Facebook behaviors associated with diurnal cortisol in adolescents: Is befriending stressful?

Psychoneuroendocrinology 63 (2016) 238–246

Contents lists available at ScienceDirect

Psychoneuroendocrinology journal homepage: www.elsevier.com/locate/psyneuen

Facebook behaviors associated with diurnal cortisol in adolescents: Is befriending stressful? Julie Katia Morin-Major a,c,d , Marie-France Marin a,c,d , Nadia Durand c , Nathalie Wan c , Robert-Paul Juster c,d,e , Sonia J. Lupien b,c,d,∗ a

University of Montreal-Department of Neurosciences, C.P. 6128, Succursale Centre-ville, Montreal, Quebec H3C 3J7, Canada University of Montreal-Department of Psychiatry, Pavillon Roger-Gaudry, C.P. 6128, Succursale Centre-ville, Montreal, Quebec H3C 3J, Canada c Center for Studies on Human Stress-Montreal Mental Health University Institute, 7401 Hochelaga Street, Montreal, Quebec H1N 3M5, Canada d Institut Universitaire en Santé Mentale de Montréal, 7401 Rue Hochelaga, Montreal, Quebec H1N 3M5, Canada e Integrated Program in Neuroscience, McGill University, Montreal Neurological Institute, 3801 University Stress, Montreal, Quebec H3A 2B4, Canada b

a r t i c l e

i n f o

Article history: Received 14 April 2015 Received in revised form 4 October 2015 Accepted 6 October 2015 Keywords: Facebook Stress Cortisol Adolescents Depression Sex differences

a b s t r a c t Facebook© is changing the way people interact and socialize. Despite great interest in psychology and sociology, little is known about Facebook behaviors in relation to physiological markers of stress. Given that the brain undergoes important development during adolescence and that glucocorticoids—a major class of stress hormones—are known to modulate its development, it is important to study psychosocial factors that may influence secretion of stress hormones during adolescence. The goal of the present study was to explore the associations between Facebook behaviors (use frequency, network size, self-presentation and peer-interaction) and basal levels of cortisol among adolescent boys and girls. Eighty-eight adolescents (41 boys, 47 girls) aged between 12 and 17 (14.5 ± 1.8) were recruited. Participants provided four cortisol samples per day for two non-consecutive weekdays. Facebook behaviors were assessed in accordance with the existing literature. Well-validated measures of perceived stress, perceived social support, self-esteem, and depressive symptoms were also included. A hierarchical regression showed that after controlling for sex, age, time of awakening, perceived stress, and perceived social support, cortisol systemic output (area under the curve with respect to ground) was positively associated with the number of Facebook friends and negatively associated with Facebook peer-interaction. No associations were found among depressive symptoms, self-esteem, and cortisol. These results provide preliminary evidence that Facebook behaviors are associated with diurnal cortisol concentrations in adolescents. © 2015 Published by Elsevier Ltd.

1. Introduction Friendships are a core source of social support for people of all ages that promote better health and well-being (House et al., 1988; Seeman 1996). Among adolescents, friendships are especially important for the normal development of social and emotional competencies (Way and Greene, 2006). A critical pathway by which friendships foster social support and can thus provide benefits on

∗ Corresponding author at: Center for Studies on Human Stress, Montreal Mental Health University Institute, 7401 Hochelaga Street, Montreal, Quebec H1N 3M5, Canada. E-mail addresses: [email protected] (J.K. Morin-Major), [email protected] (M.-F. Marin), [email protected] (N. Durand), [email protected] (N. Wan), [email protected] (R.-P. Juster), [email protected] (S.J. Lupien). http://dx.doi.org/10.1016/j.psyneuen.2015.10.005 0306-4530/© 2015 Published by Elsevier Ltd.

positive health psychology is through their influence on biological mechanisms related to stress physiology. Response to stress involves activation of the hypothalamic–pituitary–adrenal (HPA) axis that leads to the secretion of glucocorticoids (GCs; cortisol in humans) from the adrenal glands. Diurnal cortisol represents natural fluctuations secreted on a day-to-day basis by individuals in their home, work, and/or school environments at different times of the day. Such sampling methods allow researchers to assess diurnal cortisol rhythms. Among these is the ‘cortisol awakening response’ (CAR) that represents the cortisol increase observed 30–45 min after awakening (Pruessner et al., 1999). By contrast to naturalistic variations, reactive GC represents the hormonal dynamics produced in response to an acute stressor usually in a controlled laboratory environment. Chronic secretion of GCs can have damaging effects on mental health because GCs rapidly access the brain to influence learning,

J.K. Morin-Major et al. / Psychoneuroendocrinology 63 (2016) 238–246

memory, and emotional processing by binding to GC receptors in the prefrontal cortex, amygdala, and hippocampus (for a review, see (Lupien et al., 2009)). Given the slow development of the amygdala and frontal lobes during adolescence, the developing brain is especially sensitive to the effects of elevated GCs (Lupien et al., 2009). Psychoneuroendocrine studies show that high perceived social support has positive effects on the HPA axis. A large body of research reveals that social support is associated with decreased cortisol reactivity to acute stressors (Kirschbaum et al., 1995; Roy et al., 1998; Eisenberger et al., 2007) as well as to lower basal levels of cortisol in adults (Seeman and McEwen, 1996; Turner-Cobb et al., 2000; Evans and Steptoe, 2001). In adolescents, the nature of peer relationships undergoes significant change at puberty when adolescents spend increasingly more time in the company of their peers (Larson and Richards, 1991). Studies of adolescents show that the presence of a best friend buffers the effects of negative experiences on cortisol reactivity to stress (Adams et al., 2011). This suggests that integration into interpersonal networks may contribute to better health through buffering effects of social support on the biological stress system (Cobb 1976; Cohen and Wills, 1985). Over the last decade, there has been a dramatic increase in the use of social networking websites such as Facebook© among adolescents. Facebook (FB) was launched in 2004 and made available to everyone over age 13 in 2006. Since then, it has been one of the fastest-growing websites in history, attaining one billion users in 2012 (Fowler, 2012). As of 2012, 95% of adolescents worldwide are active on FB (Sterling 2013). To date, it is unclear (1) whether the integration of adolescents into virtually mediated social networks has effects on biological pathways related to stress and (2) whether these effects are similar to findings observed among more direct interpersonal networks. While no published evidence has linked FB behaviors to cortisol levels in either adolescence or adulthood, a growing number of studies have assessed the association between FB behaviors and psychological well-being. The first studies performed in this new era of cyberpsychology suggested that FB use may promote negative psychosocial wellbeing and lead to depression (O’Keeffe et al., 2011). Such studies focusing on the proposed negative association of FB use concentrated on a measure of ‘FB use frequency’ calculated as ‘time spent on FB’. The implications here are that greater time spent on FB appears to be significantly associated with greater psychological distress (Kontos et al., 2010; O’Keeffe et al., 2011; Pantic et al., 2012). Yet, not all studies report such associations (see Jelenchick et al., 2013) and none of these studies have considered the potential buffering role of social support on the association between FB use frequency and psychological well-being. In order to assess whether perceived social support is provided by FB, researchers have also assessed ‘number of FB friends’ as a measure of ‘FB network size’. In so doing, researchers found the presence of dynamic relationships between FB network size, perceived social support, and psychological well-being (Edwards et al., 1990; Kim and Lee, 2011; Manago et al., 2012; Wright, 2012). Moreover, Nabi et al. (2013) demonstrated that the number of FB friends an individual has appears to be associated with stronger perceptions of social support. In turn, social support is associated with less subjective distress, less physical illness, and strong well being (Nabi et al., 2013). Importantly, these buffering effects were minimized when interpersonal network size (number of friends in real life) was taken into consideration. This nuance is critical, as it shows the importance of considering perceived social support outside of the virtual domain when assessing the association between FB network size and psychological well being. Social support is qualitatively and quantitatively intertwined and made all the more complicated by the recent advent of social media networks that shape how we express ourselves. For example,

