The digital footprints of adolescent depression, social rejection and victimization of bullying on Facebook

Abstract

Online Social Networking Sites (SNSs) are immensely popular, especially among adolescents. Activity on these sites leaves digital footprints, which may be used to study online behavioral correlates of adolescent psychological distress and to, ultimately, improve detection and intervention efforts. In the present work, we explore the digital footprints of adolescent depression, social rejection, and victimization of bullying on Facebook. Two consecutive studies were conducted among Israeli adolescents (N = 86 and N = 162). We collected a range of Facebook activity features, as well as self-report measurements of depression, social rejection, and victimization of bullying. Findings from Study 1 demonstrate that explicit distress references in Facebook postings (e.g., "Life sucks, I want to die") predict depression among adolescents, but that such explicit distress references are rare. In Study 2, we applied a bottom-up research methodology along with the previous top-down, theory driven approach. Study 2 demonstrates that less explicit features of Facebook behavior predict social rejection and victimization of bullying. These features include 'posts by others', 'check-ins', 'gothic and dark content', 'other people in pictures', and 'positive attitudes towards others'. The potential, promises and limitations of using digital Facebook footprints for the detection of adolescent psychological distress are discussed.

Full text

1 The Digital Footprints of Adolescent Distress
The Digital Footprints of Adolescent Depression, Social Rejection and Victimization of Bullying
on Facebook
Yaakov Ophir, Christa S. C. Asterhan, & Baruch B. Schwarz
This is the pre-proof version of a paper in press for
Computers in Human Behavior

2 The Digital Footprints of Adolescent Distress
Abstract
Online Social Networking Sites (SNSs) are immensely popular, especially among adolescents.
Activity on these sites leaves digital footprints, which may be used to study online behavioral
correlates of adolescent psychological distress and to, ultimately, improve detection and
intervention efforts. In the present work, we explore the digital footprints of adolescent
depression, social rejection, and victimization of bullying on Facebook. Two consecutive studies
were conducted among Israeli adolescents (N = 86 and N = 162). We collected a range of
Facebook activity features, as well as self-report measurements of depression, social rejection,
and victimization of bullying. Findings from Study 1 demonstrate that explicit distress references
in Facebook postings (e.g., "Life sucks, I want to die") predict depression among adolescents, but
that such explicit distress references are rare. In Study 2, we applied a bottom-up research
methodology along with the previous top-down, theory driven approach. Study 2 demonstrates
that less explicit features of Facebook behavior predict social rejection and victimization of
bullying. These features include 'posts by others', 'check-ins', 'gothic and dark content', 'other
people in pictures', and 'positive attitudes towards others'. The potential, promises and limitations
of using digital Facebook footprints for the detection of adolescent psychological distress are
discussed.
Keywords: adolescents, digital footprints, depression, social rejection, bullying, social
networking sites

3 The Digital Footprints of Adolescent Distress
Online social networking technology have become a central arena for adolescent social life
today (Lenhart, Purcell, Smith, & Zickur, 2010; Statista, 2017). Adolescents around the world
use social network sites (SNSs)
1
not only for establishing and maintaining friendships, but also
for self-presentation purposes, for frustration venting and for emotional self-disclosure (Ophir,
2017; Hew & Cheung, 2012; Manago, Taylor, & Greenfield, 2012). While interacting with and
on these sites, users leave digital footprints, a set of traceable digital activities (Lambiotte &
Kosinski, 2014). Recognizing these trends, social workers in youth-at-risk centers, as well as
school teachers and counselors have leveraged these tools and established online connections
with adolescents under their care (Asterhan & Rosenberg, 2015; Rosenberg, Ophir & Asterhan,
2018). It allows them to monitor adolescents' emotional states and to detect signs of
psychological distress from their online SNS behavior.
In spite of these spontaneous practitioner initiatives, however, little is known about the extent
and the ways in which psychological distress is expressed in, and can therefore be deduced from,
adolescents’ logged SNS activities. How valid and informative are explicit online expressions of
emotional distress in SNSs? What types of distress are expected to be found online? Finally, is it
possible to detect adolescents' distress from patterns of online behavior when users choose not to
explicitly express their emotional state in postings?
Recent research on adult users has shown that the digital footprints from their SNS activities
can be used to infer personal information, such as their age, gender, certain personality traits
(e.g., Kosinski, Stillwell, & Graepel, 2013; Lambiotte & Kosinski, 2014; Schwartz et al., 2013),
and even certain mental health conditions, such as depression and anxiety (Csepeli & Nagyfi,
1
Online social networking sites are computer-based platforms such as Facebook, Twitter, or Instagram, which allow
users to present themselves, upload and share content, and establish or maintain connections with others (Obar &
Wildman, 2015).

