignificant, c’ � �0.14, 95% CI [�0.37, 0.08], SE � 0.11, once this indirect effect was included. Taken to- gether, these findings suggest that individuals felt worse after using social media, relative to those who browse the Internet without using social media, because they were more likely to make one or more upward comparisons. Self-esteem and postsession self-evaluations. Finally, we tested a moderated mediation model in which the association between dispositional self-esteem and self-evaluations at the end of the study was mediated by making one or more upward com- parisons (see Figure 7). We expected this indirect effect to be larger in the social media condition than in the no social media condition. Given the nonnormal distribution of the mediator, we used a similar analytic strategy as that described above; however, we amended this approach for a moderated mediation (Hayes, 2015). That is, we calculated conditional indirect effects at specific values of the predictor, self-esteem (i.e., 1 SD above and below the mean), and the moderator, experimental condition. These condi- tional indirect effects are reported in Table 5. Consistent with our hypothesis, the direct effect, c’ � 0.10, 95% CI [0.008, 0.19], SE � 0.05, z � 2.19, p � .03, was reduced once we accounted for the indirect effect through presence or absence of upward com- parisons. Then, we conducted an omnibus test to determine whether these conditional indirect effects at various levels of self-esteem differed between experimental conditions. This test indicated that this trend was not reliable, �2(2) � 5.84, p � .054. Nonetheless we take this trend as tentative evidence that our indirect effect was moderated by context. Taken together, these findings suggest that individuals lower in dispositional self-esteem may be more likely to make more upward comparisons than those higher in self-esteem, particularly when using social media; this greater number of comparisons on social media, in turn, is asso- ciated with lower self-evaluations for low self-esteem individuals after social media use. Discussion Study 3 demonstrates that spending 10 min on social media, relative to other online activities, increases the chances of making one or more upward social comparisons and, thus, feeling worse about the self. Furthermore, the increased likelihood of making upward comparisons mediated the link between social media use and lowered self-evaluations: Individuals using social media feel worse about themselves because they are likely to make more frequent upward comparisons. Moreover, this effect occurs despite the fact that social media comparisons involve targets no closer to the self and in domains that are no more personally relevant than comparisons made in other online contexts. This finding is theo- retically significant because it suggests that social media compar- isons may not play by the same rules as other comparisons. Past studies, including research supporting the self-evaluation mainte- nance model, indicates that comparisons to less close others are less likely to result in either a threatening contrast effect or a positive basking in reflected glory effect (e.g., Tesser, 1988). Social media comparison targets are no more close than other online comparison targets, yet they lead to more threatening up- ward social comparisons. It may be that social media contacts, while less psychologically close than individuals with whom one has more recent or in-person contact, may nevertheless represent important standards against which one measures one’s life suc- cesses. One may not feel close to a former high school classmate, but one may nevertheless feel a pang when one sees a post from the classmate highlighting their superior career accomplishments, their recent engagement, or their children’s academic successes. Our findings suggest that further research will be important to assess whether closeness is the best variable to evaluate the rele- vance of social media contacts as comparison others. This study also provides important evidence that low self- esteem individuals are especially vulnerable to self-esteem threats 15 We describe this and other analyses in greater detail in our online supplemental materials (https://osf.io/9vbw6/). Table 4 Frequencies of Participants’ Comparison Activity (Study 3) Type of comparison Comparison frequency Total no. of participantsParticipant group 1 2 3 3 Whole sample (n � 415) Upward 63 22 11 7 103 Downward 15 3 4 — 22 Lateral 23 4 — 1 28 No social media condition (n � 213) Upward 22 9 2 2 35 Downward 3 — 3 — 6 Lateral 8 1 — — 9 Social media condition (n � 202) Upward 41 13 9 5 68 Downward 12 3 1 — 16 Lateral 15 3 — 1 19 Figure 6. The association between comparison context (experimental condition: social media vs. other online contexts) and self-evaluations is mediated by greater likelihood of making one or more upward comparisons (Study 3). Statistics are reported in body of text given the nonnormal distribution of the mediator (presence/absence of upward comparisons). Figure 7. The moderated mediation model we tested. The association between self-esteem and self-evaluations is mediated by greater likelihood of making one or more upward comparisons, and this indirect effect differs depending on the comparison context (Study 3). T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 14 MIDGLEY, THAI, LOCKWOOD, KOVACHEFF, AND PAGE-GOULD http://dx.doi.org/10.1037/pspi0000336.supp http://dx.doi.org/10.1037/pspi0000336.supp https://osf.io/9vbw6/ as a result of their social comparisons. They were more likely to make upward comparisons than were high self-esteem individuals, and this self-esteem effect tended to be more pronounced among participants using social media. Contrary to our predictions, how- ever, we did not find that individuals made more extreme upward comparisons while using social media or that low self-esteem individuals made especially extreme upward comparisons. One possible explanation is that participants were reporting on all comparisons at the end of the session and, thus, were unable to identify or report subtle differences in comparison extremity. In addition, we note that many individuals in the no social media condition were viewing videos or other online material in which they may have been exposed to celebrities or other very positive content. Thus, it is possible that the relatively smaller number of comparisons made in the no-social-media contexts were as ex- tremely upward as the more numerous upward comparisons on social media. That is, it may be that upward comparisons are more extreme in all online media, whether social media, videos, or other websites. In Study 3, we asked participants about their social comparisons at the end of the 10 min that participants spent on their phones, rather than after each opportunity for comparison, as in Studies 1 and 2. This methodology reduces the likelihood that individuals were reporting many comparisons because they felt prompted to do so after each post viewed. Indeed, participants reported fewer comparisons overall in Study 3 compared with Study 1, and although this may be partially because of spending less time on social media in Study 3, it is also likely because of reduced demand characteristics. Although the total number of comparisons made was lower, however, we nevertheless observed a greater frequency of social comparisons on social media than in other contexts; this difference across contexts cannot be attributed solely to demand characteristics because both conditions received the same instructions after using their phone and before reporting any social comparisons they made. Study 4 Study 3 provides important evidence that social media compar- isons differ from those made in other technology-based contexts. It did not, however, include comparisons that occur in offline con- texts, such as face-to-face interactions. Thus, in Study 4, we conducted an experience sampling study to examine participants’ social comparison behavior across all contexts. Participants in- stalled a custom-made experience sampling app on their smart- phone for 2 weeks that prompted them six times a day to complete a short survey about any comparisons they had made since the previous report. This enabled us to: (a) examine the direction, frequency, extremity, and outcomes of social media comparisons as they occur in daily life; (b) compare social comparisons made on social media with those in all other contexts; and (c) assess whether the amount of time spent on social media predicts upward comparison frequency. Whereas the between-subjects design of Study 3 allowed us to assess whether individual differences in self-esteem predict different outcomes in different contexts (i.e., social media vs. other technology-based activities), Study 4 was designed to further investigate within-person effects of context on social comparison outcomes. That is, we were able to assess whether individuals make more frequent and more extreme up- ward comparisons when spending time on social media than they do when spending time in other contexts. Method Participants. We recruited 87 undergraduate students from a university in a large urban center to participate in a 2-week long study on daily experiences.16 Two participants were unable to participate because the smartphone app was incompatible with their devices, and six participants did not complete any experience sampling questionnaires after attending the intake session. In our final dataset, we had experience sampling data from 79 partici- pants (51 women, 26 men, and two person of other/undisclosed gender; Mage � 20.15 years, SD � 2.40 years). Only 66 partici- pants (83.54%) returned for the exit session and, thus, completed all three components of the study. Participants were invited to take part only if they owned a smartphone and reported that they used social media. We compensated participants up to $60, depending on the extent to which they participated in the study ($16 for the intake session, $2 per day that they completed four or more surveys, $3 for each completed week, and $10 for the exit session). On average, participants provided 62.51 surveys (SD � 24.78) for 12 of the 14 survey days (SD � 4.0; median response rate � 100), for a total of 4,382 completed surveys. Procedure. Participants first came to the lab for an intake session, during which they were asked to install the experience sampling app (ExperienceSampler; Thai & Page-Gould, 2018); a research assistant then instructed them on how to use the app and how to recognize and accurately report on social comparisons. Notifications were customized to participants’ sleep and wake times and included an opportunity to specify different times for weekdays and weekends. For the following 2 weeks, the app randomly prompted participants to complete a short survey six times per day, in which they indicated the time they had spent on various activities (e.g., face-to-face interactions, texting, and social media) and whether they had made a comparison since the previ- 16 At the outset of the study, we aimed to recruit as many participants as possible, up to a maximum of 100. We ended recruitment at the conclusion of the semester, which coincided with the end of the academic year. We decided to end recruitment because we anticipated being unable to follow-up with a significant portion of participants during the summer months. Table 5 Conditional Indirect Effects Through Number of Upward Comparisons on Self-Evaluations at Specific Values of Self- Esteem in Each Condition (Study 3) Self-esteem values Experimental condition Indirect effect estimate SE 95% CI Low (�1 SD) No social media 0.080 0.05 [0.018, 0.195] Social media 0.113 0.04 [0.039, 0.194] High (1 SD) No social media 0.034 0.01 [0.01, 0.057] Social media 0.099 0.04 [0.032, 0.193] Note. CI � confidence interval. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 15SOCIAL MEDIA COMPARISONS AND SELF-ESTEEM ous survey.17 Whenever participants indicated they had made a comparison, they then reported to whom they had compared them- selves (i.e., close friend, ordinary friend, acquaintance, past friend, stranger, famous person, fictional character, sibling, other family member, romantic partner, or other) and their closeness to that person on a 7-point scale with endpoints ranging from 0 (not at all close) to 6 (extremely close). Participants then indicated the pri- mary comparison domain from a list (see Table 1) and indicated the importance of that domain to them on a 7-point scale with endpoints ranging from 0 (not at all important) to 6 (extremely important). Next, they indicated in what context that comparison took place (i.e., in person/face-to-face, video call, voice-only phone call, other voice chat, e-mail, texting/SMS, social media, dating app/website, other media/online context, or in a thought/ daydream). If participants indicated the comparison was made while using social media, they were also asked to indicate the platform (e.g., Instagram, Facebook).18 Finally, participants indi- cated the direction and outcomes of the comparison, using the same questions as in Studies 1–3. At the end of the survey, regardless of comparison behavior, participants reported their cur- rent self-evaluations on a 7-point scale with endpoints ranging from �3 (much worse about myself than usual) to 3 (much better about myself than usual). After the 2-week experience sampling period, participants returned to the lab for compensation and debriefing. Results Overview of analyses. As in Studies 1 and 3, we first present analyses conducted at the individual comparison-level, contrasting social media comparisons to those made in other contexts (e.g., those made face-to-face, in a thought/daydream, or in any other computer-mediated context other than social media). We then present analyses for the overall experience sampling period, ex- amining the total number of comparisons participants made on social media compared with in other contexts.19 Individual comparisons. Domains. Consistent with Studies 1 and 3, the two most common domains of comparisons on social media were looks/ attractiveness and vacations/activities/lifestyle. In this study, un- like Study 3, looks/attractiveness was also a common domain in contexts other than social media, second only to academics/careers (see Table 1). This difference may be because of the samples involved in each study: Study 3 consisted of primarily middle-aged adults, whereas Study 4 consisted of primarily undergraduate students. Compared with middle-aged adults, undergraduate stu- dents are less likely to have established careers and be in long-term relationships, both of which may increase their interest in making comparisons in the academics/careers and looks/attractiveness do- mains. In the present study, a logistic multilevel model revealed that looks/attractiveness comparisons were 77.50% more likely to occur on social media than in other contexts, b � 1.24, SE � 0.001, z � 1311.77, p � .001, odds ratio (OR) � 3.44:1. That is, although looks/attractiveness comparisons were common across all contexts, they were especially likely when individuals were using social media. Domain importance. Relative to comparisons in other con- texts, comparisons made while using social media were in domains that participants rated as less personally important, b � �0.10, 95% CI [�0.19, �0.01], SE � 0.05, t(999.01) � �2.06, p � .039. Thus, consistent with Study 3, we found that social media com- parisons did not involve domains that were more personally im- portant to participants than comparisons made in other contexts. Closeness to target. Relative to comparisons in other contexts, comparisons made while using social media involved targets that participants rated as less close to the self, b � �0.36, 95% CI [�0.49, �0.22], SE � 0.07, t(1135.62) � �5.23, p � .001. These lower closeness targets nevertheless elicited a more negative im- pact on the self, as will be discussed next.20 Comparison direction, extremity, and self-evaluations. Next, we tested a mediational pathway such that comparisons made while browsing one’s social media news feed, relative to compar- isons made in other contexts (0 � other contexts; 1 � social media news feed), were more extreme (i.e., more upward), resulting in worse postcomparison self-evaluations. We conducted a boot- strapped 1–1–1 multilevel mediation using a similar analytic strat- egy as Studies 1 and 2 (see Figure 8). Because all predictors varied within person, we separated the predictor (context) and mediator (comparison extremity) into their between- and within-person components, but we report the within-person components only because we are primarily interested in the within-person differ- ences in context. We also included time as a covariate in our model. Finally, because the relationship between context and com- parison extremity (i.e., a path) as well as the relationship between comparison extremity and self-evaluations (i.e., b path) could vary from person-to-person, we modeled these paths as random slopes. Our analysis revealed that comparison context affected self- evaluations as a function of its relationship with comparison ex- tremity, ab � �0.26, 95% CI [�0.44, �0.13]: Comparisons made on social media news feeds, relative to other contexts, did indeed result in comparisons that were more extreme in their upwardness, a � 0.74, 95% CI [0.37, 1.25], which in turn made participants feel worse about themselves, b � �0.35, 95% CI [�0.40, �0.32]. Although the difference between comparison contexts on self- evaluations was significant, c � �0.44, 95% CI [�0.76, �0.16], the direct effect of context was not significant when the indirect path through comparison extremity was taken into account, c’ � �0.02, 95% CI [�0.31, 0.21]. The population covariance for this model was estimated to be �ab � 0, 95% CI [�0.015, 0.015]. 17 In accordance with recommendations outlined in Wheeler and Reis (1991), we opted for signal-contingent sampling, as we expected social comparisons to occur frequently (i.e., numerous times per day) and were interested in their relative distribution across a variety of domains. 18 Participants were also asked to indicate which aspect of the social media platform (i.e., public feed, private chat/message, or group chat/ message) they were using at the time of the comparison, to distinguish comparisons made in private messages from those made while browsing news feeds. Given that the news feed is a unique feature of social media that does not exist in other contexts and our focus on the effects of news feed posts on comparison behavior in Studies 1 and 2, we compared news feed comparisons to comparisons made in all other contexts. 19 In contrast to Studies 1–3, we did not have the required number of participants to test for small effects at the person-level (i.e. 85; Cohen, 1992), so we did not examine the role of self-esteem in this study. 20 As in Study 3, we did not have enough power to test the full SEM model (i.e., a potential three-way interaction between domain importance, closeness, and context). Furthermore, neither decreased closeness to the comparison target nor lower domain importance mediated the effect of context on comparison outcomes, indicating that these variables may not play as much of a role in social media comparisons as in other contexts. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 16 MIDGLEY, THAI, LOCKWOOD, KOVACHEFF, AND PAGE-GOULD This implies that the mediational model was consistent across individuals. Thus, participants felt worse about themselves after making comparisons on social media relative to comparisons in other contexts, at least in part because these comparisons were more extreme in their upwardness.21 Overall experience sampling period. Context and comparison likelihood. We first examined whether time spent on each activity predicted the likelihood of making a social comparison for any given survey using a logistic multilevel model with a random intercept, the within- and between-person components of time spent on each activity, and time as a covariate. When individuals spent more time on social media than they usually did, they were more likely to report making a social comparison when they were signaled, b � 0.01, 95% CI [0.008, 0.013], SE � 0.002, z � 5.07, p � .001. No other within-person effects were significant, zs � 1.52, ps � .12. In addition, individuals who used social media more, relative to other participants, were more likely to report making a social compari- son when they were signaled, b � 0.05, 95% CI [0.004, 0.07], SE � 0.01, z � 3.28, p � .001. No other between-person effects were significant, zs � 1.28, ps � .20. Context and comparison frequency. Next, we examined whether total time spent on each activity predicted total number of comparisons.22 On average, participants reported making 1.44 comparisons per day (SD � 1.27), of which fewer were made on social media (M � 0.22, SD � 0.42) than in other contexts (M � 1.26, SD � 1.08; t � �8.80, p � .001). However, a Poisson regression that included total time spent on various activities during the study and number of surveys completed revealed that total minutes spent on social media, b � 0.0004, 95% CI [0.0002, 0.0008], SE � 0.00004, z � 9.87, p � .001, and dating apps, b � 0.01, 95% CI [�0.02, 0.02], SE � 0.002, z � 2.99, p � .003, were positively associated with number of comparisons, whereas more time spent face-to-face, b � �0.00004, 95% CI [�0.0002, 0.00004], SE � 0.00001, z � �3.07, p � .002, and on e-mail, b � �0.003, 95% CI [�0.01, 0.01], SE � 0.0007, z � �4.37, p � .001, were negatively associated with number of comparisons. Time spent on all other activities did not predict number of comparisons, zs � 1.75, ps � .08. Furthermore, additional Poisson regressions revealed that num- ber of social media minutes were associated with a greater number of upward, b � 0.004, 95% CI [0.0001, 0.001], SE � 0.0001, z � 6.27, p � .001, lateral, b � 0.0005, 95% CI [0.00003, 0.001], SE � 0.0001, z � 4.14, p � .001, and downward comparisons, b � 0.0005, 95% CI [0.0001, 0.001], SE � 0.0001, z � 5.54, p � .001.23,24,25 Paired t tests indicated that upward comparisons were, however, more common on social media than both lateral, t(78) � 3.14, p � .002, and downward comparisons, t(78) � 3.03, p � .003. In other words, spending more time on social media was associated with making more social comparisons, and these social comparisons were most likely to be upward. The total number of comparisons made on social media was less than one might expect from Studies 1–3. It is possible that partic- ipants in this experience sampling study were simply forgetting about the less consequential comparisons made, or reporting rel- atively few to avoid completing additional questions on the app. Indeed, we allowed participants to report multiple comparisons each time they were signaled; thus, survey length increased if participants reported more comparisons. Past research has shown that participants are more likely to satisfice (i.e., impose a limit on how much effort they will apply to the survey) when responding to more time-consuming diary protocols (Barta, Tennen, & Litt, 2012). Regardless, it is noteworthy that participants made over 10% of their social comparisons on social media. Furthermore, other than spending time on dating apps (another context in which people engage in strategic self-presentation; for a review, see Finkel, Eastwick, Karney, Reis, & Sprecher, 2012), spending time on social media was the only recorded activity that predicted making more frequent comparisons. These results are consistent with our hypothesis that individuals are especially likely to make social comparisons when using social media. 21 We tested whether comparisons made on Facebook and Instagram, the two most common social media platforms, differed in extremity. When individuals reported making a comparison on Instagram, relative to a comparison on Facebook, they reported a comparison that was more extreme in its upwardness; however, this within-person effect was not significant, b � 0.66, SE � 0.39, t(149.07) � 1.67, p � .09. We also tested whether comparisons differed in terms of likelihood of being upward depending on the platform. Facebook and Instagram comparisons were equally likely to be upward, relative to other types of comparisons, b � 0.84, SE � 0.66, z � 1.29, p � .20. 22 The data for total time spent on each activity (except social media, skewness � 1.61) was highly positively skewed, skewness � 2.90. Thus, the bootstrapped confidence intervals for these coefficients may include 0. 23 Time spent on dating apps was the only other activity that predicted making more upward comparisons, b � 0.008 [�0.02, 0.02], SE � 0.002, z � 3.14, p � .002. More time spent emailing, b � �0.003, 95% CI [�0.01, 0.007], SE � 0.001, z � �3.63, p � .001, face-to-face, b � �0.00004, 95% CI [�0.0003, 0.0001], SE � 0.00002, z � �2.29, p � .02, on video calls, b � �0.0006, 95% CI [�0.004, �0.0003], SE � 0.0002, z � �3.39, p � .001, and on other voice chat, b � �0.0007, 95% CI [�.005, 0.001], SE � 0.0004, z � �2.05, p � .04, all predicted making fewer upward comparisons. No other effects were significant, zs � 0.70, ps � .49. 24 Time spent on dating apps was the only other activity that predicted making more lateral comparisons, b � 0.01, 95% CI [�0.02, 0.03], SE � 0.005, z � 2.40, p � .02. More time spent face-to-face, b � �0.0001, 95% CI [�0.0006, 0.00003], SE � 0.0001, z � �2.14, p � .03, predicted making fewer lateral comparisons. No other effects were significant, zs � 1.84, ps � .066. 25 More time spent emailing, b � �0.004, 95% CI [�0.01, 0.009], SE � 0.001, z � �3.05, p � .002, predicted making fewer downward compar- isons. No other effects were significant, zs � 1.81, ps � .07. Figure 8. The association between comparison context (social media news feeds vs. all other contexts) and self-evaluations is mediated by differences in comparison extremity (Study 4). Unstandardized regression coefficients and 95% bootstrapped confidence intervals are reported along the paths they model. Statistics reported within parentheses represent the total effect. The direct and total effects have not been parsed into within- and between-person components. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 17SOCIAL MEDIA COMPARISONS AND SELF-ESTEEM Discussion In summary, consistent with Study 3, Study 4 demonstrates that individuals are more likely to make upward than downward or lateral comparisons when using social media, relative to other contexts. Further, we found evidence that comparisons on social media are more threatening and, in turn, result in worse self- evaluations than those in other contexts, despite being to less close others and in domains that are rated as less important. Addition- ally, we demonstrated that, for any given individual, spending more time on social media was associated with making more social comparisons. This within-person finding is significant because although previous studies have demonstrated associations between time spent on social media and global, retrospective reports of comparison frequency (e.g., de Vries & Kühne, 2015), it was previously unclear whether this finding was because of the possi- bility that people who report making more comparisons also spend more time on social media. In this study, we show that, regardless of whether an individual spends more or less time on social media compared with other people, when that individual spends more time on social media, relative to her or his other daily activities, comparisons are more likely. In addition, relative to other contexts, those comparisons made on social media are more extremely upward and, thus, result in more negative self-evaluations. In summary, this study provides the first clear evidence that time spent on social media results in a greater number of upward comparisons, comparisons that are more extreme in their upward- ness, and, consequently, more negative self-evaluations. General Discussion Taken together, these studies provide the first comprehensive analysis of how individuals make and respond to social compari- sons on social media and the characteristics of these comparisons that differ from those made in other contexts. Indeed, this research provides compelling evidence that social media is associated with frequent and extreme upward comparisons, which in turn have a negative impact on individuals’ self-evaluations, mood, and life satisfaction. Upward Comparison Frequency on Social Media First, we found that social comparisons on social media are frequent, and especially likely to be upward. Although past studies identified this upward direction as a significant feature of compar- isons on social media, they did not separately assess the relative frequency of upward and downward comparisons, instead focusing only on upward comparison frequency (e.g., de Vries & Kühne, 2015; Hanna et al., 2017; Vogel et al., 2014) or general compar- ison frequency (e.g., Cramer et al., 2016; Feinstein et al., 2013; Gerson et al., 2016; Jang et al., 2016; Lee, 2014; Steers et al., 2014). The present studies directly compared the frequency of upward and downward comparisons made by individuals while browsing their own social media news feeds (Study 1), immedi- ately after using social media (Study 3), and in an experience sampling paradigm during a 2-week period (Study 4), confirming that upward comparisons consistently outnumber downward com- parisons on social media. The preponderance of upward compar- isons is consistent with past studies that have found that people generally opt for superior or upward comparison targets (Gerber et al., 2018). However, the present studies are the first to demonstrate that upward comparisons are more frequent when individuals are using social media than when they are using other online technol- ogies (Study 3) or indeed than when they are engaging in most other daily activities (Study 4). These studies are also the first to demonstrate that the more frequent upward comparisons that individuals experience while using social media are associated with more negative outcomes for the self, including worse mood, lower self-esteem, and decreased life-satisfaction. We show that the upward comparisons that indi- viduals make while using social media not only have an immediate negative impact on their self-evaluations (Studies 1– 4) but also cumulative negative effects on their self-esteem, mood, and life- satisfaction (Studies 1 and 3). In past research, investigators typ- ically have examined the outcome of a single comparison, or have compared an upward to a downward comparison on various out- comes such as self-evaluations (e.g., Morse & Gergen, 1970), motivation (e.g., Lockwood, Marshall, & Sadler, 2005; Lockwood & Pinkus, 2008; Thai, Lockwood, & Boksh, 2020) affect (e.g., Buunk, Collins, Taylor, VanYperen, & Dakof, 1990; Gibbons & Gerrard, 1989; Pinkus, Lockwood, Marshall, & Yoon, 2012; Sa- lovey & Rodin, 1984), closeness (e.g., Thai, Lockwood, Pinkus, & Chen, 2016), or domain relevance (e.g., Thai & Lockwood, 2015). In the present studies, we were able to test the cumulative effects of a series of comparisons, examining the relative impact of multiple upward and downward comparisons on state self-esteem, life satis- faction, and mood. Our studies demonstrate that upward rather than downward comparisons have the greatest impact on individuals, and that this impact is overwhelmingly negative. Indeed, any downward or lateral comparisons did little to mitigate the sting of the more prevalent upward comparisons. Further, because upward comparisons are more frequent on social media relative to other contexts, these negative outcomes are more pronounced when individuals are on social media relative to other online contexts (Study 3) or engaging in other daily activities more generally (Study 4). The present studies also provide the first evidence that, com- pared with other activities, spending time on social media in- creases the likelihood of making upward comparisons (Studies 3 and 4). This research provides new evidence to explain the nega- tive outcomes that have been associated with increased social media use (Best et al., 2014) and may also at least partially account for the associations between smartphone use and lower well-being, especially among young adults (Matar Boumosleh & Jaalouk, 2017; Twenge, 2017). When using social media, individuals are especially likely to compare themselves to superior, rather than inferior, others (Studies 1, 3, and 4), and they subsequently feel worse about themselves and less satisfied with their lives. The more time they spend on social media, the more upward compar- isons they make (Study 4), and the worse they subsequently feel about themselves (Studies 1– 4). In summary, these studies suggest that social media may be leading to changes in daily social com- parison behavior; individuals now make more upward compari- sons that are more threatening to the self. Upward Comparison Extremity on Social Media These studies provide evidence that upward comparisons are not only more frequent on social media, they are also more extreme in T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 18 MIDGLEY, THAI, LOCKWOOD, KOVACHEFF, AND PAGE-GOULD their upwardness. Because these studies allowed us to assess, for the first time, the impact of upward comparison extremity within each individual, we were able to evaluate the degree to which the upwardness of each individual comparison would determine its impact. We found that, for any given individual, a comparison that was more extremely upward (than their other comparisons) re- sulted in a more negative effect on their self-evaluations immedi- ately after the comparison (Studies 1, 2, and 4). Additionally, by comparing extremity across context, we were able to show that upward comparisons are more extreme on social media than in other daily activities (Study 4), and that this greater extremity accounts, in part, for the more negative outcomes experienced by individuals using social media. We note that the extremity results in Study 3 did not reach significance; it is possible that we did not have sufficient power to detect this effect. Alternatively, it may be that comparisons that take place in any online context are often more extreme in upwardness, given that much online content, whether on social media, dating platforms, videos, or celebrity news stories, may include carefully curated and predominantly positive content. In the future, it will be important to examine comparisons in additional contexts and evaluate their impact. Self-Esteem and Comparisons on Social Media Up to now, we have discussed the contribution of this research to understanding the negative impact of more frequent and extreme upward comparisons on social media. This research also provides valuable insights into the role of self-esteem in determining com- parison outcomes, in any context. Past research on self-esteem and comparisons has yielded mixed results: Some research indicates that lower self-esteem individuals are especially prone to making upward comparisons (e.g., Locke & Nekich, 2000), whereas other theory and research suggests that low self-esteem individuals may be especially likely to avoid upward comparisons (e.g., Wood, Giordano-Beech, Taylor, Michela, & Gaus, 1994; Wood, Michela, & Giordano, 2000). Furthermore, other research has focused pri- marily on this relationship in the reverse causal direction, suggest- ing that more upward comparisons lead to lower self-esteem (e.g., de Vries & Kühne, 2015; Hanna et al., 2017; Leahey, Crowther, & Mickelson, 2007; Vogel et al., 2014; Wood, 1989). The present studies focused on the role of self-esteem in determining comparison frequency and extremity. Specifically, we found that individuals lower in trait self-esteem made more frequent and more extreme upward comparisons, and as a result, reported greater declines in state self-esteem, life satis- faction, and self-evaluations, than individuals with higher self- esteem. Moreover, we found evidence that lower self-esteem individuals actually interpret information in a way that yields a more extreme upward comparison: Even when information in a social media post was held constant (Study 2), lower self- esteem individuals were more likely to see the poster as supe- rior, and as more extremely superior, than were higher self- esteem individuals. Finally, the more frequent upward comparisons made by low self-esteem individuals tended to lead to negative self-evaluations among individuals in social media, rather than other online, contexts (Study 3). Because social media has dramatically increased opportunities for social comparison, and upward comparison in particular, a browse through a news feed appears to pose a higher risk for individ- uals lower in self-esteem. We note that we had insufficient power to test the moderating impact of self-esteem in Study 4; accordingly, it will be important to examine how self-esteem influences reactions to social comparisons in a variety of con- texts beyond the online contexts evaluated in Study 3. Domain Importance and Closeness We found that comparisons on social media were especially distressing; this was not, however, because of the greater impor- tance of the comparison domains on social media. Social media comparisons were in domains similar in importance to those in other online contexts (Study 3) and were actually in domains less important than those in offline contexts more generally (Study 4). This suggests that social media comparisons are exerting a pow- erful impact despite the fact that they take place in less conse- quential domains. It may be that the carefully packaged informa- tion provided in social media posts packs a punch even when it is focused on relatively trivial domains such as restaurant meals or entertainment viewed, or even when posts simply showcase ran- dom (but attractively presented) selfies taken throughout the day. We note, however, that we assessed domain importance after the comparisons had been made, and a possible reaction to threatening upward comparisons is devaluing the comparison domain (Tesser, 1988; Tesser & Paulhus, 1983). Thus, future studies should exam- ine prior domain importance when examining the impact of social media comparisons on self-evaluations. Similarly, the more negative impact of social media compari- sons was not because of greater closeness of the comparison others. Indeed, although individuals compare with less close others on social media than in other contexts, they nevertheless react more negatively to these comparisons (Study 4). Social media has opened up a vast array of individuals with whom one can compare, and this broad pool of comparison targets appears to intensify rather than dilute the outcome of comparisons. Instead of making occasional comparisons to people from one’s past, at a reunion, wedding or other event, one is forced to view the achievements of less close contacts, whose positive outcomes are carefully pre- sented for maximum impact, whenever one is browsing social media. Every day has become like a high school reunion, in which one is forced to confront the apparently more successful lives of one’s former friends and acquaintances, and comparisons to these remote contacts are no less threatening for their psychological distance. Future Directions and Conclusions The present studies focused on social media comparisons in the context of Facebook (Studies 1– 4) and Instagram (Studies 1, 3, and 4), currently the most popular social media platforms world- wide (Pew Research Center, 2019).26 Facebook, in particular, is used by individuals of all ages for both social contact and business (Greenwood et al., 2016). In recent years, however, a number of 26 In this factsheet, YouTube is listed as the most popular social media channel; however, there is debate over whether it is, primarily, a social media platform (e.g., Abramovich, 2015). Furthermore, because we were interested in comparisons occurring in social media news feeds, we focused on the two most popular platforms with this feature. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 19SOCIAL MEDIA COMPARISONS AND SELF-ESTEEM other social media platforms have increased in popularity (Pew Research Center, 2019), and it will be important to examine the nature and impact of comparisons made on these other platforms. Given that all social media platforms afford individuals an oppor- tunity to carefully manage the self-image they present to the world, we would expect to find a similar preponderance of upward social comparisons. Further, to the extent that different platforms are popular among different age groups, it will be important to examine how social media use may lead to different comparison experiences and consequences among younger and older individuals. Indeed, we note that two of our four studies used first-year psychology student participants. It is possible that the type of posts that young adults see and their reactions to them differ systematically from other populations, such as older adults or those in a different stage of life. First-year undergraduates may be especially likely to compare on social media in general. They are in a period of transition (i.e., from high school to university) and, thus, may be more interested in how well they are adjusting to their new life by comparing themselves to others (Lockwood, Shaughnessy, Fortune, & Tong, 2012), evaluating what they need to do to succeed (Blakemore & Mills, 2014; Suls, Martin, & Wheeler, 2002; Wheeler, Martin, & Suls, 1997). The results of Studies 2 and 3, which involved online community samples, suggest that our findings are not limited to one age group. Nevertheless, future studies should examine social media comparisons across age groups and life stages. In addition, we note that the present studies focused on vertical or status comparisons (upward vs. downward) rather than horizon- tal comparisons (contrastive vs. connective; Locke, 2003). It is possible that some individuals will use social media to seek out lateral or similarity-based comparisons that would result in more communal or positive outcomes (Locke & Nekich, 2000). Alter- natively, it may be that opportunities for horizontal comparisons are more plentiful offline than online, as the former context may provide greater access to information about shared attributes, across multiple domains. For example, joint participation in an activity or discussion with a friend could make common interests or opinions as salient as observations about differences in appear- ance or particular skills. In future research, it will be important to consider the full range of comparisons, both vertical and horizon- tal, that may occur through social media and whether this differs from other contexts. Social media platforms offer individuals unprecedented oppor- tunities to connect with friends, stay in touch with family, share accomplishments, and feel part of a community. Indeed, past research suggests that Facebook can have positive outcomes such as helping maintain relationships (Ellison et al., 2007). The ben- efits of these platforms, however, also come with potential costs. The present studies reveal that upward comparisons on social media are commonplace and have both immediate and cumulative negative outcomes, especially for individuals with low self- esteem. With ongoing growth of social media sites, continued research on how these platforms affect their users— especially those at vulnerable ages or life stages—is essential. References Abramovich, G. (2015, March 20). CMO.com wants to know: Is YouTube a social media platform or a media play? CMO.com. Retrieved from https://www.cmo.com/features/articles/2015/3/18/cwtk-youtube-social- platform-or-media-play.html Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Thousand Oaks, CA: Sage. Alfasi, Y. (2019). The grass is always greener on my Friends’ profiles: The effect of Facebook social comparison on state self-esteem and depres- sion. Personality and Individual Differences, 147, 111–117. http://dx.doi .org/10.1016/j.paid.2019.04.032 Appel, H., Crusius, J., & Gerlach, A. L. (2015). Social comparison, envy, and depression on Facebook: A study looking at the effects of high comparison standards on depressed individuals. Journal of Social and Clinical Psychology, 34, 277–289. http://dx.doi.org/10.1521/jscp.2015 .34.4.277 Appel, H., Gerlach, A. L., & Crusius, J. (2016). The interplay between Facebook use, social comparison, envy, and depression. Current Opin- ion in Psychology, 9, 44 – 49. http://dx.doi.org/10.1016/j.copsyc.2015.10 .006 Back, M. D., Stopfer, J. M., Vazire, S., Gaddis, S., Schmukle, S. C., Egloff, B., & Gosling, S. D. (2010). Facebook profiles reflect actual personality, not self-idealization. Psychological Science, 21, 372–374. http://dx.doi .org/10.1177/0956797609360756 Barta, W. D., Tennen, H., & Litt, M. D. (2012). Measurement reactivity in diary research. In M. R. Mehl & T. S. Conner (Eds.), Handbook of research methods for studying daily life (pp. 108 –123). New York, NY: Guilford Press. Best, P., Manktelow, R., & Taylor, B. (2014). Online communication, social media and adolescent wellbeing: A systematic narrative review. Children and Youth Services Review, 41, 27–36. http://dx.doi.org/10 .1016/j.childyouth.2014.03.001 Blakemore, S. J., & Mills, K. L. (2014). Is adolescence a sensitive period for sociocultural processing? Annual Review of Psychology, 65, 187– 207. http://dx.doi.org/10.1146/annurev-psych-010213-115202 Burke, M., Marlow, C., & Lento, T. (2010). Social network activity and social well-being. Proceedings of the 28th Conference on Human Fac- tors in Computing Systems. Atlanta, GA. Buunk, B. P., Collins, R. L., Taylor, S. E., VanYperen, N. W., & Dakof, G. A. (1990). The affective consequences of social comparison: Either direction has its ups and downs. Journal of Personality and Social Psychology, 59, 1238 –1249. http://dx.doi.org/10.1037/0022-3514.59.6 .1238 Chatard, A., Bocage-Barthélémy, Y., Selimbegović, L., & Guimond, S. (2017). The woman who wasn’t there: Converging evidence that sub- liminal social comparison affects self-evaluation. Journal of Experimen- tal Social Psychology, 73, 1–13. http://dx.doi.org/10.1016/j.jesp.2017.05 .005 Chou, H-T., & Edge, N. (2012). “They are happier and having better lives than I am”: The impact of using Facebook on perceptions of others’ lives. Cyberpsychology, Behavior, and Social Networking, 15, 117–121. http://dx.doi.org/10.1089/cyber.2011.0324 Clement, J. (2020). Daily social media usage worldwide 2012–2019. Retrieved from Statista website https://www.statista.com/statistics/ 433871/daily-social-media-usage-worldwide/#statisticContainer Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155–159. http://dx.doi.org/10.1037/0033-2909.112.1.155 Collins, R. L. (1996). For better or worse: The impact of upward social comparison on self-evaluations. Psychological Bulletin, 119, 51– 69. http://dx.doi.org/10.1037/0033-2909.119.1.51 Cramer, E. M., Song, H., & Drent, A. M. (2016). Social comparison on Facebook: Motivation, affective consequences, self-esteem, and Face- book fatigue. Computers in Human Behavior, 64, 739 –746. http://dx.doi .org/10.1016/j.chb.2016.07.049 Davis, K. (2012). Friendship 2.0: Adolescents’ experiences of belonging and self-disclosure online. Journal of Adolescence, 35, 1527–1536. http://dx.doi.org/10.1016/j.adolescence.2012.02.013 T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 20 MIDGLEY, THAI, LOCKWOOD, KOVACHEFF, AND PAGE-GOULD https://www.cmo.com/features/articles/2015/3/18/cwtk-youtube-social-platform-or-media-play.html https://www.cmo.com/features/articles/2015/3/18/cwtk-youtube-social-platform-or-media-play.html http://dx.doi.org/10.1016/j.paid.2019.04.032 http://dx.doi.org/10.1016/j.paid.2019.04.032 http://dx.doi.org/10.1521/jscp.2015.34.4.277 http://dx.doi.org/10.1521/jscp.2015.34.4.277 http://dx.doi.org/10.1016/j.copsyc.2015.10.006 http://dx.doi.org/10.1016/j.copsyc.2015.10.006 http://dx.doi.org/10.1177/0956797609360756 http://dx.doi.org/10.1177/0956797609360756 http://dx.doi.org/10.1016/j.childyouth.2014.03.001 http://dx.doi.org/10.1016/j.childyouth.2014.03.001 http://dx.doi.org/10.1146/annurev-psych-010213-115202 http://dx.doi.org/10.1037/0022-3514.59.6.1238 http://dx.doi.org/10.1037/0022-3514.59.6.1238 http://dx.doi.org/10.1016/j.jesp.2017.05.005 http://dx.doi.org/10.1016/j.jesp.2017.05.005 http://dx.doi.org/10.1089/cyber.2011.0324 https://www.statista.com/statistics/433871/daily-social-media-usage-worldwide/#statisticContainer https://www.statista.com/statistics/433871/daily-social-media-usage-worldwide/#statisticContainer http://dx.doi.org/10.1037/0033-2909.112.1.155 http://dx.doi.org/10.1037/0033-2909.119.1.51 http://dx.doi.org/10.1016/j.chb.2016.07.049 http://dx.doi.org/10.1016/j.chb.2016.07.049 http://dx.doi.org/10.1016/j.adolescence.2012.02.013 de Vries, D., & Kühne, R. (2015). Facebook and self-perception: Individ- ual susceptibility to negative social comparison on Facebook. Person- ality and Individual Differences, 86, 217–221. http://dx.doi.org/10.1016/ j.paid.2015.05.029 de Vries, D. A., Möller, A. M., Wieringa, M. S., Eigenraam, A. W., & Hamelink, K. (2018). Social comparison as the thief of joy: Emotional consequences of viewing strangers’ Instagram posts. Media Psychology, 21, 222–245. http://dx.doi.org/10.1080/15213269.2016.1267647 Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The satisfaction with life scale. Journal of Personality Assessment, 49, 71– 75. http://dx.doi.org/10.1207/s15327752jpa4901_13 Dorethy, M. D., Fiebert, M. S., & Warren, C. R. (2014). Examining social networking site behaviors: Photo sharing and impression management on Facebook. International Review of Social Sciences and Humanities, 6, 111–116. Edwards, L. J., Muller, K. E., Wolfinger, R. D., Qaqish, B. F., & Scha- benberger, O. (2008). An R2 statistic for fixed effects in the linear mixed model. Statistics in Medicine, 27, 6137– 6157. http://dx.doi.org/10.1002/ sim.3429 Ellison, N. B., Steinfield, C., & Lampe, C. (2007). The benefits of Face- book “friends:” Social capital and college students’ use of online social network sites. Journal of Computer-Mediated Communication, 12, 1143–1168. http://dx.doi.org/10.1111/j.1083-6101.2007.00367.x Fardouly, J., Pinkus, R. T., & Vartanian, L. R. (2017). The impact of appearance comparisons made through social media, traditional media, and in person in women’s everyday lives. Body Image, 20, 31–39. http://dx.doi.org/10.1016/j.bodyim.2016.11.002 Fardouly, J., & Vartanian, L. R. (2015). Negative comparisons about one’s appearance mediate the relationship between Facebook usage and body image concerns. Body Image, 12, 82– 88. http://dx.doi.org/10.1016/j .bodyim.2014.10.004 Feinstein, B., Hershenberg, R., Bhatia, V., Latack, J., Meuwly, N., & Davila, J. (2013). Negative social comparison on Facebook and depres- sive symptoms: Rumination as a mechanism. Psychology of Popular Media Culture, 2, 161–170. http://dx.doi.org/10.1037/a0033111 Fairchild, A. J., & MacKinnon, D. P. (2009). A General Model for Testing Mediation and Moderation Effects. Prevention Science, 10, 87–99. http://dx.doi.org/10.1007/s11121-008-0109-6 Finkel, E. J., Eastwick, P. W., Karney, B. R., Reis, H. T., & Sprecher, S. (2012). Online dating: A critical analysis from the perspective of psy- chological science. Psychological Science in the Public Interest, 13, 3– 66. http://dx.doi.org/10.1177/1529100612436522 Geldhof, G. J., Anthony, K. P., Selig, J. P., & Mendez-Luck, C. A. (2018). Accommodating binary and count variables in mediation: A case for conditional indirect effects. International Journal of Behavioral Devel- opment, 42, 300 –308. http://dx.doi.org/10.1177/0165025417727876 Gerber, J. P., Wheeler, L., & Suls, J. (2018). A social comparison theory meta-analysis 60 years on. Psychological Bulletin, 144, 177–197. http://dx.doi.org/10.1037/bul0000127 Gerson, J., Plagnol, A. C., & Corr, P. J. (2016). Subjective well-being and social media use: Do personality traits moderate the impact of social comparison on Facebook? Computers in Human Behavior, 63, 813– 822. http://dx.doi.org/10.1016/j.chb.2016.06.023 Gibbons, F. X., & Buunk, B. P. (1999). Individual differences in social comparison: Development of a scale of social comparison orientation. Journal of Personality and Social Psychology, 76, 129 –142. http://dx .doi.org/10.1037/0022-3514.76.1.129 Gibbons, F. X., & Gerrard, M. (1989). Effects of upward and downward social comparisons on mood states. Journal of Social and Clinical Psychology, 8, 14 –31. http://dx.doi.org/10.1521/jscp.1989.8.1.14 Gilbert, D. T., Giesler, R. B., & Morris, K. A. (1995). When comparisons arise. Journal of Personality and Social Psychology, 69, 227–236. http://dx.doi.org/10.1037/0022-3514.69.2.227 Greenwood, S., Perrin, A., & Duggan, M. (2016, November 11). Social Media Update 2016. Retrieved from http://www.pewinternet.org/2016/ 11/11/social-media-update-2016/ Hanna, E., Ward, L. M., Seabrook, R. C., Jerald, M., Reed, L., Giaccardi, S., & Lippman, J. R. (2017). Contributions of social comparison and self-objectification in mediating associations between Facebook use and emergent adults’ psychological well-being. Cyberpsychology, Behavior, and Social Networking, 20, 172–179. http://dx.doi.org/10.1089/cyber .2016.0247 Hayes, A. F. (2013). Introduction to mediation, moderation, and condi- tional process analysis: A regression-based approach. New York, NY: Guilford Press. Hayes, A. F. (2015). An index and test of linear moderated mediation. Multivariate Behavioral Research, 50, 1–22. http://dx.doi.org/10.1080/ 00273171.2014.962683 Hayes, A. F., & Preacher, K. J. (2010). Quantifying and testing indirect effects in simple mediation models when the constituent paths are nonlinear. Multivariate Behavioral Research, 45, 627– 660. http://dx.doi .org/10.1080/00273171.2010.498290 Hayes, A. F., & Preacher, K. J. (2014). Statistical mediation analysis with a multicategorical independent variable. British Journal of Mathemati- cal and Statistical Psychology, 67, 451– 470. http://dx.doi.org/10.1111/ bmsp.12028 Heatherton, T. F., & Polivy, J. (1991). Development and validation of a scale for measuring state self-esteem. Journal of Personality and Social Psychology, 60, 895–910. http://dx.doi.org/10.1037/0022-3514.60.6.895 Howell, D. (2013). Statistical methods for psychology (8th ed.). Belmont, CA: Wadsworth Cengage Learning. Hwang, J., Cheong, P., & Feeley, T. (2009). Being young and feeling blue in Taiwan: Examining adolescent depressive mood and online and offline activities. New Media & Society, 11, 1101–1121. http://dx.doi .org/10.1177/1461444809341699 Jang, K., Park, N., & Song, H. (2016). Social comparison on Facebook: Its antecedents and psychological outcomes. Computers in Human Behav- ior, 62, 147–154. http://dx.doi.org/10.1016/j.chb.2016.03.082 Kalpidou, M., Costin, D., & Morris, J. (2011). The relationship between Facebook and the well-being of undergraduate college students. Cyber- psychology, Behavior, and Social Networking, 14, 183–189. http://dx .doi.org/10.1089/cyber.2010.0061 Ko, H.