239

Kim and Lee (2011) suggest that while the number of FB friends can have a positive influence on subjective well being, this association was not mediated by perceived social support. Instead, the authors focused on measuring ‘FB self-presentation behaviors’ as represented by FB behaviors where individuals present themselves to others through profile construction, status updates, photo album management, and message posting on the users’ own album (Ellison and Boyd, 2006; Strano 2008). Using this measure, they found that FB self-presentation behaviors had a significant effect on subjective well-being through increased perceived social support (Kim and Lee, 2011). This indicates that how one presents onself on FB modulates their perceived social support. Other studies assessing FB self-presentation behaviors report effects on psychological well-being. Specifically, such FB behaviors have positive effects on subjective well being by increasing self-esteem among adolescents, especially when users receive positive feedback from FB friends (Valkenburg et al., 2006; Strano, 2008). Another study found that participants who were allowed to update their profile and view their own profile during an experiment reported greater self-esteem than participants viewing other users’ profiles (Gonzales and Hancock, 2011). Interestingly, high self-esteem is associated with lower cortisol reactivity to stress in adults (Pruessner et al., 2004) and adolescents (Lindahl et al., 2005). Self-esteem therefore represents an additional personality trait that researchers must consider when assessing the positive or negative effects of FB behaviors. Beyond the aforementioned FB behaviors, FB use can allow users to engage their peers via various activities. This can include active and passive behaviors such as viewing other people’s FB page, posting comments on other people’s FB page, and adding photos of friends on their own FB page. In accordance, engagement in ‘FB peer-interaction behaviors’ may help users gather and garner social support from their FB network. As such, this is an important FB behavior to take into account given the nuances between received versus provided social support (for a review, see Abu Sadat Nurullah, 2012). For example, recent results report that women who receive much social support but provide little support to others report lower feelings of self-efficacy than women who provide high amounts of social support but who receive low amounts themselves (Jaeckela et al., 2012). Taken together, this ensemble of FB behaviors must be considered collectively in relation to psychological factors and biological markers of stress that we innovate in the current study. 1.1. Research goals The main goal of this study was to measure the association between FB behaviors and diurnal cortisol levels among adolescents. Specifically, we assessed how (1) FB use frequency, (2) FB network size, (3) FB self-presentation, and (4) FB peer-interactions would be associated with diurnal cortisol levels in adolescents that experience critical developmental changes in brain maturation of regions sub-serving the HPA axis and emotional regulation. Based on the existing FB literature, we predicted that FB behaviors would be similarly associated with diurnal cortisol profiles. Specifically, we expected that after controlling for sex, age, awakening times, perceived stress and perceived social support all known to modulate cortisol concentrations, FB use frequency would be associated with higher cortisol levels in adolescents. By contrast, we expected that FB network size, FB self-presentation behaviors, and FB peer-interaction behaviors would be associated with lower cortisol levels. Given that no study has measured the associations between FB behaviors and diurnal cortisol in adolescence, we were tentative in the proposed directionality of our hypothesis given that both hyper- and hypo-cortisolemic profiles can be functionally significant.

240

J.K. Morin-Major et al. / Psychoneuroendocrinology 63 (2016) 238–246

2. Material and methods 2.1. Participants Adolescents aged between 12 and 17 years were recruited from a larger study on family and stress conducted by the Centre for Studies on Human Stress (Montreal, Quebec, Canada). Of the 333 participants of the large study that included parents, children below the age of 12, and adolescents, we recruited the totality of adolescents for the current FB study. A total of 94 adolescents (50 girls and 44 boys) aged between 12 and 17 years old completed the study and none refused to participate in the FB study. Participating families were from middle to high socioeconomic strata, all French speaking, and all White. The mean age of the participants was 14.5 (boys: 14.9 ± 1.8; girls; 14.2 ± 1.7). Inclusion required being free of medication that may affect depressive symptoms or cortisol levels (e.g., anti-asthma medication, anxiolytics). Participants were also free of any other psychiatric, neurological, substance use, or general health conditions. Adolescents’ parents signed a consent form, while the adolescents signed an assent form. 2.2. General protocol This study was approved by the Ethics Committees of the Institut universitaire en santé mentale de Montréal and conducted in accordance with the Declaration of Helsinki (1964). A research assistant contacted the interested participants to briefly explain the study. Following verbal consent, the research assistant obtained basic demographic information and scheduled a first home visit. During this first visit, participants were explained the research project, after which they provided written informed consent. They were then briefed by the research assistant on what the study implied, and were given a study packet containing instructions guidelines and salivary collection for the study to be performed (see below). Each adolescent was given a $30 compensation for his or her participation. 2.3. Psychological measures Participants completed psychological questionnaires through a secured website at home using the Studies Web Automation Tool (SWAT) developed at the Centre for Studies on Human Stress. The SWAT is a web based platform that allows participants to answer questionnaires at home in a secure way. Participants were given an individualized secure code to access online questionnaires and asked to complete the questionnaires at any time during the study period. The system saved all completed questions, allowing participants to stop and re-log at any time to complete the various questionnaires. Participants’ answers were sent to the study database and transferred in data files for use in our statistical analyses. 2.3.1. Facebook behaviors FB use frequency was measured by asking participants how many hours per week they spend on FB. Participants had to rate the frequency on a 6-point Likert scale ranging from (1) below 1 h, (2) between 2 and 5 h, (3) between 6 and 10 h, (4) between 11 and 15 h, (5) between 16 and 20 h and (6) above 21 h. FB network size was measured by asking adolescents how many people are listed as ‘friends on their FB profile. Participants had to rate the frequency on a 5-point Likert scale ranging from (1) below 50, (2) between 50 and 100, (3) between 101 and 200, (4) between 201 and 300, (5) above 301. For comparison’s sake, offline network size was also measured by asking adolescents how many close friends they have in real life and this measure. Participants