4 The Digital Footprints of Adolescent Distress
2017; Park et al., 2014; Reece et al., 2016). In the current work, we extend existing research in
this emerging field by focusing on adolescents. The highest rates of peer bullying (Wang,
Iannotti, & Nansel, 2009) and suicide attempts are documented among adolescents and young
adults (Varnik, 2012). Moreover, parents or other caregivers are often not aware of their
children's suffering (Rey & Bird, 1991; Velting et al., 1998). Knowledge about the digital
footprints of adolescent distress will inform theory and may be used to support mental health
professionals, teachers and youth counselors in ongoing detection and prevention efforts. We
focus our search on two types of digital footprints: direct, explicit distress references in SNS
postings and additional features of SNS activity that do not include explicit references to distress
experiences. We review existing research on these two types of digital footprints separately next.
Explicit distress references
In explicit references to distress, SNS users directly and explicitly write about their personal
experiences of distress or the symptoms of their mental health condition (e.g., ""Life sucks, I
want to die "). Several studies have shown that subsets of SNS postings indeed include
explicit, direct references to distress experiences, in particular to depression (e.g., Cavazos-Rehg
et al., 2017; Moreno et al., 2011; Saleem et al., 2012). For example, Moreno et al. (2011)
developed a coding method to identify explicit references to depression based on the list of
symptoms specified in the Diagnostic and Statistical Manual of Mental Disorders (DSM). These
symptoms include, among others, depressed mood, diminished interest or pleasure, and suicide
thoughts (American Psychiatric, 2013). In addition to these classical, DSM-defined symptoms
depressed individuals have also been found to endorse maladaptive beliefs about themselves
(Beck, 1974) and ruminate about their negative experiences (Nolen-Hoeksema, Wisco, &
Lyubomirsky, 2008). Ophir, Asterhan and Schwarz (2017) found explicit depression references

5 The Digital Footprints of Adolescent Distress
among adolescents Facebook postings: Depressive posts included more references to DSM-based
depressive symptoms, included more cognitive distortions, and revealed more negative attitudes
towards the social "other" (Ophir et al., 2017).
Several explanations have been suggested for why individuals explicitly express their
personal distress on SNSs. Compared to face-to-face settings, users may feel more comfortable
to share their personal depressive feelings on online sites (Christofides, Muise, & Desmarais,
2009). SNS-based communication provides a sense of control (with privacy options to limit
access to one's profile), alongside a sense of belonging to specific networks of friends (Holleran,
2010; Pickett, Gardner, & Knowles, 2004). With the act of public, online emotional sharing,
adolescents reach broad audiences and increase the frequency of supportive feedback from their
online friends and contacts (Bazarova, Choi, Schwanda Sosik, Cosley, & Whitlock, 2015). Their
friends' replies (i.e., verbal comments, likes, or emojis) can signal attention, co-presence, and
partaking in their emotional experiences (Sosik & Bazarova, 2014), which may contribute to
their overall well-being (Burke & Kraut, 2016).
Despite these recent advances in research on online distress sharing, however, the validity of
online SNS references to psychological distress is rarely addressed. Most existing studies only
consider distress references without further cross-validating these with external criteria (i.e.,
validated diagnostic tools), whereas in others validated mental health screening tools as
associated with self-report questionnaires on SNS activities (instead of examining actual online
postings).
To date, only few studies have empirically explored the validity of (adults') explicit SNS-
based distress references using external diagnostic tools: Moreno and colleagues (2012) found a
trend approaching significance, whereby explicit, Facebook-based references to depressive