-C., & Kuo, F.-Y. (2009). Can blogging enhance subjective well- being through self-disclosure? CyberPsychology & Behavior, 12, 75–79. http://dx.doi.org/10.1089/cpb.2008.016 Krasnova, H., Wenninger, H., Widjaja, T., & Buxmann, P. (2013). Envy on Facebook: A hidden threat to users’ life satisfaction? 11th International Conference on Wirtschaftsinformatik (pp. 1–16). Leipzig, Germany: Wirtschaftsinformatik. Kross, E., Verduyn, P., Demiralp, E., Park, J., Lee, D. S., Lin, N., . . . Ybarra, O. (2013). Facebook use predicts declines in subjective well- being in young adults. PLoS ONE, 8, e69841. http://dx.doi.org/10.1371/ journal.pone.0069841 Leahey, T. M., Crowther, J. H., & Mickelson, K. D. (2007). The frequency, nature, and effects of naturally occurring appearance-focused social comparisons. Behavior Therapy, 38, 132–143. http://dx.doi.org/10.1016/ j.beth.2006.06.004 Lee, S. Y. (2014). How do people compare themselves with others on social network sites? The case of Facebook. Computers in Human Behavior, 32, 253–260. http://dx.doi.org/10.1016/j.chb.2013.12.009 Liu, D., Ainsworth, S. E., & Baumeister, R. F. (2016). A meta-analysis of social networking online and social capital. Review of General Psychol- ogy, 20, 369 –391. http://dx.doi.org/10.1037/gpr0000091 Liu, Q.-Q., Zhou, Z.-K., Yang, X.-J., Niu, G.-F., Tian, Y., & Fan, C.-Y. (2017). Upward social comparison on social network sites and depres- sive symptoms: A moderated mediation model of self-esteem and opti- T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 21SOCIAL MEDIA COMPARISONS AND SELF-ESTEEM http://dx.doi.org/10.1016/j.paid.2015.05.029 http://dx.doi.org/10.1016/j.paid.2015.05.029 http://dx.doi.org/10.1080/15213269.2016.1267647 http://dx.doi.org/10.1207/s15327752jpa4901_13 http://dx.doi.org/10.1002/sim.3429 http://dx.doi.org/10.1002/sim.3429 http://dx.doi.org/10.1111/j.1083-6101.2007.00367.x http://dx.doi.org/10.1016/j.bodyim.2016.11.002 http://dx.doi.org/10.1016/j.bodyim.2014.10.004 http://dx.doi.org/10.1016/j.bodyim.2014.10.004 http://dx.doi.org/10.1037/a0033111 http://dx.doi.org/10.1007/s11121-008-0109-6 http://dx.doi.org/10.1177/1529100612436522 http://dx.doi.org/10.1177/0165025417727876 http://dx.doi.org/10.1037/bul0000127 http://dx.doi.org/10.1016/j.chb.2016.06.023 http://dx.doi.org/10.1037/0022-3514.76.1.129 http://dx.doi.org/10.1037/0022-3514.76.1.129 http://dx.doi.org/10.1521/jscp.1989.8.1.14 http://dx.doi.org/10.1037/0022-3514.69.2.227 http://www.pewinternet.org/2016/11/11/social-media-update-2016/ http://www.pewinternet.org/2016/11/11/social-media-update-2016/ http://dx.doi.org/10.1089/cyber.2016.0247 http://dx.doi.org/10.1089/cyber.2016.0247 http://dx.doi.org/10.1080/00273171.2014.962683 http://dx.doi.org/10.1080/00273171.2014.962683 http://dx.doi.org/10.1080/00273171.2010.498290 http://dx.doi.org/10.1080/00273171.2010.498290 http://dx.doi.org/10.1111/bmsp.12028 http://dx.doi.org/10.1111/bmsp.12028 http://dx.doi.org/10.1037/0022-3514.60.6.895 http://dx.doi.org/10.1177/1461444809341699 http://dx.doi.org/10.1177/1461444809341699 http://dx.doi.org/10.1016/j.chb.2016.03.082 http://dx.doi.org/10.1089/cyber.2010.0061 http://dx.doi.org/10.1089/cyber.2010.0061 http://dx.doi.org/10.1089/cpb.2008.016 http://dx.doi.org/10.1371/journal.pone.0069841 http://dx.doi.org/10.1371/journal.pone.0069841 http://dx.doi.org/10.1016/j.beth.2006.06.004 http://dx.doi.org/10.1016/j.beth.2006.06.004 http://dx.doi.org/10.1016/j.chb.2013.12.009 http://dx.doi.org/10.1037/gpr0000091 mism. Personality and Individual Differences, 113, 223–228. http://dx .doi.org/10.1016/j.paid.2017.03.037 Lo, I. S., & McKercher, B. (2015). Ideal image in process: Online tourist photography and impression management. Annals of Tourism Research, 52, 104 –116. http://dx.doi.org/10.1016/j.annals.2015.02.019 Locke, K. D. (2003). Status and solidarity in social comparison: Agentic and communal values and vertical and horizontal directions. Journal of Personality and Social Psychology, 84, 619 – 631. http://dx.doi.org/10 .1037/0022-3514.84.3.619 Locke, K. D. (2005). Connecting the horizontal dimension of social com- parison with self-worth and self-confidence. Personality and Social Psychology Bulletin, 31, 795– 803. http://dx.doi.org/10.1177/ 0146167204271634 Locke, K. D., & Nekich, J. C. (2000). Agency and communion in natu- ralistic social comparison. Personality and Social Psychology Bulletin, 26, 864 – 874. http://dx.doi.org/10.1177/0146167200269011 Lockwood, P., & Kunda, Z. (1997). Superstars and me: Predicting the impact of role models on the self. Journal of Personality and Social Psychology, 73, 91–103. http://dx.doi.org/10.1037/0022-3514.73.1.91 Lockwood, P., Marshall, T. C., & Sadler, P. (2005). Promoting success or preventing failure: Cultural differences in motivation by positive and negative role models. Personality and Social Psychology Bulletin, 31, 379 –392. http://dx.doi.org/10.1177/0146167204271598 Lockwood, P., & Pinkus, R. T. (2008). The impact of social comparisons on motivation. In J. Y. Shah & W. L. Gardner (Eds.), Handbook of motivation science (pp. 251–264). New York, NY: Guilford Press. Lockwood, P., Shaughnessy, S. C., Fortune, J. L., & Tong, M.-O. (2012). Social comparisons in novel situations: Finding inspiration during life transitions. Personality and Social Psychology Bulletin, 38, 985–996. http://dx.doi.org/10.1177/0146167212447234 Maniaci, M. R., & Rogge, R. D. (2014). Caring about carelessness: Participant inattention and its effects on research. Journal of Research in Personality, 48, 61– 83. http://dx.doi.org/10.1016/j.jrp.2013.09.008 Matar Boumosleh, J., & Jaalouk, D. (2017). Depression, anxiety, and smartphone addiction in university students- A cross sectional study. PLoS ONE, 12, e0182239. http://dx.doi.org/10.1371/journal.pone .0182239 Morse, S., & Gergen, K. J. (1970). Social comparison, self-consistency, and the concept of self. Journal of Personality and Social Psychology, 16, 148 –156. http://dx.doi.org/10.1037/h0029862 Muise, A., Christofides, E., & Desmarais, S. (2009). More information than you ever wanted: Does Facebook bring out the green-eyed monster of jealousy? Cyberpsychology and Behavior, 12, 441– 444. http://dx.doi .org/10.1089/cpb.2008.0263 Mussweiler, T. (2003). Comparison processes in social judgment: Mech- anisms and consequences. Psychological Review, 110, 472– 489. http:// dx.doi.org/10.1037/0033-295X.110.3.472 Mussweiler, T., Rüter, K., & Epstude, K. (2004). The ups and downs of social comparison: Mechanisms of assimilation and contrast. Journal of Personality and Social Psychology, 87, 832– 844. http://dx.doi.org/10 .1037/0022-3514.87.6.832 Mussweiler, T., & Strack, F. (1999). Hypothesis-consistent testing and semantic priming in the anchoring paradigm: A selective accessibility model. Journal of Experimental Social Psychology, 35, 136 –164. http:// dx.doi.org/10.1006/jesp.1998.1364 Muthén, L. K., & Muthén, B. O. (1998 –2012). Mplus user’s guide (7th ed.). Los Angeles, CA: Author. Page-Gould, E., & Sharples, A. (2016). Bootstrapping accurate indirect effects in multilevel models. Retrieved from http://www.page-gould.com/ r/indirectmlm/ Patrick, H., Neighbors, C., & Knee, C. R. (2004). Appearance-related social comparisons: The role of contingent self-esteem and self- perceptions of attractiveness. Personality and Social Psychology Bulle- tin, 30, 501–514. http://dx.doi.org/10.1177/0146167203261891 Pew Research Center. (2019). Social media fact sheet. Retrieved from http://www.pewinternet.org/fact-sheet/social-media/ Pinkus, R. T., Lockwood, P., Marshall, T. C., & Yoon, H. M. (2012). Responses to comparisons in romantic relationships: Empathy, shared fate, and contrast. Personal Relationships, 19, 182–201. http://dx.doi .org/10.1111/j.1475-6811.2011.01347.x Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple me- diator models. Behavior Research Methods, 40, 879 – 891. http://dx.doi .org/10.3758/BRM.40.3.879 Qiu, L., Lin, H., Leung, A. K., & Tov, W. (2012). Putting their best foot forward: Emotional disclosure on Facebook. Cyberpsychology, Behav- ior, and Social Networking, 15, 569 –572. http://dx.doi.org/10.1089/ cyber.2012.0200 Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press. http://dx.doi.org/10.1515/97814 00876136 Salovey, P., & Rodin, J. (1984). Some antecedents and consequences of social-comparison jealousy. Journal of Personality and Social Psychol- ogy, 47, 780 –792. http://dx.doi.org/10.1037/0022-3514.47.4.780 Seidman, G. (2013). Self-presentation and belonging on Facebook: How personality influences social media use and motivations. Personality and Individual Differences, 54, 402– 407. http://dx.doi.org/10.1016/j.paid .2012.10.009 Stapleton, P., Luiz, G., & Chatwin, H. (2017). Generation validation: The role of social comparison in use of Instagram among emerging adults. Cyberpsychology, Behavior, and Social Networking, 20, 142–149. http:// dx.doi.org/10.1089/cyber.2016.0444 Steers, M. N., Wickham, R. E., & Acitelli, L. K. (2014). Seeing everyone else’s highlight reels: How Facebook usage is linked to depressive symptoms. Journal of Social and Clinical Psychology, 33, 701–731. http://dx.doi.org/10.1521/jscp.2014.33.8.701 Stolzenberg, R. M. (1980). The measurement and decomposition of causal effects in nonlinear andnonadditive models. Sociological Methodology, 11, 459 – 488. Suls, J., Martin, R., & Wheeler, L. (2002). Social comparison: Why, with whom, and with what effect? Current Directions in Psychological Sci- ence, 11, 159 –163. http://dx.doi.org/10.1111/1467-8721.00191 Tandoc, E. C., Jr., Ferrucci, P., & Duffy, M. (2015). Facebook use, envy, and depression among college students: Is Facebooking depressing? Computers in Human Behavior, 43, 139 –146. http://dx.doi.org/10.1016/ j.chb.2014.10.053 Tesser, A. (1988). Toward a self-evaluation maintenance model of social behavior. In L. Berkowitz (Ed.), Advances in experimental social psy- chology (Vol. 21, pp. 181–227). New York, NY: Academic Press. http://dx.doi.org/10.1016/S0065-2601(08)60227-0 Tesser, A., Millar, M., & Moore, J. (1988). Some affective consequences of social comparison and reflection processes: The pain and pleasure of being close. Journal of Personality and Social Psychology, 54, 49 – 61. http://dx.doi.org/10.1037/0022-3514.54.1.49 Tesser, A., & Paulhus, D. (1983). The definition of self: Private and public self-evaluation management strategies. Journal of Personality and So- cial Psychology, 44, 672– 682. http://dx.doi.org/10.1037/0022-3514.44 .4.672 Thai, S., & Lockwood, P. (2015). Comparing you � comparing me: Social comparisons of the expanded self. Personality and Social Psychology Bulletin, 41, 989 –1004. http://dx.doi.org/10.1177/0146167215588094 Thai, S., Lockwood, P., & Boksh, R. J. (2020). Committed to succeed: Commitment determines reactions to upward relationship comparisons. Personal Relationships, 27, 303–335. http://dx.doi.org/10.1111/pere .12316 Thai, S., Lockwood, P., Pinkus, R. T., & Chen, S. Y. (2016). Being better than you is better for us: Attachment avoidance and social comparisons T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 22 MIDGLEY, THAI, LOCKWOOD, KOVACHEFF, AND PAGE-GOULD http://dx.doi.org/10.1016/j.paid.2017.03.037 http://dx.doi.org/10.1016/j.paid.2017.03.037 http://dx.doi.org/10.1016/j.annals.2015.02.019 http://dx.doi.org/10.1037/0022-3514.84.3.619 http://dx.doi.org/10.1037/0022-3514.84.3.619 http://dx.doi.org/10.1177/0146167204271634 http://dx.doi.org/10.1177/0146167204271634 http://dx.doi.org/10.1177/0146167200269011 http://dx.doi.org/10.1037/0022-3514.73.1.91 http://dx.doi.org/10.1177/0146167204271598 http://dx.doi.org/10.1177/0146167212447234 http://dx.doi.org/10.1016/j.jrp.2013.09.008 http://dx.doi.org/10.1371/journal.pone.0182239 http://dx.doi.org/10.1371/journal.pone.0182239 http://dx.doi.org/10.1037/h0029862 http://dx.doi.org/10.1089/cpb.2008.0263 http://dx.doi.org/10.1089/cpb.2008.0263 http://dx.doi.org/10.1037/0033-295X.110.3.472 http://dx.doi.org/10.1037/0033-295X.110.3.472 http://dx.doi.org/10.1037/0022-3514.87.6.832 http://dx.doi.org/10.1037/0022-3514.87.6.832 http://dx.doi.org/10.1006/jesp.1998.1364 http://dx.doi.org/10.1006/jesp.1998.1364 http://www.page-gould.com/r/indirectmlm/ http://www.page-gould.com/r/indirectmlm/ http://dx.doi.org/10.1177/0146167203261891 http://www.pewinternet.org/fact-sheet/social-media/ http://dx.doi.org/10.1111/j.1475-6811.2011.01347.x http://dx.doi.org/10.1111/j.1475-6811.2011.01347.x http://dx.doi.org/10.3758/BRM.40.3.879 http://dx.doi.org/10.3758/BRM.40.3.879 http://dx.doi.org/10.1089/cyber.2012.0200 http://dx.doi.org/10.1089/cyber.2012.0200 http://dx.doi.org/10.1515/9781400876136 http://dx.doi.org/10.1515/9781400876136 http://dx.doi.org/10.1037/0022-3514.47.4.780 http://dx.doi.org/10.1016/j.paid.2012.10.009 http://dx.doi.org/10.1016/j.paid.2012.10.009 http://dx.doi.org/10.1089/cyber.2016.0444 http://dx.doi.org/10.1089/cyber.2016.0444 http://dx.doi.org/10.1521/jscp.2014.33.8.701 http://dx.doi.org/10.1111/1467-8721.00191 http://dx.doi.org/10.1016/j.chb.2014.10.053 http://dx.doi.org/10.1016/j.chb.2014.10.053 http://dx.doi.org/10.1016/S0065-2601%2808%2960227-0 http://dx.doi.org/10.1037/0022-3514.54.1.49 http://dx.doi.org/10.1037/0022-3514.44.4.672 http://dx.doi.org/10.1037/0022-3514.44.4.672 http://dx.doi.org/10.1177/0146167215588094 http://dx.doi.org/10.1111/pere.12316 http://dx.doi.org/10.1111/pere.12316 within romantic relationships. Journal of Social and Personal Relation- ships, 33, 493–514. http://dx.doi.org/10.1177/0265407515578826 Thai, S., & Page-Gould, E. (2018). ExperienceSampler: An open-source scaffold for building smartphone apps for experience sampling. Psycho- logical Methods, 23, 729 –739. http://dx.doi.org/10.1037/met0000151 Tien, S., & Aynsley, M. (2019, July 15). The best time to post on Facebook, Instagram, Twitter, and LinkedIn. Retrieved from Hootsuite website https://blog.hootsuite.com/best-time-to-post-on-facebook- twitter-instagram/ Twenge, J. (2017). IGen: Why today’s super-connected kids are growing up less rebellious, more tolerant, less happy—And completely unpre- pared for adulthood—And what that means for the rest of us. New York, NY: Simon & Schuster. Valkenburg, P. M., Peter, J., & Schouten, A. P. (2006). Friend networking sites and their relationship to adolescents’ well-being and social self- esteem. Cyberpsychology and Behavior, 9, 584 –590. http://dx.doi.org/ 10.1089/cpb.2006.9.584 Verduyn, P., Lee, D. S., Park, J., Shablack, H., Orvell, A., Bayer, J., . . . Kross, E. (2015). Passive Facebook usage undermines affective well- being: Experimental and longitudinal evidence. Journal of Experimental Psychology: General, 144, 480 – 488. http://dx.doi.org/10.1037/ xge0000057 Verduyn, P., Ybarra, O., Résibois, M., Jonides, J., & Kross, E. (2017). Do social network sites enhance or undermine subjective well-being? A critical review. Social Issues and Policy Review, 11, 274 –302. http://dx .doi.org/10.1111/sipr.12033 Vogel, E. A., Rose, J. P., Okdie, B. M., Eckles, K., & Franz, B. (2015). Who compares and despairs? The effect of social comparison orientation on social media use and its outcomes. Personality and Individual Dif- ferences, 86, 249 –256. http://dx.doi.org/10.1016/j.paid.2015.06.026 Vogel, E. A., Rose, J. P., Roberts, L. R., & Eckles, K. (2014). Social comparison, social media, and self-esteem. Psychology of Popular Me- dia Culture, 3, 206 –222. http://dx.doi.org/10.1037/ppm0000047 Vohs, K. D., & Heatherton, T. F. (2004). Ego threats elicits different social comparison process among high and low self-esteem people: Implica- tions for interpersonal perceptions. Social Cognition, 22, 168 –191. http://dx.doi.org/10.1521/soco.22.1.168.30983 Walther, J. B. (2007). Selective self-presentation in computer-mediated communication: Hyperpersonal dimensions of technology, language, and cognition. Computers in Human Behavior, 23, 2538 –2557. http:// dx.doi.org/10.1016/j.chb.2006.05.002 Wang, J.-L., Wang, H.-Z., Gaskin, J., & Hawk, S. (2017). The mediating roles of upward social comparison and self-esteem and the moderating role of social comparison orientation in the association between social networking site usage and subjective well-being. Frontiers in Psychol- ogy, 8, 771. http://dx.doi.org/10.3389/fpsyg.2017.00771 Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and vali- dation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070. http://dx.doi.org/10.1037/0022-3514.54.6.1063 Wayment, H. A., & Taylor, S. E. (1995). Self-evaluation processes: Mo- tives, information use, and self-esteem. Journal of Personality, 63, 729 –757. http://dx.doi.org/10.1111/j.1467-6494.1995.tb00315.x Wheeler, L., Martin, R., & Suls, J. (1997). The proxy model of social comparison for self-assessment of ability. Personality and Social Psy- chology Review, 1, 54 – 61. http://dx.doi.org/10.1207/s15327957 pspr0101_4 Wheeler, L., & Miyake, K. (1992). Social comparison in everyday life. Journal of Personality and Social Psychology, 62, 760 –773. http://dx .doi.org/10.1037/0022-3514.62.5.760 Wheeler, L., & Reis, H. T. (1991). Self-recording of everyday life events: Origins, types, and uses. Journal of Personality, 59, 339 –354. http://dx .doi.org/10.1111/j.1467-6494.1991.tb00252.x Wilson, R. E., Gosling, S. D., & Graham, L. T. (2012). A review of Facebook research in the social sciences. Perspectives on Psychological Science, 7, 203–220. http://dx.doi.org/10.1177/1745691612442904 Wood, J. V. (1989). Theory and research concerning social comparisons of personal attributes. Psychological Bulletin, 106, 231–248. http://dx.doi .org/10.1037/0033-2909.106.2.231 Wood, J. V., Giordano-Beech, M., Taylor, K. L., Michela, J. L., & Gaus, V. (1994). Strategies of social comparison among people with low self-esteem: Self-protection and self-enhancement. Journal of Person- ality and Social Psychology, 67, 713–731. http://dx.doi.org/10.1037/ 0022-3514.67.4.713 Wood, J. V., & Lockwood, P. (1999). Social comparisons in dysphoric and low self-esteem people. In R. Kowalski & M. Leary (Eds.), The social psychology of emotional and behavioral problems: Interfaces of social and clinical psychology (pp. 97–135). Washington, DC: American Psy- chological Association. http://dx.doi.org/10.1037/10320-004 Wood, J. V., Michela, J. L., & Giordano, C. (2000). Downward comparison in everyday life: Reconciling self-enhancement models with the mood- cognition priming model. Journal of Personality and Social Psychology, 79, 563–579. http://dx.doi.org/10.1037/0022-3514.79.4.563 Woods, H. C., & Scott, H. (2016). #Sleepyteens: Social media use in adolescence is associated with poor sleep quality, anxiety, depression and low self-esteem. Journal of Adolescence, 51, 41– 49. http://dx.doi .org/10.1016/j.adolescence.2016.05.008 Zhang, Z., Zyphur, M. J., & Preacher, K. J. (2009). Testing multilevel mediation using hierarchical linear models. Organizational Research Methods, 12, 695–719. http://dx.doi.org/10.1177/1094428108327450 Zhao, S., Grasmuck, S., & Martin, J. (2008). Identity construction on Facebook: Digital empowerment in anchored relationships. Computers in Human Behavior, 24, 1816 –1836. http://dx.doi.org/10.1016/j.chb .2008.02.012 Received January 22, 2019 Revision received June 24, 2020 Accepted June 25, 2020 �T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 23SOCIAL MEDIA COMPARISONS AND SELF-ESTEEM http://dx.doi.org/10.1177/0265407515578826 http://dx.doi.org/10.1037/met0000151 https://blog.hootsuite.com/best-time-to-post-on-facebook-twitter-instagram/ https://blog.hootsuite.com/best-time-to-post-on-facebook-twitter-instagram/ http://dx.doi.org/10.1089/cpb.2006.9.584 http://dx.doi.org/10.1089/cpb.2006.9.584 http://dx.doi.org/10.1037/xge0000057 http://dx.doi.org/10.1037/xge0000057 http://dx.doi.org/10.1111/sipr.12033 http://dx.doi.org/10.1111/sipr.12033 http://dx.doi.org/10.1016/j.paid.2015.06.026 http://dx.doi.org/10.1037/ppm0000047 http://dx.doi.org/10.1521/soco.22.1.168.30983 http://dx.doi.org/10.1016/j.chb.2006.05.002 http://dx.doi.org/10.1016/j.chb.2006.05.002 http://dx.doi.org/10.3389/fpsyg.2017.00771 http://dx.doi.org/10.1037/0022-3514.54.6.1063 http://dx.doi.org/10.1111/j.1467-6494.1995.tb00315.x http://dx.doi.org/10.1207/s15327957pspr0101_4 http://dx.doi.org/10.1207/s15327957pspr0101_4 http://dx.doi.org/10.1037/0022-3514.62.5.760 http://dx.doi.org/10.1037/0022-3514.62.5.760 http://dx.doi.org/10.1111/j.1467-6494.1991.tb00252.x http://dx.doi.org/10.1111/j.1467-6494.1991.tb00252.x http://dx.doi.org/10.1177/1745691612442904 http://dx.doi.org/10.1037/0033-2909.106.2.231 http://dx.doi.org/10.1037/0033-2909.106.2.231 http://dx.doi.org/10.1037/0022-3514.67.4.713 http://dx.doi.org/10.1037/0022-3514.67.4.713 http://dx.doi.org/10.1037/10320-004 http://dx.doi.org/10.1037/0022-3514.79.4.563 http://dx.doi.org/10.1016/j.adolescence.2016.05.008 http://dx.doi.org/10.1016/j.adolescence.2016.05.008 http://dx.doi.org/10.1177/1094428108327450 http://dx.doi.org/10.1016/j.chb.2008.02.012 http://dx.doi.org/10.1016/j.chb.2008.02.012 When Every Day Is a High School Reunion: Social Media Comparisons and Self-Esteem Social Media Is Associated With Threatening Social Comparisons Social Media and Upward Comparison Frequency Social Media and Upward Comparison Extremity Self-Esteem and Social Media Comparisons The Present Research Study 1 Method Participants Procedure Comparison questions Post-social media questionnaire Results Overview of analyses Individual comparisons Domains Direction, extremity, and self-evaluations Self-esteem and self-evaluations Overall session Comparison frequency and postsession outcomes Self-esteem and postsession outcomes Discussion Study 2 Method Participants Procedure Results Overview of analyses Order effects Comparison direction and extremity Self-esteem and comparison extremity Self-evaluations Self-esteem and self-evaluations Discussion Study 3 Method Participants Procedure Prescreen survey Study questionnaire Results Overview of analyses Individual comparisons Domains Domain importance Closeness to target Direction, extremity, and self-evaluations Overall session Comparison frequency Self-esteem and upward comparison frequency Postsession self-evaluations Self-esteem and postsession self-evaluations Discussion Study 4 Method Participants Procedure Results Overview of analyses Individual comparisons Domains Domain importance Closeness to target Comparison direction, extremity, and self-evaluations Overall experience sampling period Context and comparison likelihood Context and comparison frequency Discussion General Discussion Upward Comparison Frequency on Social Media Upward Comparison Extremity on Social Media Self-Esteem and Comparisons on Social Media Domain Importance and Closeness Future Directions and Conclusions References Journals/Journal 3 The Effects of Facebook on Mood in Emerging Adults Erica K. Yuen, Erin A. Koterba, Michael J. Stasio, Renee B. Patrick, Cynthia Gangi, Philip Ash, Kathleen Barakat, Virginia Greene, William Hamilton, and Briana Mansour University of Tampa Social media usage is on the rise, with the majority of American adults using Facebook. The present study examined how Facebook activity affects mood in a subset of emerging adults, specifically undergraduates attending a private 4-year university. Participants (N � 312) were randomly assigned to one of the following 20-min activities: browse the Internet, passively browse others’ Facebook profiles, actively communicate with others on Facebook via messages/posts, or update their own personal profile on Facebook. Participants also completed questionnaires assessing mood, feelings of envy, and perceived meaningfulness of their time online. The results demonstrated that using Facebook led to significantly worsened mood compared with browsing the Internet, especially when participants passively browsed Facebook. Furthermore, perceptions of meaningfulness, but not feelings of envy, mediated the relation- ship between online activity and mood. Overall, these findings add to the mounting evidence that social media use may, at times, adversely affect psychological well-being. Public Policy Relevance Statement College students reported lower mood when passively browsing Facebook compared with other online activities, possibly due in part to feelings of wasted time. Results suggest that repeated use of social media may adversely affect psychological well-being in some emerging adults. Keywords: Facebook, social media, mood, envy In the past decade, social media usage has grown dramatically. In fact, the proportion of American adults using social media rose from 7% in 2005% to 65% in 2015 (Perrin, 2015), an increase of nearly 1,000%. Although many social media sites are popular, Facebook (www.facebook.com) dominates (Duggan, Ellison, Lampe, Lenhart, & Madden, 2015), currently boasting over one billion daily and over 1.7 billion monthly users worldwide (Face- book, 2016). Furthermore, out of users of only one social media site, Facebook is the platform of choice for 79% (Duggan et al., 2015). Therefore, it is important to understand the potential psy- chological effects that using Facebook may have on its users. Although not all individuals worldwide with a Facebook profile use it regularly, the majority appear to do so. Indeed, 70% of Facebook users logged in daily in 2014 (Duggan et al., 2015). In fact, most (87%; Perrin, 2015) are using Facebook multiple times a day, logging in an average of seven times per day (Junco, 2013) at all hours of the day (Pempek, Yermolayeva, & Calvert, 2009). Most studies examining daily Facebook usage rates suggest an average of roughly 30 min of use per day (Jelenchick, Eickhoff, & Moreno, 2013; Pempek et al., 2009). Others report much higher average usage rates, ranging from roughly 60 min per day (Kal- pidou, Costin, & Morris, 2011; Skues, Williams, & Wise, 2012) to 145 min per day (Junco, 2013). Emerging adults, defined as individuals aged 18 –29 (Arnett, 2000), are particularly frequent users of social media. In fact, in 2015, 90% of emerging adults used social media (Perrin, 2015), an increase from 84% in 2013 (Duggan et al., 2015). College stu- dents, a subset of emerging adults, are often studied for their social media usage, and their use of Facebook in particular has received considerable empirical attention. Ample empirical evidence sug- gests that Facebook is ingrained into college students’ daily lives (Pempek et al., 2009; Wilson, Gosling, & Graham, 2012). This is particularly apparent in the staggering 94% of college students who have a Facebook profile (Ellison, Steinfield, & Lampe, 2007). With so many emerging adults logging on frequently, it is essential to understand how Facebook affects this population specifically. An ample literature has investigated the emotional conse- quences of Facebook use in adults and adolescents. However, no clear answer has emerged; some studies have linked Facebook use with negative outcomes, (O’Keeffe & Clarke-Pearson, 2011; Skues et al., 2012), whereas others find no such link (Jelenchick et al., 2013). Some studies even point to positive outcomes associated with Facebook use (Gentile, Twenge, Freeman, & Campbell, 2012; Gonzales & Hancock, 2011). Negative effects of Facebook use have been reported for many psychological constructs such as depression. Several studies have This article was published Online First January 18, 2018. Erica K. Yuen, Erin A. Koterba, Michael J. Stasio, Renee B. Patrick, Cynthia Gangi, Philip Ash, Kathleen Barakat, Virginia Greene, William Hamilton, and Briana Mansour, Department of Psychology, University of Tampa. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Correspondence concerning this article should be addressed to Erica K. Yuen, Department of Psychology, University of Tampa, 401 West Ken- nedy Boulevard, Box Q, Tampa, FL 33606. E-mail: eyuen@ut.edu T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. Psychology of Popular Media Culture © 2018 American Psychological Association 2019, Vol. 8, No. 3, 198 –206 2160-4134/19/$12.00 http://dx.doi.org/10.1037/ppm0000178 198 http://www.facebook.com mailto:eyuen@ut.edu http://dx.doi.org/10.1037/ppm0000178 shown that as time spent on Facebook increases, depressive symp- toms also increase (Pantic et al., 2012). Experience sampling techniques have demonstrated that being on Facebook is linked to worsened mood later in that same day (Kross et al., 2013). In fact, some researchers have gone so far as to suggest that Facebook itself creates a form of depression termed “Facebook depression” (O’Keeffe & Clarke-Pearson, 2011). In other words, spending too much time on social media leads to increased depressive symp- toms, which are in turn tied to other unhealthy behaviors such as substance abuse and risky sexual behaviors. The relationship between Facebook use and depressive symp- toms might not be as straightforward as these studies suggest, however. Several studies have reported no link between Facebook usage and negative consequences such as depression (Jelenchick et al., 2013) or well-being (Lee, Lee, & Kwon, 2011). Other studies found only indirect links between Facebook use and depression; for example, frequently using Facebook was associated with higher levels of envy, which in turn predicted depression (Tandoc, Ferrucci, & Duffy, 2015). Interestingly, Facebook users are not necessarily aware of their feelings of envy (Krasnova, Wenninger, Widjaja, & Buxmann, 2013). These findings suggest that several psychosocial processes, many of which are unknown, may mediate the causal relationship between Facebook use and depressed mood. Sagioglou and Greitemeyer (2014) suggested that the emotional effects of using Facebook are determined by users’ perception of the meaningfulness of the act of using Facebook. In their experi- ment, participants were assigned to one of three conditions: spend 20 min using Facebook, spend 20 min browsing the Internet (without accessing social media), or no additional Internet activity (control group). Participants then rated (a) how meaningful the past 20 min had been, and (b) their current mood levels on the Positive and Negative Affect Schedule (PANAS; Watson, Clark, & Tellegen, 1988). Results suggested that those who spent time on Facebook rated their time as being significantly less meaningful than those who did not spend time on Facebook. Also, the Face- book group reported a significantly lower mood compared with the other groups, which was mediated by perceived meaningfulness of activity. Thus, the authors concluded that if Facebook users be- lieve their time spent on the platform was not meaningful, then detrimental links between Facebook use and mood are present. However, two weaknesses of this study raise cause for inter- preting these results with caution. First, the authors combined the distinct positive and negative scales of the PANAS by reverse scoring the negative affect items to create an aggregate measure of positive affect rather than assessing both dimensions of mood separately. Second, the study used Mechanical Turk, which does not monitor participant activity. Therefore, the authors could not observe whether participants actually completed their assigned tasks, such as using Facebook for 20 min. The authors themselves call for an experimental laboratory study in which oversight of the study tasks can take place to address this issue (Sagioglou & Greitemeyer, 2014). An alternative explanation for the discrepant research findings regarding the effects of Facebook use might be due to the specific Facebook activities participants engage in. Facebook is a broad platform with many features that afford different types of use (Smock, Ellison, Lampe, & Wohn, 2011). For example, features include individual user profile pages (including a “Timeline” and “About Me” section) where one posts information about the self, including hobbies, interests, photos, videos, and “status updates.” In addition, users can post messages, photos, or videos to other people’s profile pages. Another popular feature of Facebook is the News Feed, where information regarding online friends appears. This section includes friends’ status updates, photos, and videos, as well as advertiser-sponsored links. Facebook has private functions as well, such as a messenger application allowing information to be shared privately between specific users. Facebook users’ behaviors can be classified as “passive” or “active” based on whether the user is actively posting information. For example, one could passively view the News Feed, or one could actively engage by posting photos and status updates or typing comments on others’ profiles. Indeed, several studies con- firm that users interact with Facebook in a myriad of ways (Ryan & Xenos, 2011) and engage with Facebook passively more often than actively (Pempek et al., 2009). Many studies approach Facebook use and its subsequent effects in a global fashion rather than parsing out the effects of different features. However, perhaps Facebook researchers should concep- tualize Facebook as a collection of separate tools rather than a global platform (Smock et al., 2011) and investigate the effects of various functions individually. It is possible that how users engage with this particular social media platform might dictate how it affects them. Specifically, perhaps whether one actively posts versus passively views information on Facebook will affect users in different ways (Nadkarni & Hofmann, 2012; Rosen, Whaling, Rab, Carrier, & Cheever, 2013). Some evidence suggests that passive Facebook use can lead to increased envy and worsened mood due to upward comparison. Social comparison theory (Festinger, 1954) suggests that we obtain valuable information about ourselves through the process of com- parison with others. Comparisons can take many forms. They can be upward (viewing another as better than the self), downward (viewing another as inferior to the self), or nondirectional. Time spent on Facebook has indeed been linked to increased social comparisons (Lee, 2014), particularly upward comparisons (John- son & Knobloch-Westerwick, 2014). Social comparisons online have been tied to increased depressive symptomatology (Steers, Wickham, & Acitelli, 2014). Interestingly, the link between com- parisons on Facebook and increased depressive symptoms holds regardless of the direction of the comparison, which is not the case in offline interactions. Both envy and rumination potentially explain this link. When Facebook comparisons induce envy, depressive symptoms in- crease (Tandoc et al., 2015) and life satisfaction decreases (Kras- nova et al., 2013). Also, engaging in comparisons on Facebook is significantly tied to rumination, which in turn significantly predicts depression (Feinstein et al., 2013). Therefore, it appears as though an indirect link between passive Facebook use and depression exists, but that link is mediated by several factors. Not all evidence suggests that online interactions are detrimental to psychological health. In fact, recent literature has documented the supportive and positive effects of actively engaging with Facebook (Oh, Ozkaya, & LaRose, 2014). For those who are actively engaging with the platform (e.g., by posting messages), it appears that loneliness is reduced (Wilson et al., 2012) and self- esteem is enhanced (Gentile et al., 2012; Gonzales & Hancock, 2011). This effect extends to receiving comments from others (Greitemeyer, Mügge, & Bollermann, 2014). These results are T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 199EFFECTS OF FACEBOOK ON MOOD particularly meaningful given that the majority of Facebook users use Facebook “actively” at least some of the time (Duggan et al., 2015; Pempek et al., 2009), even if one’s overall time on Facebook is more passive in nature. It is also possible that the source of the information posted (e.g., a good friend vs. an acquaintance) or the valence of information viewed (e.g., positive news vs. negative news) might lead to the varying mood outcomes described above. Indeed, Facebook users have a variety of online “friends” including family members, current or former friends or romantic partners, work colleagues, and more (Duggan et al., 2015). Viewing information from these sources might differentially induce mood. Furthermore, the emo- tional tone of others’ posts may be passed along to others through emotional contagion. A recent, albeit controversial, study by Fa- cebook employees (Kramer, Guillory, & Hancock, 2014) manip- ulated users’ news feeds to change the emotional tone of the majority of messages users were viewing. Completed without the knowledge of its users, this study assigned nearly 700,000 Face- book users to one of two conditions. One group’s news feeds were manipulated to include more negative posts, whereas the second group was exposed to a higher proportion of positive posts. Ex- perimenters then recorded the emotions present in participants’ subsequent posts. Indeed, exposure to more positive posts in- creased positive posts and reduced negative posts, and exposure to more negative posts increased negative posts and reduced positive posts. Therefore, it appears as though emotional contagion occurs via online means, not just in offline or in-person contexts. This study provides a compelling counterargument that users are “catching” others’ emotions, in contrast to the social comparison hypothesis regarding Facebook use. A final issue exists regarding the literature addressing the effects of Facebook. Namely, the majority of studies investigating the effects of Facebook have been questionnaire-based and nonexperi- mental. As a consequence, it is difficult to determine what actually causes the differences that are present in the literature. Therefore, the present study sought to clarify several issues inherent in the Facebook literature using an experimental design. Specifically, we aimed to examine in college students (a) the effects of active versus passive Facebook use on mood, (b) potential mediators (e.g., envy, perceived meaningfulness) of Facebook’s effect on mood, and (c) the relationship between type of information (i.e., source, valence) viewed on Facebook and mood. Participants with a Facebook account were randomly assigned to participate in one of the following 20-min activities: browse the Internet (control group), passively browse others’ Facebook profiles, actively com- municate with others on Facebook via messages and posts, or update their own personal profile on Facebook. At the end of their online activity, participants completed questionnaires assessing their mood, feelings of envy, and perceived level of meaningful- ness of their online activity. Our hypotheses were as follows: (1) Using Facebook is associated with lower mood (i.e., lower positive affect, higher negative affect); (2) Actively posting on Facebook (e.g., messaging others or updating one’s profile) results in higher mood compared with passively using Facebook; (3) Perceived meaningfulness of online activity mediates the relationship be- tween online activity and mood; (4) Envy mediates the relationship between online activity and mood; (5) Viewing information posted by friends, family, or current romantic partner is associated with higher mood; (6) Viewing information posted by a former roman- tic partner is be associated with lower mood; and (7) Viewing positive news posted by others is associated with lower mood (upward comparison). Method Participants Three hundred twelve undergraduates (79% women), ranging in age from 18 to 25 (M � 18.8), were recruited from general psychology classes to participate in the study. Inclusion criteria included having an existing Facebook account. Those who met the criteria were randomly assigned to one of four conditions, which resulted in 88 participants in the Surveillance condition, 67 in the Communicate condition, 86 in the Profile condition, and 71 in Control condition. The majority of participants were freshmen (70.5%), followed by sophomores (19.2%), juniors (5.4%), and seniors (4.8%). Most participants described their race as Caucasian (76.9%), followed by African American/Black (7.1%), Multiracial (6.7%), Asian/ Pacific Islander (4.2%), Other (4.2%), and Native American (0.3%). Participants identified their ethnicity as Hispanic (17.6%) and Non-Hispanic (81.4%). Students earned extra edit for partic- ipating. As for their social media usage, the majority of participants were frequent Facebook users and reported using the site several times a day (51%) or week (31%). Participants also reported visiting other social media sites several times a day: Instagram (77.6%), Twitter (44.9%), and YouTube (24%). Social media sites with fewer daily visits included Tumblr (9.0%), Pinterest (8.3%), Google Plus (7.7%), LinkedIn (0.6%), and Reddit (0.6%). Measures The Positive and Negative Affect Schedule. This widely established measure consists of 20 adjectives describing emotions (e.g., excited, irritable, ashamed; Watson et al., 1988). Participants indicate to what extent they are currently experiencing each emo- tion on a scale ranging from 1 (very slightly or not at all) to 5 (extremely). The measure is divided into separate subscales for positive affect (PANAS-PA; � � .92 for this sample) and negative affect (PANAS-NA; � � .82 for this sample), with scores ranging from 10 to 50 for each subscale. In our study, participants com- pleted this measure at two different points in time— baseline and directly after completing their assigned online activity. Facebook Envy Scale. This is a seven-item composite scale that measures envy on a 5-point scale, ranging from 1 (strongly agree) to 5 (strongly disagree; � � .81 for this sample; Tandoc et al., 2015). Sample items include “I wish I can travel as much as some of my friends do,” and “Many of my friends have a better life than me.” Perceived Meaningfulness of Activity. After completing their randomly assigned online activities, participants were asked to reflect on the meaningfulness of these activities (Sagioglou & Greitemeyer, 2014). The measure consisted of the following three questions (� � .82 for this sample): “How much do you feel like you have spent your time on something meaningful?”; “How much do you feel like you wasted your time?”; and “How much do you T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 200 YUEN ET AL. feel like you have done something useful?” Responses were given using a scale from 1 (not at all) to 7 (very much). Online Activity Questionnaire. Participants were asked to report on their online activities, which included the following: number of messages sent/posted, time (in minutes) spent viewing information posted by specific individuals (e.g., family members, friends, acquaintances, strangers, current romantic partner, former romantic partner), and time spent viewing positive, negative, and neutral information. Procedure The present study was approved by the university’s institutional review board, and all participants provided informed consent. Participants visited a laboratory in small groups. To minimize bias, participants were told that the purpose of the study was to inves- tigate college students’ reactions to various websites. After consent was given, participants completed pretest measures (via Survey- Monkey) consisting of a demographic questionnaire, a question- naire about frequency of social media use, and the PANAS. To mask the true purpose of the experiment, participants also com- pleted distractor questions that queried preferences for website designs/icons of well-known web browsers, as well as the impor- tance of specific website design elements (e.g., attractiveness, helpfulness, efficiency, controllability). Next, participants were randomly assigned to one of the four conditions in which they participated in an online activity for 20 min. In the Surveillance condition, participants logged into Face- book and were instructed to browse and view material but refrain from posting pictures, sending messages, or “liking” posts. In the Communicate condition, participants logged into Facebook and were instructed to actively communicate with others by posting on others’ timelines, commenting on others’ posts, and sending mes- sages. Participants assigned to the Profile condition were in- structed to stay on their own profile pages and view, edit, or add content to their “About Me” section and own timeline, as well as to “like” or respond to others’ posts on their own profile pages. In the Control condition, participants were instructed to browse the web but refrain from visiting e-mail sites, social networking sites, chatrooms, message boards, and dating sites. At the conclusion of the 20-min activity period, participants were administered posttest measures (again via SurveyMonkey) consisting of the PANAS, Facebook Envy Scale, Perceived Mean- ingfulness of Activity, and the Online Activity Questionnaire. Participants were debriefed and thanked before dismissal. Results The overall goal of the present study was to investigate the psychological consequences of Facebook use experimentally. First, the preliminary analyses are presented. Then, the results of the main analyses are organized by hypothesis. Preliminary Analyses Of the 312 participants, seven were excluded from the analyses because they did not perform their assigned activity. These in- cluded four participants who did not write any messages in the Communicate condition (final n � 63), one participant who spent zero minutes on their profile page in the Profile condition (final n � 85), and two participants who spent their time working on their schoolwork online in the Control condition (final n � 69). Preliminary analyses also revealed significant differences in baseline mood by condition, F(3, 301) � 4.32, p � .01. Partici- pants in the Communicate condition reported significantly higher baseline positive mood than those in the Surveillance and Control conditions. Due to these baseline inconsistencies, all analyses pertaining to mood effects were conducted using analyses of covariance (ANCOVAs; with baseline mood as a covariate) rather than using pretest–posttest change scores. Research comparing the statistical validity of these two strategies has demonstrated the superiority of the former strategy under circumstances in which baseline values differ by condition (Van Breukelen, 2006). We then conducted descriptive analyses on participants’ online activities to identify patterns of behavior within each condition. Participants in the surveillance condition reported spending the most time (in minutes) viewing items posted by friends, followed by organizations, family members, acquaintances, strangers, cur- rent romantic partners, and then former romantic partners. Regard- ing the valence of information, participants spent the most time viewing “neutral” and positive information (Table 1). For the communicate condition, participants spent the majority of their time actively communicating with friends, followed by family members, acquaintances, current romantic partners, orga- nizations, strangers, and then former romantic partners. The over- whelming majority of messages/posts were private messages to one other person. Regarding the valence of information, they spent the most time viewing positive and neutral posts (Table 2). Participants in the profile condition were most likely to report viewing/editing/adding to their timeline and/or photograph al- bums. The next commonly reported activity was viewing/editing/ adding to their “About Me” section (Table 3). Lastly, those participants in the control condition visited various websites. The types of websites most commonly visited, in order from most to least time spent, were as follows: search engines, commerce, entertainment, news, and other informational pages such as Wikipedia (Table 4). Table 1 Reported Facebook Activity for the Surveillance Condition Category M (no. of mins) SD Source of viewed information Friends 5.12 3.40 Organizations 3.58 4.27 Family 2.55 2.72 Acquaintances 2.38 2.51 Strangers 0.94 1.48 Current romantic partners 0.65 1.77 Former romantic partners 0.59 2.17 Valence of viewed information Neutral information/news 7.20 5.40 Specific people’s good news (e.g., posts, messages, and pictures) 