had to rate the number of close friends they have between 0 and 6 and above. This measure was used solely for descriptive purposes since it correlated with other study variables and FB behaviors. FB self-presentation behaviors were measured by indexing four questions designed to assess the extent to which participants edit, add information, and/or add photos of themselves on their FB page as respectively follows: ‘How many times per week do you’ (1) edit typed-information about yourself on FB; (2) add photos of yourself on FB; (3) add photos of yourself and others on FB; and (4) replace your FB profile picture. Participants rated the frequency of these activities on a 7-point Likert scale ranging from ‘never’ (0), to ‘once a month’ (1) to ‘twice a month’ (2), ‘once each week’ to (3) once every other day to (4) to ‘once a day’ (5) to ‘many times per day’ (6). The index of self-presentation was created by adding the score on each of these questions and we calculated a Cronbach alpha to assess its reliability that was found to be acceptable (˛ = .762). FB peer-interaction behaviors were measured by indexing three questions designed to assess the extent to which participants provided information on their friends’ FB page as follows: ‘How many times per week do you; (1) view other people’s FB pages; (2) post comments on other people’s FB page; and (3) add photos of other only—without you depicted—on other’s people FB page. Participants rated the frequency of these activities on a 7-point Likert scale ranging from ‘never’ (0), to ‘once a month’ (1) to ‘twice a month’ (2), ‘once each week’ to (3) once every other day to (4) to ‘once a day’ (5) to ‘many times per day’ (6). The index of FB peer-interactions was created by adding the score on each of these questions that was shown to be acceptably reliable for this sample (˛ = .699).

2.3.2. Depressive symptomatology Depressive symptoms were measured using the 27-item French-validated version (St-Laurent, 1999) of the Child Depression Inventory (CDI) developed for children and adolescents ages 7–17 (Kovacs, 1981; Kovacs, 1991). Each item contains three choices, ranging from 0 to 2, providing a possible score between 0 and 54. To standardize scores, our statistical analyses used t-scores transformed from the raw data. Total scores on the CDI (t-scores) served as the primary measure of self-rated depressive symptoms. Responses on the CDI for the present sample were reliable (˛ = .875). For ethical reasons, all adolescents were actively monitored by our research team for potential clinical depression. Those who scored in the clinical range (score higher than 20) of symptoms according to the known CDI cut-off points were considered in potential need of clinical intervention and were referred to a child psychologist for additional assessment and potential treatment. Adolescents and parents were informed about this procedure when they signed the consent and assent forms. As a result of this procedure, seven adolescents were referred to a child psychologist for clinically meaningful scores on the CDI. The adolescents referred to the child psychologist were allowed to pursue participation into the study and their data were included in the analyses. As such, results were analysed with and without inclusion of these data into the analyses.

2.3.3. Perceived social support Perceived social support was measured using the 20-item Perceived Social Support Scale (Procidano and Heller, 1983). This scale assesses perceived social support using a Likert-scale ranging from 1 (strongly disagree) to 4 (strong agree) to statements regarding feelings and experiences via relationships with friends at one time or another. This instrument shows high test-retest reliability over a month (r = .830) and internal consistency (˛ = .900). A sum score is used and was similarly reliable for the the present sample (˛ = .859).

J.K. Morin-Major et al. / Psychoneuroendocrinology 63 (2016) 238–246

2.3.4. Perceived stress Perceived stress was measured using the 14-item Perceived Stress Scale (Cohen et al., 1983). This instrument measures perceived stress over the last month using a 5-point Likert scale from 0 (never) to 4 (very often). Upon reverse coding of positive items, scores are summed for all items (e.g., In the last month, how often have you been upset because of something that happened unexpectedly?). Original psychometric properties revealed strong internal consistency (mean ˛ = .85), test–retest reliability (mean r = .85), and evidence of concurrent validity with depression (mean r = .71) and physical complaints (mean r = .59) among young students. Internal consistency for the current sample was high (˛ = .859). 2.3.5. Self-esteem Self-esteem was measured using the Rosenberg Self-Esteem Questionnaire (Rosenberg, 1965). The Rosenberg scale is a 10item self-report measure of overall feelings of global self-worth and self-acceptance. The items are answered on a four-point scale ranging from “strongly agree” to “strongly disagree”. The Rosenberg Self-Esteem Scale was originally developed to assess self-esteem among adolescents showing good reliability and validity across a large number of different sample groups (Rosenberg, 1965). Internal consistency for the current sample was high (˛ = .871). 2.4. Cortisol measures To assess diurnal cortisol levels, saliva samples were taken four times a day on two separate days. Participants were instructed to provide four saliva samples according to the following schedule: (1) at awakening, (2) 30 min following awakening, (3) before dinner, and (4) before going to bed on two non-consecutive weekdays (respectively named day 1 and day 2) during a three week time frame. To facilitate sampling and reduce error, each tube cap was color coded in accordance with time of day. Participants were provided with saliva tubes (Sarstedt© , tube part number 62.558.201) and oral instructions for proper collection. In addition, a video demonstrating how to properly provide samples of saliva was available to participants on the Centre for Studies on Human Stress’ website to remind participants (http://www.humanstress.ca/saliva-lab/methodology/howto-provide-a-saliva-sample.html). Prior to sampling, participants were instructed not to eat or brush their teeth to avoid contamination and to record exact sampling time in logbooks. These logbooks were also used to indicate if they experienced any difficulties or failed to follow instructions. Participants stored their samples in their home freezer until pickup from a research assistant. Compliance to saliva sampling was performed using the Medication Event Monitoring System (MEMSTM , AARDEX Ltd., Sion, Switzerland) that records sampling time for each sample. The MEMS is an electronic recording system comprised of two parts: a standard plastic container and the 45 mm MEMS 6 TrackCap (Serial number 292668–292692, Lot 117) to close the container. Once activated with the Wake-Up software (AARDEX Ltd., Sion, Switzerland), the MEMs cap registers dates and time at which the MEM’s cap is opened. Participants were instructed to put the 4 color coded saliva sampling tube in the MEMS bottle the night before sampling day. They were instructed to retrieve the appropriate tube (following the color code on the MEMS bottle) in the MEMS bottle and provide 2 mL of saliva. Once back to the lab, the MEMS log information was transferred to a computer and analyzed to detect non-compliant individuals. As well, given that time of awakening has been shown to be associated with morning cortisol levels, time of first use of the MEMS caps was used as the measure of time of