6 The Digital Footprints of Adolescent Distress
symptoms were associated with self-reported symptoms of depression, using validated screening
tools. Holleran (2010) also reported that depression can be assessed with a moderate degree of
accuracy from a person's online postings on Facebook and Myspace. These first findings on
depression in adult populations support the notion that a person's explicit references in social
media postings can provide valuable information about his/her mental health condition.
The overall goal of the current research is to determine whether and how features of
adolescents' online social network activities can be used as a basis for detecting their personal
distress. In Study 1, we extend the line of research on the predictive validity of explicit distress
references in two ways: First, we focus on adolescents, rather than on adults. Second, in addition
to depression, we include two additional types of distress, which are of a more psychosocial
nature (i.e., distress that relates to, or results from social relationships). With the transition from
childhood to adolescence, teenagers begin to spend more time with their peers, without adults'
supervision. Their social relationships grow more complex and their social status and prestige
become exceptionally dominant in their life experiences (Brown & Larson, 2009). Unfortunately,
a subsetwill experience social rejection or become victims of bullying. More than half of all
adolescents report that they have been bullied verbally or socially (Wang et al., 2009). Early
detection of psychosocial distress is crucial because low socio-metric status has been consistently
linked to adjustment difficulties (Brown & Larson, 2009). Adolescents who suffer from social
rejection or peer bullying are at risk to develop severe psychological distress, experience low
self-esteem and depression (McDougall, Hymel, Vaillancourt, & Mercer, 2001) and, in extreme
cases, may even attempt suicide (Kim & Leventhal, 2008). For these reasons, we not only focus
on depression, but also include measures of two types of adolescent psychosocial distress: social
rejection and victimization of bullying.

7 The Digital Footprints of Adolescent Distress
Non-explicit footprints of psychological distress on SNS activity
Whereas experiences of personal distress may be explicitly and intentionally communicated
in SNS postings, it may also leave non-explicit footprints on a person’s online social network
activities (Lambiotte & Kosinski, 2014). In other words, emotional distress may “leak through”
unintentionally, in seemingly unrelated online behavioral features. These features may include
social interaction markers (e.g., number of friends, number of 'likes' received to posts, number of
posts received on one’s timeline), personal interests and self-representational choices (e.g., types
of pages and groups that appear on one's personal profile), and level of engagement in social
networks (e.g., number and types of postings).
Interestingly, early studies from the pre-SNS era have already shown how bottom-up, data
mining techniques may be useful for such explorations. For example, essays of students, which
were subjected to computerized text analysis program revealed that depression is associated with
non-clinical written expressions, such as excessive usage of first person pronouns in essay
writing (Rude, Gortner, & Pennebaker, 2004). Similarly, neuroticism was found to be predicted
by a higher rate of first person pronouns usage in daily diary writing (Pennebaker & King, 1999).
More recent research has shown that personality traits can be predicted by social network activity
features (Lambiotte & Kosinski, 2014). For example, Schwartz et al. (2013) found that
introversion and extroversion leave different and distinctive digital footprints on Facebook.
Finally, a recent examination of 12 million SNS profiles has showed that SNS behaviors that
indicate offline social interactions (e.g., posting photos from offline social events), as well as
accepting online friendships requests are associated with reduced mortality rates (Hobbs, Burke,
Christakis, & Fowler, 2016).
Using a designated Facebook application to record ongoing depressive symptoms of 55 adult

8 The Digital Footprints of Adolescent Distress
Facebook users, it was found that depressive symptoms are correlated with number of friends
and location tags (Park, Lee, Kwak, Cha, & Jeong, 2013). Other digital footprints such as first
person pronouns, negative affect, and increased nighttime activity were found to be predictive of
the onset of depression disorder among adult Twitter users (De Choudhury, Gamon, Counts, &
Horvitz, 2013). Taken together, these findings about non-explicit footprints of depression and
personality traits form the basis for Study 2. In Study 2, we examine whether psychological
distress may be detected, not only by explicit distress references, but also by additional, non-
explicit features of adolescents' SNS activity.
Whereas Study 1 applies a top-down, theory-driven research approach, Study 2 combines
both top-down and bottom-up, exploratory research methods. Top-down and bottom-up are
different, yet complementary research approaches in the social sciences. In top-down
approaches, researchers apply deductive reasoning from the general theory on a given topic (e.g.,
depression) to its more specific sub-components. This method is a primary research method in
confirmatory data analyses aimed to prove or refute a given theory. Based on this method, the
search for online signs of depression should be limited to the theoretical, clinical picture of the
disorder (e.g., explicit manifestations of depressive symptoms). In the bottom-up approach,
researchers apply inductive reasoning, from specific observations to larger patterns. This method,
which is not confined to an a priori theory, is a common approach in in exploratory research. In
Study 2, we therefore search for SNS-based non-explicit footprints of adolescent distress, that a
priori do not seem directly related to the clinical picture and symptoms of the various types of
psychological distress.
In addition to its focus on adolescent users and the inclusion of two types of psychosocial
distress (i.e., social rejection and victimization), Study 2 extends the existing line of research on