5.46 3.69 General positive information 3.89 3.08 Specific people’s bad news (e.g., posts, messages, and pictures) 2.06 2.51 General negative information 1.95 2.48 T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 201EFFECTS OF FACEBOOK ON MOOD General Facebook Use and Mood First we tested Hypothesis 1 that using Facebook is associated with lower mood compared with browsing the Internet. Specifi- cally, we conducted an ANCOVA using a dichotomous condition variable (all Facebook groups combined vs. Internet browsing), and with post PANAS-PA as the dependent variable (DV) and pre PANAS-PA as the covariate. The results demonstrated that Face- book use led to significantly lower positive mood (M � 23.26, Adj. M � 22.95, SD � 9.37) compared with browsing the Internet (M � 24.26, Adj. M � 25.33, SD � 9.04), F(1, 302) � 7.18, p � .01, �p2 � .02 (Table 5). To explore effects on negative affect, a similar analysis was conducted with post PANAS-NA as the DV and pre PANAS-NA as the covariate. No significant main effect of condition was found, F(1, 302) � 1.02, p � .31. Negative mood did not differ signifi- cantly between participants using Facebook (M � 12.56, SD � 4.19) and participants browsing the web (M � 12.16, SD � 3.65). Active Versus Passive Facebook Use To test Hypothesis 2 that actively posting on Facebook (i.e., Communicate and Profile conditions) results in higher mood com- pared with passively using Facebook (i.e., Surveillance condition), we conducted an ANCOVA, with post PANAS-PA as the DV, condition (all four separated) as the independent variable, and pre PANAS-PA as the covariate. There was a significant main effect of condition, F(3, 300) � 3.25, p � .02, �p2 � .03. Simple contrasts revealed that participants in the Surveillance condition had a significantly lower positive mood compared with the control group, 95% confidence interval (CI) � [�5.28, �1.19], p � .01. In other words, viewing Facebook pages passively lead to signif- icantly lower positive mood compared with simply browsing the Internet. No other significant differences for positive mood were found (p � .05). A similar ANCOVA for negative mood revealed no significant effect, F(3, 300) � 1.51, p � .21. As a result, we found only partial support for our hypothesis that active and passive Facebook use differentially affect mood. Perceived Meaningfulness of Activity as a Mediator In addition to a significant effect on positive mood, univariate analyses yielded a significant effect of activity on perceived mean- ingfulness, F(1, 303) � 26.51, p � .01, �p2 � .09. Participants who used Facebook (regardless of the nature of usage) perceived the activity as significantly less meaningful (M � 3.50, SD � 1.58) than those who browsed the Internet (M � 4.58, SD � 1.43). Next, we sought to test Hypothesis 3 that perceived meaning- fulness of online activity mediates the relationship between online activity (using Facebook vs. browsing the Internet) and positive mood. The mediation model and results (with standardized coef- ficients) are presented in Figure 1. Multiple regression analyses (controlling for baseline positive mood) were conducted to assess each component of the model. In concordance with previous findings, condition was a significant predictor of meaningfulness, B � �1.14, t(303) � �5.43, p � .01. Also, condition [B � �2.38, t(303) � �2.68, p � .01] and meaningfulness [B � 0.96, t(303) � 4.05, p � .01] were significant predictors of positive mood, respectively. We then tested for mediation using the bootstrapping method with bias-corrected CIs (Hayes, 2009). The 95% CI for the indirect effect was obtained using 5,000 bootstrapped samples. The results (controlling for baseline mood) revealed that meaningfulness did indeed mediate the relationship between condition and positive mood, B � �1.10, CI � [�1.90, �.50]. Furthermore, the direct effect of condition on positive mood became nonsignificant when controlling for mean- ingfulness, B � �1.29, p � .16. In summary, Facebook usage (compared with browsing the Internet) was viewed as a less meaningfulness activity which, in turn, led to lowered positive mood. Table 3 Reported Facebook Activity for the Profile Condition Category M (no. of mins) SD Viewing/editing/adding to personal timeline and/or photograph albums 13.41 6.26 Viewing/editing/adding to personal “About Me” section 5.04 5.22 Viewing nonprofile pages 1.49 3.70 Table 4 Reported Online Activity for the Control Condition Category M (no. of mins) SD Search engines 4.26 4.31 Commerce 3.22 5.73 Entertainment 3.09 6.00 News 2.66 4.38 Other informational webpages (e.g., Wikipedia) 2.21 4.45 Table 2 Reported Facebook Activity for the Communicate Condition Category M SD Recipient of posts/messages (No. of mins) Friends 7.25 4.13 Family 3.81 4.20 Acquaintances 1.85 2.39 Current romantic partners 1.76 3.34 Organizations 1.68 2.43 Strangers 0.69 1.53 Former romantic partners 0.11 0.66 Valence of viewed information (No. of mins) Specific people’s good news (e.g., posts, messages, and pictures) 5.28 3.81 Neutral information/news 2.90 3.45 General positive information 2.61 2.52 Specific people’s bad news (e.g., posts, messages, and pictures) 1.47 2.28 General negative information 1.11 2.15 Type of messages/posts (No. of posts/ messages) Private (to one other individual) 13.44 28.72 Public 1.78 2.13 Group 0.71 1.77 T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 202 YUEN ET AL. Envy as a Mediator To test Hypothesis 4 that envy mediates the relationship be- tween online activity and mood, we first conducted a one-way ANOVA to determine if Facebook envy differed between condi- tions. However, no significant effect was found, F(1, 286) � 0.54, p � .46. In addition, partial correlations did not find a significant relationship between envy and mood, r � �.07, p � .25. There- fore, a mediational analysis was not conducted. Source of Viewed Information We next tested Hypothesis 5 (viewing information posted by friends, family, or current romantic partner is associated with higher mood) and Hypothesis 6 (viewing information posted by a former romantic partner is associated with lower mood). A corre- lational analysis was conducted to look for relationships between mood and the number of minutes spent viewing information by various sources (e.g., friend, acquaintance, stranger, current ro- mantic partner, former romantic partner) for the surveillance con- dition (Table 6). A partial positive correlation (controlling for baseline negative affect) was found between negative mood and viewing information posted by a former romantic partner, r � .49, p � .01. Hypothesis 5, that viewing information posted by a former partner is associated with lower mood, was supported. A partial positive correlation (controlling for baseline positive affect) was also found between positive mood and viewing infor- mation posted by a current romantic partner, r � .26, p � .02. Viewing information posted by a current partner was associated with increased positive affect (higher mood), thus supporting Hy- pothesis 6. Interestingly, those with higher baseline negative affect were more likely to view information from one’s current partner, r � .31, p � .01. Viewing information posted by acquaintances was associated with decreased positive affect, r � �.36, p � .01. No other significant partial correlations between mood and source of information were found. Valence of Viewed Information We conducted another correlational analysis to test Hypothesis 7 that viewing positive news posted by others is associated with lower mood (upward comparison; Table 7). No significant partial correlations between mood and valence of information (e.g., time spent viewing specific individuals’ good or bad news, general positive or negative information, or neutral information) were found. Discussion Of continuing interest to researchers is whether social media use, especially Facebook use, negatively impacts users’ mood. The existing literature is mixed, with some studies finding an associ- ation between Facebook use and depressed mood, and other stud- ies finding no link or even a relationship to positive well-being (see reviews by Appel, Gerlach, & Crusius, 2016; Steers, 2016). Furthermore, most studies were questionnaire based and did not use an experimental design, thus limiting causal conclusions. The present study investigated this unresolved question by using an experimental design to compare the effects of Facebook use and Internet browsing on mood among a subset of emerging adults. Differences in mood were explored for those who actively or passively engaged in Facebook. Also, special attention was paid to possible mediating factors (e.g., perceived meaningfulness) that potentially play a role in the Facebook–mood connection. Table 5 Means and Standard Deviations by Condition DV Surveillance Communicate Profile Control M SD M SD M SD M SD Positive mood (pre) 24.24 8.25 28.33 8.28 27.13 7.72 24.68 8.10 Positive mood (post) 20.67 8.44 25.52 10.04 24.26 9.26 24.26 9.04 Negative mood (pre) 14.11 4.84 14.43 5.05 14.24 4.89 14.25 4.92 Negative mood (post) 12.88 4.96 12.17 3.92 12.51 3.47 12.16 3.64 Meaningfulness 3.30 1.47 3.68 1.66 3.56 1.61 4.58 1.43 Envy 19.69 4.59 20.36 5.68 19.41 5.77 20.30 5.24 Figure 1. Standardized coefficients for the indirect effect of condition on positive mood through perceived meaningfulness. �� p � .01, all two-tailed. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 203EFFECTS OF FACEBOOK ON MOOD Our findings indicate that Facebook use may have a detrimental effect on mood for some individuals. Specifically, Facebook use leads to lower positive affect compared with browsing the Internet, among emerging adults. Despite a small effect size, this is still concerning given that many emerging adults are heavy users of Facebook (Duggan et al., 2015); hence, cumulative negative ef- fects on mood over time could be substantial. Indeed, over half of the participants in this study alone reported using Facebook several times a day, suggesting a possible elevated risk for lowered mood for these users. Our study also addressed whether active (e.g., posting) versus passive (e.g., reading news feed) engagement in Facebook would affect users differently. Although no difference in mood was detected between active and passive Facebook users, the results did indicate that passive use of Facebook in particular led to decreased positive affect compared with browsing the Internet. However, contrary to our expectations, there was no evidence to suggest that feelings of envy mediated the relationship between Facebook activity and diminished mood (cf. Tandoc et al., 2015). Participants in the surveillance condition did spend more of their time passively looking at neutral content (compared with viewing other people’s positive news), so there may have been less oppor- tunity for envy-related feelings to form via upward comparison. Several other experimental studies that did find social comparison/ envy to underlie the relationship between Facebook use and low- ered affect had participants view appealing “constructed SNS profiles” meant to induce such feelings (for a review, see Appel et al., 2016). In contrast, the present experiment allowed participants to op- erate in a more genuine social media environment that included their own friends and family, exposure to other types of more general information (e.g., news articles from organizations), and the ability to apply filters to hide the content of certain individuals that they do not wish to view. It is plausible that the everyday Facebook experience may not necessarily conjure up these envy- related feelings, at least among emerging adults. Emerging adults (relative to adolescents whose self-identities are more underdevel- oped) may spend less time engaged in social comparison, further explaining why no association was found in our study (Appel et al., 2016; Nesi & Prinstein, 2015). In addition, it is also possible that negative mood from social comparisons occurs more so when viewing the profiles of acquaintances rather than friends. Our study found that viewing information posted by acquaintances was associated with decreased positive affect; however, our partici- pants spent more time viewing information posted by friends, family, and organizations, which could explain why we did not find evidence for envy to mediate the relationship between Face- book use and mood. Interestingly, a recent review article (Appel et al., 2016) concluded that there is not yet enough evidence to support the theory that depressed mood from using Facebook is mediated by social comparison or feelings of envy. Instead, we found that Facebook use in general (vs. browsing the Internet) is perceived as a less meaningful online activity, which, in turn, decreases positive affect. Currently, there seems to be only one other study (Sagioglou & Greitemeyer, 2014) that has investigated and found perceived meaningfulness to be a mediat- ing factor. Furthermore, the present study addressed two important methodological issues present in this prior research. As mentioned previously, for example, Sagioglou and Greitemeyer (2014) used an aggregate measure of positive affect by combining the items from the PANAS-PA with reverse scored items from the PANAS- NA. However, as positive and negative affect have been found to be two separate constructs (Watson et al., 1988), we examined the negative and positive mood subscales separately. Moreover, we addressed a stated limitation/directive in the aforementioned study that concerned the need for participants’ activities to be supervised in a laboratory setting. All of our participants were monitored throughout the duration of the experiment to ensure they complied with study protocol. These modifications help to bolster the sup- porting evidence for the underlying role of perceived meaningful- ness in the connection between Facebook use and mood. We agree with Sagioglou and Greitemeyer’s (2014) assertion that it seems counterintuitive for young adults to involve them- selves in an activity that they see little value in and that ultimately results in lowered mood. These researchers found that individuals may still use Facebook because they incorrectly anticipate that this activity will make them feel better. Although this may generally be the case, there could be special instances when anticipations ring true. For example, our study found that individuals who reported a lower affective state before using Facebook were more likely to look at information posted by their current significant other. Im- portantly, reading postings from one’s existing partner was linked to elevated positive mood, so we can extrapolate that those who engaged in this particular Facebook activity correctly anticipated a mood-related boost. In contrast, viewing information posted by a former romantic partner was associated with increased negative affect. It is also important to acknowledge that emerging adults are engaging in other types of social media in addition to Facebook. We found that many individuals in our sample (51%) reported Table 7 Correlation Between Mood and Valence of Viewed Information Valence Positive mood Negative mood Specific people’s good news (e.g., posts, messages, and pictures) �.09 �.08 Specific people’s bad news (e.g., posts, messages, and pictures) .06 .06 General positive information (e.g., news article) .19 �.13 General negative information (e.g., news article) �.02 �.03 Neutral information/news �.07 .11 Note. n � 86. Table 6 Correlation Between Mood and Source of Viewed Information Source Positive mood Negative mood Organization/Group .09 �.17 Current romantic partner .26� .11 Former romantic partner .14 .49�� Family members �.09 �.04 Friends .11 �.09 Acquaintances �.36�� .01 Strangers �.19 �.15 Note. n � 86. � p � .05. �� p � .01. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 204 YUEN ET AL. visiting Facebook several times a day, but far more (78%) reported visiting Instagram several times a day. According to a Pew Re- search Center survey administered in 2014, social media sites such as Instagram and Twitter are growing in popularity; among emerg- ing adults who use the Internet, slightly over half (53%) now use Instagram and 37% use Twitter (Duggan et al., 2015). Could these more recent forms of social media better fit some emerging adults’ preferences and/or lifestyle, hence making engagement on these other sites a more meaningful, constructive avenue for boosting mood? Emerging adults often turn to Instagram (a video/photo sharing mobile application) to connect with friends, to record their social activity and to become more popular within their peer group (Sheldon & Bryant, 2016). Therefore, it may benefit future re- searchers to investigate the differential effects of various types of social media on mood. Limitations of our study include its reliance on self-report data and convenience sampling. It is conceivable that individuals may have misunderstood questions, provided socially desirable an- swers, and/or misremembered pertinent information. For example, participants reported the numbers of minutes they spent on specific Facebook activities based on memory. Also, the sample was cho- sen based on college students’ willingness to volunteer, which may have limited the generalizability of our findings (Miller, 2013). It would be interesting to explore the effects of Facebook on mood with nonstudents in the emerging adult age-group who may view and experience this type of online activity differently; it is esti- mated that nonstudents account for about half of all emerging adults (Mitchell & Syed, 2015). Future studies may also consider conducting a cross-cultural examination of the differential effects of social media on mood. Kim, Sohn, and Choi (2011) found that American college students are highly motivated to use social networking sites for entertain- ment purposes, whereas Korean students are more interested in receiving information and support from others. Perhaps these different motivations for using social media could influence how this activity is perceived (e.g., meaningfulness) and its potential impact on mood. Continued research into the many ways that using social media influences the emotional experiences of emerging adults will add to our understanding of adjustment and mental health issues during this important developmental time in life. References Appel, H., Gerlach, A. L., & Crusius, J. (2016). The interplay between Facebook use, social comparison, envy, and depression. Current Opin- ion in Psychology, 9, 44 – 49. http://dx.doi.org/10.1016/j.copsyc.2015.10 .006 Arnett, J. J. (2000). Emerging adulthood. A theory of development from the late teens through the twenties. American Psychologist, 55, 469 – 480. http://dx.doi.org/10.1037/0003-066X.55.5.469 Duggan, M., Ellison, N. B., Lampe, C., Lenhart, A., & Madden, M. (2015). Social media update 2014. Washington, DC: Pew Research Center. Ellison, N. B., Steinfield, C., & Lampe, C. (2007). The benefits of Face- book “friends:” Social capital and college students’ use of online social network sites. Journal of Computer-Mediated Communication, 12, 1143–1168. http://dx.doi.org/10.1111/j.1083-6101.2007.00367.x Facebook. (2016). Retrieved from http://newsroom.fb.com/company-info/ Feinstein, B. A., Hershenberg, R., Bhatia, V., Latack, J. A., Meuwly, N., & Davila, J. (2013). Negative social comparison on Facebook and depres- sive symptoms: Rumination as a mechanism. Psychology of Popular Media Culture, 2, 161–170. http://dx.doi.org/10.1037/a0033111 Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7, 117–140. http://dx.doi.org/10.1177/001872675400700202 Gentile, B., Twenge, J. M., Freeman, E. C., & Campbell, W. K. (2012). The effect of social networking websites on positive self-views: An experimental investigation. Computers in Human Behavior, 28, 1929 – 1933. http://dx.doi.org/10.1016/j.chb.2012.05.012 Gonzales, A. L., & Hancock, J. T. (2011). Mirror, mirror on my Facebook wall: Effects of exposure to Facebook on self-esteem. Cyberpsychology, Behavior, and Social Networking, 14, 79 – 83. http://dx.doi.org/10.1089/ cyber.2009.0411 Greitemeyer, T., Mügge, D. O., & Bollermann, I. (2014). Having respon- sive Facebook friends affects the satisfaction of psychological needs more than having many Facebook friends. Basic and Applied Social Psychology, 36, 252–258. http://dx.doi.org/10.1080/01973533.2014 .900619 Hayes, A. (2009). Beyond Baron and Kenny: Statistical mediation analysis in the new millenium. Communication Monographs, 76, 408 – 420. http://dx.doi.org/10.1080/03637750903310360 Jelenchick, L. A., Eickhoff, J. C., & Moreno, M. A. (2013). “Facebook depression?” social networking site use and depression in older adoles- cents. Journal of Adolescent Health, 52, 128 –130. http://dx.doi.org/10 .1016/j.jadohealth.2012.05.008 Johnson, B. K., & Knobloch-Westerwick, S. (2014). Glancing up or down: Mood management and selective social comparisons on social network- ing sites. Computers in Human Behavior, 41, 33–39. http://dx.doi.org/ 10.1016/j.chb.2014.09.009 Junco, R. (2013). Comparing actual and self-reported measures of Face- book use. Computers in Human Behavior, 29, 626 – 631. http://dx.doi .org/10.1016/j.chb.2012.11.007 Kalpidou, M., Costin, D., & Morris, J. (2011). The relationship between Facebook and the well-being of undergraduate college students. Cyber- psychology, Behavior, and Social Networking, 14, 183–189. http://dx .doi.org/10.1089/cyber.2010.0061 Kim, Y., Sohn, D., & Choi, S. M. (2011). Cultural difference in motiva- tions for using social network sites: A comparative study of American and Korean college students. Computers in Human Behavior, 27, 365– 372. http://dx.doi.org/10.1016/j.chb.2010.08.015 Kramer, A. D., Guillory, J. E., & Hancock, J. T. (2014). Experimental evidence of massive-scale emotional contagion through social networks. Proceedings of the National Academy of Sciences of the United States of America, 111, 8788 – 8790. http://dx.doi.org/10.1073/pnas.1320040111 Krasnova, H., Wenninger, H., Widjaja, T., & Buxmann, P. (2013, Febru- ary). Envy on Facebook: A hidden threat to users’ life satisfaction? Paper presented at the 11th International Conference on Wirtschaftsin- formatik, Leipzig, Germany. Kross, E., Verduyn, P., Demiralp, E., Park, J., Lee, D. S., Lin, N., . . . Ybarra, O. (2013). Facebook use predicts declines in subjective well- being in young adults. PLoS ONE, 8, e69841. http://dx.doi.org/10.1371/ journal.pone.0069841 Lee, G., Lee, J., & Kwon, S. (2011). Use of social-networking sites and subjective well-being: A study in South Korea. Cyberpsychology, Be- havior, and Social Networking, 14, 151–155. http://dx.doi.org/10.1089/ cyber.2009.0382 Lee, S. Y. (2014). How do people compare themselves with others on social network sites? The case of Facebook. Computers in Human Behavior, 32, 253–260. http://dx.doi.org/10.1016/j.chb.2013.12.009 Miller, S. A. (2013). Developmental research methods (4th ed.). Thousand Oaks, CA: Sage. Mitchell, L. L., & Syed, M. (2015). Does college matter for emerging adulthood? Comparing developmental trajectories of educational groups. Journal of Youth and Adolescence, 44, 2012–2027. http://dx.doi .org/10.1007/s10964-015-0330-0 T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 205EFFECTS OF FACEBOOK ON MOOD http://dx.doi.org/10.1016/j.copsyc.2015.10.006 http://dx.doi.org/10.1016/j.copsyc.2015.10.006 http://dx.doi.org/10.1037/0003-066X.55.5.469 http://dx.doi.org/10.1111/j.1083-6101.2007.00367.x http://newsroom.fb.com/company-info/ http://dx.doi.org/10.1037/a0033111 http://dx.doi.org/10.1177/001872675400700202 http://dx.doi.org/10.1016/j.chb.2012.05.012 http://dx.doi.org/10.1089/cyber.2009.0411 http://dx.doi.org/10.1089/cyber.2009.0411 http://dx.doi.org/10.1080/01973533.2014.900619 http://dx.doi.org/10.1080/01973533.2014.900619 http://dx.doi.org/10.1080/03637750903310360 http://dx.doi.org/10.1016/j.jadohealth.2012.05.008 http://dx.doi.org/10.1016/j.jadohealth.2012.05.008 http://dx.doi.org/10.1016/j.chb.2014.09.009 http://dx.doi.org/10.1016/j.chb.2014.09.009 http://dx.doi.org/10.1016/j.chb.2012.11.007 http://dx.doi.org/10.1016/j.chb.2012.11.007 http://dx.doi.org/10.1089/cyber.2010.0061 http://dx.doi.org/10.1089/cyber.2010.0061 http://dx.doi.org/10.1016/j.chb.2010.08.015 http://dx.doi.org/10.1073/pnas.1320040111 http://dx.doi.org/10.1371/journal.pone.0069841 http://dx.doi.org/10.1371/journal.pone.0069841 http://dx.doi.org/10.1089/cyber.2009.0382 http://dx.doi.org/10.1089/cyber.2009.0382 http://dx.doi.org/10.1016/j.chb.2013.12.009 http://dx.doi.org/10.1007/s10964-015-0330-0 http://dx.doi.org/10.1007/s10964-015-0330-0 Nadkarni, A., & Hofmann, S. G. (2012). Why do people use Facebook? Personality and Individual Differences, 52, 243–249. http://dx.doi.org/ 10.1016/j.paid.2011.11.007 Nesi, J., & Prinstein, M. J. (2015). Using social media for social compar- ison and feedback-seeking: Gender and popularity moderate associations with depressive symptoms. Journal of Abnormal Child Psychology, 43, 1427–1438. http://dx.doi.org/10.1007/s10802-015-0020-0 Oh, H. J., Ozkaya, E., & LaRose, R. (2014). How does online social networking enhance life satisfaction? The relationships among online supportive interaction, affect, perceived social support, sense of com- munity, and life satisfaction. Computers in Human Behavior, 30, 69 –78. http://dx.doi.org/10.1016/j.chb.2013.07.053 O’Keeffe, G. S., & Clarke-Pearson, K. (2011). The impact of social media on children, adolescents, and families. Pediatrics, 127, 800 – 804. http:// dx.doi.org/10.1542/peds.2011-0054 Pantic, I., Damjanovic, A., Todorovic, J., Topalovic, D., Bojovic-Jovic, D., Ristic, S., & Pantic, S. (2012). Association between online social net- working and depression in high school students: Behavioral physiology viewpoint. Psychiatria Danubina, 24, 90 –93. Pempek, T. A., Yermolayeva, Y. A., & Calvert, S. L. (2009). College students’ social networking experiences on Facebook. Journal of Ap- plied Developmental Psychology, 30, 227–238. http://dx.doi.org/10 .1016/j.appdev.2008.12.010 Perrin, A. (2015). Social networking usage: 2005–2015. Washington, DC: Pew Research Center. Rosen, L. D., Whaling, K., Rab, S., Carrier, L. M., & Cheever, N. A. (2013). Is Facebook creating “iDisorders”? The link between clinical symptoms of psychiatric disorders and technology use, attitudes and anxiety. Computers in Human Behavior, 29, 1243–1254. http://dx.doi .org/10.1016/j.chb.2012.11.012 Ryan, T., & Xenos, S. (2011). Who uses Facebook? An investigation into the relationship between the Big Five, shyness, narcissism, loneliness, and Facebook usage. Computers in Human Behavior, 27, 1658 –1664. http://dx.doi.org/10.1016/j.chb.2011.02.004 Sagioglou, C., & Greitemeyer, T. (2014). Facebook’s emotional conse- quences: Why Facebook causes a decrease in mood and why people still use it. Computers in Human Behavior, 35, 359 –363. http://dx.doi.org/ 10.1016/j.chb.2014.03.003 Sheldon, P., & Bryant, K. (2016). Instagram: Motives for its use and relationship to narcissism and contextual age. Computers in Human Behavior, 58, 89 –97. http://dx.doi.org/10.1016/j.chb.2015.12.059 Skues, J. L., Williams, B., & Wise, L. (2012). The effects of personality traits, self-esteem, loneliness, and narcissism on Facebook use among university students. Computers in Human Behavior, 28, 2414 –2419. http://dx.doi.org/10.1016/j.chb.2012.07.012 Smock, A. D., Ellison, N. B., Lampe, C., & Wohn, D. Y. (2011). Facebook as a toolkit: A uses and gratification approach to unbundling feature use. Computers in Human Behavior, 27, 2322–2329. http://dx.doi.org/10 .1016/j.chb.2011.07.011 Steers, M.-L. N. (2016). ‘It’s complicated’: Facebook’s relationship with the need to belong and depression. Current Opinion in Psychology, 9, 22–26. http://dx.doi.org/10.1016/j.copsyc.2015.10.007 Steers, M.-L. N., Wickham, R. E., & Acitelli, L. K. (2014). Seeing everyone else’s highlight reels: How Facebook usage is linked to de- pressive symptoms. Journal of Social and Clinical Psychology, 33, 701–731. http://dx.doi.org/10.1521/jscp.2014.33.8.701 Tandoc, E. C., Jr., Ferrucci, P., & Duffy, M. (2015). Facebook use, envy, and depression among college students: Is facebooking depressing? Computers in Human Behavior, 43, 139 –146. http://dx.doi.org/10.1016/ j.chb.2014.10.053 Van Breukelen, G. J. P. (2006). ANCOVA versus change from baseline had more power in randomized studies, more bias in nonrandomized studies [corrected]. Journal of Clinical Epidemiology, 59, 920 –925. http://dx.doi.org/10.1016/j.jclinepi.2006.02.007 Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and vali- dation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070. http://dx.doi.org/10.1037/0022-3514.54.6.1063 Wilson, R. E., Gosling, S. D., & Graham, L. T. (2012). A review of Facebook research in the social sciences. Perspectives on Psychological Science, 7, 203–220. http://dx.doi.org/10.1177/1745691612442904 Received July 20, 2017 Revision received November 20, 2017 Accepted November 28, 2017 � T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 206 YUEN ET AL. http://dx.doi.org/10.1016/j.paid.2011.11.007 http://dx.doi.org/10.1016/j.paid.2011.11.007 http://dx.doi.org/10.1007/s10802-015-0020-0 http://dx.doi.org/10.1016/j.chb.2013.07.053 http://dx.doi.org/10.1542/peds.2011-0054 http://dx.doi.org/10.1542/peds.2011-0054 http://dx.doi.org/10.1016/j.appdev.2008.12.010 http://dx.doi.org/10.1016/j.appdev.2008.12.010 http://dx.doi.org/10.1016/j.chb.2012.11.012 http://dx.doi.org/10.1016/j.chb.2012.11.012 http://dx.doi.org/10.1016/j.chb.2011.02.004 http://dx.doi.org/10.1016/j.chb.2014.03.003 http://dx.doi.org/10.1016/j.chb.2014.03.003 http://dx.doi.org/10.1016/j.chb.2015.12.059 http://dx.doi.org/10.1016/j.chb.2012.07.012 http://dx.doi.org/10.1016/j.chb.2011.07.011 http://dx.doi.org/10.1016/j.chb.2011.07.011 http://dx.doi.org/10.1016/j.copsyc.2015.10.007 http://dx.doi.org/10.1521/jscp.2014.33.8.701 http://dx.doi.org/10.1016/j.chb.2014.10.053 http://dx.doi.org/10.1016/j.chb.2014.10.053 http://dx.doi.org/10.1016/j.jclinepi.2006.02.007 http://dx.doi.org/10.1037/0022-3514.54.6.1063 http://dx.doi.org/10.1177/1745691612442904 The Effects of Facebook on Mood in Emerging Adults Method Participants Measures The Positive and Negative Affect Schedule Facebook Envy Scale Perceived Meaningfulness of Activity Online Activity Questionnaire Procedure Results Preliminary Analyses General Facebook Use and Mood Active Versus Passive Facebook Use Perceived Meaningfulness of Activity as a Mediator Envy as a Mediator Source of Viewed Information Valence of Viewed Information Discussion References Journals/journal 4 Social Media and Depression Symptoms: A Network Perspective George Aalbers University of Amsterdam Richard J. McNally Harvard University Alexandre Heeren Université Catholique de Louvain Sanne de Wit and Eiko I. Fried University of Amsterdam Passive social media use (PSMU)—for example, scrolling through social media news feeds— has been associated with depression symptoms. It is unclear, however, if PSMU causes depression symptoms or vice versa. In this study, 125 students reported PSMU, depression symptoms, and stress 7 times daily for 14 days. We used multilevel vector autoregressive time-series models to estimate (a) contemporaneous, (b) temporal, and (c) between-subjects associations among these variables. (a) More time spent on PSMU was associated with higher levels of interest loss, concentration problems, fatigue, and loneliness. (b) Fatigue and loneliness predicted PSMU across time, but PSMU predicted neither depression symptoms nor stress. (c) Mean PSMU levels were positively correlated with several depression symptoms (e.g., depressed mood and feeling inferior), but these associations disappeared when controlling for all other variables. Altogether, we identified complex relations between PSMU and specific depression symptoms that warrant further research into potentially causal relationships. Keywords: social media, depression, loneliness, stress, network analysis Supplemental materials: http://dx.doi.org/10.1037/xge0000528.supp In the past decade, social media such as Facebook and Twitter have become central to everyday life. Despite their popularity, controversy abounds regarding their impact on mental health (Twenge, 2017). Although some studies have shown that social media use is associated with beneficial effects (e.g., higher self- esteem; Gonzales & Hancock, 2011), others have identified po- tential negative effects on well-being via the promotion of stress (Meier, Reinecke, & Meltzer, 2016), loneliness (Liu & Baumeis- ter, 2016), and depression symptoms (Appel, Gerlach, & Crusius, 2016). Social media’s adverse effects may come from passive social media use (PSMU)—that is, scrolling through news feeds or brows- ing photographs of friends. Experimental research has shown that PSMU decreases affective well-being (Verduyn et al., 2015), sense of belonging (Tobin, Vanman, Verreynne, & Saeri, 2015), and life satisfaction (Wenninger, Krasnova, & Buxmann, 2014). Furthermore, cross-sectional research indicates that PSMU positively correlates with depressed mood (Frison & Eggermont, 2016). As depressed mood is a core symptom of and a strong predictor of depression (Boschloo, van Borkulo, Borsboom, & Schoevers, 2016), this obser- vation suggests that PSMU may constitute a risk factor for depression. To investigate this possibility, we assessed the link between PSMU and depression symptoms from a network perspective. According to the network perspective on psychopathology, de- pression is a complex, dynamic network of symptoms that cause each other (Borsboom, 2017). Consider the example of a divorced man who, ruminating about his romantic loss, cannot sleep for days and grows tired as time wears on. He descends into a state of hopelessness and anhedonia, causing him to withdraw from social life and alienate his loved ones. At night, he worries about his problems, causing more stress and sleep loss. As illustrated by this example, the network perspective posits that external conditions, such as stress (Fried, Nesse, Guille, & Sen, 2015), could trigger symptoms that activate other symptoms. Therefore, from a net- This article was published Online First December 3, 2018. George Aalbers, Department of Clinical Psychology, University of Amster- dam; Richard J. McNally, Department of Psychology, Harvard University; Alexandre Heeren, Department of Psychological Sciences Research Institute and Institute of Neuroscience, Université Catholique de Louvain; Sanne de Wit, Department of Clinical Psychology, University of Amsterdam; Eiko I. Fried, Department of Psychological Research Methods, University of Amster- dam. Eiko I. Fried is now at the Department of Clinical Psychology, Leiden University. We thank Payton J. Jones, Olivia Losiewicz, and Sacha Epskamp for their comments on this study. The research design and findings in this article have been discussed in the authors’ lab meetings (McNally Laboratory, Richard J. McNally; Habit Lab, Sanne de Wit) and orally presented to a student audience at the University of Amsterdam. Data, materials, and findings have not been shared online prior to resubmitting the revised article. George Aalbers devel- oped the study concept under the supervision of Richard J. McNally, Eiko I. Fried, Sanne de Wit, and Alexandre Heeren. All authors contributed to the study design. Data collection and analysis, and interpretation of results were performed by George Aalbers under the supervision of Eiko I. Fried. George Aalbers drafted the article, and all other authors provided critical revisions. All authors approved the final version of the article for submission. Correspondence concerning this article should be addressed to George Aalbers, who is now at the Nederlands Psychoanalytisch Instituut (NPI), Domselaerstraat 128, 1093MB Amsterdam, the Netherlands. E-mail: h.j.g .aalbers@gmail.com T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. Journal of Experimental Psychology: General © 2018 American Psychological Association 2019, Vol. 148, No. 8, 1454 –1462 0096-3445/19/$12.00 http://dx.doi.org/10.1037/xge0000528 1454 mailto:h.j.g.aalbers@gmail.com mailto:h.j.g.aalbers@gmail.com http://dx.doi.org/10.1037/xge0000528 work perspective, PSMU could function as a depression risk factor if it triggers individual depression symptoms (e.g., depressed mood) or conditions (e.g., stress) that trigger other depression symptoms. Previous research points to several pathways from PSMU to depression symptoms. First, PSMU could worsen symptoms (i.e., depressed mood and loss of interest) by undermining affective well-being (Verduyn et al., 2015). Furthermore, PSMU may re- duce a sense of belonging (Tobin et al., 2015) that predicts lone- liness (Mellor, Stokes, Firth, Hayashi, & Cummins, 2008). By indirectly increasing loneliness, PSMU could increase depression symptoms (Fried et al., 2015) and stress (DeBerard & Kleinknecht, 1995), and, in turn, stress may ultimately reinforce depression symptoms (Fried et al., 2015). Finally, by exposing individuals to the highly curated lives of their social media contacts—who (on average) seem happier and more popular than themselves (Bollen, Goncalves, van de Leemput, & Ruan, 2017)—PSMU could in- crease feelings of inferiority (Appel, Crusius, & Gerlach, 2015), leading to increases in depression symptoms (Blease, 2015). Conversely, depression symptoms, loneliness, and stress might increase PSMU. Longitudinal evidence demonstrates that loneli- ness predicts more social media use (Kross et al., 2013). Further- more, as individuals use the Internet to alleviate depressed mood and loneliness (LaRose, Lin, & Eastin, 2003), it is conceivable that they also use social media to do so. Although researchers have not assessed directly whether PSMU ameliorates depression symp- toms and loneliness, social media users do report that they use PSMU to reduce stress and to relieve boredom (Whiting & Wil- liams, 2013), which is positively associated with loss of interest (Goldberg, Eastwood, LaGuardia, & Danckert, 2011). Finally, repeated PSMU to reduce aversive states may become habitual. By that point, aversive states could trigger PSMU automatically and outside awareness (LaRose, 2010). Accordingly, we hypothesized that PSMU increases depression symptoms, loneliness, and stress, and vice versa. We instructed participants to report social media use, depression symptoms, loneliness, and stress seven times a day for 14 days. From this high-intensive time-series dataset, we estimated three types of network structures: contemporaneous associations, rep- resenting how variables are associated within the same timeframe (e.g., associations between depressed mood and fatigue within a timeframe of 2 hr; these likely reflect fast-moving temporal pro- cesses occurring at a time interval quicker than the sampling interval; Epskamp, Waldorp, Mõttus, & Borsboom, 2018); tempo- ral associations, representing how variables are associated from one time point to next (e.g., fatigue predicts depressed mood during the next timeframe; such temporal prediction satisfies the temporal requirement for causation—i.e., that causes must precede effects—which means that temporal associations could suggest potential causal pathways between variables); and between- subjects associations, representing how within-person mean levels of variables are associated on a larger time-scale (e.g., mean level of fatigue across participants relates to the mean level of loss of interest). Investigating networks by using this threefold framework has become standard practice, and allows complementary views of the data (e.g., Epskamp, Borsboom, & Fried, 2018; Epskamp et al., 2017). We expected that PSMU, depression symptoms, loneliness, and stress would be interconnected by positive temporal and contemporaneous associations, and included between-subjects as- sociations to explore whether, on average, participants with higher levels on PSMU were also higher on depression symptoms, lone- liness, and stress. Method Participants We recruited undergraduate psychology students (N � 132; 91 females, 41 males) via an online study participation platform. A priori power analysis has not yet been developed for the analytic approach used here, so we tried to collect as many participants as possible within the timeframe of 3 months for George Aalbers’s master’s thesis. Notably, the sample size is larger than many recently published studies using the same methodology (e.g., De- Jonckheere et al., 2017; Pe et al., 2015). Prior to data analysis, we excluded seven participants who failed to respond to a minimum number of measurements (�29 out of 98), a cut-off that we chose after consulting with an experience sampling methodology (ESM) expert (M.C. Wichers, personal communication, May 15, 2017). Included participants (N � 125; 87 females, 38 males) had a mean age of 20.44 years (SD � 1.96) and completed an average of 66.18 measurements (SD � 15.10), with a range between 29 and 92. Participants received research credits required to complete their curriculum’s mandatory study participation. Procedure At fixed times, participants received prompts on their smart- phones to complete a 12-item questionnaire (measuring PSMU, depression symptoms, loneliness, and stress) seven times daily for 14 days. We used the LifeData Company’s RealLife Exp app (https://www.lifedatacorp.com/) to prompt participants and collect data (Runyan et al., 2013). We chose to separate measurements by brief intervals (�2 hr) to investigate subtle dynamical interplay among the measured variables. During an initial one-on-one in- structional session with each participant, George Aalbers demon- strated the smartphone app and defined PSMU as: “You are using social media without commenting, posting, sharing, or chatting— that is, you are scrolling through the news feed, looking at photos, videos, and status updates shared by your social media contacts or public profiles that you follow.” This procedure was approved by the University of Amsterdam’s Institutional Review Board. Materials We constructed a 12-item questionnaire for the present ESM study. To minimize participant burden, we prioritized brevity and selected items most relevant in light of the literature on the effects of PSMU on depression symptoms. We paraphrased seven items that commonly occur in the most widely used depression ques- tionnaires (Fried, 2017) to measure depressed mood, loss of inter- est, fatigue, concentration problems, feelings of loneliness, inferi- ority, and hopelessness. To measure stress, we modified a validated one-item stress measure (Elo, Leppänen, & Jahkola, 2003). In addition to an item measuring time spent on PSMU, we included an item to measure time spent on active social media use (ASMU), thereby enabling us to disentangle the effects of passive versus active use of social media. Furthermore, we paraphrased an T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 1455SOCIAL MEDIA AND DEPRESSION SYMPTOMS https://www.lifedatacorp.com/ item from the Self-Report Habit Index (Verplanken & Orbell, 2003) to measure PSMU automaticity. However, because so many participants failed to answer the automaticity item, in light of a lot of missing data, we excluded this item from analyses. Finally, to distract participants from the study goal, we included an item measuring whether participants had received news concerning politics, public events, and issues through social media. Each item was assessed with a visual analog scale (0 � not at all; 100 � very much) to prevent restricted range. At each prompt, all items appeared in randomized order and the following statement pre- ceded the items: “Please indicate to what extent the following statements applied to you in the past 2 hours.” Data Analysis Descriptive statistics. For each participant, we calculated the mean (i.e., within-person mean) and standard deviation (i.e., within-person standard deviation) of each variable. For instance, if a participant reported depressed mood 70 times, we summed all 70 reported values and divided by 70. We repeated this procedure for all variables in all participants, resulting in a set of within-person means of all variables. From these values, we calculated a mean and standard deviation for each variable (reported in Table 1, first column). The same procedure was used to calculate the within- person standard deviation, which resulted in a set of within-person standard deviations for each variable. We calculated the mean and standard deviation of these values (reported in Table 1, second column). Assumption checks. We used Kolmogorov–Smirnov tests to check whether each variable was normally distributed. A require- ment for the statistical model estimated in the present study (i.e., multilevel vector auto-regression [VAR]) is that stationarity holds. This means that the mean and variance of a variable do not change as a function of time. For each variable of each participant, we tested for stationarity using the Kwiatkowksi-Phillips-Schmidt- Shin unit root test (following Bringmann, 2016; Kwiatkowski, Phillips, Schmidt, & Shin, 1992). For both assumption checks, we applied Bonferroni correction to adjust p values for multiple test- ing. Network estimation and visualization. Using the R package mlVAR (Epskamp et al., 2018), we estimated contemporaneous correlations, temporal correlations, between-subjects correlations, and between-subjects partial correlations among depression symp- toms, stress, PSMU, and ASMU. Contemporaneous correlations represent how variables are associated within the same timeframe, and represent associations that remain after partialing out all other variables in the network within the same timeframe, and after partialing out temporal associations among variables. For instance, a positive contemporaneous correlation between PSMU and lone- liness indicates that, within the same timeframe of 2 hr, higher levels of PSMU co-occur with feeling lonelier, after controlling for all other contemporaneous and temporal relationships. Although a direction of effect cannot be established in these undirected net- works, associations likely come from temporal relationships, and especially relationships that occur at a shorter timeframe than sampled in the present study (minutes, not hours) will end up in the contemporaneous network structure. Temporal correlations indi- cate how a variable is predicted by all other variables (including itself) at a previous timeframe; these are “partial” correlations because they represent an association after controlling for all other temporal effects. The multilevel VAR model has intercepts for each item, which represents the mean item level across time. Each variable in each individual has a mean level, and we can calculate between-subjects correlations between these mean levels. A positive between- subjects correlation between PSMU and loneliness indicates that participants who on average spend more time on PSMU also tend to have a higher level of loneliness. In the between-subjects partial correlations network, associations represent correlations between mean levels of variables while controlling for all other variables in the network. The package mlVAR estimates the contemporaneous, temporal, and between-subjects correlations in two steps. In the first step, mlVAR estimates temporal and between-subjects associations. In the present study, this is done by estimating 10 multiple regression equations. In each equation, one of the variables in our ESM study is predicted by all variables (including itself) at a previous time- frame (t�1). Consider the following example equation: depressed moodt � �0 � depressed moodt�1 ��1 � PSMUt�1 ��2. In this equation, �2 represents the partial correlation between PSMU at a previous timeframe (t�1) and depressed mood at a subsequent timeframe (t), after controlling for depressed mood at a previous timeframe (t�1). The intercept of these equations (�0) represents the value of depressed mood at time t when depressed moodt�1 and PSMUt�1 are equal to zero. Because mlVAR estimates a random intercept, we can obtain an intercept for each variable in each participant—the mean value of the variable across 2 weeks. mlVAR uses these person-specific intercepts (means) to estimate the between-subjects associations, which are partial correlations between the person-specific means of all variables. The model estimated in the first step does not fit the data perfectly; stated differently, the original data points are not always equal to the values predicted by the model. In the second step, the residuals of this model (i.e., the differences between the original data points and predicted values) are used to estimate partial contemporaneous associations. mlVAR does this by estimating how the residuals of one variable are predicted by the residuals of all other variables at the same timeframe. To make the model computationally tractable, we forced random effects of temporal and contemporaneous associations to be orthog- Table 1 Means and Standard Deviations of Within-Person Means and Within-Person Standard Deviations for All Variables ESM variable M (SD) SD (SD) Stress 20.55 (14.53) 16.84 (7.42) Depressed mood 13.01 (11.23) 12.51 (7.48) Loss of interest 25.18 (14.71) 20.44 (8.09) Fatigue 37.07 (17.37) 23.49 (6.67) Concentration problems 26.56 (14.38) 21.00 (7.19) Loneliness 11.90 (11.07) 10.88 (7.22) Feeling inferior 10.36 (10.53) 9.10 (6.53) Feeling hopeless 11.85 (10.89) 10.84 (6.66) PSMU 31.27 (14.08) 25.03 (6.03) ASMU 21.27 (15.38) 18.85 (7.76) Note. ESM � experience sampling methodology; PSMU � Passive social media use; ASMU � active social media use. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 1456 AALBERS, MCNALLY, HEEREN, DE WIT, AND FRIED onal (i.e., random slopes and intercepts were uncorrelated). In the undirected (i.e., contemporaneous and between-subjects) networks, the model predicts A by B and B by A, resulting in two p values for the association between the variables. To avoid estimating false pos- itive associations, we used the conservative AND rule, which means we only included associations in the model if both coefficients were significant at a level of p � .05. The AND rule does not apply to temporal associations, because these are estimated only once. mlVAR deals with missing data by removing all measurement moments that include at least one missing observation. mlVAR returned a between-subjects correlation matrix with many implausibly high correlations. This might be because mlVAR esti- mates between-subjects partial correlations and then calculates between-subjects correlations from these associations, which can lead to unstable results (S. Epskamp, personal communication, May 29, 2017). To resolve this issue, we consulted with the package author (S. Epskamp, personal communication, May 29, 2017) who suggested a different way to calculate these associations (to calculate within- person means for each variable, and then estimate the correlations between them). We followed this procedure and report these corre- lations in the Results section. Using the R package qgraph, we visualized all aforementioned associations as networks. These graphs comprise nodes, which represent the variables, and edges, which represent the associations between the variables. We plotted all networks with the same layout, which we determined by averaging over the layout of all networks based on the Fruchterman-Reingold algorithm (Fruch- terman & Reingold, 1991). We only included correlations in the graphs with p values smaller than .05. All adjacency matrices are available in the online supplementary materials. Results Descriptive Statistics For all ESM variables that we included in the analysis, Table 1 contains means and standard deviations of within-person means and within-person standard deviations. Responses ranged from 0 to 100 for all variables. Assumption Checks Kolmogorov–Smirnov tests indicated that no variable was nor- mally distributed (p � .001). Distributions indicated bimodality for some variables (fatigue, concentration problems, loss of inter- est, stress, PSMU, and ASMU) and right-skew for others (de- pressed mood, feelings of inferiority, loneliness, and hopeless- ness). Within-person mean levels were normally distributed for fatigue, concentration problems, and loss of interest (p � .05), but not for depressed mood, stress, PSMU, ASMU, and feelings of loneliness, inferiority, and hopelessness (p � .001). Kwiatkowski- Phillips-Schmidt-Shin unit root tests suggested stationary data for all variables in all participants. Contemporaneous Network The contemporaneous network in Figure 1 shows the direct associations between the variables within the same timeframe after controlling for all other temporal and contemporaneous relations. PSMU is positively associated with concentration problems, loss of interest, fatigue, and loneliness, but unrelated to stress and feeling inferior, hopeless, or depressed. Furthermore, there are positive associations between ASMU and feelings of inferiority, and ASMU and concentration problems. Moreover, depression symptoms are positively associated. For instance, loss of interest is positively associated with fatigue and concentration problems. Finally, several depression symptoms, such as concentration prob- lems and depressed mood, are positively associated with stress. Temporal Network The temporal network in Figure 2 demonstrates how PSMU, ASMU, depression symptoms, loneliness, and stress predict each other from one timeframe to the next. Fatigue, loneliness, and Figure 1. Partial contemporaneous correlation network of depression symptoms, stress and social media use. Solid lines represent positive partial correlations. Thicker lines represent stronger partial correlations. The thickest line (between active social media use [ASMU] and Passive social media use [PSMU]) represents a partial correlation of 0.33; the thinnest line (between PSMU and loneliness) represents a partial correlation of 0.03. We only plotted correlations with p-values � .05. See the online article for the color version of this figure. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 1457SOCIAL MEDIA AND DEPRESSION SYMPTOMS http://dx.doi.org/10.1037/xge0000528.supp ASMU positively predict PSMU, but PSMU positively predicts only AMSU. Furthermore, loneliness and stress positively predict ASMU, and ASMU negatively predicts fatigue. Moreover, several depression symptoms predict each other bidirectionally across time. For instance, depressed mood positively predicts loneliness, and vice versa. However, not all depression symptoms directly predict each other. There is no temporal association between, for example, depressed mood and concentration problems. Finally, stress positively predicts and is predicted by concentration prob- lems, depressed mood, and feeling hopeless. Loneliness positively predicts stress, but not vice versa. Between-Subjects Network The between-subjects network in Figure 3 depicts the corre- lations between intraindividual mean levels of PSMU, ASMU, depression symptoms, loneliness, and stress. As Figure 3 shows, mean levels of PSMU correlated positively with mean levels of ASMU, depressed mood, and feelings of loneliness, hopelessness, and inferiority. ASMU correlated positively with the same symptoms as well as with stress and concentration problems. This means that, for instance, students who on aver- age spent more time on PSMU tended to have a higher mean level of loneliness. However, as can be seen in the between-subjects partial corre- lations network (see Figure 3), PSMU showed only one direct relation with another variable: ASMU. Thus, the positive zero- order correlation between PSMU and other nodes decreased upon controlling for all other items in the network, and became nonsig- nificant (feeling inferior predicted by PSMU, p � .09; PSMU predicted by feeling inferior, p � .20; depressed mood predicted by PSMU, p � .81; PSMU predicted by depressed mood, p � .48; loneliness predicted by PSMU, p � .89; PSMU predicted by loneliness, p � .75; hopelessness predicted by PSMU, p � .19; PSMU predicted by hopelessness, p � .25). One interpretation is that these relationships cease to exist at the level of partial corre- lations; another is that the present study lacked sufficient power to detect small edge coefficients in the partial correlation network. Finally, we see several positive partial correlations among depres- sion symptoms, loneliness, and stress; for instance, stress and fatigue feature unique positive associations, as do depressed mood and loneliness. Discussion Summary of Findings In an experience sampling study of 125 students, with seven prompts per day, engaging in PSMU did not predict depression symptoms, loneliness, or stress. Instead, previous fatigue and loneliness predicted PSMU, indicating that these symptoms might lead participants to scroll through social media pages. Within the same timeframe, PSMU co-occurred with loss of interest, concen- tration problems, fatigue, and loneliness. These contemporaneous relations have been commonly interpreted in the literature as indicative of fast-moving causal processes (Epskamp et al., 2018), but the lack of temporal precedence does not allow for insights into the direction of the effects, that is, if PSMU leads to depression symptoms, vice versa, or both. Finally, we found that participants who spent more time passively using social media also experi- enced higher mean levels of depressed mood, loneliness, hopeless- ness, and feeling inferior. However, when controlling for all vari- ables in this network structure, PSMU was unrelated to all variables except for active social media usage. This means either that there are no partial correlations between PSMU and depres- Figure 2. Temporal network of depression symptoms, stress and social media use. An arrow from variable A to variable B represents partial temporal correlations between variable A at t�1 and variable B at t. Solid arrows represent positive temporal correlations, and the dotted arrow represents a negative temporal correlation. Thicker arrows represent stronger correlations. The thickest arrow (from depressed mood to loneliness) represents a partial correlation of 0.09; the thinnest arrow (from stress to feelings of hopelessness) represents a partial correlation of 0.03. For interpretability, we did not plot auto-regressive correlations (see online supplementary materials for a graph including all auto-regressive correlations). We only plotted correlations with p-values � .05. See the online article for the color version of this figure. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 1458 AALBERS, MCNALLY, HEEREN, DE WIT, AND FRIED http://dx.doi.org/10.1037/xge0000528.supp http://dx.doi.org/10.1037/xge0000528.supp sive symptoms, or that these relations are too weak for detection in the present analysis.1 Importance of Findings We believe this is the first study to show that PSMU is con- temporaneously associated with concentration problems, fatigue, loneliness, and loss of interest. Given the undirected nature of these associations, we do not know if PSMU causes these symp- toms, or vice versa, or both. However, the observation that PSMU is associated with concentration problems aligns with research demonstrating that individuals who spend more time on PSMU tend to have lower attentional control (Alloway & Alloway, 2012). Possibly, people with poor attentional control tend to get distracted and are unable to inhibit habitual checking of Facebook. As we found this effect in the contemporaneous but not in the temporal network, the present study suggests that this effect occurs on a small timescale. Furthermore, because loss of interest reflects reduced positive affect (Nutt et al., 2007), the present study offers one potential explanation regarding the way PSMU may decrease affective well-being (Kross et al., 2013; Verduyn et al., 2015). The positive contemporaneous association between PSMU and fatigue is in line with research suggesting that social media use might cause fatigue in individuals who feel overwhelmed by social media (e.g., because they receive too many messages; Lee, Son, & Kim, 2016). However, a different interpretation can be derived by looking at the temporal network, which indicates that PSMU could be part of a (beneficial) self-regulatory feedback loop: fatigue ¡ � PSMU ¡ � ASMU ¡ � fatigue. Additionally, whereas Lee, Son, and Kim (2016) hypothesize that the effect of social media on fatigue is mediated by stress, the present investigation finds no contemporaneous association between social media and stress, no temporal relation from social media use to stress and no (direct) temporal relation from stress to fatigue. One potential explanation is that the contemporaneous and temporal relations in the present study pertain to a narrow time window, whereas Lee et al. (2016) analyzed cross-sectional survey data, which encompass a larger time window, like the between-subjects partial correlations network in the present study. However, this network is inconsistent with findings by Lee et al. (2016): Although fatigue and stress are positively associated, social media use and stress are not. Alto- gether, these findings call into question the hypothesis that social media causes fatigue via stress caused by information overload. Consistent with Kross et al.’s (2013) findings, we found that loneliness predicted PSMU and ASMU, but not vice versa, which suggests a unidirectional relationship between social media use and loneliness. Unlike Kross et al. (2013; and Verduyn et al., 2015), we did not find that more time on social media predicts lower affective well-being across time. This discrepancy might be explained by differences in operationalization of affective well- being (momentary affect in previous research vs. mood measures in the present study) and statistical analysis. Kross et al. (2013) and Verduyn et al. (2015) estimated the correlation between momen- tary affect and time spent on social media in the past 2 hr, which were measured at the same moment. Possibly, this procedure could have led to recall bias (e.g., when a person feels negative, they 1 In a sensitivity analysis (see online supplementary materials), total time spent on social media (i.e., sum of PSMU and ASMU) was positively related to feeling inferior in the between-subjects partial correlation net- work. Figure 3. Between-subjects correlations (left panel), and between-subjects partial correlation network (right panel), representing the correlations between mean levels of depression symptoms, stress, and social media use. Solid lines represent positive correlations; dotted lines represent negative correlations. Thicker lines indicate stronger associations. In the left panel, the thickest line (between loneliness and depressed mood) represents a correlation of 0.94; the thinnest line (between loss of interest and Passive social media use [PSMU]) represents a correlation of 0.10. In the right panel, the thickest line (between loneliness and depressed mood) represents a partial correlation of 0.69; the thinnest line (between fatigue and feeling inferior) represents a partial correlation of 0.18. We only plotted associations with p-values � .05. See the online article for the color version of this figure. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 1459SOCIAL MEDIA AND DEPRESSION SYMPTOMS http://dx.doi.org/10.1037/xge0000528.supp might overestimate the time they spent on social media in the past two hours). This issue is less likely to occur in the present study, because temporal associations pertain to relationships between variables at different measurement moments. A second reason for this discrepancy could be that Kross et al. (2013) and Verduyn et al. (2015) examined the effects of social media in an American student sample, whereas we did so in a European student sample. This could suggest that social media use has a different effect on American students than on European students. A final possibility is that social media use does predict depressed mood and loss of interest, but when controlling for all other variables in the network, these are statistically nonsignificant. However, when we reran our analyses including only social media use, depressed mood, and loss of interest, we found no temporal associations from social media use to mood items. Furthermore, we did not find evidence that social media causes individuals to compare themselves with their (ostensibly) superior contacts, which could cause depression symptoms by increasing feelings of inferiority (Blease, 2015). Although research shows that social media exposes individuals to the highly curated lives of social media users who (on average) seem happier and more popular (Bollen et al., 2017), present findings suggest that—in a student population—there is no direct influence from PSMU to feelings of inferiority. Finally, this study adds to a growing body of research demon- strating that individual depression symptoms are differentially associated with nonsymptom variables, such as psychosocial func- tioning (Fried & Nesse, 2014). Thereby, it underscores the impor- tance of modeling individual depression symptoms in research instead of sum scores or diagnoses that obfuscate crucial informa- tion (Fried & Nesse, 2015). Strengths and Limitations To the best of our knowledge, this study is the first to apply a network perspective to the link between social media and mental health. Our work extends prior work (e.g., Kross et al., 2013; Verduyn et al., 2015) in that we went beyond affect, including more detailed mental health facets such as fatigue and concentra- tion problems, and applied a recently developed statistical analytic procedure. An important quality of the present study is that its experience sampling protocol was almost twice as intensive as those applied in previous studies on social media and psycholog- ical well-being (Kross et al., 2013; Verduyn et al., 2015). Experi- ence sampling is considered an approach with high ecological validity because it monitors people in daily life. We make all ESM data, R-syntax, model output, and the correlation matrix of the data available in the online supplementary materials. Our study does have several limitations. First, as this was a student sample, mean levels of depression symptoms, loneliness, and stress were fairly low. Further, items might have been inter- preted differently by this nonclinical sample than they would have been in a clinical sample. For instance, to individuals without depression, endorsing “I had little interest in doing anything” could mean they felt bored, whereas in individuals with depression, this could represent anhedonia. As social media’s adverse effects ap- pear to be demonstrably stronger in depressed than in nonde- pressed individuals (Appel, Crusius, & Gerlach, 2015), it remains possible that clinical samples show stronger and more (temporal) relations between social media use and depression symptoms. Therefore, future extensions of the present work in clinical sam- ples is needed. Finally, although our ESM questionnaire included a distractor item for social media use (i.e., news), we did not include additional items to distract from negative affect items. As a consequence, some participants might have been aware of our study’s purpose, which could have influenced their responses in the direction of our hypotheses (e.g., reporting greater loss of interest when also reporting more PSMU). This will be an inter- esting challenge moving forward in the emerging field of ESM studies, with the goal to balance the need of focusing on few items to reduce participant burden with including sufficient distractors. This is particularly important when such research has an intensive sampling protocol as in the present study. One possibility is that future studies include one or two positive or neutral affect items to counterbalance the aforementioned potential effects, or include a social desirability question. Second, as PSMU may be initiated with minimal awareness when performed habitually, self-reports may not always accurately reflect actual PSMU. Conversely, it is possible that individuals sometimes overestimate how much time they spend on social media (Junco, 2013). Using self-tracking applications that can accurately estimate time spent on social media might solve this problem. We consider this an important direction for future work on this topic. Third, although tests suggested that the present data met the assumption of stationarity, sensitivity analyses suggested that this assumption might have been violated. One way to deal with this issue is to detrend nonstationary individual time series data before running group level analyses. Furthermore, our data did not meet all assumptions of multivariate normality. This is not unusual in psychology, but as it is unclear at present how robust the employed methods are to such violations, results need to be interpreted with caution. Following a reviewer’s suggestion, we log transformed the data and reran analyses (reported in the online supplementary materials as sensitivity analysis), which did not affect the overall pattern of results. As an additional assumption check, suggested by two reviewers, we also tested if the residuals of the contempora- neous network followed a normal distribution. We found that this was not the case for any of the variables, violating some of the model assumptions. Because some variables were bimodally dis- tributed, future research into this issue might benefit from logistic rather than linear regression. However, to this date, logistic regres- sion approaches have not been made readily available in common multilevel analysis routines of ESM data, and these and related challenges of non-normal residuals require urgent attention in future methodological research. Finally, network models in cross- sectional data have greatly benefitted from recent investigations into the accuracy and stability of network parameters such as edge weights and centrality estimates (Epskamp et al., 2018). Unfortu- nately, such checks are not yet available for mlVAR models as estimated in R, and we look forward to methodological develop- ments in this field. Fourth, the present modeling framework is limited to estimating linear relationships, and cannot capture higher-order interactions among variables. For instance, it is possible that PSMU only predicts feelings of inferiority when individuals feel very lonely. We are looking forward to statistical developments that would allow to tackle these issues. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 1460 AALBERS, MCNALLY, HEEREN, DE WIT, AND FRIED http://dx.doi.org/10.1037/xge0000528.supp http://dx.doi.org/10.1037/xge0000528.supp http://dx.doi.org/10.1037/xge0000528.supp Constraints on Generality The sample of participants is representative of Dutch undergrad- uate psychology students at a large university. The present study included participants from a relatively narrow age band. Because social media use is highly differentiated by age, it is important to note that our results are more likely to generalize to young than to old individuals. Replication studies, using the present study’s ESM questionnaire (see online supplementary materials), could be con- ducted in other student populations, age groups, and in clinical populations, such as depressed individuals. The present findings depend on algorithms that social media use to determine news feed content. Changes in these algorithms might lead to different re- sults. A Network Approach to Social Media and Psychopathology These limitations notwithstanding, we believe the network per- spective provides important insights into the complex, dynamic relation between social media and psychopathology. Until re- cently, this perspective has focused primarily on symptom net- works (Fried et al., 2017); however, several studies have now estimated networks that include nonsymptom variables (e.g., Bern- stein, Heeren, & McNally, 2017; Heeren & McNally, 2016). These studies and our own study align with the recently proposed ex- panded network approach, which aims to uncover the network structure of all variables—symptoms and nonsymptoms—that could be causally relevant to psychopathology (Fried & Cramer, 2017; Jones, Heeren, & McNally, 2017). We believe the present study illustrates the utility of this approach and we hope it encour- ages researchers to investigate the network structure of symptoms and beyond. Context of the Research The general idea for this study was developed by combining ideas and findings in clinical psychology, psychological research methods, and communication science. We consider network anal- ysis an important tool to quantitatively integrate empirical research from different fields of study, and the present study aimed to do so for clinical psychology and communication science. Our findings are best situated in the authors’ novel research program—the expanded network approach (Fried & Cramer, 2017; Jones et al., 2017)—which examines the network structure of variables that are causally relevant to mental health. Future extensions of the present research are direct replications in clinical samples (e.g., in indi- viduals with depression, or body image disorders), and other cultures, and studies that include symptoms of other psychiatric syndromes (e.g., compulsively checking nonexistent or minor flaws in physical appearance in individuals with body dysmorphic disorder). References Alloway, T. P., & Alloway, R. G. (2012). The impact of engagement with social networking sites (SNSs) on cognitive skills. Computers in Human Behavior, 28, 1748 –1754. http://dx.doi.org/10.1016/j.chb.2012.04.015 Andreassen, C. S., Pallesen, S., & Griffiths, M. D. (2017). The relationship between addictive use of social media, narcissism, and self-esteem: Findings from a large national survey. Addictive Behaviors, 64, 287– 293. http://dx.doi.org/10.1016/j.addbeh.2016.03.006 Appel, H., Crusius, J., & Gerlach, A. L. (2015). Social comparison, envy, and depression on Facebook: A study looking at the effects of high comparison standards on depressed individuals. Journal of Social and Clinical Psychology, 34, 277–289. http://dx.doi.org/10.1521/jscp.2015 .34.4.277 Appel, H., Gerlach, A. L., & Crusius, J. (2016). The interplay between Facebook use, social comparison, envy, and depression. Current Opin- ion in Psychology, 9, 44 – 49. http://dx.doi.org/10.1016/j.copsyc.2015.10 .006 Bernstein, E. E., Heeren, A., & McNally, R. J. (2017). Unpacking rumi- nation and executive control: A network perspective. Clinical Psycho- logical Science, 5, 816 – 826. http://dx.doi.org/10.1177/216770 2617702717 Blease, C. R. (2015). Too many ‘friends’, too few ‘likes’? Evolutionary psychology and ‘Facebook depression’. Review of General Psychology, 19, 1–13. http://dx.doi.org/10.1037/gpr0000030 Bollen, J., Goncalves, B., van de Leemput, I., & Ruan, G. (2017). The happiness paradox: Your friends are happier than you. EPJ Data Sci- ence, 6, 4. http://dx.doi.org/10.1140/epjds/s13688-017-0100-1 Borsboom, D. (2017). A network theory of mental disorders. World Psy- chiatry, 16, 5–13. Boschloo, L., van Borkulo, C. D., Borsboom, D., & Schoevers, R. A. (2016). A prospective study on how symptoms in a network predict the onset of depression. Psychotherapy and psychosomatics, 85, 183–184. http://dx.doi.org/10.1159/000442001 Bringmann, L. F. (2016). Dynamical networks in psychology: More than a pretty picture? (master’s thesis). University of Groningen, Groningen, the Netherlands. DeBerard, M. S., & Kleinknecht, R. A. (1995). Loneliness, duration of loneliness, and reported stress symptomatology. Psychological Reports, 76, 1363–1369. http://dx.doi.org/10.2466/pr0.1995.76.3c.1363 Dejonckheere, E., Bastian, B., Fried, E. I., Murphy, S. C., & Kuppens, P. (2017). Perceiving social pressure not to feel negative predicts depres- sive symptoms in daily life. Depression and Anxiety, 34, 836 – 844. http://dx.doi.org/10.1002/da.22653 Elo, A. L., Leppänen, A., & Jahkola, A. (2003). Validity of a single-item measure of stress symptoms. Scandinavian Journal of Work, Environ- ment & Health, 29, 444 – 451. http://dx.doi.org/10.5271/sjweh.752 Epskamp, S., Borsboom, D., & Fried, E. I. (2018). Estimating psycholog- ical networks and their accuracy: A tutorial paper. Behavior Research Methods, 50, 195–212. http://dx.doi.org/10.3758/s13428-017-0862-1 Epskamp, S., van Borkulo, C. D., van der Veen, D. C., Servaas, M. N., Isvoranu, A. M., . . . Cramer, A. O. (2017). Personalized network modeling in psychopathology: The importance of contemporaneous and temporal connections. Clinical Psychological Science, 6, 416 – 427. Epskamp, S., Waldorp, L. J., Mõttus, R., & Borsboom, D. (2018). The Gaussian graphical model in cross-sectional and time-series data. Mul- tivariate Behavioral Research. Advance online publication. http://dx.doi .org/10.1080/00273171.2018.1454823 Fried, E. I. (2017). The 52 symptoms of major depression: Lack of content overlap among seven common depression scales. Journal of Affective Disorders, 208, 191–197. http://dx.doi.org/10.1016/j.jad.2016.10.019 Fried, E. I., Bockting, C., Arjadi, R., Borsboom, D., Amshoff, M., Cramer, A. O., . . . Stroebe, M. (2015). From loss to loneliness: The relationship between bereavement and depressive symptoms. Journal of Abnormal Psychology, 124, 256 –265. http://dx.doi.org/10.1037/abn0000028 Fried, E. I., & Cramer, A. O. J. (2017). Moving forward: Challenges and directions for psychopathological network theory and methodology. Perspectives on Psychological Science, 12, 999 –1020. http://dx.doi.org/ 10.1177/1745691617705892 T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 1461SOCIAL MEDIA AND DEPRESSION SYMPTOMS http://dx.doi.org/10.1037/xge0000528.supp http://dx.doi.org/10.1016/j.chb.2012.04.015 http://dx.doi.org/10.1016/j.addbeh.2016.03.006 http://dx.doi.org/10.1521/jscp.2015.34.4.277 http://dx.doi.org/10.1521/jscp.2015.34.4.277 http://dx.doi.org/10.1016/j.copsyc.2015.10.006 http://dx.doi.org/10.1016/j.copsyc.2015.10.006 http://dx.doi.org/10.1177/2167702617702717 http://dx.doi.org/10.1177/2167702617702717 http://dx.doi.org/10.1037/gpr0000030 http://dx.doi.org/10.1140/epjds/s13688-017-0100-1 http://dx.doi.org/10.1159/000442001 http://dx.doi.org/10.2466/pr0.1995.76.3c.1363 http://dx.doi.org/10.1002/da.22653 http://dx.doi.org/10.5271/sjweh.752 http://dx.doi.org/10.3758/s13428-017-0862-1 http://dx.doi.org/10.1080/00273171.2018.1454823 http://dx.doi.org/10.1080/00273171.2018.1454823 http://dx.doi.org/10.1016/j.jad.2016.10.019 http://dx.doi.org/10.1037/abn0000028 http://dx.doi.org/10.1177/1745691617705892 http://dx.doi.org/10.1177/1745691617705892 Fried, E. I., & Nesse, R. M. (2014). The impact of individual depressive symptoms on impairment of psychosocial functioning. PLoS ONE, 9, e90311. http://dx.doi.org/10.1371/journal.pone.0090311 Fried, E. I., & Nesse, R. M. (2015). Depression sum-scores don’t add up: Why analyzing specific depression symptoms is essential. BMC Medi- cine, 13, 72. http://dx.doi.org/10.1186/s12916-015-0325-4 Fried, E. I., Nesse, R. M., Guille, C., & Sen, S. (2015). The differential influence of life stress on individual symptoms of depression. Acta Psychiatrica Scandinavica, 131, 465– 471. http://dx.doi.org/10.1111/ acps.12395 Fried, E. I., van Borkulo, C. D., Cramer, A. O., Boschloo, L., Schoevers, R. A., & Borsboom, D. (2017). Mental disorders as networks of prob- lems: A review of recent insights. Social Psychiatry and Psychiatric Epidemiology, 52, 1–10. http://dx.doi.org/10.1007/s00127-016-1319-z Frison, E., & Eggermont, S. (2016). Exploring the relationships between different types of Facebook use, perceived online social support, and adolescents’ depressed mood. Social Science Computer Review, 34, 153–171. http://dx.doi.org/10.1177/0894439314567449 Fruchterman, T. M., & Reingold, E. M. (1991). Graph drawing by force- directed placement. Software, Practice & Experience, 21, 1129 –1164. http://dx.doi.org/10.1002/spe.4380211102 Goldberg, Y. K., Eastwood, J. D., LaGuardia, J., & Danckert, J. (2011). Boredom: An emotional experience distinct from apathy, anhedonia, or depression. Journal of Social and Clinical Psychology, 30, 647– 666. http://dx.doi.org/10.1521/jscp.2011.30.6.647 Gonzales, A. L., & Hancock, J. T. (2011). Mirror, mirror on my Facebook wall: Effects of exposure to Facebook on self-esteem. Cyberpsychology, Behavior, and Social Networking, 14, 79 – 83. http://dx.doi.org/10.1089/ cyber.2009.0411 Heeren, A., & McNally, R. J. (2016). An integrative network approach to social anxiety disorder: The complex dynamic interplay among atten- tional bias for threat, attentional control, and symptoms. Journal of Anxiety Disorders, 42, 95–104. http://dx.doi.org/10.1016/j.janxdis.2016 .06.009 Jones, P. J., Heeren, A., & McNally, R. J. (2017). Commentary: A network theory of mental disorders. Frontiers in Psychology, 8, 1305. http://dx .doi.org/10.3389/fpsyg.2017.01305 Junco, R. (2013). Comparing actual and self-reported measures of Face- book use. Computers in Human Behavior, 29, 626 – 631. http://dx.doi .org/10.1016/j.chb.2012.11.007 Kross, E., Verduyn, P., Demiralp, E., Park, J., Lee, D. S., Lin, N., . . . Ybarra, O. (2013). Facebook use predicts declines in subjective well- being in young adults. PLoS ONE, 8, e69841. http://dx.doi.org/10.1371/ journal.pone.0069841 Kwiatkowski, D., Phillips, P. C. B., Schmidt, P., & Shin, Y. (1992). Testing the null hypothesis of stationarity against the alternative of a unit root. Journal of Econometrics, 54, 159 –178. http://dx.doi.org/10.1016/ 0304-4076(92)90104-Y LaRose, R. (2010). The problem of media habits. Communication Theory, 20, 194 –222. http://dx.doi.org/10.1111/j.1468-2885.2010.01360.x LaRose, R., Lin, C. A., & Eastin, M. S. (2003). Unregulated internet usage: Addiction, habit, or deficient self-regulation? Media Psychology, 5, 225–253. http://dx.doi.org/10.1207/S1532785XMEP0503_01 Lee, A. R., Son, S. M., & Kim, K. K. (2016). Information and communi- cation technology overload and social networking service fatigue: A stress perspective. Computers in Human Behavior, 55, 51– 61. http://dx .doi.org/10.1016/j.chb.2015.08.011 Liu, D., & Baumeister, R. F. (2016). Social networking online and per- sonality of self-worth: A meta-analysis. Journal of Research in Person- ality, 64, 79 – 89. http://dx.doi.org/10.1016/j.jrp.2016.06.024 Meier, A., Reinecke, L., & Meltzer, C. E. (2016). Facebocrastination? Predictors of using Facebook for procrastination and its effects on students’ well-being. Computers in Human Behavior, 64, 65–76. http:// dx.doi.org/10.1016/j.chb.2016.06.011 Mellor, D., Stokes, M., Firth, L., Hayashi, Y., & Cummins, R. (2008). Need for belonging, relationship satisfaction, loneliness, and life satis- faction. Personality and Individual Differences, 45, 213–218. http://dx .doi.org/10.1016/j.paid.2008.03.020 Nutt, D., Demyttenaere, K., Janka, Z., Aarre, T., Bourin, M., Canonico, P. L., . . . Stahl, S. (2007). The other face of depression, reduced positive affect: The role of catecholamines in causation and cure. Journal of Psychopharmacology, 21, 461– 471. http://dx.doi.org/10.1177/0269 881106069938 Pe, M. L., Kircanski, K., Thompson, R. J., Bringmann, L. F., Tuerlinckx, F., Mestdagh, M., . . . Gotlib, I. H. (2015). Emotion-network density in major depressive disorder. Clinical Psychological Science, 3, 292–300. http://dx.doi.org/10.1177/2167702614540645 Runyan, J. D., Steenbergh, T. A., Bainbridge, C., Daugherty, D. A., Oke, L., & Fry, B. N. (2013). A smartphone ecological momentary assess- ment/intervention “app” for collecting real-time data and promoting self-awareness. PLoS ONE, 8, e71325. http://dx.doi.org/10.1371/journal .pone.0071325 Tobin, S. J., Vanman, E. J., Verreynne, M., & Saeri, A. K. (2015). Threats to belonging on Facebook: Lurking and ostracism. Social Influence, 4510, 1–12. Twenge, J. M. (2017). iGen: Why today’s super-connected kids are grow- ing up less rebellious, more tolerant, less happy—and completely un- prepared for adulthood. New York, NY: Atria. Verduyn, P., Lee, D. S., Park, J., Shablack, H., Orvell, A., Bayer, J., . . . Kross, E. (2015). Passive Facebook usage undermines affective well- being: Experimental and longitudinal evidence. Journal of Experimental Psychology: General, 144, 480 – 488. http://dx.doi.org/10.1037/xge 0000057 Verplanken, B., & Orbell, S. (2003). Reflections on past behavior: A self-report index of habit strength. Journal of Applied Social Psychol- ogy, 33, 1313–1330. http://dx.doi.org/10.1111/j.1559-1816.2003 .tb01951.x Wenninger, H., Krasnova, H., & Buxmann, P. (2014, June). Activity matters: Investigating the influence of Facebook on life satisfaction of teenage users. Proceedings of the European Conference on Information Systems (ECIS) 2014, Tel Aviv, Israel. Whiting, A., & Williams, D. (2013). Why people use social media: A uses and gratifications approach. Qualitative Market Research, 16, 362–369. http://dx.doi.org/10.1108/QMR-06-2013-0041 Received March 16, 2018 Revision received September 20, 2018 Accepted September 23, 2018 � T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 1462 AALBERS, MCNALLY, HEEREN, DE WIT, AND FRIED http://dx.doi.org/10.1371/journal.pone.0090311 http://dx.doi.org/10.1186/s12916-015-0325-4 http://dx.doi.org/10.1111/acps.12395 http://dx.doi.org/10.1111/acps.12395 http://dx.doi.org/10.1007/s00127-016-1319-z http://dx.doi.org/10.1177/0894439314567449 http://dx.doi.org/10.1002/spe.4380211102 http://dx.doi.org/10.1521/jscp.2011.30.6.647 http://dx.doi.org/10.1089/cyber.2009.0411 http://dx.doi.org/10.1089/cyber.2009.0411 http://dx.doi.org/10.1016/j.janxdis.2016.06.009 http://dx.doi.org/10.1016/j.janxdis.2016.06.009 http://dx.doi.org/10.3389/fpsyg.2017.01305 http://dx.doi.org/10.3389/fpsyg.2017.01305 http://dx.doi.org/10.1016/j.chb.2012.11.007 http://dx.doi.org/10.1016/j.chb.2012.11.007 http://dx.doi.org/10.1371/journal.pone.0069841 http://dx.doi.org/10.1371/journal.pone.0069841 http://dx.doi.org/10.1016/0304-4076%2892%2990104-Y http://dx.doi.org/10.1016/0304-4076%2892%2990104-Y http://dx.doi.org/10.1111/j.1468-2885.2010.01360.x http://dx.doi.org/10.1207/S1532785XMEP0503_01 http://dx.doi.org/10.1016/j.chb.2015.08.011 http://dx.doi.org/10.1016/j.chb.2015.08.011 http://dx.doi.org/10.1016/j.jrp.2016.06.024 http://dx.doi.org/10.1016/j.chb.2016.06.011 http://dx.doi.org/10.1016/j.chb.2016.06.011 http://dx.doi.org/10.1016/j.paid.2008.03.020 http://dx.doi.org/10.1016/j.paid.2008.03.020 http://dx.doi.org/10.1177/0269881106069938 http://dx.doi.org/10.1177/0269881106069938 http://dx.doi.org/10.1177/2167702614540645 http://dx.doi.org/10.1371/journal.pone.0071325 http://dx.doi.org/10.1371/journal.pone.0071325 http://dx.doi.org/10.1037/xge0000057 http://dx.doi.org/10.1037/xge0000057 http://dx.doi.org/10.1111/j.1559-1816.2003.tb01951.x http://dx.doi.org/10.1111/j.1559-1816.2003.tb01951.x http://dx.doi.org/10.1108/QMR-06-2013-0041 Social Media and Depression Symptoms: A Network Perspective Method Participants Procedure Materials Data Analysis Descriptive statistics Assumption checks Network estimation and visualization Results Descriptive Statistics Assumption Checks Contemporaneous Network Temporal Network Between-Subjects Network Discussion Summary of Findings Importance of Findings Strengths and Limitations Constraints on Generality A Network Approach to Social Media and Psychopathology Context of the Research References Journals/journal 5 An Investigation Into Audiences’ Reactions to Transgressions by Liked and Disliked Media Figures Mu Hu and James Young West Virginia Wesleyan College Jun Liang and Yuntao Guo Anhui University The present study investigates audiences’ parasocial relationship (PSR) reduction, parasocial breakup (PSB), attribution of causes, and forgiveness of their liked and disliked media figures for the figures’ transgressions. Using a 2 (media figure: liked or disliked) � 2 (transgression severity: minor or major) between-subjects factorial design, an experiment was conducted. There were significant main effects of media figure and transgression severity on PSR reduction, attribution of causes, and forgive- ness. Attribution of causes partially mediated the relationship between PSR and forgiveness. Public Policy Relevance Statement Media figures’ transgressions undermine people’s liking of them. Furthermore, people would feel more hurt by their liked media figures’ transgressions than by their disliked figures’ transgressions. However, people were found to be more likely to forgive their liked media figures. They tended to attribute their liked figures’ transgressions to situational factors but their disliked figures’ transgressions to dispositional factors. Such an attribution bias partly influences the previously mentioned relationship between liking and forgiveness. Keywords: parasocial relationship, parasocial breakup, forgiveness, fundamental attri- bution error Parasocial interaction (PSI) refers to audi- ences’ illusive interaction with media figures or characters (Horton & Wohl, 1956). These fig- ures and characters are called “personae.” PSI is illusive because it “characteristically is one- sided, nondialectical, controlled by the per- former, and not susceptible to mutual develop- ment” (p. 215). Despite its illusive nature, audiences may interpret personae’s behavior as genuine and reciprocal. For instance, audiences sometimes feel that the personae on screen are directly addressing them and then respond to the personae verbally and nonverbally (Auter, 1992; Auter & Davis, 1991; Hartmann & Gold- hoorn, 2011; Schramm & Wirth, 2010). PSI occurs within audiences’ viewing processes, and over time PSI can evolve into a parasocial relationship (PSR). PSR can exist outside of audiences’ viewing processes and influence var- ious aspects of audiences’ life (Dibble, Hart- mann, & Rosaen, 2015). Audiences may keep following certain personae, actively seek news about them, and miss them if they don’t see them for a long time. One of the recent advances in PSI and PSR research is parasocial breakup (PSB) studies. PSB is defined as audiences’ negative reactions when their PSRs with liked personae come to an This article was published Online First May 1, 2017. Mu Hu and James Young, Department of Communica- tion, West Virginia Wesleyan College; Jun Liang and Yun- tao Guo, School of Journalism and Communication, Anhui University. This research was supported in part by a grant from the Public Sentiment and Regional Image Research Center of Anhui University, China. Preparation and revision of this article was supported in part by a Visiting Professor’s Grant from Anhui University, China. Correspondence concerning this article should be ad- dressed to Mu Hu, Department of Communication, West Virginia Wesleyan College, 59 College Avenue, Buck- hannon, WV 26201. E-mail: hu_m@wvwc.edu T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. Psychology of Popular Media Culture © 2017 American Psychological Association 2018, Vol. 7, No. 4, 484 – 498 2160-4134/18/$12.00 http://dx.doi.org/10.1037/ppm0000146 484 mailto:hu_m@wvwc.edu http://dx.doi.org/10.1037/ppm0000146 end (Cohen, 2003). Earlier PSB research fo- cuses on involuntary PSB due to hypothetical or real incidents when audiences’ liked personae are taken off the air (Cohen, 2003, 2004; Eyal & Cohen, 2006; Lather & Moyer-Guse, 2011), whereas in the past few years researchers have investigated audiences’ voluntary PSB caused by personae’s transgressions (Cohen, 2010; Hu, 2016). PSB studies have formed a well-documented line of research, but a few weaknesses impede its further development. First, almost all the PSB research focuses on audiences’ favorite or liked personae but PSB with disliked personae is barely touched. Second, current PSB litera- ture seldom investigates audiences’ forgiveness of personae’s transgressions, although people often consider the forgiveness decision after transgressions and it has been well studied in interpersonal relationship research. Third, few if any studies have examined the mechanism of audiences’ forgiveness of personae for their transgressions. Therefore, there are three goals of the present study. First, it explores audiences’ PSB with liked and disliked media figures. Second, it compares audiences’ forgiveness of their liked and disliked media figures’ transgressions. Third, it investigates how fundamental attribu- tion error (FAE) influences audiences’ reactions to media figures’ transgressions. It needs to be clarified that “media figures” in this study refer to those who are direct representations of real people (e.g., actors or actresses, sports stars, singers, and newscasters, etc.) but not fictional media characters (e.g., “Rachel” in Friends). PSB and Media Figures’ Transgressions Early PSB research primarily centers on in- voluntary PSB. Researchers either investigate audiences’ imagined PSB toward hypothetical disappearance of their favorite TV personalities (Cohen, 2003, 2004) or examine audiences’ ac- tual PSB toward real PSR termination incidents (Eyal & Cohen, 2006; Lather & Moyer-Guse, 2011). Later researchers have attempted to study voluntary PSB caused by media figures’ transgressions. Transgressions, such as former president Bill Clinton’s sex scandal, refer to the incidents committed by actors that violate observers’ ex- pectations of how the actors should behave (Thompson et al., 2005). Media figures’ trans- gressions can erode audiences’ interest in them and lead to PSR dissolution (Cohen, 2003; Eyal & Cohen, 2006). Cohen (2010) compared peo- ple’s anticipated reductions in closeness to friends and media figures as a result of moral, trust, and social expectancy violations. Respon- dents reported a greater sense of closeness to their friends than to media figures. However, for both major and minor moral violations, antici- pated closeness reductions were greater for PSRs than for friendships. The researcher also found that the closeness reductions were related to the types of media figures. For major moral violations, reduction in closeness was greatest for sports figures than for other types of figures (actors, singers, and TV hosts). For social vio- lations, TV hosts scored highest in anticipated reduction in closeness. The four types of figures did not differ in closeness reduction of minor moral violations and trust violations. Hu (2016) found that a scandal involving a celebrity un- dermined audiences’ PSR with him. Further- more, audiences’ PSR with the celebrity was positively related to PSB. However, audiences’ PSI with the celebrity in a talk show and a character played by the celebrity in a movie was not affected by the scandal. Transgressions and Forgiveness Forgiveness is the next step that people con- sider after being hurt by transgressions (Finkel, Rusbult, Kumashiro, & Hannon, 2002). Al- though the empirical explorations of forgive- ness have flourished in interpersonal communi- cation research (Thompson et al., 2005), it remains unknown in PSB literature why some audiences forgive media figures for their trans- gressions whereas some others don’t. As far as the authors of the present study know, only one work has been published that specifically exam- ines forgiveness in the parasocial domain. Sand- erson and Emmons (2014) analyzed the post- ings regarding the baseball star Josh Hamilton’s scandal of alcohol addiction. The researchers discovered that some people expressed their forgiveness and justification of Hamilton’s be- havior. In comparison, some others did not for- give Hamilton but blamed him for lack of will power, trying to seek attention, and insincerity. Despite the illuminating results of this study, it provides limited evidence showing why people 485REACTIONS TO TRANSGRESSIONS BY MEDIA FIGURES T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. are different in extending or withholding for- giveness. In order to investigate why audiences are different in terms of their forgiveness for media figures’ transgressions, we need to point out a couple of characteristics of forgiveness in the present study. First, forgiveness in this study is regarded as a continuum but not an all-or- nothing dichotomy. In many transgression cases, complete forgiveness is difficult and thus partial or conditional forgiveness is granted. Finkel et al. (2002) proposed that although com- plete forgiveness is the logical endpoint of the forgiveness dimension, the magnitude of for- giveness should be considered. Second, forgive- ness suggests the reduction of avoidance ten- dency. According to McCullough et al. (1998), interpersonal forgiveness in close relationships is reflected through the reduction of two mo- tives: revenge and avoidance. As to revenge, because audiences are not the direct victims of media figures’ transgressions, it is not conceiv- able that the figures’ transgressions would trig- ger the audiences’ motive of revenge. Accord- ingly, it is not conceivable either that audiences’ forgiveness would include the reduction of the revenge motive. In comparison, people may try to avoid perpetrators even though they are not the direct victims. They don’t want to be asso- ciated with the perpetrators at the risk of stig- matization (Newman, 1987). Similarly, if audi- ences find their PSRs unsatisfying they are free to withdraw at any moment (Horton & Wohl, 1956). Therefore, the reduction of avoidance motive seems to be applicable to forgiveness in PSR. On the basis of the discussion above, we view forgiveness in this study as the degree to which media figures’ transgressions are ex- cused, suggesting audiences’ inclination to con- tinue PSRs with the figures and their unwilling- ness to withdraw from PSRs. PSI, PSR, and PSB With Disliked Media Figures Relationships are built upon not only per- sonal preference, but also physical locations (e.g., neighbor), work environment (e.g., col- leagues and professional connections), and bio- logical bonds (e.g., family). Therefore, people sometimes have to engage in involuntary rela- tionships with disliked others (Thibaut & Kel- ley, 1959/1986). In order to alleviate the dis- comfort brought by involuntary relationships, people adopt distancing behaviors such as ex- pressing detachment, avoiding involvement, and showing antagonism (Hess, 2000). In con- trast to romantic relationships and friendships, relationships with disliked others are character- ized by conflicts during formation and aggres- sion and avoidance during maintenance (Card, 2007). Given the prevalence of mass media, it is quite possible that people are exposed to both liked and disliked personae in media. However, PSB research, similar to PSI and PSR research, has predominantly revolved around audiences’ liked or favorite personae. However, research- ers point out that not only audiences’ reactions toward “positive” or “attractive” personae but also their reactions to “negative” or “ugly” personae should be studied (Hoorn & Konijn, 2003; Konijn & Hoorn, 2005). Hartmann, Stuke, and Daschmann (2008) argued that PSR contains cognitive components and thus PSR research should include both positive PSRs and hostile PSRs. Because positive PSR has been found to be similar to friendships or romantic relationships, negative PSRs should resemble relationships of hatred and disgust. In PSI literature, a number of studies have been conducted to examine PSI with disliked personae. Tian and Hoffner (2010) asked their research participants to select their liked, neu- tral, or disliked characters from the ABC drama Lost. The researchers measured the participants’ perceived similarity, identification, PSI with the characters, and the extent to which they had tried to change their life to be more like the characters. The scores of the four variables were higher for liked and neutral characters than for disliked characters. PSI was higher for liked characters than for neutral characters. In com- parison, Dibble and Rosaen (2011) focused on audiences’ PSI experience during viewing. It was found that audiences can have the “conver- sational give and take” originally described by Horton and Wohl (1956) with both liked and disliked characters. However, audiences re- ported stronger PSI experience with liked char- acters than with disliked characters. Interpersonal relationship literature has shown that when learning of others’ transgres- sions, people may be exposed to information that is dissonant with their previous assump- tions about the others, and such dissonance can 486 HU, YOUNG, LIANG, AND GUO T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. cause profound distress (Thompson et al., 2005). This may explain why media figures’ transgressions can lead to audiences’ PSR re- duction and PSB (Cohen, 2010; Hu, 2016). Moreover, Hu’s study (2016) showed that audi- ences who had a stronger PSR with a media figure experienced higher levels of PSB caused by the figure’s transgressions. According to Heider (1958), people expect their liked others to do good things and their disliked others to perform bad actions. Therefore, in contrast to the disliked people’s transgressions, the liked people’s transgressions should cause more dis- sonance. It is reasonable to predict that when audiences learn that their liked media figures have committed transgressions, they should ex- perience more PSR reduction and stronger PSB. The first two hypotheses of the present study are as follows. H1: In contrast to the disliked media fig- ures’ transgressions, liked media figures’ transgressions would cause greater PSR reduction. H2: In contrast to the disliked media fig- ures’ transgressions, liked media figures’ transgressions would cause greater PSB. Finkel et al. (2002) pointed out that although forgiveness is an individualized reaction toward transgressions, it is influenced by two factors: relational history and severity of the transgres- sions (as cited in Sanderson & Emmons, 2014). As to relational history, it has been found that forgiveness is positively related to the level of closeness to perpetrators (McCullough et al., 1998). Likable offenders are perceived as more morally responsible and less capable of future transgressions and hence are more likely to be forgiven (Miller & Vidmar, 1981). In compari- son, people tend to expect more punishment for unlikable or unattractive offenders (Tedeschi & Felson, 1994). This difference can be explained from the perspective of dissonance as well. Transgressions cause dissonance between what audiences expect the media figures to do and what the figures actually do. When people are distressed by dissonance, they are motivated to seek strategies to reduce it (Festinger, 1957). Forgiveness can be such a strategy because it involves the motivation to continue the relation- ships with the perpetrators (Beatty, 1970; Sand- erson & Emmons, 2014). According to disso- nance theory, the magnitude of dissonance is positively related to the motivations to reduce the dissonance (Festinger, 1957). As mentioned earlier, liked media figures’ transgressions cause greater dissonance in audiences. There- fore, the audiences should be more motivated to reduce the dissonance caused by their liked media figures’ transgressions and thus more willing to forgive their liked media figures. Our third hypothesis is as follows. H3: Audiences are more likely to forgive liked media figures’ transgressions than to forgive disliked media figures’ transgressions. As to transgression severity, it refers to the degree to which transgression consequences are negative and enduring. More severe transgres- sions have more profound and irreversible con- sequences and thus are more difficult to forgive (Bennett & Earwaker, 1994; Girard & Mullet, 1997; Newman, 1987). Hu (2016) doubted the previous PSB research findings based on a sin- gle transgression incident, and Cohen’s (2010) study shows that audiences’ reactions to media figures’ transgressions vary by transgression se- verity. Interpersonal dissolution literature may shed some light on revealing the relationship between severity and forgiveness in PSRs. Boon and Sulsky (1997) and Girard and Mullet (1997) examined their respondents’ reactions toward a series of transgression scenarios, and both studies found an inverse relationship be- tween forgiveness and transgression severity. In contrast, Finkel et al. (2002), McCullough et al. (1998), and McCullough, Fincham, and Tsang (2003) asked people to recall real life experi- ences of being hurt by transgressions and found that transgression severity was negatively asso- ciated with the extent to which participants for- gave the perpetrators. Based on these research findings, we propose that— H4: The severity of a media figure’s trans- gression is negatively related to audiences’ forgiveness of the media figure. Attribution of Transgressions Causes and FAE People constantly try to find out the causes of others’ behavior because of their innate desire to understand and control the external world (Heider, 1958). This process is called attribu- 487REACTIONS TO TRANSGRESSIONS BY MEDIA FIGURES T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. tion. Attribution plays an important role in de- termining how they react to others’ behavior (Kelley & Michela, 1980). During the attribu- tion process, an observer focuses on either the internal factors (e.g., an actor’s dispositions or personality) or external factors (e.g., environ- ment or situation, Reeder, 1982). Although attribution theories demonstrate that human beings are logical creatures, the process of attribution can sometimes be biased. One of the common attribution biases is FAE. FAE refers to people’s tendency to overestimate the influence of internal factors while underes- timating the influence of external factors during attribution processes (Amabile, Ross, & Stein- metz, 1977; Jones & Harris, 1967; Ross, 1977). FAE can occur for a couple of reasons. First, people are cognitive misers and they tend to seek quick and easy answers to the causes of others’ behavior. Therefore, they pay more at- tention to the more stable and easily accessible information such as an actor’s dispositions, rather than the more fragmented and less obvi- ous information such as situations (Forgas, 1998). Second, people may lack necessary knowledge of situations (Gilbert & Malone, 1995). An observer may know very little about the context of an actor’s behavior and thus rely primarily on the actor’s dispositional factors to explain the behavior. Observers’ liking of an actor influences whether they attribute the actor’s behavior to internal factors or external factors (Kelley & Michela, 1980). When observers’ liked people do good things and disliked people do bad things, the observers treat the actions as typical actions for the actors and thus make internal attributions. In contrast, when there is a “mis- match” between the actor’s actions and the ob- servers’ expectations, the observers tend to be- lieve that such actions are forced by the situations and thus make external attributions (Regan, Strauss, & Fazio, 1974). Research has shown that people’s attribution of media figures’ behaviors is similar to their attribution of their real social relations’ behav- iors (Perse & Rubin, 1989; Tal-Or & Papirman, 2007). Given the above-mentioned two causes of FAE, audiences’ attributions of media fig- ures’ transgressions may be particularly suscep- tible to FAE. First, usually PSRs are less strong than interpersonal relationships because people invest less in PSRs (Branch, Wilson, & Agnew, 2013; Cohen, 2010). Therefore, people are less motivated to make efforts in finding out media figures’ transgression causes. Lack of such a motivation restricts audiences from actively collecting and analyzing the complex situa- tional cues surrounding the transgressions. Rather, they tend to take a cognitive “shortcut” by primarily focusing on the relatively stable and easily comprehensible information, such as the transgressors’ dispositional traits and the previous knowledge of the transgressors. Sec- ond, in real social relationships, people have opportunities to ask transgressors or other wit- nesses about situational information. However, in PSRs, audiences usually rely on media re- ports to obtain the situational cues of media figures’ transgressions. PSR’s one-sided nature prevents people from obtaining further situa- tional information of a particular transgression. Consequently, they tend to rely on what they already know about the media figures, usually the internal and stable factors (e.g., personali- ties, dispositions, and behavioral patterns), to make inferences about the causes of transgres- sions. Therefore, we propose the following hy- pothesis. H5: Audiences tend to attribute their liked media figures’ transgressions to external factors while their disliked media figures’ transgressions to internal factors. Attribution may further influence audiences’ emotional and behavioral reactions toward me- dia figures’ transgressions (Kepplinger, Geiss, & Siebert, 2012). In the above-mentioned Sand- erson and Emmons’s (2014) study, those who granted forgiveness made external attributions and defended the transgressor, whereas those who withheld forgiveness made internal attribu- tions and called for punishment. Forgiveness, as an important reaction to transgressions, some- times is granted because of positive relational history (e.g., liking) and less blameful interpre- tations of transgressions (Finkel et al., 2002). Finkel and colleagues’ (2002) research shows that the cognitive interpretation of betrayals partially mediates the relationship between re- lational commitment and interpersonal forgive- ness. If people’s liking of transgressors is high, they are more motivated to attend to the envi- ronments and come up with interpretations that are more favorable to the transgressors. These 488 HU, YOUNG, LIANG, AND GUO T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. interpretations in turn facilitate their willingness to grant forgiveness. Therefore, our sixth hy- pothesis of the present study is as follows. H6: Audiences’ attribution of the causes (internal vs. external) of media figures’ transgressions mediates the relationship between their PSRs with the figures and their forgiveness of the figures. Method Sample A total of 137 participants were recruited from a Midwestern liberal arts college for this study. The sample ranged in age from 17 to 41 (M � 19.70, SD � 2.52). Thirty-one percent were male (N � 42) and 68% were female (N � 93). Two participants did not disclose their gen- der information. Eighty percent of the partici- pants were Caucasian (N � 80), 9% were Af- rican Americans (N � 12), .7% were Latin Americans (N � 1), 2% were Asian Americans (N � 3), and 7% (N � 10) of the participants identified their ethnicity as “others.” Two par- ticipants did not reveal their ethnicity informa- tion. Procedure A 2 (media figure: liked or disliked) � 2 (transgression severity: minor or major) be- tween-subjects factorial design was employed in the present study. We chose George Clooney as the liked media figure and Charlie Sheen as the disliked media figure. George Clooney has been rated as one of the top five celebrities with the best reputations (“Jennifer Lawrence has the best celebrity reputation, just as we suspected,” 2014). We compiled a biographical profile of Clooney which highlighted the awards he had received, the charity activities he had organized, and the funds he had launched for the refugees of international terrorism. In comparison, Char- lie Sheen, despite the success of his perfor- mance career, has long been suffering from scandals. We also compiled a profile of Charlie Sheen but it focused on his imprudent private life characterized by drug abuse, physical vio- lence, damaging public properties, and sex scandals. We tested our hypotheses with two newspa- per articles. These two articles reported two transgressions of minor and major severity re- spectively. The participants were told that the news articles were published in The New York Times but they were actually written by the researchers. The participants were also told that the transgressions occurred just at the night before they participated in the study. Each arti- cle was printed on a page with The New York Times logo on its top margin, which looked like a news page from The New York Times website. In the minor transgression story, the actor who was in a hurry (either Clooney or Sheen, de- pending on the conditions) opened a door in a bar without looking back and the door hit a person behind him in the face. This incident left the victim a few bruises. In the major transgres- sion story, the actor cut into the front of the line to enter a bar. When a man in the line com- plained, the actor struck him in the face. The victim was sent to Ronald Reagan UCLA Med- ical Center. The participants were randomly assigned to four conditions. In Condition 1 (liked figure and minor transgression news, N � 33) and Condi- tion 2 (liked figure and major transgression news, N � 32), the participants first read Cloo- ney’s profile, and filled out a questionnaire test- ing their PSR with the actor as a pretest. Then the participants in these two conditions read the minor and major transgression news articles, respectively. After reading the articles, the par- ticipants were instructed to complete another questionnaire including a question asking how serious the participants thought the transgres- sion was, a posttest PSR Scale, a PSB Scale measuring their reactions to the transgression news, a scale of attribution of causes, and a scale of forgiveness. In Condition 3 (disliked figure and minor transgression news, N � 36) and Condition 4 (disliked figure and major transgression news, N � 36), the procedure was the same except that the participants read the profile of Charlie Sheen and the transgressor in the news was Charlie Sheen. Measures In addition to such demographic variables as age, gender, and ethnicity, PSR, severity of transgressions, PSB, attribution of causes of transgressions, and forgiveness were measured. PSR. Both the pretest and the posttest of PSR were measured with the 10-item PSI Scale 489REACTIONS TO TRANSGRESSIONS BY MEDIA FIGURES T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. devised by Rubin and Perse (1987). Over de- cades this scale has been adapted to measure PSR with a wide range of personae and consis- tently proved to be valid and reliable. We slightly modified the wording of the scale to make it better fit the present study. The PSR measure used in this study included such items as “I look forward to seeing him on TV or movies,” “I see him as a natural, down-to-earth person,” and “He makes me feel comfortable, as if I am with a friend.” The items were measured on a 5-point Likert-type scale (1 � strongly disagree, 5 � strongly agree). The Cronbach’s alpha of both the pretest scale and the posttest scale was .90. Severity. Severity was measured with a question asking the participants how severe the incidents were. This question was measured on a 5-point Likert-type scale (1 � not severe at all, 5 � very severe). PSB. PSB was measured with 5 items se- lected from the 13-item PSB Scale devised by Cohen (2003). The original scale was used to measure audiences’ reactions to involuntary PSB (e.g., favorite characters were off the air), whereas we intended to measure audiences’ vol- untary PSB after reading the transgression news. Therefore, we selected 5 items that could be applied to this study. The items measured how angry, sad, disappointed, betrayed, and lonely the respondents felt when they read the transgression news. The items were measured on a 5-point Likert-type scale (1 � strongly disagree, 5 � strongly agree). The scale’s Cronbach’s alpha was .78. Attribution of causes. Attribution of causes was measured with the locus of causality subscale of the Causal Dimension Scale devised by Russell (1982). This subscale measures whether a certain behavior is attributable to external or internal factors. It was composed of three items asking whether the cause of the incident is the situation or the actor himself, outside of the actor or inside of the actor, and something about himself or something about others (reversely coded). The items were mea- sured on a 7-point semantic differential scale. Higher scores suggested higher tendency to at- tribute the incident to internal factors. The scale’s Cronbach’s alpha was .72. Forgiveness. A PSR is different from a real interpersonal relationship because it is an illu- sive, one directional, and nondialectical rela- tionship. Therefore, those widely used forgive- ness scales developed from interpersonal relationship research cannot be directly used in the current study (e.g., McCullough et al.’s Transgression-Related Interpersonal Motiva- tions Inventory or TRIM). In this study, forgive- ness was viewed as the degree to which media figures’ transgressions are excused, reflecting audiences’ willingness to continue PSRs and unwillingness of avoidance. Therefore, we used three items to measure forgiveness, “I would forgive this actor,” “I would remain loyal to this actor,” and “I would not follow this actor any- more” (reversely coded). The items were mea- sured on a 5-point Likert-type scale (1 � strongly disagree, 5 � strongly agree). The Cronbach’s alpha of this scale was .77. Results Manipulation Check Before testing the hypotheses, we conducted manipulation checks to test the validity of the experimental manipulations. First, we per- formed a t test to compare the means of PSR pretests with the liked actor (Condition 1 and 2 in which the participants read the profile of George Clooney) and the disliked actor (Con- dition 3 and 4 in which the participants read the profile of Charlie Sheen). The pretest of PSR with the liked actor (M � 3.31, SD � .64) was significantly higher than the pretest of PSR with the disliked actor (M � 2.13, SD � .61), t(135) � 11.05, p � .001. The effect size for this analysis (d � 1.89) was found to exceed Cohen’s (1988) convention for a large effect (d � .80). Therefore, our manipulation of liking was successful. Second, we conducted another t test to compare the means of severity of the minor and major transgressions. The severity score in the major transgression (M � 3.19, SD � .55) was significantly higher than that in the minor transgression (M � 2.24, SD � .52), t(1,133) � 5.66, p � .001. The effect size for this analysis (d � 1.78) was also found to exceed Cohen’s (1988) convention for a large effect (d � .80). Therefore, our manipulation of minor and major transgressions was also suc- cessful. 490 HU, YOUNG, LIANG, AND GUO T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. Gender Difference Both the liked and disliked media figures in the present study were male, so it is necessary to examine whether there are any gender differ- ence between male and female participants in terms of their pretest of PSR, posttest of PSR, PSB, attribution of causes, and forgiveness. We compared the means of these variables in the liked actor conditions (Condition 1 and 2) and disliked actor conditions (Condition 3 and 4) respectively. In the liked actor conditions, there were no significant gender differences on these variables. In the disliked actor conditions, fe- males’ pretest and posttest PSR scores were significantly lower than males’ scores. These results suggest that the female respondents’ PSRs with Charlie Sheen were weaker than the male respondents’ PSRs both before and after they learned of his transgression stories. There were no gender differences on attribution, PSB, and forgiveness (see Table 1 for details). Hypotheses 1, 2, 3, 4, and 5 In order to test the first five hypotheses, a 2 � 2 multivariate analysis of variance (MANOVA) factorial model was employed in the present study with media figure (liked or disliked) and transgression severity (minor or major) as the independent variables and PSR reduction, PSB, forgiveness, and attribution of causes as the dependent variables. PSR reduction score was the result of the PSR pretest score minus the PSR posttest score, so higher scores represented more reduction of PSR. We ran a series of t tests comparing the PSR pretests and posttests in the four conditions respectively. As is shown in Table 2, for the liked actor (Condition 1 and Condition 2), the PSR posttest scores were sig- nificantly lower than the pretest scores in both minor and major transgression conditions. In contrast, for the disliked actor (Condition 3 and Condition 4), the posttest score was signifi- cantly lower than the pretest score only in the major transgression condition. The results of MANOVA revealed significant multivariate effects for media figure, Wilks’ � � .63, F(4, 128) � 18.91, p � .001, and transgression severity, Wilks’ � � .66, F(4, 128) � 16.40, p � .001. There was also a significant interaction effect, Wilks’ � � .90, F(4, 128) � 1.91, p � .01. In examining the multivariate effect for me- dia figure, three significant univariate tests were observed, including PSR reduction, forgiveness, and attribution of causes. PSR reduction in the liked media figure conditions was significantly higher than that in the disliked media figure conditions. Forgiveness in the liked media fig- ure conditions was significantly higher than that in the disliked media figure conditions. As to attribution of causes, the attribution scores in the disliked media figure conditions were sig- nificantly higher than the scores in the liked media figure conditions, suggesting that the dis- liked media figure’s transgressions were more likely than liked media figures’ transgressions to be attributed to internal factors. Therefore, our first, third, and fifth hypothesis was sup- ported. However, the univariate test for PSB was not significant, so there was no difference between PSB toward the liked media figure’s Table 1 Mean Comparisons of PSR Pretest, PSR Posttest, Attribution of Causes, PSB, and Forgiveness Between Female and Male Participants in the Liked Actor Conditions and the Disliked Actor Conditions Variable Liked actor conditions t-values and p-values Disliked actor conditions t-values and p-valuesFemale Male Female Male PSR pretest 3.29 3.35 t(63) � �.40, p � .69 2.04 2.41 t(68) � �2.35, p � .05 PSR posttest 3.07 3.02 t(63) � .29, p � .78 1.97 2.34 t(68) � �2.15, p � .05 Attribution of causes 4.01 3.89 t(63) � .35, p � .73 4.33 4.77 t(68) � �1.05, p � .30 PSB 2.17 2.26 t(63) � �.43, p � .67 2.13 2.19 t(68) � �.32, p � .75 Forgiveness 3.65 3.58 t(63) � .41, p � .68 2.68 3.07 t(68) � �1.76, p � .08 Note. PSR � parasocial relationship; PSB � parasocial breakup. PSR pretest, PSR posttest, PSB, and forgiveness scale items were measured on a 5-point Likert scale (1 � strongly disagree, 5 � strongly agree). Attribution of causes scale items were measured on a 7-point semantic differential scale (higher scores suggested higher tendency to attribute the incident to the internal factors). 491REACTIONS TO TRANSGRESSIONS BY MEDIA FIGURES T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. and disliked media figure’s transgressions. Therefore, the first, third, and fifth hypotheses were supported but the second hypothesis was not. The results of these analyses are reported in Table 3. In examining the multivariate effect for trans- gression severity, four significant univariate tests were observed, including PSR reduction, PSB, forgiveness, and attribution of causes. PSR reduction in the minor transgression con- ditions was significantly lower than that in the major transgression conditions. PSB in the mi- nor transgression conditions were significantly lower than the major transgression conditions. Forgiveness in the minor transgression condi- tions was significantly higher than that in the major transgression conditions. Therefore, our fourth hypothesis regarding forgiveness and se- verity of transgression was supported. As to attribution of causes, attribution scores in the minor transgression conditions were signifi- cantly lower than those in the major transgres- sion conditions. The results of these analyses are reported in Table 4. In examining the multivariate interaction ef- fect, the univariate test for PSB was significant, F(1, 131) � 4.30, p � .05. The other three univariate tests were not significant, PSR reduc- tion, F(1, 131) � 3.75, p � .06, forgiveness, F(1, 131) � 2.56, p � .11, and attribution of causes, F(1, 131) � 3.21, p � .08. Hypothesis 6 In the sixth hypothesis, we anticipated that attribution of causes would mediate the rela- tionship between PSR and forgiveness. In order to test this mediating effect, we followed Baron and Kenny’s procedure (1986) and ran a series of regression models. In the first step, we ex- amined whether the independent variable (PSR pretest) had a direct effect on the dependent variable (forgiveness). The effect was con- firmed because PSR pretest was significantly related to forgiveness, � � .61, p � .001. In the second step, we tested whether the independent variable (PSR pretest) had an effect on the me- diating variable (attribution). The relationship was also confirmed, � � �.19, p � .05. In the third step, forgiveness (dependent variable) was set as the criterion variable and both PSR pretest (independent variable) and attribution (mediat- ing variable) were used as predictors in a re- gression equation. Attribution had a significant effect on forgiveness after PSR pretest was con- trolled, � � �.45, p � .001, and the equation of regressing forgiveness onto both attribution and PSR reached significance, F(2, 134) � 83.23, p � .001, R2 � .56. In the fourth step, the effect of PSR pretest on forgiveness controlling for attribution was examined. After the mediating variable (attribution) was controlled, the size of the direct effect identified in the first step (be- tween PSR pretest and forgiveness) was re- duced from � � .61 to �’ � .52, suggesting that attribution partially mediated the relationship Table 2 Mean Comparisons of PSR Pretest and Posttest by Condition Condition PSR pretest PSR posttest t-values and p-values Condition 1 3.26 3.18 t(32) � 2.12, p � .05 Condition 2 3.36 2.92 t(31) � 4.76, p � .001 Condition 3 2.19 2.18 t(34) � .13, p � .90 Condition 4 2.09 1.96 t(34) � 2.19, p � .05 Note. PSR � parasocial relationship. Both PSR pretest and PSR posttest scale items were measured on a 5-point Likert scale (1 � strongly disagree, 5 � strongly agree). Table 3 Univariate Tests of PSR Reduction, PSB, Forgiveness, and Attribution of Causes of Transgressions Between the Liked Media Figure and the Disliked Media Figure Variable Liked media figure Disliked media figure F values and p values PSR reduction .26 .06 F(1, 131) � 10.15, p � .001 PSB 2.21 2.15 F(1, 131) � .25, p � .62 Forgiveness 3.62 2.79 F(1, 131) � 51.16, p � .001 Attribution of causes 3.98 4.45 F(1, 131) � 4.86, p � .05 Note. PSR � parasocial relationship; PSB � parasocial breakup. 492 HU, YOUNG, LIANG, AND GUO T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. between PSR pretest and forgiveness (see Fig- ure 1). Therefore, the sixth hypothesis was par- tially supported. Discussion The present study’s contribution is reflected in the following aspects. First, it extends the current literature by comparing audiences’ re- actions to their liked media figures’ and disliked media figures’ transgressions. In contrast to a disliked figure’s transgressions, audiences’ liked figure’s transgressions caused greater PSR reduction. Second, it explores audiences’ for- giveness and fills the void of research on the reactions following PSR reduction and PSB. Although a liked media figure’s transgressions resulted in greater PSR reduction, audiences were more likely to forgive him. Third, it ex- amines the mechanism of audiences’ forgive- ness of media figures for their transgressions. It discovers the influence of FAE on audiences’ attribution of transgression causes and their for- giveness of the media figures. The audiences attributed the liked actor’s transgressions to ex- ternal factors but the disliked actor’s transgres- sions to internal factors. Such an attribution partially mediated the relationship between peo- ple’s PSR and their forgiveness. These results are similar to the findings of interpersonal rela- tionship studies, suggesting that the conse- quences of the dissolution process in a PSR resemble those in an interpersonal relationship. This resemblance further implies that media are a core element in people’s life. Mediated expe- rience and real experience are so interwoven with each other that the boundaries between the two experiences are blurring. Although PSRs are not as strong as interpersonal relationships (Cohen, 2010), media frequently and routinely bring media figures into people’s daily life and mold these figures into the substance of their living experiences (Piccirillo, 1986). It is intriguing that our first hypothesis on PSR reduction was supported but the second hypothesis regarding PSB was not. This may be due to the fact that we assigned two media figures to the participants rather than let the participants choose their own liked and disliked figures. This approach influenced people’s PSB scores with the liked actor in particular. The mean score of the PSR pretest with Clooney (Condition 1 and Condition 2) was 3.31 on a 5-point scale, suggesting that the audiences’ PSR with him might not be as strong as their PSRs with the liked personae of their choosing. Table 4 Univariate Tests of PSR Reduction, PSB, Forgiveness, and Attribution of Causes of Transgressions by Transgression Severity Variable Minor Major F values and p values PSR reduction .04 .29 F(1, 135) � 15.64, p � .001 PSB 1.99 2.37 F(1, 135) � 9.76, p � .01 Forgiveness 3.53 2.87 F(1, 135) � 31.90, p � .001 Attribution of causes 3.51 4.93 F(1, 135) � 44.18, p � .001 Note. PSR � parasocial relationship; PSB � parasocial breakup. β = -.19 β = -.45 β = .61, β’ = .52 (after attribution was controlled) Attribution PSR Forgiveness Figure 1. Mediating effect of attribution on the relationship between PSR and forgiveness. � � .61, �’ � .52 (after attribution was controlled). 493REACTIONS TO TRANSGRESSIONS BY MEDIA FIGURES T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. However, PSB measured pretty strong feelings such as anger, sadness, disappointment, sense of being betrayed, and loneliness. Therefore, Cloo- ney’s transgressions, although caused signifi- cant PSR reduction, might not arouse such strong PSB that could be differentiated from the PSB caused by Sheen’s transgressions. This study also revealed the effects of trans- gression severity. The main effects of the trans- gression severity (minor and major) on PSR reduction, PSB, attribution, and forgiveness in- dicate that increasing severity of transgressions would cause greater PSR reduction, higher PSB, more likeliness of internal attribution, and less forgiveness. This was consistent with the above-reviewed research examining severity of transgression in interpersonal relationships and PSRs. It is worth noting that for the liked media figure, there was significant reduction of PSR in both minor and major transgressions, whereas for the disliked media figure only major trans- gression caused significant reduction of PSR. These results may confirm our proposition that transgressions yield dissonance in audiences. As to their liked figures, people have high ex- pectations of what the figures ought to do, so when they learn of their even minor transgres- sions, they would experience some dissonance. In contrast, people’s expectations of disliked figures may be low due to the figures’ negative behavioral patterns in the past. Therefore, their minor transgression may not cause much disso- nance in the audiences. Another illuminating finding of this study is the partial mediating effect of attribution on the relationship between PSR and forgiveness. This finding indicates that people’s forgiveness is not completely explained by relational history. Rather, how they interpret the transgressions, especially in terms of attribution of causes, in- fluences the relationship between PSR and for- giveness. This finding is consistent with Finkel et al.’s (2002) research on forgiveness of be- trayal in close relationships, which shows that the association between relational commitment and forgiveness was partially mediated by attri- bution of causes. Despite the contribution, however, the results should be interpreted with caution given the limitations of this study. First, the media figures we used in the present study were actors. How- ever, there are other types of media figures, such as politicians, athletes, religious leaders, news- casters, and so forth. People’s expectations of me- dia figures’ behavior may vary by figure type. For example, people may expect a religious leader to be more prudent than a pop star in his lifestyle. Therefore, a lifestyle transgression committed by a religious leader may be perceived as more se- vere and harder to forgive than such a transgres- sion by a pop star. Future researchers need to explore how media figure type influences audi- ences’ reactions toward media figures’ transgres- sions. Second, people often make apologies after transgressions, and media figures often use apologies as a strategy of crisis management. Interpersonal forgiveness literature shows that apology plays an important role in the process of forgiveness (Baumeister, Exline, & Sommer, 1999; McCullough et al., 1998). Apology is a constructive step to repair media figures’ public images and restore their relationships with au- diences. Audiences’ reactions to media figures’ transgression are influenced by the figures’ at- titudes to the transgressions and their apology strategies (Sanderson & Emmons, 2014). How- ever, we did not take apology into consideration in this study, and future parasocial studies should further explore this topic. Third, in the majority of PSI or PSR studies, the research participants are asked to name their favorite or liked media figure either in general or from a certain TV program, whereas in the present study the participants didn’t get to choose their own media figures. Rather, in each condition, the researchers assigned a media fig- ure to them and then asked them to report their reactions toward a transgression committed by the figure. Therefore, the transgression stories may not have resulted in stronger feelings. However, it would be difficult to test the influ- ence of transgressions on people’s reactions to- ward a media figure of their choosing. It is unknown when people’s favorite media figures would commit transgressions and thus it is chal- lenging to capture the PSR reduction caused by the transgressions. If we employ the present study’s approach of using concocted news sto- ries, people may not believe the stories because people are probably very familiar with the me- dia figures of their choosing and know about the most recent news about those figures. If we use hypothetical transgression stories and ask peo- ple to imagine their reactions, this approach compromises response accuracy because there 494 HU, YOUNG, LIANG, AND GUO T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. may be differences between what people predict they would do and what they actually would do. Fourth, each participant in this study only read a single news report of an actor’s trans- gression. In addition, the participants were asked to report their reactions immediately after reading the news report. However, in real life a media figure’s transgression may be reported by a series of news in sequence. Sometimes the news reported later is different from or even contradictory to the news released earlier. Be- cause audiences keep learning about new pieces of the transgression over time (e.g., O. J. Simp- son’s case, Clinton-Lewinsky scandal, etc.), their reactions to media figures’ transgressions may fluctuate over time as well. Finkel et al.’s study (2002) shows that people’s immediate reactions to transgressions are more negative than their delayed reactions. Therefore, the dif- ference between how audiences were exposed to transgression news in this study and how they are exposed to such news in real life threatens the external validity of the present study. Future research should study the process of audiences’ sequential reactions to media figures’ transgres- sions. Fifth, during the time when this study was conducted, Charlie Sheen’s HIV-positive status was disclosed by news media. Although this was not mentioned in Charlie Sheen’s profile in the present study, the participants were proba- bly aware of the news. Research has shown that HIV-infected people risk discrimination and stigmatization (Kelso et al., 2014; Liamputtong, 2013). Therefore, Charlie Sheen’s HIV-positive status news may taint the way some participants viewed this study’s fictional news stories about him because they might already have prejudice against the disease. Sixth, although this study focuses on media figures’ transgressions, audiences’ reactions to- ward fictional media characters should not be ignored. In Cohen’s survey (2003), when asked to name their favorite personae, nearly two thirds of the respondents chose hosts of news, current events, or talk shows. Meanwhile, over one third of them chose fictional characters from TV series or movies. Despite the fictional nature of the characters, audiences may develop very intimate PSRs with them. As mentioned above in Tal-Or and Papirman’s (2007) study, fictional characters’ traits were so compelling that audiences believed that the actors who play the characters actually possessed such traits. Not only the incidents that occur on media figures but also those that happen to fictional characters can cause strong and profound reac- tions from audiences. Eyal and Cohen (2006) pointed out that successful TV series and movie sequels can last years even decades. If audi- ences keep following these TV series and movie sequels, they can “keep in touch” with the char- acters and hence develop long-term PSRs. The length of relationships predicts the distress over relationship dissolution (Simpson, 1987), so the long-term audiences are particularly vulnerable to the impact of PSR dissolution with the char- acters (Eyal & Cohen, 2006; Lather & Moyer- Guse, 2011). Therefore, it is worth examining these audiences’ reactions toward fictional me- dia characters’ transgressions. Seventh, we used five items from Cohen’s PSB Scale (2003) to measure PSB in this study. This PSB Scale was originally used to measure people’s negative reactions toward the termina- tion of PSRs with their favorite TV personali- ties. However, we used the same five items from the same scale to measure PSB toward both liked media figures and disliked media figures. In future research involving PSB toward liked and disliked media figures’ transgressions, re- searchers should go beyond simply comparing the extent of PSB. Rather, they should further explore the nature and components of PSB to- ward the liked media figures and disliked media figures respectively. For example, is it possible that PSB toward a liked figure’s transgression includes more surprise, disappointment, and re- gret while a disliked figure’s transgression may cause more anger, disgust, and even gloat in- stead? If so, whether is it the difference of the nature and the components of PSB that influ- ences audiences’ forgiveness of their liked and disliked media figures’ transgressions? References Amabile, T. M., Ross, L., & Steinmetz, J. L. (1977). Social roles, social control, and biases in social- perception processes. Journal of Personality and Social Psychology, 35, 485– 494. http://dx.doi.org/ 10.1037/0022-3514.35.7.485 Auter, P. J. (1992). TV that talks back: An experimental validation of a Parasocial Interaction Scale. Journal of Broadcasting and Electronic Media, 36, 173–181. http://dx.doi.org/10.1080/08838159209364165 495REACTIONS TO TRANSGRESSIONS BY MEDIA FIGURES T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. http://dx.doi.org/10.1037/0022-3514.35.7.485 http://dx.doi.org/10.1037/0022-3514.35.7.485 http://dx.doi.org/10.1080/08838159209364165 Auter, P. J., & Davis, D. M. (1991). When characters speak directly to viewers: Breaking the fourth wall in television. The Journalism Quarterly, 68, 165–171. http://dx.doi.org/10.1177/107769909106800117 Baron, R. M., & Kenny, D. A. (1986). The modera- tor-mediator variable distinction in social psycho- logical research: Conceptual, strategic, and statis- tical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. http://dx.doi .org/10.1037/0022-3514.51.6.1173 Baumeister, R. F., Exline, J. J., & Sommer, K. L. (1999). The victim role, grudge theory, and two dimensions of forgiveness. In E. L. Worthington, Jr. (Ed.), Dimensions of forgiveness: Psychologi- cal research and theological perspectives (pp. 79– 104). Philadelphia, PA: Templeton Foundation Press. Beatty, J. (1970). Forgiveness. American Philosoph- ical Quarterly, 7, 246–252. Bennett, M., & Earwaker, D. (1994). Victims’ re- sponses to apologies: The effects of offender re- sponsibility and transgression severity. Journal of Social Psychology, 134, 457– 464. Boon, S. D., & Sulsky, L. M. (1997). Attributions of blame and forgiveness: A policy-capturing study. Journal of Social Behavior and Personality, 12, 19– 44. Branch, S. E., Wilson, K. M., & Agnew, C. R. (2013). Committed to Oprah, Homer, or House: Using the Investment Model to understand paraso- cial relationships. Psychology of Popular Media Culture, 2, 96–109. http://dx.doi.org/10.1037/ a0030938 Card, N. A. (2007). ‘I hated her guts!’: Emerging adults’ recollections of the formation, mainte- nance, and termination of antipathetic relation- ships during high school. Journal of Adolescent Research, 22, 32–57. http://dx.doi.org/10.1177/ 0743558406295783 Cohen, E. (2010). Expectancy violations in relation- ships with friends and media figures. Communica- tion Research Reports, 27, 97–111. http://dx.doi .org/10.1080/08824091003737836 Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erl- baum. Cohen, J. (2003). Parasocial breakups: Measuring individual differences in responses to the dissolu- tion of parasocial relationships. Mass Communica- tion and Society, 6, 191–202. http://dx.doi.org/10 .1207/S15327825MCS0602_5 Cohen, J. (2004). Parasocial breakup from favorite television characters: The role of attachment styles and relationship intensity. Journal of Social and Personal Relationships, 21, 187–202. http://dx.doi .org/10.1177/0265407504041374 Dibble, J. L., Hartmann, T., & Rosaen, S. F. (2015). Parasocial interaction and parasocial relationship: Conceptual clarification and a critical assessment of measures. Human Communication Research, 42, 21– 44. http://dx.doi.org/10.1111/hcre.12063 Dibble, J. R., & Rosaen, S. F. (2011). Parasocial interaction as more than friendship: Evidence for parasocial interaction with disliked media figures. Journal of Media Psychology, 23, 122–132. http:// dx.doi.org/10.1027/1864-1105/a000044 Eyal, K., & Cohen, J. (2006). When good friends say goodbye: A parasocial breakup study. Journal of Broadcasting and Electronic Media, 50, 502–523. http://dx.doi.org/10.1207/s15506878jobem5003_9 Festinger, L. (1957). A theory of cognitive disso- nance. Stanford, CA: Stanford University Press. Finkel, E. J., Rusbult, C. E., Kumashiro, M., & Han- non, P. A. (2002). Dealing with betrayal in close relationships: Does commitment promote forgive- ness? Journal of Personality and Social Psychol- ogy, 82, 956–974. http://dx.doi.org/10.1037/0022- 3514.82.6.956 Forgas, J. P. (1998). On being happy and mistaken: Mood effects on the fundamental attribution error. Journal of Personality and Social Psychology, 75, 318–331. http://dx.doi.org/10.1037/0022-3514.75 .2.318 Gilbert, D. T., & Malone, P. S. (1995). The corre- spondence bias. Psychological Bulletin, 117, 21– 38. http://dx.doi.org/10.1037/0033-2909.117.1.21 Girard, M., & Mullet, E. (1997). Propensity to for- give in adolescents, young adults, older adults, and elderly people. Journal of Adult Development, 4, 209–220. http://dx.doi.org/10.1007/BF02511412 Hartmann, T., & Goldhoorn, C. (2011). Horton and Wohl revisited: Exploring viewers’ experience of parasocial interaction. Journal of Communication, 61, 1104–1121. http://dx.doi.org/10.1111/j.1460- 2466.2011.01595.x Hartmann, T., Stuke, D., & Daschmann, G. (2008). Positive parasocial relationships with drivers affect suspense in racing sports spectators. Journal of Media Psychology, 20, 24–34. http://dx.doi.org/10 .1027/1864-1105.20.1.24 Heider, F. (1958). The psychology of interpersonal relations. New York, NY: Wiley. http://dx.doi.org/ 10.1037/10628-000 Hess, J. A. (2000). Maintaining nonvoluntary rela- tionships with disliked partners: An investigation into the use of distancing behaviors. Human Com- munication Research, 26, 458– 488. http://dx.doi .org/10.1111/j.1468-2958.2000.tb00765.x Hoorn, J. F., & Konijn, E. A. (2003). Perceiving and experiencing fictional characters: An integrative account. The Japanese Psychological Research, 45, 250–268. http://dx.doi.org/10.1111/1468-5884 .00225 Horton, D., & Wohl, R. R. (1956). Mass communication and para-social interaction: Observations on intimacy at 496 HU, YOUNG, LIANG, AND GUO T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. http://dx.doi.org/10.1177/107769909106800117 http://dx.doi.org/10.1037/0022-3514.51.6.1173 http://dx.doi.org/10.1037/0022-3514.51.6.1173 http://dx.doi.org/10.1037/a0030938 http://dx.doi.org/10.1037/a0030938 http://dx.doi.org/10.1177/0743558406295783 http://dx.doi.org/10.1177/0743558406295783 http://dx.doi.org/10.1080/08824091003737836 http://dx.doi.org/10.1080/08824091003737836 http://dx.doi.org/10.1207/S15327825MCS0602_5 http://dx.doi.org/10.1207/S15327825MCS0602_5 http://dx.doi.org/10.1177/0265407504041374 http://dx.doi.org/10.1177/0265407504041374 http://dx.doi.org/10.1111/hcre.12063 http://dx.doi.org/10.1027/1864-1105/a000044 http://dx.doi.org/10.1027/1864-1105/a000044 http://dx.doi.org/10.1207/s15506878jobem5003_9 http://dx.doi.org/10.1037/0022-3514.82.6.956 http://dx.doi.org/10.1037/0022-3514.82.6.956 http://dx.doi.org/10.1037/0022-3514.75.2.318 http://dx.doi.org/10.1037/0022-3514.75.2.318 http://dx.doi.org/10.1037/0033-2909.117.1.21 http://dx.doi.org/10.1007/BF02511412 http://dx.doi.org/10.1111/j.1460-2466.2011.01595.x http://dx.doi.org/10.1111/j.1460-2466.2011.01595.x http://dx.doi.org/10.1027/1864-1105.20.1.24 http://dx.doi.org/10.1027/1864-1105.20.1.24 http://dx.doi.org/10.1037/10628-000 http://dx.doi.org/10.1037/10628-000 http://dx.doi.org/10.1111/j.1468-2958.2000.tb00765.x http://dx.doi.org/10.1111/j.1468-2958.2000.tb00765.x http://dx.doi.org/10.1111/1468-5884.00225 http://dx.doi.org/10.1111/1468-5884.00225 a distance. Psychiatry, 19, 215–229. http://dx.doi .org/10.1080/00332747.1956.11023049 Hu, M. (2016). The influence of a scandal on para- social relationship, parasocial interaction, and parasocial breakup. Psychology of Popular Media Culture, 5, 217–231. “Jennifer Lawrence has the best celebrity reputation, just as we suspected.” (2014). Retrieved from http:// www.huffingtonpost.com/2014/02/12/jennifer- lawrence-celebrity-reputation_n_4768307.html Jones, E. E., & Harris, V. A. (1967). The attribution of attitudes. Journal of Experimental Social Psy- chology, 3, 124. http://dx.doi.org/10.1016/0022- 1031(67)90034-0 Kelley, H. H., & Michela, J. L. (1980). Attribution theory and research. Annual Review of Psychology, 31, 457–501. http://dx.doi.org/10.1146/annurev.ps .31.020180.002325 Kelso, G. A., Cohen, M. H., Weber, K. M., Dale, S. K., Cruise, R. C., & Brody, L. R. (2014). Crit- ical consciousness, racial and gender discrimina- tion, and HIV disease markers in African Ameri- can women with HIV. AIDS and Behavior, 18, 1237–1246. http://dx.doi.org/10.1007/s10461-013- 0621-y Kepplinger, H. M., Geiss, S., & Siebert, S. (2012). Framing scandals: Cognitive and emotional me- dia effects. Journal of Communication, 62, 659– 681. http://dx.doi.org/10.1111/j.1460-2466.2012 .01653.x Konijn, E. A., & Hoorn, J. F. (2005). Some like it bad: Testing a model for perceiving and experiencing fictional characters. Media Psychology, 7, 107–144. http://dx.doi.org/10.1207/S1532785XMEP0702_1 Lather, J., & Moyer-Guse, E. (2011). How do we react when our favorite characters are taken away? An examination of temporary parasocial breakup. Mass Communication and Society, 14, 196–215. http://dx.doi.org/10.1080/15205431003668603 Liamputtong, P. (Ed.). (2013). Stigma, discrimina- tion and living with HIV/AIDS: A cross-cultural perspective. New York, NY: Springer Science � Business Media. McCullough, M. C., Fincham, F. D., & Tsang, J. A. (2003). Forgiveness, forbearance, and time: The temporal unfolding of transgression-related inter- personal motivations. Journal of Personality and Social Psychology, 84, 540–557. McCullough, M. E., Rachal, K. C., Sandage, S. J., Worthington, E. L., Jr., Brown, S. W., & Hight, T. L. (1998). Interpersonal forgiving in close rela- tionships: II. Theoretical elaboration and measure- ment. Journal of Personality and Social Psychol- ogy, 75, 1586–1603. http://dx.doi.org/10.1037/ 0022-3514.75.6.1586 Miller, D., & Vidmar, N. (1981). The social psychol- ogy of punishment reactions. In S. Lerner & M. Lerner (Eds.), The justice motive in social behav- ior (pp. 145–167). New York, NY: Plenum Press. http://dx.doi.org/10.1007/978-1-4899-0429-4_8 Newman, L. E. (1987). The quality of mercy: On the duty to forgive in the Judaic tradition. The Journal of Religious Ethics, 15, 155–172. Perse, E. M., & Rubin, R. B. (1989). Attribution in social and parasocial relationships. Communica- tion Research, 16, 59–77. http://dx.doi.org/10 .1177/009365089016001003 Piccirillo, M. S. (1986). On the authenticity of tele- visual experience: A critical exploration of para- social closure. Critical Studies in Mass Communi- cation, 3, 337–355. http://dx.doi.org/10.1080/ 15295038609366658 Reeder, G. D. (1982). Let’s give the fundamental attribution error another chance. Journal of Per- sonality and Social Psychology, 43, 341–344. http://dx.doi.org/10.1037/0022-3514.43.2.341 Regan, D. T., Strauss, E., & Fazio, R. (1974). Liking and the attribution process. Journal of Experimen- tal Psychology, 10, 385–397. Ross, L. (1977). The intuitive psychologist and his shortcomings. Distortions in the attribution pro- cess. In L. Berkowitz (Ed.), Advances in experi- mental social psychology (Vol. 10, pp. 174–177). New York, NY: Academic Press. http://dx.doi.org/ 10.1016/S0065-2601(08)60357-3 Rubin, A. M., & Perse, E. M. (1987). Audience activity and soap opera involvement: A uses and effects investigation. Human Communication Re- search, 14, 246–292. http://dx.doi.org/10.1111/j .1468-2958.1987.tb00129.x Russell, D. (1982). The Causal Dimension Scale: A measure of how individuals perceive causes. Jour- nal of Personality and Social Psychology, 42, 1137–1145. http://dx.doi.org/10.1037/0022-3514 .42.6.1137 Sanderson, J., & Emmons, B. (2014). Extending and withholding forgiveness to Josh Hamilton: Explor- ing forgiveness within parasocial interaction. Communication and Sport, 2, 24– 47. http://dx.doi .org/10.1177/2167479513482306 Schramm, H., & Wirth, W. (2010). Testing a univer- sal tool for measuring parasocial interactions across different situations and media: Findings from three studies. Journal of Media Psychology, 22, 26–36. http://dx.doi.org/10.1027/1864-1105/ a000004 Simpson, J. A. (1987). The dissolution of romantic relationships: Factors involved in relationship sta- bility and emotional distress. Journal of Personal- ity and Social Psychology, 53, 683– 692. http://dx .doi.org/10.1037/0022-3514.53.4.683 Tal-Or, N., & Papirman, Y. (2007). The fundamental at- tribution error in attributing fictional figures’ character- istics to the actors. Media Psychology, 9, 331–345. http://dx.doi.org/10.1080/15213260701286049 497REACTIONS TO TRANSGRESSIONS BY MEDIA FIGURES T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. http://dx.doi.org/10.1080/00332747.1956.11023049 http://dx.doi.org/10.1080/00332747.1956.11023049 http://www.huffingtonpost.com/2014/02/12/jennifer-lawrence-celebrity-reputation_n_4768307.html http://www.huffingtonpost.com/2014/02/12/jennifer-lawrence-celebrity-reputation_n_4768307.html http://www.huffingtonpost.com/2014/02/12/jennifer-lawrence-celebrity-reputation_n_4768307.html http://dx.doi.org/10.1016/0022-1031%2867%2990034-0 http://dx.doi.org/10.1016/0022-1031%2867%2990034-0 http://dx.doi.org/10.1146/annurev.ps.31.020180.002325 http://dx.doi.org/10.1146/annurev.ps.31.020180.002325 http://dx.doi.org/10.1007/s10461-013-0621-y http://dx.doi.org/10.1007/s10461-013-0621-y http://dx.doi.org/10.1111/j.1460-2466.2012.01653.x http://dx.doi.org/10.1111/j.1460-2466.2012.01653.x http://dx.doi.org/10.1207/S1532785XMEP0702_1 http://dx.doi.org/10.1080/15205431003668603 http://dx.doi.org/10.1037/0022-3514.75.6.1586 http://dx.doi.org/10.1037/0022-3514.75.6.1586 http://dx.doi.org/10.1007/978-1-4899-0429-4_8 http://dx.doi.org/10.1177/009365089016001003 http://dx.doi.org/10.1177/009365089016001003 http://dx.doi.org/10.1080/15295038609366658 http://dx.doi.org/10.1080/15295038609366658 http://dx.doi.org/10.1037/0022-3514.43.2.341 http://dx.doi.org/10.1016/S0065-2601%2808%2960357-3 http://dx.doi.org/10.1016/S0065-2601%2808%2960357-3 http://dx.doi.org/10.1111/j.1468-2958.1987.tb00129.x http://dx.doi.org/10.1111/j.1468-2958.1987.tb00129.x http://dx.doi.org/10.1037/0022-3514.42.6.1137 http://dx.doi.org/10.1037/0022-3514.42.6.1137 http://dx.doi.org/10.1177/2167479513482306 http://dx.doi.org/10.1177/2167479513482306 http://dx.doi.org/10.1027/1864-1105/a000004 http://dx.doi.org/10.1027/1864-1105/a000004 http://dx.doi.org/10.1037/0022-3514.53.4.683 http://dx.doi.org/10.1037/0022-3514.53.4.683 http://dx.doi.org/10.1080/15213260701286049 Tedeschi, J., & Felson, R. (1994). Violence, aggres- sion, and coercive actions. Washington, DC: APA. http://dx.doi.org/10.1037/10160-000 Thibaut, J. W., & Kelley, H. H. (1986). The social psychology of groups. New Brunswick, NJ: Trans- action Books. (Original work published 1959) Thompson, L. Y., Snyder, C. R., Hoffman, L., Mi- chael, S. T., Rasmussen, H. N., Billings, L. S., . . . Roberts, D. E. (2005). Dispositional forgiveness of self, others, and situations. Journal of Personality, 73, 313–360. http://dx.doi.org/10.1111/j.1467- 6494.2005.00311.x Tian, Q., & Hoffner, C. (2010). Parasocial interaction with liked, neutral, and disliked characters on a popular TV series. Mass Communication & Soci- ety, 13, 250–269. Received July 17, 2016 Revision received November 22, 2016 Accepted December 7, 2016 � Correction to Skopp et al. (2018) In the article “Positive and Negative Aspects of Facebook Use by Service Members During Deployment to Afghanistan: Associations With Perceived Social Support” by Nancy A. Skopp, Cynthia L. Alexander, Tracy Durham, and Valerie Scott (Psychology of Popular Media Culture, 2018, Vol. 7, No. 3, pp. 297–307. http://dx.doi.org/10.1037/ppm0000123), the scale name Social Media Use and Integration Scale that appears in the abstract and in the Measures section should appear instead as Social Media Use Integration Scale. The online version of this article has been corrected. http://dx.doi.org/10.1037/ppm0000214 498 HU, YOUNG, LIANG, AND GUO T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. http://dx.doi.org/10.1037/10160-000 http://dx.doi.org/10.1111/j.1467-6494.2005.00311.x http://dx.doi.org/10.1111/j.1467-6494.2005.00311.x http://dx.doi.org/10.1037/ppm0000123 http://dx.doi.org/10.1037/ppm0000214 PSB and Media Figures’ Transgressions Transgressions and Forgiveness PSI, PSR, and PSB With Disliked Media Figures Attribution of Transgressions Causes and FAE Method Sample Procedure Measures PSR Severity PSB Attribution of causes Forgiveness Results Manipulation Check Gender Difference Hypotheses 1, 2, 3, 4, and 5 Hypothesis 6 Discussion References Correction to Skopp et al. (2018) Journals/journal 6 BRIEF REPORT How People With Serious Mental Illness Use Smartphones, Mobile Apps, and Social Media John A. Naslund, Kelly A. Aschbrenner, and Stephen J. Bartels Dartmouth College Objective: Research shows that people with serious mental illness are increasingly using mobile devices. Less is known about how these individuals use their mobile devices or whether they access social media. We surveyed individuals with serious mental illness to explore their use of these technologies. Method: Individuals with serious mental illness engaged in lifestyle interventions through community mental health centers completed a survey about their use of mobile and online technologies. Responses were compared with data from the general population. Results: Among respondents (n � 70), 93% owned cellphones, 78% used text messaging, 50% owned smartphones, and 71% used social media such as Facebook. Most respondents reported daily use of text messaging, mobile apps, and social media. Technology use was comparable to the general population, though smartphone ownership was lower. Conclusions and Implications for Practice: These findings can inform future interventions that fully leverage this group’s use of popular digital technologies. Keywords: community mental health center, serious mental illness, smartphone, social media, technology More than 3.5 billion people globally use mobile devices, about 3 billion access the Internet, and more than 2 billion are active social media users (Kemp, 2015). Researchers and cli- nicians are developing cutting edge interventions using emerg- ing digital, mobile, and social technologies to address health disparities including the significantly reduced life expectancy and elevated chronic disease burden impacting individuals liv- ing with serious mental illness (Naslund, Marsch, McHugo, & Bartels, 2015; Walker, McGee, & Druss, 2015). Prior studies have documented increasing use of mobile devices among in- dividuals living with serious mental illness (Firth et al., 2016), however little is known about how these individuals actually use their own mobile devices or whether they access popular social media such as Facebook. The success of future technology-based interventions will depend largely on how the target population of people living with serious mental illness use and access services through their mobile devices. It is necessary to understand whether and how often these individuals use features such as text messaging, mobile apps, social media, or connecting with others to inform future interventions that can fully leverage this group’s use of mobile and online technology. The purpose of this study was to explore how people living with serious mental illness who receive services through community mental health centers use these technologies. Method Participants Participants were age 21 or older and had a serious mental illness defined by an axis I diagnosis of schizophrenia, schizo- affective disorder, major depressive disorder, or bipolar disor- der (based on the Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders, fourth edition, text revision [DSM–IV]). Participants were excluded if they were residing in a nursing home or other institution, had cognitive impairment defined as a Mini Mental Status Exam (Folstein, Folstein, & McHugh, 1975) score �24, had an active substance use disorder, and were unable to speak English. Participants were recruited from three different community mental health This article was published Online First June 16, 2016. John A. Naslund, Health Promotion Research Center at Dartmouth, and The Dartmouth Institute for Health Policy and Clinical Practice, Dartmouth College; Kelly A. Aschbrenner, Health Promotion Research Center at Dartmouth, and The Dartmouth Institute for Health Policy and Clinical Practice, and Department of Psychiatry, Geisel School of Medicine, Dart- mouth College; Stephen J. Bartels, Health Promotion Research Center at Dartmouth, The Dartmouth Institute for Health Policy and Clinical Prac- tice, Department of Psychiatry, and Department of Community and Family Medicine, Geisel School of Medicine, Dartmouth College. This study was supported by grants from the National Institute of Mental Health (R01 MH089811), the Agency for Health care Research and Quality (K12 HS021695-01), and from the United States Centers for Disease Control and Prevention Health Promotion and Disease Prevention Re- search Center (Cooperative Agreement U48 DP005018). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Each of the authors contributed to the data collection, analysis, and preparation of this brief report. Correspondence concerning this article should be addressed to John A. Naslund, 46 Centerra Parkway, Lebanon, NH 03766. E-mail: john.a .naslund@gmail.com T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. Psychiatric Rehabilitation Journal © 2016 American Psychological Association 2016, Vol. 39, No. 4, 364 –367 1095-158X/16/$12.00 http://dx.doi.org/10.1037/prj0000207 364 mailto:john.a.naslund@gmail.com mailto:john.a.naslund@gmail.com http://dx.doi.org/10.1037/prj0000207 centers located in urban areas in New Hampshire to participate in lifestyle intervention studies. The lifestyle interventions fo- cused on promoting healthy eating, exercise, and weight loss, and were delivered within community mental health settings. Participants completed a mobile health technology survey as part of their baseline assessment. A trained research interviewer met with participants in person and administered the surveys in a community mental health center setting. The mobile health survey consisted of questions about participants’ use of the Internet, mobile devices, text messaging, and social media, as well as specific questions about how they use these different mobile and online technologies. Participants were compensated for completing the assessment. Committees for the Protection of Human Subjects at Dartmouth College and the New Hamp- shire Department of Health and Human Services approved all study procedures. Data Analysis Descriptive statistics were calculated for participants’ demo- graphic and clinical characteristics. Survey responses were tab- ulated and compared to national data available for the general population published by the Pew Research Center in 2015 (Anderson, 2015; Duggan, Ellison, Lampe, Lenhart, & Madden, 2015; Duggan & Page, 2015). This approach of comparing data on mobile technology use from a community sample of people living with serious mental illness and from the general popu- lation was similarly employed in a recent study (Glick, Druss, Pina, Lally, & Conde, 2015). Data collection was completed in 2014 and 2015, and data analysis was completed in 2015. Results Participants (N � 70) had a mean age of 47.1 years (SD � 12.4), and were mostly female (60%) and predominantly non- Hispanic white (96%). Most participants (81%) lived indepen- dently, 17% lived with family, many were never (46%) or previously (44%) married, 45% had a high school diploma or less education, and 80% were unemployed. About one quarter (26%) of participants had a schizophrenia spectrum disorder, 41% had major depressive disorder, and 33% had bipolar dis- order. Table 1 highlights characteristics of participants’ use of their mobile devices and social media, including the type of device they use, the frequency of use, and who they contact using these technologies. The majority (93%) of survey respondents re- ported owning cellphones, 78% used text messaging, 50% owned smartphones, and 71% used popular social media such as Facebook. Most participants reported daily use of text messag- ing, mobile apps or social media. About 30% of participants with a smartphone or tablet indicated that they had used mobile apps for health or wellness purposes, and about one quarter of participants who use social media reported posting (24%) or searching for health related information (26%) on these popular platforms. Participants reported mainly using text messaging and social media to connect with family or friends. Figure 1 shows that respondents with serious mental illness engaged in lifestyle interventions through community mental health centers showed comparable rates of mobile and online connectivity as the general population. Rates of cellphone own- ership, and Internet, text messaging, and social media use were comparable with the general population. Participants living with serious mental illness reported lower rates of smartphone ownership (50%) when compared with the general population (68%). Discussion Our findings are consistent with prior surveys and reviews that have demonstrated trends of increasing mobile device ownership and digital technology use among people living with serious mental illness recruited through community mental health settings (Ben-Zeev, Davis, Kaiser, Krzsos, & Drake, 2013; Firth et al., 2016; Glick et al., 2015; Miller, Stewart, Schrimsher, Peeples, & Buckley, 2015). A recent study of a community sample of people living with serious mental illness found similarly high rates of text messaging (78%), but lower rates of smartphone ownership (37%; Glick et al., 2015). Al- though half of our sample reported owning smartphones, this was considerably lower than smartphone ownership in the gen- eral population (68%; Anderson, 2015). Research suggests that this gap in smartphone ownership, which has been largely attributed to the high costs of these devices, is rapidly closing as the devices and data plans become more affordable and widely available (Firth et al., 2016). Importantly, our study expands on prior surveys by exploring how this high-risk group uses popular social media such as Facebook or Twitter among a middle-age sample of adults with serious mental illness. One recent survey of inpatients and outpatients with schizophrenia and mean age of 41 years found that close to half (47%) used social media, of which most (79%) reported using these websites at least once each week (Miller et al., 2015). We found that a higher proportion (71%) of our sample used social media, which may be reflective of the fact that our respondents were community dwelling and did not have impaired cognitive functioning. We also found that among respondents who reported using social media, they used these websites frequently (87% reported using these websites weekly) and several reported posting or searching for health related information. Another recent survey of youth ages 12 to 21 with serious mental illness recruited from inpatient units and outpa- tient departments at a New York hospital found that more than 97% were social media users with Facebook as the most popular (94%), followed by Instagram (61%), Twitter (45%), YouTube (39%), and Tumblr (36%; Birnbaum, Rizvi, Correll, & Kane, 2015). Our survey also expands on recent studies that have suggested that people with serious mental illness are increas- ingly turning to popular social media to connect with others, to seek advice, and to share their illness experiences (Highton- Williamson, Priebe, & Giacco, 2015; Naslund, Grande, Asch- brenner, & Elwyn, 2014). Additionally, characterizing the use of social media among community samples of people with serious mental illness is critical to inform efforts to leverage these online networks for delivering interventions aimed at promoting the mental and physical wellbeing of people living with serious mental illness (Naslund, Aschbrenner, Marsch, & Bartels, 2016). T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 365TECHNOLOGY USE BY PEOPLE WITH SERIOUS MENTAL ILLNESS Limitations There are several limitations with our study that should be considered when interpreting these findings. Given the small sample size and lack of racial or ethnic diversity, our findings may not be representative of the broader population of people with serious mental illness receiving services through public mental health settings. Further, our respondents were recruited through community mental health centers located in urban areas from New Hampshire, and therefore the findings may not generalize to other settings or geographic regions. Lastly, our survey respondents were enrolled in lifestyle interventions, suggesting that they were already interested in health and wellness. Therefore, our respondents may have been more likely to post or search for health information on social media or use mobile apps for exercise, diet, or weight loss. Despite these limitations, our findings offer new insights about how these individuals use their mobile devices and whether they use social media, text messaging, and mobile apps. Conclusions and Implications for Practice This study contributes to a growing body of evidence highlight- ing the potential of using smartphone technologies to reach people living with serious mental illness and to support illness self- management (Ben-Zeev et al., 2014), symptom monitoring (Alvarez-Jimenez et al., 2014), and health promotion efforts tar- geting this high-risk group (Aschbrenner, Naslund, Barre, et al., 2015; Aschbrenner, Naslund, Shevenell, Mueser, & Bartels, 2015; Naslund, Aschbrenner, & Bartels, 2016). The successful develop- Table 1 How People Living With Serious Mental Illness Use Mobile Devices and Social Media Survey responses Participants, n (%) Among the 63 (93%) participants who own a cellphone: Use a cellphone to access the Internet 25 (40%) Frequency of sending or receiving text messages Daily 31 (49%) Weekly 12 (19%) Less than once per week 6 (10%) Use text messaging to contact Family 38 (60%) Partner or spouse 10 (16%) Friend 35 (56%) Among the 41 (59%) participants who own a smartphone or tablet: Type of mobile device Android Smartphone 23 (56%) iPhone 5 (12%) Tablet 10 (24%) Other smartphone 3 (8%) Use mobile device to access apps 39 (95%) Frequency of using mobile apps Daily 30 (73%) Weekly 6 (15%) Less than once per week 4 (10%) Use mobile apps to connect with family and friends 25 (61%) Amount willing to pay for a mobile app Will not pay, only download free apps 26 (63%) Will pay only $.99 for an app 2 (5%) Will pay more than $.99 for an app 12 (29%) Use of mobile apps for health Use mobile apps for exercise 12 (29%) Use mobile apps for diet 14 (34%) Use mobile apps for weight loss 11 (27%) Use mobile apps for quitting smoking 1 (2%) Among the 50 (71%) participants who use social media: Device used to access social media Smartphone 24 (48%) Tablet 16 (32%) Computer 32 (64%) Frequency of using social media Daily 38 (79%) Weekly 4 (8%) Less than once per week 6 (13%) Use social media to connect with Family 37 (74%) Partner or spouse 2 (4%) Friend 42 (84%) Ever posted personal health information on social media 12 (24%) Ever searched for health information on social media 13 (26%) T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 366 NASLUND, ASCHBRENNER, AND BARTELS ment and dissemination of mobile health interventions likely will be dependent on whether these interventions can be seamlessly integrated into the patterns of daily mobile device use among the target group of people with serious mental illness. For example, mobile health interventions must capitalize on the ways in which people with serious mental illness use their devices, further high- lighting the importance of understanding how these individuals use mobile and online technologies. Our findings can inform efforts that take full advantage of the different features of these technol- ogies and could support the design of tailored interventions to extend the reach and quality of services delivered through com- munity mental health settings. References Alvarez-Jimenez, M., Alcazar-Corcoles, M. A., González-Blanch, C., Ben- dall, S., McGorry, P. D., & Gleeson, J. F. (2014). Online, social media and mobile technologies for psychosis treatment: A systematic review on novel user-led interventions. Schizophrenia Research, 156, 96 –106. http://dx.doi.org/10.1016/j.schres.2014.03.021 Anderson, M. (2015). Technology device ownership: 2015. Retrieved from http://www.pewinternet.org/files/2015/10/PI_2015-10-29_device- ownership_FINAL Aschbrenner, K. A., Naslund, J. A., Barre, L. K., Mueser, K. T., Kinney, A., & Bartels, S. J. (2015). Peer health coaching for overweight and obese individuals with serious mental illness: Intervention development and initial feasibility study. Translational Behavioral Medicine, 5, 277– 284. http://dx.doi.org/10.1007/s13142-015-0313-4 Aschbrenner, K. A., Naslund, J. A., Shevenell, M., Mueser, K. T., & Bartels, S. J. (2015). Feasibility of behavioral weight loss treatment enhanced with peer support and mobile health technology for individuals with serious mental illness. Psychiatric Quarterly, 87, 401– 415. http:// dx.doi.org/10.1007/s11126-015-9395-x Ben-Zeev, D., Brenner, C. J., Begale, M., Duffecy, J., Mohr, D. C., & Mueser, K. T. (2014). Feasibility, acceptability, and preliminary efficacy of a smartphone intervention for schizophrenia. Schizophrenia Bulletin, 40, 1244 –1253. http://dx.doi.org/10.1093/schbul/sbu033 Ben-Zeev, D., Davis, K. E., Kaiser, S., Krzsos, I., & Drake, R. E. (2013). Mobile technologies among people with serious mental illness: Oppor- tunities for future services. Administration and Policy in Mental Health and Mental Health Services Research, 40, 340 –343. http://dx.doi.org/ 10.1007/s10488-012-0424-x Birnbaum, M. L., Rizvi, A. F., Correll, C. U., & Kane, J. M. (2015). Role of social media and the Internet in pathways to care for adolescents and young adults with psychotic disorders and non-psychotic mood disor- ders. Early Intervention in Psychiatry. Advance online publication. http://dx.doi.org/10.1111/eip.12237 Duggan, M., Ellison, N. B., Lampe, C., Lenhart, A., & Madden, M. (2015). Social media update 2014. Retrieved from http://www.pewinternet.org/ files/2015/01/PI_SocialMediaUpdate20144 Duggan, M., & Page, D. (2015). Mobile messaging and social media 2015. Retrieved from http://www.pewinternet.org/files/2015/08/Social-Media- Update-2015-FINAL2 Firth, J., Cotter, J., Torous, J., Bucci, S., Firth, J. A., & Yung, A. R. (2016). Mobile phone ownership and endorsement of “mHealth” among people with psychosis: A meta-analysis of cross-sectional studies. Schizophre- nia Bulletin, 42, 448 – 455. Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). “Mini-mental state”: A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research, 12, 189 –198. http://dx .doi.org/10.1016/0022-3956(75)90026-6 Glick, G., Druss, B., Pina, J., Lally, C., & Conde, M. (2015). Use of mobile technology in a community mental health setting. Journal of Telemedi- cine and Telecare, 22, 430 – 435. http://dx.doi.org/10.1177/ 1357633X15613236 Highton-Williamson, E., Priebe, S., & Giacco, D. (2015). Online social networking in people with psychosis: A systematic review. International Journal of Social Psychiatry, 61, 92–101. http://dx.doi.org/10.1177/ 0020764014556392 Kemp, S. (2015). Digital, social and mobile worldwide in 2015. Retrieved from http://wearesocial.com/uk/special-reports/digital-social-mobile- worldwide-2015. Miller, B. J., Stewart, A., Schrimsher, J., Peeples, D., & Buckley, P. F. (2015). How connected are people with schizophrenia? Cell phone, computer, email, and social media use. Psychiatry Research, 225, 458 – 463. http://dx.doi.org/10.1016/j.psychres.2014.11.067 Naslund, J. A., Aschbrenner, K. A., & Bartels, S. J. (2016). Wearable devices and smartphones for activity tracking among people with serious mental illness. Mental Health and Physical Activity, 10, 10 –17. http:// dx.doi.org/10.1016/j.mhpa.2016.02.001 Naslund, J. A., Aschbrenner, K. A., Marsch, L. A., & Bartels, S. J. (2016). The future of mental health care: Peer-to-peer support and social media. Epidemiology and Psychiatric Sciences, 25, 113–122. http://dx.doi.org/ 10.1017/S2045796015001067 Naslund, J. A., Grande, S. W., Aschbrenner, K. A., & Elwyn, G. (2014). Naturally occurring peer support through social media: The experiences of individuals with severe mental illness using YouTube. PLoS ONE, 9, e110171. http://dx.doi.org/10.1371/journal.pone.0110171 Naslund, J. A., Marsch, L. A., McHugo, G. J., & Bartels, S. J. (2015). Emerging mHealth and eHealth interventions for serious mental illness: A review of the literature. Journal of Mental Health, 24, 321–332. http://dx.doi.org/10.3109/09638237.2015.1019054 Walker, E. R., McGee, R. E., & Druss, B. G. (2015). Mortality in mental disorders and global disease burden implications: A systematic review and meta-analysis. Journal of the American Medical Association Psy- chiatry, 72, 334 –341. http://dx.doi.org/10.1001/jamapsychiatry.2014 .2502 Received January 27, 2016 Revision received May 18, 2016 Accepted May 20, 2016 � Figure 1. Technology use among people living with serious mental illness compared with the general population. Values represent proportions of the entire community sample (n � 70) of people living with serious mental illness surveyed. National survey data were obtained from the Pew Research Center reports from 2015 on Internet, smartphone, and social media use (Anderson, 2015; Duggan et al., 2015; Duggan & Page, 2015). T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 367TECHNOLOGY USE BY PEOPLE WITH SERIOUS MENTAL ILLNESS http://dx.doi.org/10.1016/j.schres.2014.03.021 http://www.pewinternet.org/files/2015/10/PI_2015-10-29_device-ownership_FINAL http://www.pewinternet.org/files/2015/10/PI_2015-10-29_device-ownership_FINAL http://dx.doi.org/10.1007/s13142-015-0313-4 http://dx.doi.org/10.1007/s11126-015-9395-x http://dx.doi.org/10.1007/s11126-015-9395-x http://dx.doi.org/10.1093/schbul/sbu033 http://dx.doi.org/10.1007/s10488-012-0424-x http://dx.doi.org/10.1007/s10488-012-0424-x http://dx.doi.org/10.1111/eip.12237 http://www.pewinternet.org/files/2015/01/PI_SocialMediaUpdate20144 http://www.pewinternet.org/files/2015/01/PI_SocialMediaUpdate20144 http://www.pewinternet.org/files/2015/08/Social-Media-Update-2015-FINAL2 http://www.pewinternet.org/files/2015/08/Social-Media-Update-2015-FINAL2 http://dx.doi.org/10.1016/0022-3956%2875%2990026-6 http://dx.doi.org/10.1016/0022-3956%2875%2990026-6 http://dx.doi.org/10.1177/1357633X15613236 http://dx.doi.org/10.1177/1357633X15613236 http://dx.doi.org/10.1177/0020764014556392 http://dx.doi.org/10.1177/0020764014556392 http://wearesocial.com/uk/special-reports/digital-social-mobile-worldwide-2015 http://wearesocial.com/uk/special-reports/digital-social-mobile-worldwide-2015 http://dx.doi.org/10.1016/j.psychres.2014.11.067 http://dx.doi.org/10.1016/j.mhpa.2016.02.001 http://dx.doi.org/10.1016/j.mhpa.2016.02.001 http://dx.doi.org/10.1017/S2045796015001067 http://dx.doi.org/10.1017/S2045796015001067 http://dx.doi.org/10.1371/journal.pone.0110171 http://dx.doi.org/10.3109/09638237.2015.1019054 http://dx.doi.org/10.1001/jamapsychiatry.2014.2502 http://dx.doi.org/10.1001/jamapsychiatry.2014.2502 How People With Serious Mental Illness Use Smartphones, Mobile Apps, and Social Media Method Participants Data Analysis Results Discussion Limitations Conclusions and Implications for Practice References Journals/Journal 7 BRIEF REPORT The Quality of Social Networks Predicts Age-Related Changes in Cardiovascular Reactivity to Stress Bert N. Uchino, Robert G. Kent de Grey, and Sierra Cronan University of Utah Although existing life span models suggest that positivity in relationships should benefit the health of older adults, much less is known about how relationships that contain both positive and negative aspects (i.e., ambivalent ties) might influence age-associated cardiovascular risk. Given the increased interper- sonal stress associated with ambivalent ties, the SAVI model would predict that older adults might be more negatively influenced given age-related changes in physiological flexibility. In this study, the quality of an individual’s social network (i.e., supportive, ambivalent, aversive) was used to predict cardiovascular reactivity during laboratory stress across a 10-month follow-up period in 108 participants between the ages 30 to 70. Results revealed evidence that the number of ambivalent network ties predicted greater increases in diastolic blood pressure reactivity. Importantly, there was an Age � Ambivalent Ties interaction in which the number of ambivalent ties was related to greater increases in systolic blood pressure reactivity primarily in older adults. These data are discussed in terms of the health implications of social networks across the life span. Keywords: ambivalent ties, cardiovascular reactivity, social support The quality of one’s social relationships is a robust predictor of physical health outcomes. In one meta-analysis, supportive rela- tionships predicted lower mortality rates even when considering initial health status and geographic region (Holt-Lunstad, Smith, & Layton, 2010). The link between social support and health shown in the meta-analysis was comparable with, if not stronger than, widely accepted risk factors, including smoking and physical ac- tivity. Importantly, the chronic conditions associated with greater mortality develop slowly over time. Hence life span relationship processes may help explain age-associated risk for chronic condi- tions such as cardiovascular disease (Antonucci, Ajrouch, & Birditt, 2014; Charles, 2010; Uchino, Ong, Queen, & Kent de Gray, 2016). Consistent with epidemiological findings, most major life span models emphasize the benefits that accrue in the social networks of older adults. Socioemotional selectivity theory (SST) highlights the importance of emotional goals as individuals age which lead older individuals to actively select or prioritize emotionally close social ties to more peripheral ones (Carstensen, Isaacowitz, & Charles, 1999). Thus, the social networks of older adults should contain emotionally positive relationships who would serve as rich sources of social support. The strength and vulnerability integra- tion (SAVI) model is a more general perspective built on SST and explains why older adults often have better well-being and takes into account both perceptions of time left but also experience gained over time (Charles, 2010). With experience, individuals have learned to avoid negative situations and/or de-escalate exist- ing negativity with close others by using selective strategies, some of which are activated after the fact (Blanchard-Fields, 2007; Charles, 2010). By incorporating aspects of SST, the predictions of SAVI regarding the importance of positive ties for health are similar. One important aspect of the SAVI model is that it also high- lights conditions under which one might expect age-associated differences in the links between relationship processes and health- related outcomes. That is, the SAVI model predicts that older adults may be more negatively influenced by interpersonal stres- sors given that older adults have less physiological flexibility (e.g., stiffer arteries, Charles, 2010; Ferrari, Radaelli, & Centola, 2003). This is an important point because many close relationships con- tain both negative and positive aspects and create significant interpersonal distress (Campo et al., 2009). Life span work on intergenerational ties (e.g., parents and their adult offspring) suggests that ambivalence in such relationships is quite common and grounded in specific interpersonal processes such as unsolicited advice, personality differences, child rearing, and past relationship problems (Birditt, Miller, Fingerman, & Lefkowitz, 2009; Pillemer et al., 2007). More generally, ambiva- Bert N. Uchino, Robert G. Kent de Grey, and Sierra Cronan, Department of Psychology and Health Psychology Program, University of Utah. This research was generously supported by James A. Shannon Director’s Award 1-R55-AG13968 from the National Institute on Aging. We thank David Lozano, Daniel Litvack, John T. Cacioppo, Robert Kelsey, and William Guethlein for their expert technical assistance and for providing us with copies of their data acquisition and reduction software (i.e., ANS suite and ENSCOREL). Correspondence concerning this article should be addressed to Bert N. Uchino, Department of Psychology, University of Utah, 380 S. 1530 E., Rm. 502, Salt Lake City, UT 84112-0251. E-mail: bert.uchino@psych.utah.edu T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. Psychology and Aging © 2016 American Psychological Association 2016, Vol. 31, No. 4, 321–326 0882-7974/16/$12.00 http://dx.doi.org/10.1037/pag0000092 321 mailto:bert.uchino@psych.utah.edu http://dx.doi.org/10.1037/pag0000092 lent ties appear to exert detrimental influences on health, above and beyond relationship positivity and negativity, as a result of increased interpersonal stress and reduced support seeking (Holt- Lunstad, Uchino, Smith, & Hicks, 2007; Reblin, Uchino, & Smith, 2010). In addition, ambivalent ties are typically close relationships and the existing positivity makes it more difficult to avoid such ties, at least compared with aversive relationships (Uchino, Holt- Lunstad, Uno, & Flinders, 2001). As a result, ambivalent ties are independent predictors of worse health outcomes including greater inflammation and coronary artery calcification (Uchino et al., 2013; Uchino, Smith, & Berg, 2014). Consistent with the importance of examining positive and neg- ative network qualities, one relevant study utilized such a compre- hensive assessment and examined age differences in cardiovascu- lar reactivity (ages 30 to 70) to a stress reactivity protocol (Uchino et al., 2001). This cross-sectional study found that both positivity and ambivalence in network ties predicted cardiovascular reactiv- ity during stress (Uchino et al., 2001). For instance, a low number of supportive ties and a high number of ambivalent ties predicted greater heart rate reactivity as a function of age. However, no study to date has examined such associations longitudinally. This is important because social networks may influence older adults for better or worse thereby influencing their ability to cope with stress in a laboratory setting over time. Such laboratory stress assess- ments are thought to reflect how individuals cope with stress in their daily life and is a predictor of future cardiovascular disease risk in otherwise healthy populations (Chida & Steptoe, 2010). The main goal of this study was thus to examine the prediction of changes in cardiovascular reactivity over a 10-month period via social network quality and its potential moderation by age. Based on prior work, it was predicted that the number of supportive and ambivalent ties would predict changes in cardiovascular reactivity: a higher number of supportive ties should predict lower cardio- vascular reactivity over time whereas a higher number of ambiv- alent ties should predict greater cardiovascular reactivity over time. Although our prior work suggests more limited influences for aversive ties, consistent with the SAVI model it was predicted that the link between aversive and ambivalent ties would be moderated by age as a higher number of such ties should be related to greater increases in cardiovascular reactivity primarily in older adults. Method Participants The original study tested 64 men and 69 women between the ages of 30 and 70 (see Uchino et al., 2001). Approximately equal numbers of men and women were recruited from each decade group (e.g., 30 to 39) through advertisements placed in local newspapers. Individuals were paid $35.00 for approximately 2.5 hours of participation. The following self-reported inclusion crite- ria were used to select healthy participants: no existing hyperten- sion, no cardiovascular prescription medication use, no past his- tory of chronic disease with a cardiovascular component (e.g., diabetes), no recent history of psychological disorder (e.g., major depressive disorder), no tobacco use, and no consumption of more than 10 alcoholic beverages a week (Cacioppo et al., 1995). In the follow-up, individuals were rescreened according to the inclusion criteria listed above and paid $35.00 for approximately 2.5 hours of participation; 108 (81%) of the original sample were retested on average 10 months later (SD � 1.6, range of 7 to 16 months). The main reasons for attrition were related to our inabil- ity to contact and in a small minority of cases problems in sched- uling individuals. Importantly, there were no significant differ- ences in demographic characteristics between participants who were lost at the follow-up in comparison to participants who were retested (see Uchino, Holt-Lunstad, Bloor, & Campo, 2005). Procedure Individuals were first recontacted via telephone and rescreened according to the inclusion criteria detailed above. Qualifying in- dividuals were scheduled for an appointment and participants’ self-reports were again checked for reliability against the inclusion criteria. Upon arrival, participants completed an informed consent document and a demographic background questionnaire. Follow- ing completion of these questionnaires, the participant’s height and weight were obtained using a standard medical scale from which body mass index was calculated (i.e., weight in kg/height2 in meters). Participants were then escorted to a separate sound attenuated room where an occluding cuff of appropriate size was also placed on the upper left arm. Individuals were seated in a comfortable chair and instructed to relax for the next 12 min while resting measures of cardiovascular function were obtained. During the final 5 min of the resting assessment, cardiovascular assessments of heart rate, SBP, and DBP were obtained once every 90 seconds. Following the resting assessments, participants performed a speech and mental arithmetic protocol (Cacioppo et al., 1995). The same basic stressors tasks were used for the time two reactivity protocol with slight changes to minimize habituation (e.g., different speech topic and different set of serial subtractions). The order of the stressors was counterbalanced and all verbal instructions were standardized (see Uchino et al., 2005). Upon completion of both psychological stressors, participants were debriefed, compensated, and thanked for their participation. Although impedance measures of cardiac output, preejection period, and respiratory sinus arrhyth- mia were collected as part of the larger protocol, these measures did not elucidate the underlying determinants of blood pressure and heart rate reported below so they are not considered further. Assessments Social Relationships Index (SRI). The SRI was completed at Time 1 and instructed individuals to list the initials of up to 10 important network members with whom they have contact (Campo et al., 2009). These network members were then rated in terms of how helpful and upsetting they were (1 � not at all, 2 � a little, 3 � sometimes, 4 � moderately, 5 � very, 6 � extremely) when the participant needed emotional, tangible, and informational sup- port. These helpful and upsetting ratings for aspects of support have been shown to load on a general positivity and negativity factor (Campo et al., 2009). For the present study, the internal consistencies for the SRI helpful and upset ratings were compara- ble to our pilot work (alphas of .76 to .87). To increase reliability through aggregation, for each network member an index of posi- tivity and negativity was calculated by averaging across support components within helpful and upset ratings. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 322 UCHINO, KENT DE GREY, AND CRONAN Consistent with our prior work, we operationalized different cate- gories of social relationships as the number of individuals in one’s network who were only sources of positivity (i.e., supportive ties), only sources of negativity (i.e., aversive ties), or sources of both positivity and negativity (i.e., ambivalent ties). Thus, a socially sup- portive network member was an individual rated as a “2” or greater on helpful and only a “1” on upset and an aversive network tie was an individual rated as only a “1” on helpful and a “2” or greater on upset. An ambivalent network member was rated as a “2” or greater on both helpful and upset ratings. In our prior work, these network measures were temporally stable with significant 3-month test–retest correla- tions of r � .61 (p � .001) for the number of supportive ties, r � .30 (p � .001) for the number of aversive ties, and r � .68 (p � .001) for the number of ambivalent ties. The SRI also has good convergent (e.g., perceived support) and discriminant (e.g., personality) validity (Campo et al., 2009). Cardiovascular measures. A Minnesota Impedance Cardio- graph Model 304B was used to measure heart rate, whereas a Di- namap Model 8100 monitor (Critikon corporation, Tampa, Florida) was used to measure blood pressure. The Dinamap used the oscillo- metric method to estimate blood pressure (see Gorback, Quill, & Lavine, 1991 for validation). Mean SBP, DBP, and heart rate for each epoch was averaged across minutes to increase the reliability of these assessments (Kamarck et al., 1992). Statistical Analyses For primary analyses, moderated regression analyses with pre- dictors centered at the grand mean were used to examine the separate links between the number of supportive, aversive, and ambivalent ties at Time 1 and subsequent age-related changes in cardiovascular reactivity (i.e., task minus baseline). Age was treated as a continuous factor in all analyses to increase statistical power. To conduct these tests, the Age � Social Tie cross-product terms (based on the centered main effects) were entered into the model after the respective main effects (Aiken & West, 1991). Main effects of social ties on cardiovascular reactivity were ex- amined without the cross-product term. Effects sizes for significant associations were calculated based on f2 (Aiken & West, 1991). Consistent with prior work, all analyses of Time 2 cardiovascular reactivity measures controlled for age, sex, changes in body mass index, Time 1 resting measures of cardiovascular function and cardiovascular reactivity, as well as Time 2 resting measures of cardiovascular function. In addition, all analyses statistically con- trolled for race/ethnicity (White, Other), education, and income. Simple slope analyses were carried out to examine the form of any significant Age � Social Tie interactions by testing for the signif- icance of the slope one standard deviation above and below the mean for age (Aiken & West, 1991). Finally, ancillary analyses examined if any significant associations between positive, nega- tive, or ambivalent network ties were relatively independent of each other by statistically controlling for the other relationship categories. Results Sample Characteristics The average age of the follow-up sample was 47.9 (SD � 10.9, range 30 to 70). Over 84% of the sample was non-Hispanic White. The median education and yearly income were partial college/ graduate college and $20,000 to $29,000, respectively. Participants also listed an average of 9.34 important network members (out of 10). The average number of ambivalent ties was 6.1, the average number of supportive ties was 3.0, and the average number of aversive ties was 0.2. Finally, as detailed in our prior publication with this sample, age was related to greater increases in stress reactivity (e.g., SBP reactivity) over time (see Uchino et al., 2005). Main Analyses Given the importance of positive network ties in life span models, analyses first examined whether the number of supportive network ties predicted age-related changes in cardiovascular reac- tivity. In these models, there were no significant main effects of supportive ties nor any Age � Supportive ties interaction on any measures of cardiovascular reactivity (see Table 1).1 Analyses focusing on the number of ambivalent ties showed one main effect as a relatively high number of ambivalent ties at Time 1 predicted greater increases in DBP reactivity over time, b � .41, 95% CI [.10, .73], � � .22, p � .01, f 2 � .06. Importantly, there was also an Age � Ambivalent Ties interaction for changes in SBP reactivity, b � .05, 95% CI [.01, .09], � � .16, p � .02, f 2 � .075.2 These represent small to moderate effect sizes (.02 to .15, Aiken & West, 1991). As shown in Figure 1, simple slope analyses for SBP reactivity one standard deviation above and below the mean for age showed that the number of ambivalent ties predicted greater increases in reactivity in older (p � .03) but not younger (p � .24) participants. Consistent with our prior work on aversive ties, no significant main effects or interactions were found on any measures of cardiovascular reactivity. Finally, analyses were run to examine whether the results above for ambivalent ties were independent of the other categories. First of all, the number of ambivalent ties remained an independent predictor of DBP reactivity even when statistically controlling for the number of supportive ties and aversive ties (ps � .02). The Age � Ambivalent Ties interaction on SBP reactivity was also still significant while statistically controlling for the Age � Supportive Ties and Age � Aversive Ties interactions (ps � .05). Discussion The main goal of this study was to examine whether age moderated the link between relationship quality and cardiovascular reactivity during stress over time. The SAVI model would predict that older adults in particular might have problems coping with interpersonal stress because of age-related decreases in physiolog- ical flexibility (Charles, 2010). In partial support of this model, the 1 There was a marginally significant Age � Supportive Ties interaction on changes in heart rate reactivity during acute stress, b � .03, 95% CI [�.002, .06], � � .12, p � .07. Exploratory simple slope analyses within age groups found that the only simple effect to approach significance was within the younger group, where the number of supportive ties predicted a greater decrease in heart rate reactivity over time (p � .09). 2 There was a marginal Age � Ambivalent Ties interaction on heart rate reactivity, b � �.02, 95% CI [�.05, .001], � � �.12, p � .07. Exploratory simple slope analyses showed that the number of ambivalent ties primarily predicted greater increases in heart rate reactivity in the younger (p � .09) group. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 323AGE, SOCIAL TIES, AND CARDIOVASCULAR CHANGES number of ambivalent ties predicted greater increases in DBP reactivity and greater increases in SBP reactivity over time pri- marily in older adults. Although the relatively short follow-up precludes strong inferences about cardiovascular changes, these findings are potentially important because SBP and DBP reactivity assessed in laboratory setting are strong predictors of future car- diovascular risk such as incident hypertension (Chida & Steptoe, 2010). Most of the prior work in the area emphasizes the importance of relationship positivity in protecting individuals from adverse health outcomes (Holt-Lunstad et al., 2010). This study provided little evidence for the benefits of supportive ties in predicting age-related changes in cardiovascular reactivity. The primacy of ambivalent ties over supportive ties in our study might be attrib- utable to a negativity bias in which stimuli which contain negative properties have more robust associations to physiological out- comes (Taylor, 1991). Of course, the sample size was relatively small which resulted in a few marginally significant findings (see Footnote 1 and 2) so future work would be needed to examine such associations with greater statistical power. The SAVI model also highlights the potential problems older adults might have in coping with negative social ties (Charles, 2010). Such negative influences were not detected in regards to aversive (negative) ties. However, such ties might be easier to cope with given that they are less close and hence individuals more readily discount or ignore them (Uchino et al., 2001). In addition, older adults are more likely to use selectivity as a coping mecha- nism and hence avoid such ties in their daily life (Carstensen et al., 1999). However, the results for ambivalent ties appear consistent with the model. The SAVI model highlights the potential difficul- ties older adults have in coping with high arousal situations (Charles, 2010; also see Labouvie-Vief, 2008). This argument assumes that interactions with ambivalent ties are relatively high in arousal compared to interactions with aversive relationships. In- deed, coping with ambivalent ties appears taxing (Fingerman, Pitzer, Lefkowitz, Birditt, & Mroczek, 2008) and results in higher levels of ambulatory blood pressure during everyday life compared with aversive ties (Holt-Lunstad, Uchino, Smith, Olson-Cerny, & Nealey-Moore, 2003). This might tax the coping abilities of older adults who are already showing age-related declines in cardiovas- cular functioning such that they are less able to cope with stressors more generally as indexed by our laboratory protocol (Charles, 2010). Future work that directly models important mediators (e.g., level of arousal) postulated by existing life span models will be key to integrating empirical data in this domain. There are several limitations of this study. One important lim- itation is that the time frame for the follow-up was relatively short, so whether these results reflect enduring changes or shorter fluc- Table 1 Main Results for Moderated Regression Models Predicting Changes in SBP, DBP, and Heart Rate Over the 10-Month Period Predictor Systolic blood pressure Diastolic blood pressure Heart rate b SE p � 95% CI b SE p � 95% CI b SE p � 95% CI Ambivalent ties T1 baseline .19 .07 .007 .26 .05, .32 .29 4.28 �.001 .52 .16, .43 .14 .04 �.001 .30 .06, .22 T2 baseline �.09 .07 .204 �.12 �.22, .05 �.29 �4.29 �.001 �.52 �.42, �.15 �.15 .04 �.001 �.32 �.23, �.07 T1 reactivity .73 .08 �.001 .66 .57, .89 .46 5.46 �.001 .48 .29, .62 .60 .05 �.001 .76 .49, .71 Age .18 .06 .004 .23 .06, .30 .03 .65 .515 .06 �.05, .11 .01 .03 .798 .02 �.05, .07 Sex �1.52 1.20 .208 �.09 �3.91, .86 1.09 1.20 .234 .11 �.71, 2.88 1.13 .70 .108 .11 �.25, 2.51 Ethnicity .25 .83 .765 .02 �1.39, 1.89 �.72 �1.20 .234 �.11 �1.92, .47 .91 .48 .060 .13 �.04, 1.87 Income �.26 .34 .442 �.06 �.94, .41 .34 1.40 .166 .12 �.14, .82 �.16 .18 .377 �.06 �.53, .20 Education .61 .46 .184 .09 �.30, 1.52 .37 1.12 .268 .10 �.29, 1.02 .11 .27 .672 .03 �.42, .65 BMI change 8.74 274.94 .975 �.01 �537, 555 139.09 197.26 .483 .06 �253, 531 27.87 151.48 .854 .01 �273, 329 Ambivalent ties T1 .17 .22 .429 .05 �.26, .60 .41 2.63 .010 .22 .10, .73 .04 .12 .729 .02 �.20, .29 Age � T1 Ambivalent ties .05 .02 .022 .16 .01, .09 .01 .59 .554 .05 �.02, .04 �.02 .01 .066 �.13 �.05, .001 Supportive ties T1 baseline .20 .07 .006 .27 .06, .34 .30 .07 �.001 .53 .16, .44 .15 .04 �.001 .31 .07, .23 T2 baseline �.10 .07 .161 �.13 �.24, .04 �.28 .07 �.001 �.51 �.42, �.15 �.15 .04 �.001 �.32 �.23, �.07 T1 reactivity .75 .08 �.001 .68 .59, .92 .47 .09 �.001 .50 .30, .64 .59 .05 �.001 .75 .49, .70 Age .16 .06 .012 .20 .04, .28 .01 .04 .767 .03 �.07, .09 .01 .03 .688 .03 �.05, .07 Sex �1.41 1.23 .254 �.08 �3.86, 1.03 1.10 .92 .236 .11 �.73, 2.93 1.12 .70 .111 .11 �.26, 2.50 Ethnicity .27 .86 .754 .02 �1.43, 1.97 �.69 .62 .268 �.10 �1.93, .54 .93 .48 .059 .13 �.04, 1.89 Income �.17 .35 .634 �.04 �.85, .52 .39 .24 .110 .14 �.09, .08 �.17 .18 .352 �.06 �.53, .19 Education .55 .48 .256 .08 �.40, 1.50 .38 .34 .277 .10 �.31, 1.06 .03 .28 .906 .01 �.52, .58 BMI change �103.0 290.35 .724 �.03 �679, 473 72.91 207.04 .726 .03 �338, 484 107.0 155.41 .493 .05 �202, 416 Supportive ties T1 �.04 .27 .869 �.01 �.57, .48 �.31 .19 .112 �.14 �.69, .07 .03 .14 .810 .02 �.25, .32 Age � T1 supportive ties �.03 .03 .355 �.07 �.08, .03 �.01 .02 .581 �.05 �.05, .03 .03 .01 .072 .12 �.002, .06 Aversive ties T1 baseline .20 .07 .004 .28 .07, .34 .28 .07 �.001 .50 .14, .42 .14 .04 �.001 .31 .06, .23 T2 baseline �.11 .07 .097 �.15 �.25, .02 �.26 .07 �.001 �.48 �.40, �.13 �.15 .04 �.001 �.33 �.24, �.07 T1 reactivity .74 .08 �.001 .67 .58, .91 .45 .09 �.001 .48 .28, .62 .57 .05 �.001 .72 .46, .68 Age .16 .06 .012 .21 .04, .29 .01 .04 .808 .02 �.07, .09 .02 .03 .563 .04 �.04, .08 Sex �1.17 1.24 .346 �.07 �3.63, 1.29 1.27 .93 .175 .13 �.58, 3.11 1.18 .72 .105 .12 �.25, 2.60 Ethnicity .11 .84 .898 .01 �1.57, 1.78 �.61 .62 .325 �.09 �1.83, .61 .99 .49 .045 .14 .02, 1.95 Income �.21 .35 .547 �.05 �.91, .48 .31 .25 .223 .11 �.19, .81 �.22 .19 .236 �.08 �.60, .15 Education .43 .46 .358 .07 �.49, 1.35 .28 .34 .412 .07 �.39, .94 .13 .28 .650 .03 �.42, .67 BMI change �21.56 284.21 .940 �.01 �586, 543 180.12 202.66 .376 .08 �222, 582 66.51 154.22 .667 .03 �240, 373 Aversive ties T1 �.61 1.25 .625 �.04 �3.09, 1.86 �.22 .91 .810 �.03 �2.02, 1.58 �.41 .70 .556 �.05 �1.79, .97 Age � T1 Aversive ties �.10 .15 .528 �.05 �.40, .21 �.13 .11 .230 �.13 �.36, .09 �.01 .09 .873 �.01 �.19, .16 Note. Ambivalent, supportive, and aversive ties main effects statistics are based on models without the cross-product term. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 324 UCHINO, KENT DE GREY, AND CRONAN tuations is unknown. How quickly or slowly cardiovascular reac- tivity to stress changes over time in older adults and its relevance to health is difficult to determine as there is little longitudinal work on the issue (Uchino et al., 2005). However, there is predictive validity to such assessments at any given point in time as higher SBP and DBP reactivity to laboratory stress predicts incident hypertension many years later (Chida & Steptoe, 2010). Future longitudinal work that follows individuals over a longer period of time will be needed to test for slower changes that might result in smaller effect sizes that the present study was underpowered to detect. In a related point, we did not assess the SRI at both time points, so whether relationships are changing over this time period is unknown. Although there is both stability and change in the relationships of aging adults (Carstensen et al., 1999), ambivalent ties may be relatively stable as they are important familial ties (e.g., parents, adult offspring; van Gaalen, Dykstra, & Komter, 2010) and can be maintained as both explicit and implicit relation- ship representations (Petty, Tormala, Briñol, & Jarvis, 2006). Nevertheless, this remains an important issue as most prior work has not examined changes in ambivalent relationships across the life span. Finally, although the sample was relatively large for work examining cardiovascular reactivity, future work will be needed using larger samples sizes to increase confidence in the pattern of results. Nevertheless, these data are also among the first evidence link- ing ambivalent ties to age-related increases in cardiovascular re- activity over time. Although the associations detected in this study are closer to a small effect size, it is important to highlight that greater blood pressure reactivity assessed in similar laboratory settings are linked to increased cardiovascular disease risk (Chida & Steptoe, 2010; Uchino, Birmingham, & Berg, 2010). In fact, these effect sizes are comparable with those found in a meta- analysis linking increased cardiovascular reactivity to cardiovas- cular disease (Chida & Steptoe, 2010). Although research has not yet tested whether intervention-related reductions in stress reac- tivity can decrease cardiovascular risk over time, these data high- light the potential importance of future work in the area. To this point, interventions can be aimed at helping older adults cope with such ambivalent ties using established (e.g., family or marital therapy) or potentially unique (e.g., meditation) approaches that appear successful in reducing biological reactivity (Ditzen, Hahl- weg, Fehm-Wolfsdorf, & Baucom, 2011; Nyklíček, Mommer- steeg, Van Beugen, Ramakers, & Van Boxtel, 2013). References Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park: Sage. Antonucci, T. C., Ajrouch, K. J., & Birditt, K. S. (2014). The Convoy Model: Explaining social relations from a multidisciplinary perspective. The Gerontologist, 54, 82–92. http://dx.doi.org/10.1093/geront/gnt118 Birditt, K. S., Miller, L. M., Fingerman, K. L., & Lefkowitz, E. S. (2009). Tensions in the parent and adult child relationship: Links to solidarity and ambivalence. Psychology and Aging, 24, 287–295. http://dx.doi.org/ 10.1037/a0015196 Blanchard-Fields, F. (2007). Everyday problem solving and emotion. Cur- rent Directions in Psychological Science, 16, 26 –31. http://dx.doi.org/ 10.1111/j.1467-8721.2007.00469.x Cacioppo, J. T., Malarkey, W. B., Kiecolt-Glaser, J. K., Uchino, B. N., Sgoutas-Emch, S. A., Sheridan, J. F., . . . Glaser, R. (1995). Heteroge- neity in neuroendocrine and immune responses to brief psychological stressors as a function of autonomic cardiac activation. Psychosomatic Medicine, 57, 154 –164. Retrieved from http://www.ncbi.nlm.nih.gov/ pubmed/7792374 Campo, R. A., Uchino, B. N., Holt-Lunstad, J., Vaughn, A., Reblin, M., & Smith, T. W. (2009). The assessment of positivity and negativity in social networks: The reliability and validity of the social relationships index. Journal of Community Psychology, 37, 471– 486. http://dx.doi .org/10.1002/jcop.20308 Carstensen, L. L., Isaacowitz, D. M., & Charles, S. T. (1999). Taking time seriously. A theory of socioemotional selectivity. American Psycholo- gist, 54, 165–181. http://dx.doi.org/10.1037/0003-066X.54.3.165 Charles, S. T. (2010). Strength and vulnerability integration: A model of emotional well-being across adulthood. Psychological Bulletin, 136, 1068 –1091. http://dx.doi.org/10.1037/a0021232 Chida, Y., & Steptoe, A. (2010). Greater cardiovascular responses to laboratory mental stress are associated with poor subsequent cardiovas- cular risk status: A meta-analysis of prospective evidence. Hypertension, 55, 1026 –1032. http://dx.doi.org/10.1161/HYPERTENSIONAHA.109 .146621 Ditzen, B., Hahlweg, K., Fehm-Wolfsdorf, G., & Baucom, D. (2011). Assisting couples to develop healthy relationships: Effects of couples relationship education on cortisol. Psychoneuroendocrinology, 36, 597– 607. http://dx.doi.org/10.1016/j.psyneuen.2010.07.019 Ferrari, A. U., Radaelli, A., & Centola, M. (2003). Invited review: Aging and the cardiovascular system. Journal of Applied Physiology, 95, 2591– 2597. http://doi.org/10.1152/japplphysiol.00601.2003 Fingerman, K. L., Pitzer, L., Lefkowitz, E. S., Birditt, K. S., & Mroczek, D. (2008). Ambivalent relationship qualities between adults and their parents: Implications for the well-being of both parties. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 63, 362–371. http://dx.doi.org/10.1093/geronb/63.6.P362 Gorback, M. S., Quill, T. J., & Lavine, M. L. (1991). The relative accu- racies of two automated noninvasive arterial pressure measurement devices. Journal of Clinical Monitoring, 7, 13–22. http://www.ncbi.nlm .nih.gov/pubmed/1999694. http://dx.doi.org/10.1007/BF01617893 Holt-Lunstad, J., Smith, T. B., & Layton, J. B. (2010). Social relationships and mortality risk: A meta-analytic review. PLoS Medicine, 7, e1000316. http://dx.doi.org/10.1371/journal.pmed.1000316 Holt-Lunstad, J., Uchino, B. N., Smith, T. W., & Hicks, A. (2007). On the importance of relationship quality: The impact of ambivalence in friend- 8 10 12 14 16 18 20 Low Amb. Ties High Amb. Ties Young Older S B P R ea ct iv ity (m m H g) Figure 1. Predicted changes in SBP reactivity over the 10-month period one standard deviation above and below the mean for age and the number of ambivalent ties. T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 325AGE, SOCIAL TIES, AND CARDIOVASCULAR CHANGES http://dx.doi.org/10.1093/geront/gnt118 http://dx.doi.org/10.1037/a0015196 http://dx.doi.org/10.1037/a0015196 http://dx.doi.org/10.1111/j.1467-8721.2007.00469.x http://dx.doi.org/10.1111/j.1467-8721.2007.00469.x http://www.ncbi.nlm.nih.gov/pubmed/7792374 http://www.ncbi.nlm.nih.gov/pubmed/7792374 http://dx.doi.org/10.1002/jcop.20308 http://dx.doi.org/10.1002/jcop.20308 http://dx.doi.org/10.1037/0003-066X.54.3.165 http://dx.doi.org/10.1037/a0021232 http://dx.doi.org/10.1161/HYPERTENSIONAHA.109.146621 http://dx.doi.org/10.1161/HYPERTENSIONAHA.109.146621 http://dx.doi.org/10.1016/j.psyneuen.2010.07.019 http://doi.org/10.1152/japplphysiol.00601.2003 http://dx.doi.org/10.1093/geronb/63.6.P362 http://www.ncbi.nlm.nih.gov/pubmed/1999694 http://www.ncbi.nlm.nih.gov/pubmed/1999694 http://dx.doi.org/10.1007/BF01617893 http://dx.doi.org/10.1371/journal.pmed.1000316 ships on cardiovascular functioning. Annals of Behavioral Medicine, 33, 278 –290. http://dx.doi.org/10.1007/BF02879910 Holt-Lunstad, J., Uchino, B. N., Smith, T. W., Olson-Cerny, C., & Nealey- Moore, J. B. (2003). Social relationships and ambulatory blood pressure: Structural and qualitative predictors of cardiovascular function during everyday social interactions. Health Psychology, 22, 388 –397. http://dx .doi.org/10.1037/0278-6133.22.4.388 Kamarck, T. W., Jennings, J. R., Debski, T. T., Glickman-Weiss, E., Johnson, P. S., Eddy, M. J., & Manuck, S. B. (1992). Reliable measures of behaviorally-evoked cardiovascular reactivity from a PC-based test battery: Results from student and community samples. Psychophysiol- ogy, 29, 17–28. http://dx.doi.org/10.1111/j.1469-8986.1992.tb02006.x Labouvie-Vief, G. (2008). Dynamic integration theory: Emotion, cogni- tion, and equilibrium in later life. In V. Bengtson, M. Silverstein, N. Putney, & D. Gans (Eds.), Handbook of theories of aging (pp. 277–293). New York, NY: Springer. Nyklíček, I., Mommersteeg, P. M. C., Van Beugen, S., Ramakers, C., & Van Boxtel, G. J. (2013). Mindfulness-based stress reduction and phys- iological activity during acute stress: A randomized controlled trial. Health Psychology, 32, 1110 –1113. http://dx.doi.org/10.1037/a0032200 Petty, R. E., Tormala, Z. L., Briñol, P., & Jarvis, W. B. G. (2006). Implicit ambivalence from attitude change: An exploration of the PAST model. Journal of Personality and Social Psychology, 90, 21– 41. http://dx.doi .org/10.1037/0022-3514.90.1.21 Pillemer, K., Suitor, J. J., Mock, S. E., Sabir, M., Pardo, T. B., & Sechrist, J. (2007). Capturing the complexity of intergenerational relations: Ex- ploring ambivalence within later-life families. Journal of Social Issues, 63, 775–791. http://dx.doi.org/10.1111/j.1540-4560.2007.00536.x Reblin, M., Uchino, B. N., & Smith, T. W. (2010). Provider and recipient factors that may moderate the effectiveness of received support: Exam- ining the effects of relationship quality and expectations for support on behavioral and cardiovascular reactions. Journal of Behavioral Medi- cine, 33, 423– 431. http://dx.doi.org/10.1007/s10865-010-9270-z Taylor, S. E. (1991). Asymmetrical effects of positive and negative events: The mobilization-minimization hypothesis. Psychological Bulletin, 110, 67– 85. http://dx.doi.org/10.1037/0033-2909.110.1.67 Uchino, B. N., Birmingham, W., & Berg, C. A. (2010). Are older adults less or more physiologically reactive? A meta-analysis of age-related differences in cardiovascular reactivity to laboratory tasks. Journals of Gerontology: Series B: Psychological Sciences and Social Sciences, 65, 154 –162. http://doi.org/10.1093/geronb/gbp127 Uchino, B. N., Bosch, J. A., Smith, T. W., Carlisle, M., Birmingham, W., Bowen, K. S., . . . O’Hartaigh, B. (2013). Relationships and cardiovas- cular risk: Perceived spousal ambivalence in specific relationship con- texts and its links to inflammation. Health Psychology, 32, 1067–1075. http://dx.doi.org/10.1037/a0033515 Uchino, B. N., Holt-Lunstad, J., Bloor, L. E., & Campo, R. A. (2005). Aging and cardiovascular reactivity to stress: Longitudinal evidence for changes in stress reactivity. Psychology and Aging, 20, 134 –143. http:// dx.doi.org/10.1037/0882-7974.20.1.134 Uchino, B. N., Holt-Lunstad, J., Uno, D., & Flinders, J. B. (2001). Heter- ogeneity in the social networks of young and older adults: Prediction of mental health and cardiovascular reactivity during acute stress. Journal of Behavioral Medicine, 24, 361–382. http://dx.doi.org/10.1023/A: 1010634902498 Uchino, B. N., Ong, A., Queen, T. L., & Kent de Grey, R. G. (2016). Theories of social support in health and aging. In R. Bengtson & V. Settersten (Eds.), Handbook of theories of aging (3rd. ed.). New York, NY: Springer. Uchino, B. N., Smith, T. W., & Berg, C. A. (2014). Spousal relationship quality and cardiovascular risk: Dyadic perceptions of relationship am- bivalence are associated with coronary-artery calcification. Psychologi- cal Science, 25, 1037–1042. http://dx.doi.org/10.1177/09567 97613520015 van Gaalen, R. I., Dykstra, P., & Komter, A. E. (2010). Where is the exit? Intergenerational ambivalence and relationship quality in high contact ties. Journal of Aging Studies, 24, 105–114. http://dx.doi.org/10.1016/j .jaging.2008.10.006 Received September 4, 2015 Revision received March 30, 2016 Accepted April 5, 2016 � T hi s do cu m en t is co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia ti on or on e of it s al li ed pu bl is he rs . T hi s ar ti cl e is in te nd ed so le ly fo r th e pe rs on al us e of th e in di vi du al us er an d is no t to be di ss em in at ed br oa dl y. 326 UCHINO, KENT DE GREY, AND CRONAN http://dx.doi.org/10.1007/BF02879910 http://dx.doi.org/10.1037/0278-6133.22.4.388 http://dx.doi.org/10.1037/0278-6133.22.4.388 http://dx.doi.org/10.1111/j.1469-8986.1992.tb02006.x http://dx.doi.org/10.1037/a0032200 http://dx.doi.org/10.1037/0022-3514.90.1.21 http://dx.doi.org/10.1037/0022-3514.90.1.21 http://dx.doi.org/10.1111/j.1540-4560.2007.00536.x http://dx.doi.org/10.1007/s10865-010-9270-z http://dx.doi.org/10.1037/0033-2909.110.1.67 http://doi.org/10.1093/geronb/gbp127 http://dx.doi.org/10.1037/a0033515 http://dx.doi.org/10.1037/0882-7974.20.1.134 http://dx.doi.org/10.1037/0882-7974.20.1.134 http://dx.doi.org/10.1023/A:1010634902498 http://dx.doi.org/10.1023/A:1010634902498 http://dx.doi.org/10.1177/0956797613520015 http://dx.doi.org/10.1177/0956797613520015 http://dx.doi.org/10.1016/j.jaging.2008.10.006 http://dx.doi.org/10.1016/j.jaging.2008.10.006 The Quality of Social Networks Predicts Age-Related Changes in Cardiovascular Reactivity to Stress Method Participants Procedure Assessments Social Relationships Index (SRI) Cardiovascular measures Statistical Analyses Results Sample Characteristics Main Analyses Discussion References Journals/journal 8 BRIEF REPORTS Aging and Social Satisfaction: Offsetting Positive and Negative Effects William von Hippel University of Queensland Julie D. Henry and Diana Matovic University of New South Wales Social satisfaction in late adulthood originates from competing sources. Older adults tend to be more positive and less negative than younger adults, but social contact and working memory often decrease with age, both of which might limit older adults’ social functioning. In the current study of younger and older adults, these socially facilitative vs. socially debilitative changes were found to underlie stasis in social satisfaction. These findings show that the lack of an overall effect for age can mask competing changes in social functioning in late adulthood, as the sources of social satisfaction might change even if the outcome does not. Keywords: social satisfaction, social activities, working memory, hassles, uplifts Social satisfaction in late adulthood originates from competing sources. On the one hand, life experiences and the wisdom that accompanies them can lead to more harmonious social relation- ships. Thus, socioemotional models of aging propose that social functioning in late adulthood does not follow the same course of decline as cognitive and biological aging (Antonucci, 2001; Charles & Carstensen, 2007). Evidence also suggests that older adults are biased toward positive information (Carstensen & Mikels, 2005; Gross et al., 1997; but see Grühn, Smith, & Baltes, 2005). This bias might reflect motivational shifts driven by time perspective that lead to prioritization of emotion-related goals (Charles & Carstensen, 2007). This bias might also be a mecha- nism that explains older adults’ generally more positive approach to conflict resolution (Blanchard-Fields, 2007; Carstensen, Gott- man, & Levenson, 1995) as well as their reports of fewer unde- sirable daily events (Almeida & Horn, 2004) and less interpersonal tension (Birditt, Fingerman, & Almeida, 2005). Competing with these facilitative effects, however, are other changes in late adulthood that are likely to have a detrimental impact on social functioning and on consequent social satisfaction. First and foremost, although older adults appear to intentionally cull their peripheral social partners to focus on closer and more meaningful relationships (Charles & Carstensen, 2007; Lansford, Sherman, & Antonucci, 1998), they also suffer unintended social losses brought about by poor health, retirement, mobility con- straints, widowhood, and so on, that can lead to reduced social satisfaction (Jang, Mortimer, Haley, & Borenstein Graves, 2004; Pinquart & Sörensen, 2001; van Tilburg, 1998). Thus, although intentional reductions in social network size are theorized not to cause decreased social satisfaction, unintended reductions in social contact have the potential to do so. Compounding the effects of reduced social contact, cognitive losses associated with normal adult aging might also affect older adults’ social functioning by limiting their ability to negotiate complex social relationships. Indeed, social reasoning skills im- pose substantial demands on various aspects of cognitive function- ing, such as mental flexibility and inhibitory control (German & Hehman, 2006), and there is considerable evidence that links cognitive decline to reduced social functioning in older adulthood. For example, cognitive deficits have been linked to socially insen- sitive behaviors in the context of normal adult aging, such as increased difficulty in taking another’s perspective (Bailey & Henry, in press), off-target and verbose speech (Pushkar et al., 2000), and prejudicial and other socially inappropriate comments (von Hippel & Dunlop, 2005; von Hippel, Silver, & Lynch, 2000). Whether older adults show increased or decreased social satis- faction, or if indeed they show any change at all, depends on the relative magnitude of these socially facilitating versus socially debilitating changes. This lack of reliable change might mask important and competing changes in social functioning in late adulthood, as the sources of social satisfaction might change even if the outcome does not (cf. Kunzmann, Little, & Smith, 2000). The goal of this article is to test this possibility. In service of this goal, we examined several social and cognitive factors that might account for competing changes in social satis- faction in later life. To assess social factors that might impact social satisfaction, we asked older and younger adults to indicate how much time they spend alone and how frequently they engage in various social activities. To assess cognitive factors that might impact social satisfaction, we asked older and younger adults to complete measures of working memory and inhibitory control. Based on previous research, we expected older adults to spend more time alone, to engage in fewer social activities, and to have decreased working memory capacity, and we expected these age differences to have a negative impact on social satisfaction. We also expected older adults to show increased cognitive disinhibi- tion compared to younger adults, but it was unclear whether this William von Hippel, School of Psychology, University of Queensland, St. Lucia, Australia; Julie D. Henry and Diana Matovic, School of Psy- chology, University of New South Wales, Sydney, Australia. This research was supported by Australian Research Council Grant DP0774268. Correspondence concerning this article should be addressed to William von Hippel, School of Psychology, University of Queensland, St Lucia QLD 4072, Australia. E-mail: billvh@psy.uq.edu.au Psychology and Aging Copyright 2008 by the American Psychological Association 2008, Vol. 23, No. 2, 435– 439 0882-7974/08/$12.00 DOI: 10.1037/0882-7974.23.2.435 435 T hi s do cu m en t i s co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . T hi s ar tic le is in te nd ed s ol el y fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . effect would impact social satisfaction. On one hand, poorer cognitive inhibition can be associated with social disinhibition (see Pushkar et al., 2000; von Hippel & Dunlop, 2005; von Hippel & Gonsalkorale, 2005; von Hippel et al., 2000) and thus might predict reduced social satisfaction if social disinhibition is causing relationship problems for older adults. On the other hand, inhibitory control is unlikely to play as broad of a role in social functioning as does working memory (see Feldman Barrett, Tugade, & Engle, 2004) and by itself may not be a significant predictor of social satisfaction. On the basis of the age-related positivity effect (Carstensen & Mikels, 2005), however, we expected older adults to experience their social events as more positive and less negative than younger adults. As a consequence, once the suppressing effects of time alone, reduced social activity levels, and reduced working memory were controlled, we expected this increased positivity to lead older adults to experience greater social satisfaction than younger adults. Thus, we expected older adults who did not spend a great deal of time alone, who maintained high social activity levels, and who retained good working memory to experience greater social satis- faction than younger adults. Furthermore, if increased emphasis on the positive (or decreased emphasis on the negative; Grühn et al., 2005) leads to enhanced social satisfaction, then older adults should experience more uplifts and fewer hassles from their daily activities and relationships than should younger adults. This effect, in turn, should mediate the increased residual social satisfaction that emerges with age after controlling for time alone, decreased social activity levels, and reduced working memory. Method Participants Thirty-eight younger adults aged 18 –30 years (M � 23.8, SD � 3.2; 26 female) and 40 older adults aged 66 –91 years (M � 74.4, SD � 7.5; 25 female) participated in this study. Participants were paid $20 Australian (�US$16). Older and younger adults had similar levels of education, with average responses from both groups that ranged between some university study and completed university degree, �2(7, N � 78) � 7.23, p � .40. Participants were community-dwelling, and older and younger adults were recruited in a similar manner from social clubs, apartment com- plexes, and churches in the Sydney metropolitan area. All older adults had normal mental status as indexed by the Mini-Mental State Exam (MMSE; range � 27–30, M � 29.40, SD � 0.87; Folstein, Folstein, & McHugh, 1975). Procedure Participants first completed a brief demographic form on which they indicated their age, gender, and education level, and then they completed several measures unrelated to the current research. Next, because measures of working memory can be fatiguing and sometimes disheartening for older participants and thus these measures have the potential to influence subsequent measures, participants completed either a working memory measure followed by a Hassles and Uplifts Scale (DeLongis, Folkman, & Lazarus, 1988) followed by a Stroop test or a Stroop test followed by a Hassles and Uplifts Scale followed by a working memory measure. The working memory measure was adopted from Daneman and Carpenter (1980). The measure required participants to read a series of sentences aloud and then to recall the last word of each sentence. Participants began by reading three sets of two sentences each, followed by three sets of three sentences, and they worked their way up to three sets of five sentences. Working memory was scored as the total number of final words that participants recalled, summed across the series of different lengths. A shortened form of the Hassles and Uplifts Scale was used to assess the degree to which daily experiences and relationships were experienced as hassles and uplifts. Participants rated a series of 43 items (e.g., family-related obligations, friends, and recreation) on the degree to which the items were experienced as hassles and uplifts over the last week, with responses to both questions provided on 4-point scales ranging from not at all (0) to a great deal (3). Participants were also given the option to indicate that an item was not applicable. For the Stroop test, participants first named aloud the ink colors of 42 color blocks and then named aloud the ink colors of 42 color words whose letters were printed in a color inconsistent with most of the word meanings (e.g., red printed in green ink). The reading times for the page of 42 color blocks and the page of 42 color words were measured with a stopwatch, and Stroop Color–Word Interference scores were computed as the difference in time taken to read color words and color blocks divided by the time taken to read color blocks. Next, participants were asked to indicate how many hours each day they spend alone. They then completed a slightly modified version of the Prosocial subscale of the Social Functioning Scale (SFS–P; Birchwood, Smith, Cochrane, Wetton, & Copestake, 1990). The standard version of this measure assesses level of participation in 22 different social activities (e.g., being visited by/visiting relatives, eating out), but in the present study a 23rd activity (volunteering) was added, as this was considered to be a social activity in which older adults were particularly likely to be engaged. Participants were asked to provide an estimate of the frequency with which they had engaged in each activity in the last month. The SFS–P assesses involvement in activities that one generally does in the company of others (e.g., being visited by or visiting relatives, eating out), as opposed to activities one may be more likely to do alone (e.g., reading, gardening, or knitting). Responses to the 23 items were averaged to form a total score, with higher scores indicative of more active social functioning—the range in the current sample was 0.23 to 12.52. After participants completed the SFS–P, they responded to the single item “How satisfied are you with your social life?” on a scale that ranged from 0 (very dissatisfied) to 10 (very satisfied). Results The first step in the analyses was to examine age differences in the primary measures to assess whether the expected age differ- ences emerged. Consistent with prior research, a significant mul- tivariate analysis of variance, F(5, 70) � 15.72, p � .001, followed up by univariate analyses of variance revealed that older adults spent more hours per day alone (M � 7.04, SD � 4.55) than did younger adults (M � 3.36, SD � 2.30), F(1, 75) � 19.95, MSE � 13.09, p � .001. The scores of older adults on the SFS–P, � � .78, indicated that they engaged in fewer social activities during the prior month (M � 1.99, SD � 1.42) than did younger adults (M � 4.04, SD � 2.46), F(1, 76) � 20.50, MSE � 3.99, p � .001. Older adults showed greater Stroop interference (M � 1.17, SD � 0.45) than did younger adults (M � 0.82, SD � 0.30), F(1, 75) � 15.53, 436 BRIEF REPORTS T hi s do cu m en t i s co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . T hi s ar tic le is in te nd ed s ol el y fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . MSE � 0.15, p � .001, and older adults recalled fewer of the words on the working memory task (M � 27.96, SD � 6.53) than did younger adults (M � 33.32, SD � 4.19), F(1, 76) � 18.39, MSE � 30.42, p � .001. Despite these age-related deficits, older adults showed high levels of social satisfaction (M � 7.85, SD � 1.99) that were comparable to those of younger adults, (M � 7.78, SD � 1.98), F(1, 76) � 0.02, MSE � 3.93, p � .85. The Hassles and Uplifts Scale was then examined to see if younger and older adults responded to the same items with the same frequencies. This analysis revealed that some of the items were differentially likely to be chosen as not applicable by older or younger adults (e.g., items that referred to children were not applicable to most younger adults, whereas items that referred to parents were not applicable to most older adults). To ensure that the analyses were not unduly influenced by age differences in the likelihood of a participant’s responses to particular items, we created subscales that included only the 23 items that older and younger adults were equally—and greater than 75%—likely to respond to the item, � � .70 for hassles, � � .82 for uplifts. Analyses of these subscales revealed that older adults reported fewer hassles (M � 0.55, SD � 0.38) than did younger adults (M � 0.88, SD � 0.41), F(1, 76) � 13.63, MSE � 0.16, p � .001; but older adults reported no differences in uplifts (M � 1.54, SD � 0.43) compared to younger adults (M � 1.46, SD � 0.49), F(1, 76) � 0.64, MSE � 0.21, p � .40. It should be noted that analyses of the remaining 20 items for which there were differential re- sponses by age group revealed that older adults again reported fewer hassles, (M � 0.29, SD � 0.33), than did younger adults (M � 0.92, SD � 0.51), F(1, 76) � 42.21, MSE � 0.18, p � .001, and in this case greater uplifts, (M � 1.70, SD � 0.56) than younger adults (M � 1.24, SD � 0.53), F(1, 76) � 14.25, MSE � 0.30, p � .001. The analyses thus far had suggested that older adults were neither more nor less satisfied with their social relationships than were younger adults, but as noted above, the absence of an age effect might mask underlying competing effects. That is, a sup- pression effect might have emerged whereby age led to positive effects on social satisfaction via one mechanism and to negative effects on social satisfaction via a different mechanism. To test this possibility, we estimated a regression-based causal model follow- ing the procedures outlined in Baron and Kenny (1986). The correlation coefficients for all of the variables in the model, and also for education and MMSE, are presented in Table 1. In the first step of the regression analysis, social satisfaction was regressed on age and gender (with male coded as 0 and female coded as 1) because social satisfaction correlated with gender (see Table 1).1 As shown in Figure 1, and consistent with the analysis of variance results presented above, age had no direct effect on social satisfaction. In the second step of the analysis, social satis- faction was regressed on age, time spent alone, social activity levels, working memory, and Stroop performance. After control- ling for these four factors, age exerted a substantial and positive direct effect on social satisfaction (see Figure 1). In the third step of the analysis, self-reported hassles and uplifts were included as additional mediators between age and social satisfaction (i.e., social satisfaction was regressed on age, time spent alone, social activity levels, working memory, Stroop performance, and self- reported hassles and uplifts). This analysis revealed that self- reported uplifts but not hassles partially mediated the residual effect of age on social satisfaction. In the final step of the analysis, social activity levels were included as a mediator between age and self-reported hassles and uplifts (i.e., hassles and uplifts were regressed on age and social activity levels). This analysis revealed the presence of an additional suppression effect, whereby the direct effect of age on self-reported uplifts was not significant until social activity levels were included in the analysis. Thus, it seems that decreased social activity levels with age also accounted for the fact that no age differences emerged in self-reported uplifts. Older adults who did not experience reduced social activity levels re- ported more uplifts than younger adults (see Figure 1). To determine whether these mediational pathways were signif- icant, we computed indirect effects from unstandardized regression weights with 10,000 bootstrap resamples to obtain accurate con- fidence limits by following the syntax provided by Preacher and Hayes (2007). Three sets of indirect effects and associated confi- dence intervals were computed in analyses that controlled for the effect of gender on social satisfaction. First, analyses revealed that increased time alone, decreased social activity levels, and de- 1 The results are essentially unchanged when gender is not included in the analyses. Table 1 Correlations Between Demographic Variables and Indices of Psychosocial Functioning Variable 1 2 3 4 5 6 7 8 9 10 11 1. Age — 2. Time alone .43*** — 3. SFS–P �.43*** �.39*** — 4. Working memory �.49*** �.28* .05 — 5. Stroop .47*** .19 �.09 �.43*** — 6. Uplifts .10 .07 .41*** �.16 .00 — 7. Hassles �.42*** �.25* .33** .23* �.26* .17 — 8. MMSE (n � 40) �.20 �.19 �.01 .55*** �.02 �.26 .23 — 9. Gender �.12 �.02 �.03 .03 �.16 .07 .10 .13 — 10. Education .15 .00 .05 .17 �.14 .05 .08 .09 �.14 — 11. Social satisfaction .06 �.28* .26* .11 �.01 .40*** .09 .15 .30** �.01 — Note. Correlations of gender and education were calculated with Spearman’s rho. All other correlations were calculated with Pearson’s r. SFS–P � Social Functioning Scale—Prosocial; Stroop � Stroop Color–Word Interference Test; MMSE � Mini-Mental State Exam. * p � .05. ** p � .01. *** p � .001. 437BRIEF REPORTS T hi s do cu m en t i s co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . T hi s ar tic le is in te nd ed s ol el y fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . creased working memory all significantly mediated the negative effect of aging on social satisfaction (indirect effect via time alone � �.008, SE � .004, 95% CI � �.018, �.001; indirect effect via social activity levels � �.011, SE � .004, 95% CI � �.022, �.004; indirect effect via working memory � �.009, SE � .005, 95% CI � �.022, �.001). Consistent with the lack of a relationship between the Stroop and social satisfaction, Stroop performance was not a significant mediator in this analysis (indi- rect effect � �.001, SE � .004, 95% CI � �.009, .008). Second, analyses revealed that when self-reported uplifts and hassles were added to the model, uplifts significantly mediated the residual positive effect of aging on social satisfaction (indirect effect � .010, SE � .006, 95% CI � .001, .023), but hassles did not (indirect effect � .000, SE � .002, 95% CI � �.003, .006). Third, analyses revealed that uplifts also significantly mediated the pos- itive effect of social activity levels on social satisfaction (indirect effect � .117, SE � .060, 95% CI � .016, .251). Finally, when these mediational analyses were conducted with the full Hassles and Uplifts Scale or with only those items in the Hassles and Uplifts Scale that were explicitly social in nature, the results were equivalent to those depicted in Figure 1. Discussion The results of this study provide a picture of aging in which social costs are offset by social gains. A community-based sample of older adults showed neither greater nor lesser social satisfaction than a community-based sample of younger adults. Additionally, older adults in this sample reported neither greater nor fewer uplifts from the experiences that they had in common with younger adults. Yet this apparent stability in social experience masked underlying counter-currents whereby age-related losses suppressed the effect of age-related gains. On the loss ledger, older adults spent more time alone, engaged in fewer social activities, and had poorer working memory than did younger adults. All three of these factors played a mediating role in decreased social satisfaction among older adults. When these factors were included in the model, aging was associated with residual increases in social satisfaction and self- reported uplifts. Furthermore, the residual increase in social satisfac- tion was itself partially mediated by the residual increase in uplifts. That is, once age-related decreases in social activity levels were controlled, older adults reported greater social satisfaction than did younger adults, which was partially accounted for by the degree to which older adults also reported increased uplifts. Results also revealed that older adults considered their daily activities to be less of a hassle than did younger adults. This age effect on self-reported hassles appears to be of lesser importance than the effect on uplifts, however, as uplifts and not hassles played a mediating role in increased residual social satisfaction among older adults. Thus, it seems to be the case that increased positivity but not decreased negativity played a role in maintaining social satisfaction in late adulthood. The fact that age had a significant residual effect on social satisfaction after self-reported hassles and uplifts had been accounted for, however, suggests that other age-related changes also lead to increased social satisfaction. One possibility is that other aspects of increased positivity and decreased negativity are important determinants of increased so- cial satisfaction with age, as it is highly unlikely that hassles and uplifts capture the full experience of increased positivity and decreased negativity with age. Additionally, a variety of other factors might be at play in the residual increase in social satisfac- tion that emerges in late adulthood. For example, increased social satisfaction might derive from changes in anger management strat- egies (Phillips, Henry, Hosie, & Milne, 2006, in press) or perhaps from older adults’ tendency to make increased use of reappraisal to regulate their emotions (Gross et al., 1997; John & Gross, 2004). Future research might attempt to gain a more complete picture of how all of these factors contribute simultaneously to change and stability in social satisfaction across the lifespan. The fact that the current study is cross-sectional and reliant on self-report limits the causal conclusions that can be drawn. Clearly, longitudinal work with a larger sample and multiple behavioral or physiological indicators would represent an important supplement Age Time Alone Social Activities Working Memory Satisfac w/ Social Life -.32** .25* .43** -.42** -.49*** .48** .35** Gender .27** .09 Panel A Age Time Alone Social Activities Working Memory Hassles Uplifts Satisfac w/ Social Life.35* -.32** .16 .31* .25* -.03 .43** -.42** -.49*** .34** .55*** -.42*** .10 Gender .27** Panel B Figure 1. Mediational models of the effect of age on time spent alone, social activity levels, working memory, hassles, uplifts, and social satis- faction are depicted. Path coefficients are standardized beta weights. In Panel A, two direct effects of age on social satisfaction are included in the model. The coefficient above the path is from a model that did not include any mediators. The coefficient below the path is from a model that included time spent alone, social activity levels, working memory, and Stroop performance as mediators. Stroop performance is not depicted because it was not a significant mediator. In Panel B, two direct effects are included in the model for the impact of age on uplifts. The coefficient above the path is from a model that did not include social activity levels as a mediator. The coefficient below the path is from a model that included social activity as a mediator. The nonsignificant path from social activity levels to hassles is not depicted for the sake of expositional clarity. Satisfac w/ � satisfaction with. * p � .05. ** p � .01. *** p � .001. 438 BRIEF REPORTS T hi s do cu m en t i s co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . T hi s ar tic le is in te nd ed s ol el y fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . to the current findings. Additionally, it should be noted that the order of the questions might have increased the impact of time spent alone and social activity levels on social satisfaction. That is, because these items were assessed directly before social satisfac- tion was measured, their momentary accessibility might have increased their impact on this overall judgment (Strack, Martin, & Schwarz, 1988). Although future research should counterbalance these questions to ensure that time spent alone and social activity levels predict social satisfaction independent of question order, it should be noted that prior research has found that time spent alone and social activity levels are related to various indices of well- being that include social and life satisfaction (Hawkley et al., 2008; Iecovich et al., 2004; Pinquart & Sörensen, 2001, 2003; Routasalo, Savikko, Tilvis, Strandberg, & Pitkälä, 2006). References Almeida, D. M., & Horn, M. C. (2004). Is daily life more stressful during middle adulthood? In C. D. Ryff & R. C. Kessler (Eds.), A portrait of midlife in the United States (pp. 425– 451). Chicago: University of Chicago Press. Antonucci, T. C. (2001). Social relations: An examination of social net- works, social support, and sense of control. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology of aging (5th ed., pp. 427– 453). San Diego, CA: Academic Press. Bailey, P. E. & Henry, J. D. (in press). Growing less empathic with age: Disinhibition of the self-perspective. Journal of Gerontology: Psycho- logical Sciences. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Birchwood, M., Smith, J., Cochrane, R., Wetton, S., & Copestake, S. (1990). The Social Functioning Scale: The development and validation of a new scale. British Journal of Psychiatry, 157, 853– 859. Birditt, K. S., Fingerman, K. L., & Almeida, D. M. (2005). Age differences in exposure and reactions to interpersonal tensions: A daily diary study. Psychology and Aging, 20, 330 –340. Blanchard-Fields, F. (2007). Everyday problem solving and emotion: An adult developmental perspective. Current Directions in Psychological Science, 16, 26 –31. Carstensen, L. L., Gottman, J. M., & Levenson, R. W. (1995). Emotional behavior in long-term marriage. Psychology and Aging, 10, 140 –149. Carstensen, L. L., & Mikels, J. A. (2005). At the intersection of emotion and cognition: Aging and the positivity effect. Current Directions in Psychological Science, 14, 117–121. Charles, S. T., & Carstensen, L. L. (2007). Emotion regulation and aging. In J. J. Gross (Ed.), Handbook of emotion regulation (pp. 307–327). New York: Guilford Press. Daneman, M., & Carpenter, P. A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19, 450 – 466. DeLongis, A., Folkman, S., & Lazarus, R. S. (1988). The impact of daily stress on health and mood: Psychological and social resources as medi- ators. Journal of Personality and Social Psychology, 54, 486 – 495. Feldman Barrett, L., Tugade, M. M., & Engle, R. W. (2004). Individual differences in working memory capacity and dual-process theories of mind. Psychological Bulletin, 130, 553–573. Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). Mini-Mental State: A practical method for grading cognitive state of patients for the clinician. Journal of Psychiatry Research, 12, 189 –198. German, T. P., & Hehman, J. A. (2006). Representational and executive selection resources in ‘theory of mind’: Evidence from compromised belief-desire reasoning in old age. Cognition, 101, 129 –152. Gross, J. J., Carstensen, L. L., Pasupathi, M., Tsai, J., Skorpen, C. G., & Hsu, A. Y. C. (1997). Emotion and aging: Experience, expression, and control. Psychology and Aging, 12, 590 –599. Grühn, D., Smith, J., & Baltes, P. B. (2005). No aging bias favoring memory for positive material: Evidence from a heterogeneity– homogeneity list paradigm using emotionally toned words. Psychology and Aging, 20, 579 –588. Hawkley, L. C., Hughes, M. E., Waite, L. J., Masi, C. M., Thisted, R. A., & Cacioppo, J. T. (2008). Who are the lonely people II: The Chicago health, aging, and social relations study. Manuscript submitted for publication. Iecovich, E., Barasch, M., Mirsky, J., Kaufman, R., Avgar, A., & Kol- Fogelson, A. (2004). Social support networks and loneliness among elderly Jews in Russia and Ukraine. Journal of Marriage and Family, 66, 306 –317. Jang, Y., Mortimer, J. A., Haley, W. E., & Borenstein Graves, A. R. (2004). The role of social engagement in life satisfaction: Its significance among older individuals with disease and disability. Journal of Applied Gerontology, 23, 266 –278. John, O. P., & Gross, J. J. (2004). Healthy and unhealthy emotion regu- lation: Personality processes, individual differences, and life span de- velopment. Journal of Personality, 72, 1301–1334. Kunzmann, U., Little, T. D., & Smith, J. (2000). Is age-related stability of subjective well-being a paradox? Cross-sectional and longitudinal evi- dence from the Berlin aging study. Psychology and Aging, 15, 511–526. Lansford, J. E., Sherman, A. M., & Antonucci, T. C. (1998). Satisfaction with social networks: An examination of socioemotional selectivity theory across cohorts. Psychology and Aging, 13, 544 –552. Phillips, L. H., Henry, J. D., Hosie, J., & Milne, A. B. (2006). Age, anger regulation and well-being. Aging & Mental Health, 10, 250 –256. Phillips, L. H., Henry, J. D., Hosie, J. A., & Milne, A. B. (in press). Adult aging and the experience, expression and regulation of negative affect. Journal of Gerontology: Psychological Sciences. Pinquart, M., & Sörensen, S. (2001). Influences of loneliness in older adults: A meta-analysis. Basic and Applied Social Psychology, 23, 245–266. Pinquart, M., & Sörensen, S. (2003). Risk factors for loneliness in adulthood and old age: A meta-analysis. In S. P. Shohov (Ed.), Advances in Psychol- ogy Research, Volume 19 (pp. 111–143). Hauppauge, NY: Nova Science. Preacher, K. J., & Hayes, A. F. (2007). Asymptotic and resampling strategies for assessing and comparing indirect effects in simple and multiple mediator models. Manuscript under review. Pushkar, D., Basevitz, P., Arbuckle, T., Nohara-LeClair, M., Lapidus, S., & Peled, M. (2000). Social behavior and off-target verbosity in elderly people. Psychology and Aging, 15, 361–374. Routasalo, P. E., Savikko, N., Tilvis, R. S., Strandberg, T. E., & Pitkälä, K. H. (2006). Social contacts and their relationship to loneliness among aged people: A population-based study. Gerontology, 52, 181–187. Strack, F., Martin, L. L., & Schwarz, N. (1988). Priming and communi- cation: Social determinants of information use in judgments of life satisfaction. European Journal of Social Psychology, 18, 429 – 442. van Tilburg, T. (1998, November). Losing and gaining in old age: Changes in personal network size and social support in a four-year longitudinal study. Journal of Gerontology: Series B: Psychological Sciences and Social Sciences, 53B, S313–S323. von Hippel, W., & Dunlop, S. M. (2005). Aging, inhibition, and social inappropriateness. Psychology and Aging, 20, 519 –523. von Hippel, W., & Gonsalkorale, K. (2005). “That is bloody revolting!” Inhibitory control of thoughts better left unsaid. Psychological Science, 16, 497–500. von Hippel, W., Silver, L. A., & Lynch, M. E. (2000). Stereotyping against your will: The role of inhibitory ability in stereotyping and prejudice among the elderly. Personality and Social Psychology Bulletin, 26, 523–532. Received June 20, 2007 Revision received January 11, 2008 Accepted January 21, 2008 � 439BRIEF REPORTS T hi s do cu m en t i s co py ri gh te d by th e A m er ic an P sy ch ol og ic al A ss oc ia tio n or o ne o f i ts a lli ed p ub lis he rs . T hi s ar tic le is in te nd ed s ol el y fo r t he p er so na l u se o f t he in di vi du al u se r a nd is n ot to b e di ss em in at ed b ro ad ly . Journals/journal 9 Psychology and Aging I W T . V o f . r " - '12. No. 3. 433-443 Copyright 1997 by the Americ rt Psychological Association, Inc. 0882-7974/97/S3.0C Everyday Functioning and Successful Aging: The Impact of Resources Margret M. Baltes and Frieder R. Lang Free University Berlin The goal of this article is to examine differential aging in everyday functioning between resource- rich and resource-poor older adults. Four groups of older adults were identified on the basis of 2 distinct resource factors: a Sensorimotor-Cognitive factor and a Social-Personality factor. The resource-richest group consisted of those participants who were above the median in both factors; those falling below the median in both factors comprised the resource-poorest group; and 2 additional groups consisted of older adults who were above the median in either 1 of the 2 factors. At the level of mean differences, the 4 groups differed in the length of the waking day, the.variability in activities, the frequency of intellectual-cultural and social-relational activities, and resting times. Considering age differences there are more and larger negative age effects in the resource-poorest group than in the resource-richest one. The metamodel of selective optimization with compensation is used to interpret the findings. Effective functioning in everyday life is a developmental task of old age, yielding autonomy and independent living. Losses in different life domains threaten everyday functioning and, thus, require adaptation. Effective mastery of this developmental task is an integral part of successful aging (M. Baltes & Carstensen, 1996; P. Baltes & M. Baltes, 1990; Marsiske, Lang, Baltes, & Baltes, 1995). How well individuals can adapt to functional loss depends on the availability of resources in the sensorimotor, cognitive, personality, and social domains of functioning. This is not to say that individuals who are rich in resources will not experience functional loss at all, but they will begin with loss experiences at higher levels of functioning (see Lindenberger & The present study is part of the Berlin Aging Study (BASE), which has been conducted by the interdisciplinary working group Aging and Societal Development (AGE) in cooperation with institutes and research units of the Free University Berlin, the Humboldt University of Berlin, and the Max Planck Institute for Human Development and Education in Berlin, Germany. AGE was constituted in 1987 by the former Academy of Sciences of Berlin (West) and has been continued since 1994 by the Berlin-Brandenburg Academy of Sciences (BBAW). BASE was finan- cially supported between 1989 and 1991 by the German Ministry of Research and Technology (No. TA O i l + 13 TA O i l / A ) ; since 1992 financial support has been received from the German Ministry of Family, Seniors, Women, and Youth. Primary responsibility for BASE is shared by Paul B. Baltes, Karl Ulrich Mayer (Max Planck Institute for Human Development and Education), Hanfried Helmchen (Free University Berlin), and Elisabeth Steinhagen-Thiessen (Humboldt University Ber- lin). This steering committee is joined by a central project coordinator, R. Nuthmann (Max Planck Institute for Human Development and Education). We thank Fredda Blanchard-Fields, Chris Hertzog, Ulman Lindenber- ger, Heiner Meier, and Michael Marsiske for their valuable comments on a version of this article. Correspondence concerning this article should be addressed to either Margret M. Baltes or Frieder R. Lang, Research Unit Psychological Gerontology, Department of Gerontopsychiatry, Free University Berlin, Nussbaumallee 38, 14050 Berlin, Germany. Electronic mail may be sent via Internet to baltesma@zedat.fu-berlin.de. Baltes, 1997). In other words, the aging process might reflect the same slope of differences or decline in functioning, but functional deterioration will manifest itself later because still- available resources may be used to compensate and optimize with or select from in the face of real or anticipated losses. Consequently, negative aging effects might be less pronounced in individuals who are rich in resources. In summary, we argue that sensorimotor, cognitive, personality, and social resources have imminent importance for successful aging because they facilitate the interplay between three adaptive processes: selec- tion, compensation, and optimization; the more resources a per- son has, the easier it is to anticipate, confront, and adapt to aging losses. Typically, in gerontology, aging theories have put forth differ- ent criteria or outcomes that characterize and indicate successful aging (M. Baltes & Carstensen, 1996; P. Baltes & M. Baltes, 1990). For instance, subjective well-being or life satisfaction has played a key role as an indicator of successful aging (e.g., Rowe & Kahn, 1987). This criterion for evaluating modes of aging well was most notably applied in the disengagement- activity debate. The disengagement theory (Gumming & Henry, 1961) posits a gradual and mutual withdrawal between the indi- vidual and society in old age that represents a symbolic prepara- tion for death. This model views success as acceptance of and reconciliation with the loss of power endemic in old age. Some- what similarly, Buhler's (1933) five-stage model views success- ful aging as the realization and acceptance of age-related physi- cal limitations and the reconciliation of conflicting goals. In contrast, activity theory (Havighurst & Albrecht, 1953; Maddox, 1963) argues that social withdrawal in old age is not natural, but that it is imposed by traditional societal practices such as retirement and intractable age-related problems, like poor health. According to activity theory, successful aging demands the maintenance of activity, replacement of lost roles with new ones, and involvement in society and interpersonal relationships. The more recent continuity theory (Atchley, 1989) suggests that a continuation of previous life patterns best ensures adaptation and success in late life. 433 434 BALTES AND LANG Another set of criteria has sprung from the emphasis on per- sonality growth. In this sense, for Jung (1931,1934), successful aging involves life review, psychological interiorization, and expansion beyond gender constraints toward full humanity and wisdom. Erikson's stage model (1959, 1968, 1982; Erikson, Erikson, & Kivnick, 1986), arguably the most popular model of adult personality development, suggests that the task of old age is acceptance and resolution of one's life. Described as a crisis between ego integrity and despair, the challenge in old age is to accept oneself and one's deeds, both good and bad, in order to obtain psychological peace and acceptance of death. Theoretical models of positive mental health have also contrib- uted to the discussion of successful aging criteria. For example, models of self-actualization (Maslow, 1968), the healthy per- sonality (Jahoda, 1958), and the trusting, meaningful life (Rog- ers, 1961) all tacitly suggest that successful aging is the realiza- tion of one's full human potential. In this context of personality and positive mental health models, the more recent work by Ryff (1982, 1989a, 1989b) is perhaps the most ambitious. Ryff has proposed an integrative model of successful aging that is based on developmental, clinical, and mental health criteria. Her model includes six dimensions of positive functioning: self- acceptance, positive relations with others, autonomy, environ- mental mastery, purpose in life, and personal growth. In summary, when considering these criteria of successful aging, most require a high level of functioning that in turn requires the availability of resources in the biological, psycho- logical, and the social domains of functioning. In fact, although not a criterion in the above-indicated models, cognitive function- ing as an important general-resource factor supporting success- ful aging has been an argument in life span psychology. Specifi- cally, recent data point to the importance of sensorimotor func- tioning as a resource (Lindenberger & Baltes, 1994). In conclusion, both objective and subjective resources help main- tain and optimize functioning and should cover as broad a con- tent base as possible. In this sense, the purpose of the present article is to examine the interrelationship of a number of resources necessary for the attainment of high functioning, such as sensorimotor, cognitive, personality, and social resources, their effect on everyday func- tioning as well as their relationship to negative age differences in everyday functioning. Our hypothesis states that participants scoring high on resources will show less age differences in everyday functioning when compared with people who score low. This means that successful aging will manifest itself in the reduction of the magnitude of cross-sectional age differences in everyday functioning. Consequently, the question to be answered in this article is the following: Are age differences in everyday functioning less pronounced in older people who have many resources (score high on a variety of resource indicators) avail- able? To answer this question, first, resource-rich versus re- source-poor older adults are identified on a total of 11 indicators from four domains of resources; namely, sensorimotor, cogni- tive, personality, and social resources. Second, to examine the effects of these resources on everyday functioning, 21 indicators of everyday functioning, which were assessed by the activities reported in a Yesterday Interview (YI) by the older participants, are inspected. Third, we analyze age differences in these re- source effects on everyday functioning. Existing age-differential effects are discussed using the metamodel of selective optimiza- tion with compensation (M. Baltes & Carstensen, 1996; P. B. Baltes, 1993; P. B. Baltes & M. M. Baltes, 1990; Marsiske et al., 1995). Method Participants In the Berlin Aging Study (BASE), 516 community-dwelling and institutionalized West Berlin residents age 70-103 years (M = 84.9 years, SD = 8.7) took part in an intensive 14-session interview schedule, the content of which addressed four disciplines: psychology, sociology, psychiatry, and internal medicine (see P. B. Baltes, Mayer, Helmchen, & Steinhagen-Thiessen, 1993). The sessions were distributed over 4.5 months (SD = 1.9). Participants were identified through probability sampling from the local registration office (in Germany each citizen must he registered) and stratified by age and sex. Twenty-seven percent of those contacted took part in all 14 sessions of BASE. In an analysis of selectivity and representativity of this final BASE sample compared with the original sample of 1,908 participants, Lindenberger, Gilberg, Potter, Little, and Baltes (1996) did not find any substantial evidence for sample biases. However, there was indication that the sample of 516 had a lower mortality rate after 1 year of testing, 5% as compared with 14% in the general population, and all other variables inspected exhibited the expected, albeit nonsignificant, direction in the differences (see Lin- denberger et al., 1996). There was also no indication that covariances and variances of variables in the final sample differed from a sample of participants who have completed an intake assessment of basic vari- ables from all disciplines (i.e., 49% of the parent sample; see Lindenber- ger et al., 1996). Of the 516 participants, 86% lived in the community and 14% lived in long-term care institutions (i.e., sheltered housing, nursing homes, and hospitals for the chronically ill), and 21% were diagnosed with dementia according to Diagnostic and Statistical Manual of Mental Disorders (3rd ed.. rev.; American Psychiatric Association, 1987) crite- ria. Twenty-nine percent of the participants were married, 55% were widowed, and 15% were divorced or have never been married. On the average, participants had 10.8 years of education (SD - 2.4): 54% had 8 to 10 years, 40% had between 10 and 13 years, and 7% had more than 13 years. Measures Everyday functioning. The YI (Moss & Lawton. 