241

awakening and entered into our statistical analyses as a covariate. At the end of the study, samples of saliva were retrieved by research assistants, saliva samples were stored in freezers at −20 ◦ C at the Centre for Studies on Human Stress (www.humanstress.ca) until determination using a high sensitivity enzyme immune assay kit (Salimetrics® State College, PA, Catalogue No. 1-3102). Frozen samples were brought to room temperature to be centrifuged at 15,000 × g (3000 rpm) for 15 min. The range of detection for this assay is between 0.012 and 3 ␮g/dL. Upon receiving duplicate assay values for each sample, we averaged these values together. The two cortisol samples taken at each testing session were averaged to account for intra- and inter-individual variability during group testing (Lupien et al., 2001). This protocol was employed to minimize the potentially confounding influence of extraneous factors that can distort the representation of a single cortisol measurement.

2.5. Statistical analysis All cortisol data were inspected to detect potential outliers. Raw cortisol values that were more than three standard deviations above or below the mean were considered outliers. Moreover, a participant who had more than three raw cortisol values over the two days of sampling was considered an outlier and consequently, was excluded from the analyses. Given that cortisol was measured on two different weekdays, a mean was computed for each time point. In the case of an individual outlier, the mean was replaced by the non-outlier variable. In order to control for stress that may have been caused by novelty of sampling, a two-way repeated measure ANOVA was conducted to compare the effect of day of sampling (day 1 and day 2) on cortisol levels. Greenhouse–Geisser corrections were applied when the assumption of sphericity was violated (Greenhouse and Geisser, 1959). Following these analyses, mean cortisol levels were calculated for each of the sampling times by averaging the two sampling days for each sampling time. Beyond raw and aggregated data, all four samples across the day were used to calculate the area under the curve with respect to ground (AUCg) based on the trapezoid formula to represent systemic output throughout the day (Pruessner et al., 2003). Likewise, smaller diurnal time-points like cortisol awakening response (CAR) were similarly calculated in supplemental analyses. Preliminary analyses used bivariate correlations to explore associations among study variables. Our main analysis employed a hierarchical regression with cortisol systemic output (AUCg) as the criterion included in three sequential models. In order to assess whether effects were driven by the cortisol awakening response or PM cortisol levels, the analyses were re-analyzed using CAR and PM levels. As well, all regression analyses were performed with and without inclusion of the seven adolescents with cut-off scores on the CDI. In each regression model, sex, age, and awakening time were first entered as covariates (Model 1). Second, perceived stress and perceived social support were next entered in order to account for the adolescents’ psychosocial contexts (Model 2). Third, FB behaviors were entered into the regression equation (Model 3). Beyond statistical significance set at ˛ = .05, we focused on effect magnitude changes across models using change statistics (R2 effect sizes) and 95% confidence intervals (CI). Due to intermitting missing psychometric and biometric data, degrees of freedom vary in our analyses. All statistical analyses employed the Statistical Package for the Social Sciences© Version 22 for Macintosh for Macintosh.

242

J.K. Morin-Major et al. / Psychoneuroendocrinology 63 (2016) 238–246

Table 1 Descriptive statistics and correlation matrix of all study variables. Variable

M (SD)

1. Sex, ♂ = 1; ♀ = 2 2. Age, years 3. Awakening time, hours/day 4. Perceived stress, sum 5. Depressive symptoms, sum 6. Self-esteem, sum 7. Perceived social support, sum 8. Offline network size, N 9. FB use frequency 10. FB network size, N 11. FB self-presentation behaviors 12. FB peer-interaction behaviors 13. Cortisol systemic output (AUCg) 14. Cortisol awakening response

1.53 (0.502) 14.5 (1.76) 7:28 (1:17) 23.59 (8.46) 10.32 (7.14) 31.22 (5.95) 23.18 (8.42) 3.62 (1.39) 3.75 (1.6) 124 (1.26) 3.24 (2.85) 7.23 (4.26) 193.61 (76.23) 0.585 (0.69)

Correlations 1

2

3

4

5

6

7

8

9

10

11

12

13

– −.215* −.168 .351*** −.082 −.194† .164 −.039 .334** .084 .163 .305** .244* .212*

– .205† .095 .064 −.005 −.093 −.115 .116 .145 −309** −.251* .260* −.131

– −.153 .065 .116 −.038 −.041 .025 .003 −.118 −.116 −.316** −.263*

– .120 −.596*** −.330** .008 .224** .233* .048 .147 .343** .277**

– −.073 −.105 −.093 −.097 −.031 −.064 −.064 .049 .042

– .409** .004 −.069 .183 .098 −.031 −.206† −.031

– .387** .221* .183 .368** .279* .232* .007

– .110 .40** .353* 0.463*** .172 −.069

-– .296** .199† .252* .180 .041

– .347*** .459*** .378*** .230*

– .718*** .199 .054

– .081 .130

– .400***

Abbreviations: AUCg = area under the curve with respect to ground; ♂ = boys; ♀ = girls. * p < 0.05. ** p < 0.01. *** p < 0.001. † p < 0.10.