9 The Digital Footprints of Adolescent Distress
non-explicit digital footprints as follows: We examine Facebook activity rather than public
Twitter posts, which is the focus of most existing work in this field (e.g., Coppersmith, Dredze,
Harman, Hollingshead, & Mitchell, 2015; Reece et al., 2016; Tsugawa et al., 2015). Non-
explicit, digital footprints of psychosocial distress may be more dominant on Facebook than on
Twitter. Twitter usage is usually driven by interests in real-time news and by the need to be
updated on "hot" topics, people, and conversations (Hargittai & Litt, 2011), whereas Facebook
usage is focused on social interaction and relationships (Hew, 2011). Facebook users tend to
form online connections with people they know from their offline environments and thereby
increase their "bonding social capital", strong relationships that provide emotional kinship and
social support (Phua, Jin, & Kim, 2017). Twitter usage, in contrast, has been linked to increased
"bridging social capital", more weak and distant relationships (Phua et al., 2017). Since social
relationships are intertwined in individuals' mental health, we expect that Facebook activities
would serve as a natural source for digital footprints of psychological distress.
The current work and hypotheses
Relying on both theory-driven and exploratory research methods, we present two consecutive
studies that examine the relation between adolescents' online SNS behaviors on Facebook and
external, traditional validators of emotional distress. The first study focuses on direct, explicit
distress references in adolescent SNS postings (i.e., "I feel depressed"). Based on previous
research that has examined the validity of explicit references to depression among young adults
(Holleran, 2010; Moreno et al., 2012), we hypothesize that explicit references to distress in
Facebook postings, and specifically to depressive symptoms, will predict self-reported
depression among adolescents (hypothesis #1). In study 2, we adopt an open, exploratory search
for non-explicit Facebook behavior features that distinguish between adolescents that do and do

10 The Digital Footprints of Adolescent Distress
not suffer from emotional distress. Based on findings from research on adult users of social
media, it is expected that emotional distress leaves discernable digital footprints in adolescents'
Facebook activities (hypothesis #2). Finally, in light of the fact that Facebook serves as a
platform for social interactions for many adolescents, we hypothesize that two types of
psychosocial distress, namely social rejection and victimization of bullying, will be correlated
with, and expressed in, features of teenagers' Facebook activity (hypothesis #3).
Study 1
The main goal of Study 1 is to examine the predictive validity of explicit references to
personal distress in adolescents' Facebook postings. In order to increase the frequency of
encountering self-reported, actual distress, we did not limit recruitment of participants from the
general adolescent population only, but also specifically recruited participants from youth-at-risk
associations. In this study, the occurrence of explicit references to personal distress on Facebook
was compared with external, well-established, self-report measures of psychological distress
(i.e., depression, social rejection, and victimization of bullying).
Method
Participants and procedure. The procedure was approved by the Chief Scientist's office at
the [BLINDED] Ministry of Education (protocol number #8335) and by the Ethical Committee
for the Use of Human Subjects in Research of the School of Education at the [BLINDED]
University. Participation criteria included age (between 13 and 18), language (Hebrew speakers),
and having a Facebook profile. Recruitment was conducted in informal education settings, such
as in local, popular youth movements, and other after-school activities. Youth leaders,
counsellors and educators were contacted to recruit participants and conduct the data collection
on their facilities.