1982; see also M. Baltes, Wahl, & Schmid-Furstoss, 1990), an instrument that provides a minute-to-minute reconstruction of the sequence, duration, frequency, the geographical and social context of activities engaged in during the day preceding the interview, was used to assess everyday functioning. The YI was part of the seventh session in the intensive data protocol and was conducted by trained interviewers. The interviewer asks the participant first to recount all activities in the sequence of their occur- rence during the preceding day from waking up to falling asleep. Once activities are recounted, the interviewer starts anew with the first one reported and asks for the duration of each activity mentioned before including where and with whom they occurred. All reported activities were recorded as verbatim as much as possible. Trained coders coded these activities into 44 activity codes (see M. M. Baltes, Maas. Wilms, & Borchelt, 1996). Kappa values of intercoder reliability were above .80 (see M. M. Baltes, Mayr, Borchelt, Maas, & Wilms, 1993). On average, participants reported 28 (SD — 7.3) activities during an average waking day of 968.5 rain (SD = 109.7). In the present article these 44 activity categories were summarized into eight classes of activities differentiating between self-care or basic EVERYDAY FUNCTIONING AND SUCCESSFUL AGING 435 activities, those related to household chores, social engagement, physical leisure, intellectual-cultural leisure, television viewing, and resting and sleeping. Table 1 presents these eight activity classes. Information about the social context was classified along 3 categories into time spent alone, time spent with family members, and time spent with others. Furthermore, we included the variability or variety in activities indicating how many different activity types the older participants engaged in and the length of the waking day as additional indicators of everyday func- tioning. In all, a total of 21 parameters of everyday functioning was examined. There is information about satisfactory reliability measures from Law- ton and colleagues (Moss & Lawton, 1982). The YI appears to be robust; that is, participants with diagnosis of mild dementia are able to reconstruct their previous day at least on the basis efface validity. Only. 31 or 6% of interviews had to be excluded because of too many missing values. Of these 31 people, 27 (87%) had a diagnosis of dementia. This attrition in the sample explains why all subsequent analyses and tables contain a sample of 485 instead of 516. Resources indicators. For biological resources we used sensorimotor indicators instead of more direct health indicators such as medical diag- noses because of two reasons. First, there is a high interrelation between sensorimotor functioning and physical health in BASE (see Steinhagen- Thiessen & Borchelt, 1996). Second, sensorimotor functioning has been shown in BASE to have a significant effect on a number of domains, such as cognitive functioning, well-being, and self-care (Lindenberger & Bakes, 1994; Marsiske, Klumb, & Baltes, 1997). 1. Sensorimotor resources are indicated by three constructs: visual acuity, auditory acuity, and balance-gait. Visual acuity consists of a composite construct of three measurements, (a) Distance visual acuity was assessed binocularly with a reading table placed at least 2.5 m from the participant. Close visual acuity was assessed separately (b) for the left eye and (c) for the right eye with a reading table placed at reading distance. Auditory acuity was assessed with a Bosch ST-20-1 pure-tone audiometer by using headphones at the participant's residence or in the clinic of a university medical school (for a detailed description of the measures for visual acuity and auditory acuity, see Lindenberger & Baltes, 1994). For the assessment of balance-gait, participants were asked to perform a 360° turn as fast as they could without risking a fall. The number of steps needed to complete the circle was counted (for a detailed description of this measure, see M. M. Baltes et al., 1993). For psychosocial resources, cognitive, personality, and social re- sources are included. The cognitive resources included are closely re- lated to the concept of general intelligence (Lindenberger, Mayr, & Kliegl, 1993). As personality resources we used two measures that reflect protective factors (i.e., resources) rather than risk factors. For social resources we used information gathered with the social relation- ship questionnaire in BASE, which is in part based on the circle diagram (Kahn & Antonucci, 1980; Lang, 1996; Lang & Carstensen, 1994). 2. Cognitive resources relate to a measure of reasoning consisting of three tests: Figural Analogies, Letter Series, and Practical Problems (Lindenberger et al., 1993). All tests were assessed within the cognitive battery of the BASE study with a Macintosh SE30 personal computer system equipped with a Micro Touch Systems touch-sensitive screen for stimulus presentation and data collection (for a detailed description of the cognitive battery used in BASE, see Lindenberger et al., 1993). 3. Personality resources relate to the personality dimension of extra- version and goal strength. The personality construct of extraversion con- sists of a 6-item German adaptation (Borkenau & Ostendorf, 1990; see also Smith & Baltes, 1993) of the Extraversion subscale of the NEO- Personality Questionnaire (Costa & McCrae, 1985). Participants rated the degree to which each of the items described themselves on a 5-point scale. Internal consistency was indicated by an alpha coefficient of .64. Goal strength, the second personality resource, represented the average of the perceived investment in 10 goal domains (e.g., health, hobbies, friends, family, and death and dying; Staudinger, Smith, & Freund, 1992). 4. Social resources are indicated with three measures concerning the availability of social and relational resources, (a) Perceived social sup- port in the network is measured as the number of different instances of social support received during the past 3 months (e.g., being cheered up or being supported in times of sorrow; Lang, 1996). (b) Role variety in the social network reflected the availability of different role relation- ships. If an older person mentioned, for instance, the presence of a grandchild in the social network, this was scored as an instance for the role of grandparent. For a detailed description of the social network measure used in the BASE study, see also Lang and Carstensen (1994; Lang, 1996); The third indicator was assessed by the sociologists in BASE and referred to social status, (c) As an indicator of social status, occupational status was assessed with the Wegener Prestige Scale (Weg- ener, 1988). Married and widowed women received scores reflecting the occupational status of their husbands (i.e., if the husband's status was higher than that of the wife). For unmarried and single women, their own occupational status was determined by Iheir own last occupation. Overview of Statistical Procedures To explore mechanisms of late-life adaptivity, such as selection, opti- mization, and compensation, a comparison of age-related differences in everyday functioning among resource-rich and resource-poor adults ultimately requires a longitudinal approach. However, first insights can also be gained from cross-sectional analysis, especially when the focus is on the comparison of age differences in some dependent variable Table 1 Description of Eight Classes of Everyday Activities Activity Description Self-care Household chores/housekeeping Physical leisure Intellectual-cultural leisure Television viewing Social engagement Resting Sleeping Basic activities of daily living (e.g., getting up, toileting, and eating) Instrumental activities of daily living (e.g., shopping, cleaning banking, and getting medical care) Activities that include any kind of physical efforts or demands (e.g., exercising, gardening, walking, and traveling) Activities refering to intellectual efforts or cultural interest (e.g., visiting the museum, painting, visiting the theater, writing, reading, playing, listening to radio or music, and religious activities) Separate leisure category Activities that explicitly engage social partners (e.g., talking to others, visiting others, telephoning, helping others, and political activities) Passive phases during the day (not sleeping) Sleeping during daytime 436 BALTES AND LANG between resource-defined groups of older adults (e.g., Hertzog, 1985). We present four steps of data analysis. In Step 1, four major domains of resources were differentiated. These were sensorimotor, cognitive, personality, and social resources that were assessed with 11 criteria, with 3 criteria in each domain except for only 2 in the personality domain (see p. 435). Unit-weighted composites for each domain of resource indicators were aggregated (i.e., factor scores) as a result of four independent exploratory factor analyses. Table 2 displays the factor loadings, communalities, and the total explained variances for each of the four resource factors. In Step 2, these four domains of resources were analyzed to detect groups of resource-rich and resource-poor older adults. Because these resources showed differential age correlations, an exploratory factor analysis on the four domains of resources was conducted with age par- tialed out. These residualized factors represent resources not directly associated with age. Thus, partialing age allowed us to identify relatively age-heterogeneous groups for further age comparisons concerning every- day functioning. Two factors emerged: a Sensorimotor-Cognitive re- source factor and a Social-Personality resource factor. In Step 3, these two resulting age-residualized resource factors were dichotomized to identify four resource-defined groups of older adults. Given the quality of our measurement variables in terms of validity and reliability, a me- dian-split approach is justified and reduces the complexity of further tests for interaction effects between age and resources, despite the known problems of such procedures (for a discussion of this problem, see, for example. Maxwell & Delaney, 1993; McClelland & Judd, 1993). In Step 4, differential age effects on parameters of everyday functioning between the four groups of older adults were compared by using hierar- chical regression analyses. Results Exploratory Factor Analyses on Resource Indicators First, to examine the relationship between the four resource domains—sensorimotor, cognitive, personality, and social inte- gration—four exploratory factor analyses were conducted on the subsets of the 11 success criteria (3 criteria in each domain save for only 2 in the personality domain). Table 2 shows the factor loadings for each resource factor. As Table 2 also shows, most of the four resource indicators were correlated with age. Sensorimotor resources were highly negatively related to age (r — -.73, p < .001), cognitive resources were moderately negatively related (r = —.51, p < .001), personality resources were only mildly correlated with age (r = -.23, p < .001), and social integration resources did not show any significant age association (r = -.09, ns). Next, to be able to make age comparisons between resource-rich and resource-poor adults, age was partialed from the four resource factors by using age- residualized factor scores. Remember that partialing out age allowed us to extract in a second step relatively age-heteroge- neous factors, thus making further age comparisons between resource-rich versus resource-poor older adults possible. Second, an exploratory factor analysis with oblique rotation was conducted on the age-residualized resource indicators, re- vealing two relatively distinct factors: one capturing the variance of sensorimotor and cognitive resources, and one capturing the variance of the social and personality resources. The factor anal- ysis accounted for 66% of the total variance (see Table 3). The Sensorimotor-Cognitive factor correlated mildly with the Social-Personality factor (r = .22). Third, to identify groups of resource-rich versus resource- poor participants, the Sensorimotor-Cognitive and Social-Per- sonality factor scores were dichotomized by using a median split resulting in four groups of participants: (a) 95 participants were above the median on both resource factors (resource-rich group), (b) 140 participants were above the median in the Sensorimotor-Cognitive resource factor but not in Social-Per- sonality resources, (c) 146 participants were above the median in the Social-Personality resource factor but not in the Sensori- motor-Cognitive factor, and (d) 104 participants were below the median on both resource factors (resource-poor group). Table 4 gives an overview of some of the central descriptors of the four resource-defined groups of older adults. As a consequence of having partialed out age earlier, the four groups did not necessarily differ in respect to age. However, Table 2 Four Exploratory Factor Analyses on Unit-Weighted Resource Indicators Within Four Domains: Factor Loadings, Communalities, Explained Variances, and Age Correlations Factor analyses Factor loading h2 Total explained variance (%) Age correlation Sensorimotor resources Auditory acuity Visual acuity Gait Cognitive resources Figural Analogies test Letter Series tesl Practical Problems test Personality resources Extraversion Goal strength Social resources Role variety of network Received social support Occupational status .78 .80 .78 .86 .88 .86 .80 .80 .76 .75 .53 62 .61 .65 .60 75 .74 .77 .75 63 .64 .64 47 .57 .56 .28 -.73*** -.51*** -.23*** -.09 Note. N = 516. A2 represents factor communality. ***p < .00!. EVERYDAY FUNCTIONING AND SUCCESSFUL AGING 437 Table 3 Results of an Exploratory Factor Analysis With Age- Residualized Resource Factors Factor loadings Resource factors Sensoriraotor Cognitive Social Personality Total explained variance (%) Factor 1: Sensorimotor- Cognitive .80 .81 .27 -.16 4J.1 Factor 2: Personality - Social -.03 .03 .64 .90 24.7 Factor communality (A2) .63 .68 .68 .77 groups differed significantly in respect to marital status, living arrangement, and dementia; no gender difference was found. Compared with all other groups, participants below the median in both factors (resource-poor group) were more likely not to be married, to live in an institution, and to be demented. Note that the 146 participants above the median in the Social -Person- ality factor only were also more likely to be demented than those in the other two groups; namely, the participanls above the median in both factors (resource-rich group) as well as those above the median in the Sensorimotor-Cognitive factor only. Before investigating the effect of Age x Group interaction effects on everyday functioning, correlations between age and the two resource factors within each of the four groups had to be checked. Differential correlations within the groups could account for differential age correlations between the four groups. With two exceptions, no significant correlations be- tween age and the resource factors were found within the four groups. The Sensorimotor-Cognitive factor was positively cor- related with age within the resource-poor group (below the median in both factors; r = .34, p < .01) as well as in the group above the median in the Personality-Social resource factor only (r = .19, p < .01). These correlations were not significant in the two other groups (—.02 within the resource-rich group and —.11 in the group above the median in sensorimotor-cognitive resources only). Description of Everyday Functioning in the Four Resource-Defined Groups Table 5 displays means and standard deviations of the indica- tors of everyday functioning in the four groups of older adults as well as the total explained variances and significance of group comparisons. As can be seen in Table 5, groups differed significantly in respect to the length of the waking day, the variability of activities, the frequencies of intellectual and social activities, and resting or contemplation during the day. Group differences explained only small portions of the total variances with respect to the frequencies of housekeeping activities and physical leisure activities and to the duration of resting phases. Most of these observed group differences were related to differences between the resource-poor group and the other three groups. The resource-poor group reported shorter days, a smaller variety of activities, fewer physical and intellectual- cultural activities, fewer social activities, and more as well as longer resting times during the day. When excluding the re- source-poor group, none of the reported differences, save the one for frequency of social activities, remained significant. The older adults above the median in only sensorimotor-cognitive resources reported also fewer social activities than the resource- rich older adults (3% of total explained variance; p < .01). When contrasting the resource-rich versus the resource-poor group, the first one reported a greater variety of activities— housekeeping, physical, intellectual, and social-relational— more television viewing, and less resting or passive times during Table 4 Descriptors of the Four Resource-Defined Groups Descriptor Age M SD Women (%) Married {%) Widowed (%) Never married-divorced (%) Demented (%) Living in institutions (%) Sensorimotor-cognitive resources T-value mean SD Social-personality resources T-value mean SD Resource rich Sensorimotor-cognitive (H = 95) resources (n = 140) 84.1 9.2 42.1 43.2 46.3 10.5 5.3 2.1 58.6 6.1 56.7 5.7 84.2 8.8 43.6 27.9 55.0 17.1 7.9 8.6 58.8 6.7 42.0 6.8 Social —personality resources (n - 146) 85.0 7.9 54.1 34.2 53.4 12.3 24.0 11.6 41.9 6.6 59.1 6.2 Resource poor (n = 104) F(3, 481) x2<3, N = 485) 84.6 0.3 8.7 0.3 52.9 0.6 21.2 59.6 14.0* 19.2 29.8 34.8*** 23.1 3.4*** 43.4 268.7*** 5.3 43.0 254.0*** 6.4 Note, T values represent a T transformation. *p< .05. ***p < .001. 438 BALTES AND LANG Table 5 Means of Everyday Activity Parameters for the Four Resource-Defined Groups Sensorimotor— cognitive Resource rich Activity parameters Length of waking day (min) Time spent alone (min) Time spent with family (min) Time spent with others (min) Variability of activities (T values) Frequency Self-care Household chores Physical leisure Intellectual-cultural leisure Television viewing Social engagement Resting Sleeping during daytime Weighted duration61 Self-care Household chores Physical leisure Intellectual -cultural leisure Television viewing Social engagement Resting Sleeping during daytime (n M 975 590 169 123 53.1 7.8 5.5 4.0 3.5 1.8 2.0 1.6 0.5 19.5 24.9 26.5 55.1 89.6 32.7 43.5 35.1 = 95) SD 96 311 234 147 9.1 1.7 3.8 3.5 1.8 1.3 1.8 1.4 0.6 6.7 18.4 24.2 34.1 67.7 33.3 47.2 47.2 resources (« = M 991 611 219 132 50.9 8.0 5.4 3.9 3.2 1.7 1.5 2.1 0.5 18.7 25.1 24.0 51.1 93.1 35.2 42.9 38.6 140) SD 102 340 289 197 9.4 1.9 3.2 3.3 2.0 1.2 1.4 2.1 0.6 5.9 16.5 23.0 32.3 67.8 40.4 42.0 55.7 Social - personality resources (» = M 972 580 235 153 51.0 8.0 5.3 3.4 2.9 1.8 1.8 2.3 0.6 19.3 24.8 23.7 47.8 87.4 30.1 48.0 41.8 146) SD 107 334 308 227 9.7 1.8 3.4 3.2 2.1 1.3 1.8 2.0 0.7 6.3 16.6 29.4 32.6 67.1 39.3 38.7 53.9 Resource poor (n M 928 603 249 149 44.4 7.7 4.1 2.7 2.3 1.4 1.2 2.6 0.5 19.5 22.4 23.1 51.8 82.6 34.8 60.1 36.1 = 104) SD 125 310 292 222 10 1.9 3.7 2.8 2.0 1.2 1.5 2.2 0.8 7.0 23.3 33.5 42.6 81.6 59.6 45.2 51.6 «• 4.3*** 0.2 1.6 0.5 9.3*** 0.3 2.1* 2.3» 44* *t 1.6 3.8**» 2.8** 0.3 0.3 0.3 0.2 0.5 0.3 0.2 2.3* 0.2 Note. T value represents a T transformation. a R2 X 100 indicates the proportion of explained variance in the respective category. Significance relates to F tests of univariate analyses of variance. b Weighted duration of activities relates to the average duration of one single activity in the respective activity category (in minutes). *p<.Q5. **p < .01. ***p<.001. the day. There were no differences between these two groups in respect to frequency of self-care and sleeping during the day. Overall, everyday functioning of resource-rich older adults as compared with resource-poor older adults is characterized by more activity in different activity domains. Age Differences in Everyday Functioning To test for differential age effects on everyday functioning among the resource-poor and resource-rich groups of older adults, we compared the regression coefficients of age on the indicators of everyday functioning. Table 6 provides the differ- ential age correlations within all four resource-defined groups. AX2 in the right column of Table 6 indicates the size of effects (percentage of explained variance) and significance of the dif- ferential slopes when comparing the resource-rich and the re- source-poor groups of older adults, Age differences bet\veen the resource-rich and the resource- poor groups. There were consistent differences between the resource-rich and the resource-poor groups with respect to the correlations between everyday activities and age. First, in com- parison to resource-poor older adults, resource-rich older adults had fewer negative age differences. Second, age correlations were significantly reduced in the resource-rich group compared with the resource-poor group with respect to six variables: length of day, variety of activities, time spent alone, number and duration of housekeeping activities, and duration of intellectual - cultural activities. Figure 1 displays the scatterplots for the sig- nificant differential age trends in the resource-rich and the re- source-poor groups on length of waking day (Figure la), vari- ability of activities (Figure I b ) , and frequency of social-rela- tional activities (Figure Ic). In addition, there were also differences between both groups in respect to correlations between age and duration indicators of everyday activities. Figure 2 displays the scatterplots for the significant differential age trends in the resource-rich and the resource-poor groups on duration of time spent alone (Figure 2a), duration of housekeeping activities (Figure 2b), and dura- tion of intellectual activities (Figure 2c). Not included in Figures 1 and 2 is the interaction effect of Age X Group on time spent with others (cf. Table 6). Most of this effect was related to the fact that with increasing age the resource-poor older adults were more likely to live in an institu- tion. That is, very old (above 85 years) compared with young- old ( below 85 years) older people in the resource-poor group reported more time spent with professional helpers and room- mates. When helpers were excluded from the analysis, the effect disappears. In contrast, such a relationship with age was not found among the resource-rich older adults. Age differences between all four resource groups. Differen- tial age effects on everyday functioning between the resource- EVERYDAY FUNCTIONING AND SUCCESSFUL AGING 439 Table 6 Age Effects for Everyday Activities Within Each Group: The Resource Rich, the Sensorimotor-Cognitive Resources Only, the Social-Personality Resources Only, and the Resource Poor Regression coefficients of age on everyday activities within Activity parameters Length of waking day (min) Time spent alone (min) Time spent with family (min) Time spent with others (min) Variability of activities Frequency Self-care Household chores Physical leisure Intellectual -cultural leisure Television viewing Social engagement Resting Sleeping during daytime Weighted duration Self-care Household chores Physical leisure Intellectual-cultural leisure Television viewing Social engagement Resting Sleeping during daytime Resource rich (n = 95) -.23* .16 -.17 -.10 -.23* .09 -.15 -.17 -.28** -.11 -.34* .24* .27** .17 -.06 -.11 .19 .01 -.17 .12 .29** Sensorimotor -cognitive resources (n = 140) -.19* .03 -.02 -.14 -.35** .02 -.19* -.41** -.31** -.20* -.13 .38** .16 .15 .02 -.11 -.11 .02 -.05 .29** .15 Social-personality resources („ = 146) -.02 .09 -.28** .16 -.35** -.06 -.19** -.20* -.25* -.20* -.13 .30 .09 .24* -.12 -.17* -.08 -.07 -.06 .31** .09 groups Resource poor (n = 104) -.48** -.19* -.22* .23* -.44** -.02 -.41** -.28** -.22* -.35** -.03 .16 .05 .22* -.30** -.06 -.19* -.19* -.03 .36** .04 Resource rich vs. poor group A/?2" 3.2* 3.0* 0.0 3.1* 1.4* 0.3 2.1* 0.1 0.0 1.3 2.6* 0.0 0.8 0.1 1.8* 0.0 3.7* 1.2 0.1 1.4 1.2 a AW2 X 100 indicates the increase of explained variance in the respective category accounted for by the Age X Group interaction effect. * ; > < .05. * * p < . 0 1 . rich and the resource-poor groups were also present in compari- sons of these two groups with each of the two one-factor-only resource groups (Table 6). First, comparison of the resource- poor group with the sensoricognitive group revealed signifi- cantly different age associations with respect to length of waking day (AR2 = 3.1', p < .05), time spent with others (A/{2 = 3.5, p < .01), and duration as well as frequency of household- related activities (A/?2 = 3.4, p < .01, respectively, 1.8, p < .05). When comparing the social—personality group with the resource-poor group significantly different age associations were apparent in the length of waking day (AS2 = 5.4, p < .01) and time spent alone (AR2 = 1.8, p < .05). Second, when comparing the resource-rich group with the sensoricognitive group, significant Age X Group interaction ef- fects on everyday functioning were found with respect to dura- tion of intellectual activities (AW 2 = 2.4, p < .05), frequency of social-relational activities (AR2 = 1.7, p < .05), and dura- tion of passive times (AR2 = 1.5, p < .05). Comparison with the social-personality group revealed significant Age X Group interaction effects with regard to time spent with others (AR 2 = 1.6, p < .05) and duration of intellectual activities (AR2 = 1.8, p < .05). This pattern of findings underscores that both social-personality resources as well as sensoricognitive re- sources contribute to the pattern of age differences in everyday functioning found for the resource-rich group. Being rich in only one of the two resource factors alone could not account for the pattern of findings reported above. Discussion In this research we tested the assumption that the resource- rich participants exhibit less negative age trends than the resource-poor participants. In this sense, resources would have protective or buffering functions against negative aging effects on everyday functioning. This buffering or protective function can be explained with the model of selective optimi- zation with compensation (SOC; P. B. Baltes & M. M. Baltes, 1990). On the basis of the SOC model, we argue that resource-rich people will age more successfully because they can make more frequent use of the three processes—selection, optimization, and compensation—and thereby delay decline. In the context of aging, selection is defined as actively or passively reducing the number of goals and domains in order to free and conserve energy and motivation for more important goals or to select new goals in the service of new developmental tasks (e.g., awareness of one's own finitude); compensation is defined as searching for and using alternate means to reach a goal once old means are lost or blocked. Optimization is defined as the refinement of means and resources that are necessary to reach a goal and to excel in selected domains, thereby maximizing the quantity and quality of one's life (for extensive discussions of the pro- cesses see M. Baltes & Carstensen, 1996; P. B. Baltes & M. M. ' AR, is reported after multiplying by 100. 440 BALTES AND LANG Baltes, 1990; Marsiske et al., 1995). Accordingly, the model treats the success criteria as resources or protective factors (Rut- ter, 1987; Staudinger, Marsiske, & Baltes, 1995): The more I have in terms of resources the better I can engage in selection, compensation, and optimization strategies to replace or readjust goals, the more likely will I age successfully—that is, reach those goals that 1 have selected as important for me—and the less I will experience aging decline. In a cross-sectional design, we used a factor analytic ap- proach, first, to describe the relationship between the 11 re- source indicators that resulted in four domains of resources— sensorimotor, cognitive, personality, and social resources—and, second, to identify groups differing in resources. In the factor analytic procedure, age was partialed out to allow the detection of differential age effects in the identified groups. Two age- residualized factors emerged; a Sensorimotor-Cognitive re- sources factor and a Personality-Social resources factor. Using a median-split procedure, we identified four groups: a resource- rich group identified as above the median in both factors, and a resource-poor group identified as below the median in both factors. These two groups were considered to represent the most successful and the least successful aging adults. Two additional groups were identified as possessing resources in only one of the factors; that is, those that were above the median in either one of the two resource factors. As was shown in the data, these groups clearly fall in between the most and least successful aging group with regard to everyday functioning. The findings permit some general and specific statements. In general, comparing the mean level of the 21 indicators of every- day functioning, the resource-rich older adults (i.e., the most successful aging adults) exhibited a higher level of activity. This higher level refers almost exclusively to higher frequency scores and to greater variety in activities. No differences in activity level were found, however, with regard to self-care activities, television watching, and sleeping during the daytime. Although there was a significant difference between the groups as to length of waking day, there were no significant differences between the groups when looking at the duration or time spent in activi- ties. There was a difference, however, in resting time that was longer in the resource-poor group than the other groups. At first glance, it might seem surprising that neither frequency nor duration of self-care activities differed significantly between the four groups. Several explanations come to mind. First of all, there is a finite number of self-care activities that can and need to be accomplished per day. Usually you dress yourself or undress yourself once a day. Second, most of these self-care activities are highly routinized and automatized and thus might be the least or the last affected by aging losses. Third, significant differences in time spent with self-care across age groups (see Figure 2) suggested that when difficulties do arise in the execu- tion of these activities, help and assistance is forthcoming. Therefore, time spent with these activities would not increase, if anything perhaps even decrease. In summary, looking at group differences, older people with only few resources seem to be less active and engaged in less different activities. This is true particularly for the least success- ful aging adults—the resource-poor group. Most of the signifi- cant differences disappeared when the resource-poor group was deleted from the comparison. This seems to signify the follow- ing : the more resources the better, at least with regard to every- day functioning. When inspecting the age effects on parameters of daily func- tioning we found first that age was a risk factor in all four groups: \bung older people did better than the older people in all groups. As expected, however, the most successful aging, resource-rich group showed fewer and less dramatic age effects when compared with the resource-poor group. The negative difference with regard to variability in activities between young and old older people in both groups can be interpreted as the result of selection strategies that help concentrate efforts and energies on those goals that are most important to the older person. Significant Group X Age interaction effects related to the length of the waking day, variety of activities, frequency and duration of household activities, and duration of intellectual activities, all of which showed significantly greater negative age effects in the resource-poor group. Only with regard to time spent alone or time spent with others and frequency of social - relational activities did the resource-rich group show larger neg- ative age effects. This discrepancy can be explained by selection and compensation processes. We have already alluded to the assistance by others; that is, professional helpers, which explains the finding that the resource-poor, least successful aging older adults exhibited less negative age differences with regard to time spent in social-relational activities than the resource-rich, most successful adults (see Figure 2 ) . Unable to perform basic activities any longer, the resource-poor, least successful aging older adults became dependent on others to execute those activi- ties and thus spend more time more often in the presence of others (M. M. Baltes, 1995). They did compensate for losses in respect to self-care activities with the help of others. In contrast, the negative age differences in social-relational activities for the most successful aging group could be a selec- tion effect as postulated, for instance, by the socioemotional selectivity theory by Carstensen (1991, 1993). According to a series of studies by Carstensen (1993) one would expect nega- tive age differences in social-relational activities that reflect a preference for emotionally meaningful or gratifying interactions with very close social partners, whereas interactions with other less close social partners are reduced (Lang & Carstensen, 1994; Lang, Staudinger, & Carstensen. 1996). Furthermore, we found within the resource-rich group that the young-old older adults spent less time in intellectual—cultural activities than the old-old older adults. This could be the product of either optimizing or compensating strategies. That is, the very old resource-rich older adults might compensate in that they allow themselves more time to execute these tasks or they might optimize intellectual performance and, therefore, put more effort and time into these tasks. When considering, however, that the Sensoricognitive-factor-only group did not show such a positive difference between young and old, optimization seems to be the more fitting interpretation. Similarly, among the resource-rich older adults, the old old showed increasingly more and longer sleeping periods than the young old. Within SOC, the resource- rich old old use sleeping to revitalize, and thus sleeping becomes a compensatory strategy for a lowered energy level. In general, these findings support our hypothesis that re- EVERYDAY FUNCTIONING AND SUCCESSFUL AGING 441 Length of Waking Day O O • • 89 94 Age Cohort 99 104 (b) 80 ~ 60 - 50 - 40 - 30 - 20 O Resource Rich (O) 69 74 79 84 89 94 Age Cohort 99 104 (c) 10 -| Frequency of Social Activities Resource Poor (•) Resource Rich (O) 74 79 84 89 94 Age Cohort 99 104 Figure !. Significant age differences in three indicators of everyday functioning—length of waking day, variability of activities, and fre- quency of social activities—shown separately for the resource-rich group and for the resource-poor group. sources help maximize the quantity and quality of life in old and very old age. Resources provide fertile ground for the interplay between the three processes of selection, optimization, and com- pensation and thereby buffer and reduce the magnitude of cross- sectional age differences in the resource-rich group. In sum- mary, the cross-sectional findings speak for the assumption that resource-rich older adults when compared with resource-poor older adults will demonstrate more investment of time and effort in selected domains of everyday functioning. Among resource- poor adults, everyday functioning seems to be characterized by a general concentration on basic levels of everyday competence. (a) Duration of Time Alone J 1200- 1000- 800- 600- 400- 200- 0 Resource Rich (O)£L5£g Resource (b) 69 74 79 84 89 94 Age Cohort 99 104 Duration of Housekeeping Activities £ 90-1 O 70 - 50 - Resource Rich (O) Resource Poor(.) (C) 69 74 79 84 89 94 99 104 Age Cohort Duration of Intellectual Activities 74 79 84 89 94 99 Age Cohort 104 Figure 2. Significant age differences in three indicators of everyday functioning—duration of time alone, duration of household chores- housekeeping, and duration of intellectual activities—shown separately for the resource-rich group and the resource-poor group. 442 BALTES AND LANG These specific interpretations as well as the general relationship between protective factors and their buffering effect toward de- cline or negative aging changes will be validated with longitudi- nal data that will be available for the Berlin Aging Study in the near future. References American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3rd ed., rev.). Washington, DC: Author. Atchley, R. C. (1989). A continuity theory of normal aging. Gerontolo- gist, 29, 183-190. Baltes, M. M. (1995). Dependency in old age: Gains and losses. Current Directions in Psychological Science, 4, 14-19. Baltes, M. M., & Carstensen, L. L. (1996). The process of successful aging. Ageing and Society, 16, 397—422. flaltes, M. M., Maas, I., Wilms, H.-U, & Borchelt, M. (1996). Alltags- kompetenz im Alter: Theoretische Oberlegungen und empirische Be- funde [Everyday competence in old age: Theoretical considerations and empirical findings]. In K. U. Mayer & P. B. Baltes (Eds,), Die Berliner Altersstudie (pp. 525-542). Berlin: Akademie Verlag. Baltes, M. M., Mayr, U., Borchelt, M., Maas, I., & Wilms, H.-U. (1993). Everyday competence in old and very old age: An inter-disciplinary perspective. Ageing and Society, 13, 657-680. Baltes, M. M., Wahl, H. W, & Schmid-Furstoss, U. (1990). The daily life of elderly Germans: Activity patterns, personal control, and func- tional health. Journals of Gerontology: Psychological Sciences, 45, P173-P179. Baltes, P. B. (1993). The aging mind: Potential and limits. Gemmolo- gist, 33, 580-594. Baltes, P. B., & Baltes, M. M. (1990). Psychological perspectives on successful aging: The model of selective optimization with compensa- tion. In P. B. Baltes & M. M. Baltes (Eds.), Successful aging: Per- spectives from the behavioral sciences (pp. 1-34). New Tfork: Cam- bridge University Press. Baltes, P. B., Mayer, K. U., Helmchen, H., & Steinhagen-Thiessen, E. (1993). The Berlin Aging Study (BASE): Overview and design. Age- ing and Society, 13, 483-515. Borkenau, P., & Ostendorf, R (1990). Comparing explanatory and con- firmatory factor analysis: A study on the five-factor model of personal- ity. Personality and Individual Differences, 11, 515-524. Biihler, C. (1933). Der menschliche Lebenslauf als psychologisches Problem [The human life course as psychological problem]. Leipzig, Germany: Hirzel. Carstensen, L. L. (1991). Socioemotional selectivity theory: Social ac- tivity in life-span context. Annual Review of Gerontology and Geriat- rics, 11, 195-217. Carstensen, L. L. (1993). Motivation for social contact across the life span: A theory of socio-emotional selectivity. In J. Jacobs (Ed.), Nebraska Symposium on Motivation (Vol. 40, pp. 209-254). Lincoln: University of Nebraska Press. Costa, P. T, & McCrae, R. R. (1985). The NEO-Personality Inventory: Manual Form S and Form R. Odessa, FL: Psychological Assessment Resources. Cumming, E., & Henry, W. E. (1961). Growing old: The process of disengagement. New 'tork: Basic Books. Erikson, E. H. (1959). Identity and the life cycle. Psychological Issues (Monograph 1). New York: International University Press. Erikson, E. H. (1968). Life cycle. In D. L. Sills (Ed.), International encyclopedia of the social sciences (pp. 286-292). New ^fork: Mac- millan/Free Press. Erikson, E. H. (1982). The life cycle completed. New York: Norton. Erikson, E. H., Erikson, J. M., & Kivnick, H. (1986). Vital involvement in old age: The experience of old age in our time. London: Norton. Havighurst, R. J., & Albrecht, R. (1953). Older people. New York: Longmans. Hertzog, C. (1985). An individual differences perspective. Research on Aging, 7, 7-45. Jahoda, M. (1958). Current concepts of positive mental health. New York: Basic Books. Jung, C. G. (1931). Die Lebenswende [The life transition]. In C. G. Jung (Ed.), Seelenprobleme der Gegenwarl (pp. 248-274). Zurich, Switzerland: Rascher. Jung, C. G. (1934). Seele und Tod [Psyche and death]. In C. G. Jung (Ed.), Wirklichkeit der Seele (pp. 212-230). Zurich, Switzerland: Rascher. Kahn, R. L., & Antonucci, T. C. (1980). Convoys over the life course: Attachment, roles and social support. In P. B. Baltes & O. G. Brim, Jr. (Eds.), Life-span development and behavior (pp. 254-283). New York: Academic Press. Lang, F. R. (1996). Social support relationships of parents and nonpar- ents in old and very old age. In H. Mollenkopf (Ed.), Elderly people in industrialised societies (pp. 41-51). Berlin: Edition Sigma. Lang, F. R., & Carstensen, L. L. (1994). Close emotional relationships in late life: Further support for proactive aging in the social domain. Psychology and Aging, 9, 315-324. Lang, F. R., Staudinger, U. M., & Carstensen, L. L. (1996). Socioemo- tional selectivity in late life: How personality and social context do (and do not) make a difference. Unpublished manuscript, Free Univer- sity Berlin. Lindenberger, U., & Baltes, P. B. (1994). Sensory functioning and intel- ligence in old age: A strong connection. Psychology and Aging, 9, 339-355. Lindenberger, U, & Baltes, P. B. (1997). Intellectual functioning in old and very old age: Cross-sectional results from the Berlin Aging Study. Psychology and Aging, 12, 395-409. Lindenberger, U., Gilberg, R., Potter, U., Little, T. D., & Baltes, P. B. (1996). Stichproblenselektivitat und Generalisierbarkeit der Ergeb- nisse in der Berliner Altersstudie [Sample selectivity and generaliz- ability of the results in the Berlin Aging Study]. In K. U. Mayer & P. B. Baltes (Eds.), Die Berliner Altersstudie (pp. 85-108). Berlin: Akademie Verlag. Lindenberger, U., Mayr, U, & Kliegl, R. (1993). Speed and intelligence in old age. Psychology and Aging, 8, 207-220. Maddox, G. L. (1963). Activity and morale: A longitudinal study of selected elderly subjects. Social Forces, 42, 195-204. Marsiske, M., Klumb, P. L., & Baltes, M. M. (1997). Everyday activity patterns and sensory functioning in old age. Psychology and Aging, 12, 444-457. Marsiske, M., Lang, F. R., Baltes, P. B., & Baltes, M. M. (1995). Selec- tive optimization with compensation: Life-span perspectives on suc- cessful human development. In R. Dixon & L. Backman (Eds.), Psy- chological compensation: Managing losses and promoting gains (pp. 35-79). Hillsdale, NJ: Erlbaum. Maslow, A. (1968). Toward a psychology of being (2nd ed.). Princeton, NJ: Van Nostrand. Maxwell, S. E., & Delaney, H. D. (1993). Bivariate median splits and spurious statistical significance. Psychological Bulletin, 113, 181- 190. McClelland, G. H., & Judd, C. M. (1993). Statistical difficulties of detecting interactions and moderator effects. Psychological Bulletin. 1/4, 376-390. Moss, M., & Lawton, M. P. (1982). Time budgets of older people: A window on four lifestyles. Journal of Gerontology, 37, 576-582. Rogers, C. R. (1961). On becoming a person. Boston: Houghton Mifflin. Rowe, J. W, & Kahn. R. L. (1987). Human aging: Usual and successful. Science, 237, 143-149. Rutter, M. (1987). Resilience in the face of adversity. Protective factors EVERYDAY FUNCTIONING AND SUCCESSFUL AGING 443 and resistance to psychiatric disorder. British Journal of Psychiatry, 147, 598-611. Ryff, C. D. (1982). Successful aging: A developmental approach. Ger- onlologisl, 22, 209-214. Ryff, C. D. (1989a). In the eye of the beholder: Views of psychological well-being among middle-aged and older adults. Psychology and Aging, 4, 195-210. Ryff, C. D. (I989b). Beyond Ponce de Leon and life satisfaction: New directions in quest of successful aging. International Journal of Be- havioral Development, 12, 35-55. Smith, J., & Baltes, P. B. (1993). Differential psychological ageing: Profiles of the old and very old. Ageing and Society, 13, 551-587. Staudinger, U. M., Marsiske, M., & Baltes, P. B. (1995). Resilience and reserve capacity in later adulthood: Potentials and limits of develop- ment across the life span. In D. Cicchetti & D. J. Cohen (Eds.), Developmental psychopathology (Vol. 2, pp. 801-847). New York: Wiley. Staudinger, U. M., Smith, J., & Freund, A. (1992, November). The goal system: A facet of the resilient self in old age? Paper presented at the 45th annual scientific meeting of the Gerontological Society of America, Washington, DC. Steinhagen-Thiessen, E., & Borchelt, M. (1996). Morbiditat, Medika- tion und Funktionalitat im Alter [Morbidity, medication, and func- tional capacity in old age]. In K. U. Mayer & P. B. Baltes (Eds.), Die Berliner Altersstudie (pp. 151-183). Berlin: Akademie Verlag. Wegener, B. (1988). Kritik des Prestiges [Critique of prestige], Opla- den, Germany: Westdeutscher Verlag. Received June 26, 1996 Revision received January 10, 1997 Accepted January 10, 1997 • Sternberg Appointed Editor of Contemporary Psychology (APA Review of Books), 1999-2004 The Publications and Communications Board of the American Psychological Association announces the appointment of Robert J. Sternberg, Yale University, as editor of Contempo- rary Psychology (APA Review of Books) for a 6-year term beginning in 1999. Sternberg, at the request of the Publications and Communications Board, as well as many readers, will be embarking on a program to make the journal more timely, more interesting, and more relevant to psychologists during his editor-elect year in 1998. Some of the changes envisioned include fewer but longer and more thoughtful reviews of books, reviews only of "new" books (with a few noteworthy exceptions), comparative textbook reviews at strategic times of the year, and changes in publication frequency and pricing. Steinberg welcomes suggestions for improving the journal and serving reader needs. E-mail: sterobj@yalevm.cis.yale.edu Publishers should note that books should not be sent to Sternberg. Publishers should continue to send two copies of books to be considered for review plus any notices of publication to PsycINFO Services Department, APA Attn: Contemporary Psychology Processing 750 First Street, NE Washington, DC 20002-4242 (Please note that all reviews are written by invitation.)