3. Results 3.1. Preliminary analysis Four participants were not active on Facebook, one was unable to provide enough saliva, while another participant was considered an outlier based on the aforementioned criteria. These participants were thus excluded from statistical analyses. The final sample was composed of 88 participants aged between 12 and 17 years old. Mean age was 14.5 years old, 52% were girls (47 girls and 41 boys). Participants mean body mass indices (BMI) were 20.5 (15.2–32.3). A repeated-measure 2-way ANOVA with Samples (4) and Day of Testing (2) was first conducted to assess any differences between sampling days. No main or interaction effects between Day 1 and Day 2 samples were found (F(1,142.81) = 0.17, p = .685). Consequently, data from the two days were justifiably collapsed across all subsequent analyses as averages. Table 1 reports the descriptive statistics and correlation matrix for all study variables used for exploratory purposes. Girls showed positive associations with perceived stress, FB use frequency, FB prosocial behaviors, and cortisol systemic output (AUCg). Age was negatively correlated with being a boy, FB self-presentation behaviors, and FB peer-interaction behaviors, but positively correlated with cortisol systemic output that was in turn negatively correlated with awakening time. These associations justify our inclusion of sex, age, and awakening time as covariates in our main analysis. Additional associations of interest were identified for psychosocial contexts. Specifically, perceived stress was negatively correlated with perceived social support while positively correlated with self-esteem, FB use frequency, FB network size and cortisol systemic output. Across the board, perceived social support was positively correlated with self-esteem, FB use frequency, offline network size, FB self presentation, FB peer interaction behaviors, and cortisol systemic output. Critically, depressive symptoms were not correlated with a single study variable and therefore dropped from any subsequent analyses. Self-esteem was negatively correlated with perceived stress and positively correlated with perceived social support and FB self-presentation behaviors but was not correlated with cortisol systemic output. Exclusion of the seven adolescents with cut-off scores on the CDI did not change these results. These preliminary associations demonstrate significant associations between perceived stress, social support, and systemic

cortisol output. As such, we were obliged to account for psychosocial contexts in our main analysis so that any findings related to FB behaviors and cortisol levels would be above and beyond those already explained according to covariates, perceived stress, and perceived social support. 3.2. Main analysis A hierarchical regression predicting cortisol systemic output was executed in three sequential models: (1) sex, age, and awakening time as covariates (Model 1); (2) perceived stress and perceived social support (Model 2); and lastly (3) FB use frequency, FB network size, FB self-presentation behaviors, and FB peer-interactions (Model 3). Tests of multi-collinearity identified variance inflation factors ranging from 1.05 to 2.83. In supplemental analyses, we re-ran our regressions with each FB behavior separately to reduce sources of multi-collinearity present in the complete main analysis that follows. Table 2 reports all regression coefficient information. Model 1 with covariates entered was significant (F(3,64) = 7.81, p < .001, R2 = 0.268): cortisol systemic output was positively associated with being a girl and older age, while negatively associated with awakening time. Entering psychosocial factors in Model 2 (F(5,62) = 8.88, p < .001, R2 = 0.417) provided significantly better fit (p = .001, R2 = .149): both perceived stress (Fig. 1A) and perceived social support (Fig. 1B) were positively associated with cortisol systemic output. Lastly, entering FB behaviors in Model 3 (F(9,58) = 6.42, p < .001, R2 = 0.499) added additional explained variance (p = .063, R2 = .082). Here, cortisol systemic output was positively associated with FB network size (Fig. 1C) and negatively associated with FB peer-interactions (Fig. 1D). In supplemental analyses aimed at reducing multi-collinearity that employed the same regressions four times with each FB behavior entered alone, only FB network size remained significantly positively associated to cortisol systemic output (t = 2.1, p = .043, CI: .43, 26.52) Sensitivity analyses were conducted to ascertain which diurnal cortisol timeframe was most strongly associated with FB behaviors. To do so, we re-ran the same regression models with three AUCg scores as outcomes representing: (1) the cortisol awakening response (CAR) incorporating changes from awakening to +30 min after; (2) the PM decline incorporating changes from supper until bedtime; and (3) the maximum AM to minimum PM decline incorporating changes from +30 min after awakening to supper time to

J.K. Morin-Major et al. / Psychoneuroendocrinology 63 (2016) 238–246

243

Table 2 Regression coefficients predicting cortisol systemic output. Model

Predictors

B

SE

t

p

95% CI

1

Sex Age Awakening time

40.77 18.92 −.001

18.1 5.2 .002

2.26 3.64 −3.23

.027 .001 .002

4.69, 76.85 8.54, 29.30 −.010, −.002

2

Perceived stress Perceived social support

3.18 2.87

1.1 .79

3.0 3.65

.004 .001

1.03, 5.34 1.30, 4.44

3

FB use frequency FB network size FB self-presentation FB peer-interaction

−.41 2.58 .90 -2.12

.684 .012 .372 .038

−13.84, 9.15 4.27, 33.75 −3.83, 10.1 −11.34, −.33

−2.35 19.01 3.12 −5.83

5.7 7.4 3.5 2.7

Fig. 1. Partial regression scatterplots showing cortisol systemic output in association with (A) perceived stress, (B) perceived social support, (C) Facebook friends, and (D) Facebook peer-interactions. Due to standardized scaling, values closer to zero on both the x-axis and y-axis represent individual values closer to the sample’s mean. Abbreviation: AUCg = area under the curve with respect to ground.

bedtime. Again, only FB network size was significantly positively associated with the CAR (t = 2.24, p = .029, 95% CI: .07, 1.23) and as a trend with the maximum AM to minimum PM decline (t = 1.78, p = .08, 95% CI: −525, 72, 9148.67). No associations were found for PM declines from supper time to bedtime. To summarize approximate effect sizes, covariates explained 27%, psychosocial context explained an additional 15%, and FB behaviors explained an additional 8% of the variance in cortisol systemic output. Taken together, the final regression model explained nearly 50% of the variance in cortisol systemic output. Note that FB

network size alone explained approximately 4% of the additional variance. 4. Discussion This study demonstrates that FB network size and FB peerinteractions together are positively associated with diurnal cortisol profiles in adolescents. Due to the hierarchical nature of our regression analysis, these findings are over and above the variance explained by covariates (sex, age, awakening time) and the psychosocial contexts (perceived stress and perceived social sup-

244

J.K. Morin-Major et al. / Psychoneuroendocrinology 63 (2016) 238–246

port) of the adolescents investigated. In analyses focusing on FB behaviors in isolation, only FB network size was associated with cortisol systemic output and more precisely morning concentrations that included the CAR. As the first psychoneuroendocrine study of FB behaviors and stress physiology, our discussion will focus on possible mechanisms in the hopes of guiding continued research assessing the pathways whereby the stress of social media might ‘get over the screen’ to eventually ‘get under the skin and skull’. Our main finding links more FB friends to more diurnal cortisol production. With respect to directionality, we had originally proposed that FB use frequency would be associated with higher cortisol levels in adolescents, while FB network size, FB selfpresentation, and FB peer-interactions would be associated with lower cortisol levels. This would have been consistent with existing psychoneuroendocrine research showing that social support buffers against stress pathophysiology. Our results are contrary to these hypotheses. While we did not find any significant correlations between FB use frequency and cortisol concentrations, we must be cautious in our interpretations of high versus low cortisol systemic output during this complex developmental period marked by numerous challenges in social and emotional adaptation. The functional significance of our findings must be demarcated further in future studies that take into account other overlapping factors known to modulate the HPA axis. The relation between HPA axis profiles and depression are complex in adolescence. For example, peaks in morning cortisol (higher than the 80th percentile) increase by 2.9 times the risk of subsequent major depressive disorders in adolescents at risk for major depression (Goodyer et al., 2000). Elevated morning salivary cortisol at 13 years old has been shown to predict elevated depressive symptoms at 16 years old (OR = 1.37) over and above key confounders (Halligan et al., 2007). In a large study from a community sample of 17–18 year-old adolescents, Adam et al. (2010) reported that amplified CAR significantly increased the risk of major depressive disorder (OR = 2.96) one year later. Based on these results, it is possible that the high levels of cortisol reported in adolescents as a function of FB network size marks an increased vulnerability toward stress-related disorder later in life that we cannot answer in the current cross-sectional study. Critically for the interpretation of our findings, we found no FB associations with depressive symptoms that were in turn not correlated with HPA axis functioning. This would have provided insights into the link between FB behaviors and cortisol profiles. In comparison to the cyberpsychology literature, our results are in conflict with those that report significant negative associations between FB use frequency and well-being (Kross et al., 2013), self-esteem (Kalpidou et al., 2011; Chen and Lee, 2013), while being positively associated with depression (Pantic et al., 2012; Jelenchick et al., 2013). It is important to emphasize that these psychological studies are based on college students and/or adults while our study was performed in adolescents. Mixed results could therefore be due to cohort effects. Indeed, FB use was less frequent among adolescents when the first studies emerged, and it is thus possible that those adolescents who used FB at the time were those presenting depressive symptomatology. Nowadays, the vast majority of adolescents are on FB, and this could have prevented us from detecting associations between FB behaviors and depressive symptoms that would have helped us interpret the directionality of our cortisol findings. Another possibility is that there are major age effects that mediate the association between FB behaviors, psychological well being, and physiological correlates like hyper- and/or hypo-cortisolism. Indeed, the average time from the onset of chronic stress in children to the emergence of clinical depression is 11.5 years, with the first major episode occurring during late adolescence (Widom et al., 2007). We can therefore

speculate that there could be protracted effects linking cortisol to psychological outcomes that we could not detect with our sample’s mean age of 14. FB behaviors were strongly related to each other and with various psychological factors. Specifically, FB network size was positively correlated with FB use frequency, FB self-presentation, and FB peer-interactions. In endeavoring to understand why FB friends and cortisol are positively correlated, it is possible that more FB friends may compel adolescents to be more active on FB in order to keep track of all these friendships. Indeed, it is interesting to note that FB peer-interactions were negatively associated with diurnal HPA axis functioning only when other FB factors were included in the regression analysis, especially FB network size. This represents multi-collinearity among constructs that must be further teased apart. Given also that FB network size was positively associated with perceived stress but not perceived social support is consistent with the notion that FB behaviors are synergistic and co-dependent. Recent research reports a curvilinear association between number of FB friends and social support. Specifically, the number of FB friends has a positive association with perceived social support only up to a certain point, after which negative associations between number of FB friends and social support become apparent (Tong et al., 2008; Kim and Lee, 2011). This is an interesting finding in line with psychoneuroendocrine studies showing that threat to the ego/self is a major determinant of HPA axis reactivity to socioevaluative threat (Dickerson and Kemeny, 2002). Moreover, there exists a theoretical inverted-U shape function between circulating cortisol concentrations and adaptation (Lupien and McEwen, 1997). Interestingly, both perceived stress and perceived social support were positively associated with cortisol systemic output in our study. Perhaps having too many friends during adolescence may switch social support toward social pressure and lead to increased levels of cortisol. In order to test this postulate, future studies must delineate the nature of FB friendships (e.g., close friends, acquaintances, strangers) and associations with cortisol levels at different ages. Based on the literature, we also predicted that FB selfpresentation behaviors would be associated with increased self-esteem and lower cortisol levels. Although other studies confirm that FB self-presentation behaviors are positively associated with self-esteem (Valkenburg et al., 2006, Gonzales and Hancock, 2011), we did not find correlations with basal cortisol levels. Interestingly, FB self-presentation behaviors were negatively correlated with age, showing that older adolescents presented a reduced tendency to enact FB self-presentation behaviors. FB self-presentation behaviors seem important however given that they were positively correlated with perceived social support. In synergy, it is thus possible that FB behaviors, personality traits, and psychosocial contexts interact in a complex fashion that the current study was underpowered to precisely link to specific HPA axis profiles. Given that FB is an interactive social network, it also allows one to engage in peer-interaction behaviors. Ours is the first study to show that engagement in FB peer-interactions is negatively associated with cortisol levels in adolescents when controlling for personal (sex, age and time of awakening) and psychosocial (perceived stress and perceives social support) factors. Critically, however, is the inclusion of other FB behaviors in the regression that allowed us to unmask this effect that was otherwise in the opposite direction in bivariate correlations and were ultimately wiped out when assessed by itself. This speaks to the need to develop stronger independent FB constructs with discriminate validity. In so doing, it will be essential to tease apart different effects observed vis-à-vis the provision versus reception of social support and HPA axis functioning (Jaeckela et al., 2012).

J.K. Morin-Major et al. / Psychoneuroendocrinology 63 (2016) 238–246

4.1. Limitations Because this study was the first to address potential associations between FB behaviors and basal cortisol levels in adolescents, some limitations need to be addressed. One major limitation is that our measures of FB behaviors were self-reported. With rapidly changing technological advancements, it is now possible to monitor participants’ Facebook activity pending their consent. At the time that we performed this study, these approaches were not available to us. Future studies should therefore include objective measures by using FB monitoring. Moreover, today’s technologies allow us to examine the time allotted to different FB activities, type of activities performed, and even detect the emotional valence of content posted on FB allowing for the detection of emotional contagion effects (Coviello et al., 2014). Future studies should take advantage of new technologies and other stress biomarkers. Another avenue that the present study did not assess are behaviors related to FB bullying. Bullying is ubiquitous on social networks and frighteningly prevalent among adolescent victims who are at greater risk for depression (Kaltiala-Heino and Frojd, 2011). It is important to also note that our inclusion/exclusion of adolescents according to depressive symptoms and psychometric cut-offs cannot substitute clinical assessments of adolescents at high risk of stress-related problems that include bullying among others. Relatedly, the equal distribution of boys and girls in our sample and statistical adjustment for sex does not preclude important interaction effects related to various adverse contexts (e.g., socio-cultural gender differences in bullying). Additionally, the results of this study can only be generalized to White adolescents from North America. It is likely that important cross-cultural diversity modulate FB behaviors, biomarkers of stress, and psychopathological trajectories that will need to be investigated in future studies. 5. Conclusions Despite these limitations, this study represents the first attempt to delineate the associations between FB behaviors and diurnal cortisol profiles in healthy adolescents. We show that high basal cortisol levels—especially morning concentrations and the CAR—were associated with more FB friends and less engagement in peer-interactions towards these friends. The preliminary nature of our findings will require refined measurement of FB behaviors in relation to physiological functioning. In addition, future studies must determine whether similar associations exist in younger children and older adults. Developmental analyses will allow us to understand how social media can modulate the neurobiological processes related to adaptation and whether virtual stress can indeed ‘get over the screen’ before ‘getting under the skin’. Conflict of interest All authors report having no conflict of interest. Funding source Funding sources were not involved in the writing of the report; and in the decision to submit the article for publication. Contributors This study was conceived and designed by Sonia Lupien, with the collaboration of Julie Katia Morin-Major, Nadia Durand and Nathalie Wan. Nadia Durand, Nathalie Wan and Julie Katia MorinMajor performed the experiments. Julie Katia Morin-Major, Robert Paul Juster and Sonia Lupien analyzed the data. Julie Katia Morin-

245

Major, Marie-France Marin, Robert Paul Juster and Sonia Lupien wrote the paper. All authors have contributed significantly to the manuscript and they all consent to have their names listed on it. Acknowledgements This study was supported by a grant from the Canadian Institutes for Health Research (Foundation Grant #331786) and by a Senior Investigator Chair from the Canadian Institutes of Health Research, Institute of Gender and Health to S.J.L. J.K.M.M was supported by a Master Scholarship from University of Montreal (Department of Physiology and Faculty of Medicine) and by a Master Scholarship Award from the Fonds Recherche Québec Santé. The authors would also like to thank Helen Findlay at the Centre for Studies on Human Stress laboratory for conducting the cortisol assays. References Abu Sadat Nurullah, M.A., 2012. Received and provided social support: a review of current evidence and future directions. Am. J. Health Stud. 27, 173–188. Adam, E.K., Doane, L.D., Zinbarg, R.E., Mineka, S., Craske, M.G., Griffith, J.W., 2010. Prospective prediction of major depressive disorder from cortisol awakening responses in adolescence. Psychoneuroendocrinology 35 (6), 921–931. Adams, R.E., Santo, J.B., Bukowski, W.M., 2011. The presence of a best friend buffers the effects of negative experiences. Dev. Psychol. 47 (6), 1786–1791. Chen, W., Lee, H., 2013. Sharing, liking, commenting, and distressed? The pathway between Facebook interaction and psychological distress. Cyberpsychol. Behav. Soc. Netw. 16 (10), 728–734. Cobb, S., 1976. Social support as a moderator of life stress. Psychosom. Med. 38, 300–314. Cohen, S., Kamarck, T., Mermelstein, R., 1983. A global measure of perceived stress. J. Health Soc. Behav. 24 (4), 385–396. Cohen, S., Wills, T.A., 1985. Stress, social support and the buffering hypothesis. Psychol. Bull. 98, 301–357. Coviello, M., Sohn, Y., Kramer, A.D.I., Marlow, C., Franceschetti, M., Christakis, N.S., Fowler, J.H., 2014. Detectin emotional contagion in massive social networks. PLoS One 9, 1–6. Dickerson, S.S., Kemeny, M.E., 2002. Acute stressors and cortisol reactivity: a meta-analytic review. Psychosom. Med. 54, 105–123. Edwards, E., Harkins, G., Wright, F., 1990. Effects of bilateral adrenalectomy on the induction of learned helplessness behavior. Neuropsychopharmacology 3 (2), 109–114. Eisenberger, N.I., Taylor, S.E., Gable, S.L., Hilmert, C.J., Lieberman, M.D., 2007. Neural pathways link social support to attenuated neuroendocrine stress responses. Neuroimage 35 (4), 1601–1612. Ellison, N.B., Boyd, D., 2006. Social network sites: definition, history, and scholarship. J. Comput. Med. Commun. 13, 210–230. Evans, O., Steptoe, A., 2001. Social support at work, heart rate, and cortisol: a self-monitoring study. J. Occup. Health Psychol. 6 (4), 361–370. Fowler, G.A., 2012. Facebook: One Billion and Counting Retrieved 20-03-2015, 2015, from http://www.wsj.com/articles/ SB10000872396390443635404578036164027386112. Gonzales, A.L., Hancock, J.T., 2011. Mirror, mirror on my Facebook wall: effects of exposure to Facebook on self-esteem. Cyberpsychol. Behav. Soc. Netw. 14 (1–2), 79–83. Goodyer, I.M., Herbert, J., Tamplin, A., Altham, P.M., 2000. Recent life events, cortisol, dehydroepiandrosterone and the onset of major depression in high-risk adolescents. Br. J. Psychiatry 177, 499–504. Greenhouse, S.W., Geisser, S., 1959. On methods in the analysis of profile data. Psychometrika 24, 95–112. Halligan, S.L., Herbert, J., Goodyer, I., Murray, L., 2007. Disturbances in morning cortisol secretion in association with maternal postnatal depression predict subsequent depressive symptomatology in adolescents. Biol. Psychiatry 62 (1), 40–46. House, J.S., Landis, K.R., Umberson, D., 1988. Social relationships and health. Science 241 (4865), 540–545. Jaeckela, D., Seigerb, C.P., Orha, U., Wiese, B.S., 2012. Social support reciprocity and occupational self-efficacy beliefs during mothers’ organizational re-entry. J. Vocat. Behav. 80, 390–399. Jelenchick, L.A., Eickhoff, M.A., Moreno, M.A., 2013. Facebook depression? social networking site use and depression in older adolescents. J. Adolesc. Health 52 (1), 128–130. Kalpidou, M., Costin, D., Morris, J., 2011. The relationship between Facebook and the well-being of undergraduate college students. Cyberpsychol. Behav. Soc. Netw. 14 (4), 183–189. Kaltiala-Heino, R., Frojd, S., 2011. Correlation between bullying and clinical depression in adolescent patients. Adolesc. Health Med. Ther. 2, 37–44. Kim, J., Lee, E., 2011. The Facebook paths to happiness: effects of the number of Facebook friends and self-presentation on subjective well-being. Cyberpsychol. Behav. Soc. Netw. 14 (6), 359–364.

246

J.K. Morin-Major et al. / Psychoneuroendocrinology 63 (2016) 238–246

Kirschbaum, C., Klauer, T., Filipp, S.H., Hellhammer, D.H., 1995. Sex-specific effects of social support on cortisol and subjective responses to acute psychological stress. Psychosom. Med. 57 (1), 23–31. Kontos, E.Z., Emmons, K.M., Puleo, E., Viswanath, K., 2010. Communication inequalities and public health implications of adult social networking site use in the United States. J. Health Commun. 15 (Suppl. 3), 216–235. Kovacs, M., 1981. Rating scales to assess depression in school aged children. Acta Paedopsychiatr. 46, 305–315. Kovacs, M., 1991. Children’s Depression Inventory (CDI). M.-H.S. Inc., New York. Kross, E., Verduyn, P., Demiralp, E., Park, J., Lee, D.S., Lin, N., Shablack, H., Jonides, J., Ybarra, O., 2013. Facebook use predicts declines in subjective well-being in young adults. PLoS One 8 (8), e69841. Larson, R.W., Richards, M.H., 1991. Boredom in the middle school years: blaming schools versus blaming students. Am. J. Educ. 99, 418–443. Lindahl, M., Theorell, F., Lindblad, F., 2005. Test performance and self-esteem in relation to experienced stress in Swedish sixth and ninth graders—saliva cortisol levels and psychological reactions to demands. Acta Paediatr. 94 (4), 489–495. Lupien, S.J., King, S., Meaney, M.J., McEwen, B.S., 2001. Can poverty get under your skin? Basal cortisol levels and cognitive function in children from low and high socioeconomic status. Dev. Psychopathol. 13, 651–674. Lupien, S.J., McEwen, B.S., 1997. The acute effects of corticosteroids on cognition: integration of animal and human model studies. Brain Res. Brain Res. Rev. 24 (1), 1–27. Lupien, S.J., McEwen, B.S., Gunnar, M.R., Heim, C., 2009. Effects of stress throughout the lifespan on the brain, behaviour and cognition. Nat. Rev. Neurosci. 10 (6), 434–445. Manago, A.M., Taylor, T., Greenfield, P.M., 2012. Me and my 400 friends: the anatomy of college students’ Facebook networks, their communication patterns, and well-being. Dev. Psychol. 48 (2), 369–380. Nabi, R.L., Prestin, A., So, J., 2013. Facebook friends with (health) benefits? Exploring social network site use and perceptions of social support, stress, and well-being. Cyberpsychol. Behav. Soc. Netw. 16 (10), 721–727. O’Keeffe, G.S., Clarke-Pearson, K., Council on Communications and Media, 2011. The impact of social media on children, adolescents, and families. Pediatrics 127 (4), 800–804. Pantic, I., Damjanovic, A., Todorovic, J., Topalovic, D., Bojovic-Jovic, D., Ristic, S., Pantic, S., 2012. Association between online social networking and depression in high school students: behavioral physiology viewpoint. Psychiatr. Danub. 24 (1), 90–93. Procidano, M.E., Heller, K., 1983. Measures of perceived social support from friends and from family: three validation studies. Am. J. Community Psychol. 11, 1–24. Pruessner, J.C., Hellhammer, D.H., Kirschbaum, C., 1999. Burnout, perceived stress, and cortisol responses to awakening. Psychosom. Med. 61 (2), 197–204.

Pruessner, J.C., Kirschbaum, C., Meinlschmid, G., Hellhammer, D.H., 2003. Two formulas for computation of the area under the curve represent measures of total hormone concentration versus time-dependent change. Psychoneuroendocrinology 28 (7), 916–931. Pruessner, J.C., Lord, C., Meaney, M., Lupien, S., 2004. Effects of self-esteem on age-related changes in cognition and the regulation of the hypothalamic-pituitary-adrenal axis. Ann. N. Y. Acad. Sci. 1032, 186–194. Rosenberg, M., 1965. Society and the Adolsecent Self-image. Princeton University Press, Princeton, NJ. Roy, M.P., Steptoe, A., Kirschbaum, C., 1998. Life events and social support as moderators of individual differences in cardiovascular and cortisol reactivity. J. Pers. Soc. Psychol. 75 (5), 1273–1281. Seeman, T.E., 1996. Social ties and health: the benefits of social integration. Ann. Epidemiol. 6 (5), 442–451. Seeman, T.E., McEwen, B.S., 1996. Impact of social environment characteristics on neuroendocrine regulation. Psychosom. Med. 58 (5), 459–471. St-Laurent, L., 1999. Adaptation francaise du Children’s Depression Inventory de Maria Kovacs. M. Toronto. Ontario. Sterling, G., 2013. Pew: 94% Of Teenagers Use Facebook, Have 425 Facebook Friends, But Twitter & Instagram Adoption Way Up. Retrieved 20-03-2015, from http://marketingland.com/pew-the-average-teenager-has-425-4facebook-friends-44847. Strano, M.M., 2008. User descriptions and interpretations of self-presentation through Facebook profile images. CyberPsychol. J. Psychsoc. Res. Cyberspace 2, 1–18. Tong, S.T., Van Der Heide, B., Langwell, L., 2008. Too much of a good thing? The relationship between number of friends and interpersonal impressions on Facebook. J. Comput. Med. Commun. 13, 531–549. Turner-Cobb, J.M., Sephton, S.E., Koopman, C., Blake-Mortimer, J., Spiegel, D., 2000. Social support and salivary cortisol in women with metastatic breast cancer. Psychosom. Med. 62 (3), 337–345. Valkenburg, P.M., Peter, J., Schouten, A.P., 2006. Friend networking sites and their relationship to adolescents’ well-being and social self-esteem. Cyberpsychol. Behav. 9 (5), 584–590. Way, N., Greene, M.L., 2006. Trajectories of perceived friendship quality during adolescence: the patterns and contextual predictors. J. Res. Adolesc. 16, 293–320. Widom, C.S., DuMont, K., Czaja, S.J., 2007. A prospective investigation of major depressive disorder and comorbidity in abused and neglected children grown up. Arch. Gen. Psychiatry 64 (1), 49–56. Wright, K., 2012. Emotional support and perceived stress among college students using Facebook. com: an exploration of the relationship between source perceptoins and emotional support. Commun. Res. Rep. 29, 175–184.