11 The Digital Footprints of Adolescent Distress
Adolescents received information about the general goal and importance of the study and
were informed in detail about the study's procedure, including the request to download data from
their personal Facebook activity log. The adolescents were provided with parental consent forms,
which were attached to written descriptions of the study, and were informed of the reward for
participation (i.e., a pizza raffle). A total of 101 adolescents achieved their parents' consent and
agreed to participate in the study. Eight participants dropped out without completing the self-
report questionnaires and seven participants were omitted from the final sample because they did
not meet the language criterion. Altogether, the ultimate sample used for analyses in Study 1
consisted of 86 (51.2% girls) Hebrew speaking adolescents, aged 13 to 18 yrs (M = 15.98, SD =
1.3). Close to forty percent of the sample (N = 34) were recruited from after-school programs
organized by youth-at-risk organization.
Participants were instructed to find a quiet and private spot where they could read the
instructions, sign the consent form and take their time to honestly complete three self-report
measures of distress: social rejection, victimization of bullying, and depression. Subsequently,
participants downloaded the last three months of their personal Facebook account's "activity
log". The activity log is a feature of Facebook that allows users to review their own postings and
actions, as well as the posting that others have posted on the participant's Facebook "wall".
Specifically, participants were asked to download three sections from this activity log: (1) 'Your
posts', which include Facebook posts that were created ('Owner-Created') or shared ('Owner-
Shared') by the user, (2) 'Posts by others', which include posts that were created by other people
but had a reference to the participant ('Other-Tagged') or that were attached to the participant's
own timeline ('Other-Created'), and (3) a screen shot of the 'About' page (i.e., a Facebook page
that allows users to share basic personal details about themselves, including hobbies, movies, and

12 The Digital Footprints of Adolescent Distress
music preferences). The About section also shows the total number of Facebook friends of the
user. Technical assistance was provided when needed. The data was transferred with a flash
drive to a secured computer that was authorized for use by the authors and a research assistant
only.
Instruments and content analysis. Three self-report measures of distress were used in both
studies: Depression was measured using the well-established Beck depression inventory-II
(Beck, Steer, Ball, & Ranieri, 1996). Social rejection was measured with four items from the
social problems sub-scale of the widely used Youth Self Report protocol (YSR; Achenbach
1991). Finally, victimization of bullying was measured using six items drawn from the Peer
Relations Questionnaire (PRQ), a well-validated 33 items scale (Rigby, 1998). All self-report
scales demonstrated good internal consistency across both studies (Cronbach's α of depression,
social rejection, and victimization ranged from .90-.91, .77-.85, and .85-.86, respectively). A
detailed description for each scale is offered in the supplementary material.
Content analysis. Each Facebook profile was evaluated whether it contained (one or more)
explicit references to distress. The content of all the posts published by fifteen (out of 86)
different Facebook users was rated by two trained, independent judges, the first author and a
research assistant (N = 142 posts in total), who were blind to the participants' self-report measure
outcomes. Interrater reliability was satisfactory (κ = .73, Fisher's Exact Test = 14.74, p = .01).
Specifically, only one profile was coded differently between the judges, with two versus three
explicit references to personal distress. After the rating of all 86 profiles, all posts that were
marked as containing explicit references to personal distress (26 posts altogether), were
subjected to further classification by both judges. The two judges classified the type of distress in
each post upon three nominal variables: (1) posts with references to depressive symptoms, (2)

13 The Digital Footprints of Adolescent Distress
posts with references to psychosocial distress, and (3) posts with undefined or other type of
distress. Inter-rater reliability was high for the primary type of distress (κ = .86) as well as for the
secondary type of distress (κ = .91). Three disagreements (two primary, and one secondary, type
of distress) were discussed among the judges until a final score was decided (Further description
of the content analysis is offered in a supplementary material).
Results
Self-report measures of distress. Descriptive statistics and zero order correlations of the
self-reported measures are presented in Table 1. Consistent with the literature on adolescents'
distress, positive correlations were found between social rejection, victimization of bullying, and
depression. The average sum score
2
of the BDI was 10.23 (SD = 8.9).
Explicit references to personal distress. Altogether, the 86 participants published a total of
1,168 posts on their timeline in the three months prior to the date of data collection. Of these, 26
posts contained explicit references to personal distress. The 26 distress posts were published by
10 different individuals (12% from the total sample), three individuals from the mainstream
sample (5.7%) and seven individuals from the youth-at-risk sample (20.6%). In 15 cases (58% of
the explicit postings) the primary distress reference related to depression, nine (35%) to
psychosocial distress, and two (8%) referred to undefined distresses. Seven posts (26.9%)
included a secondary type of distress: four (15%) references to depressive symptoms and three
(12%) references to psychosocial distress.
Explicit references to personal distress and depressive symptoms. An independent-samples
t-test was conducted to compare depression scores for adolescents with (N = 10) and without
explicit references to any type of distress (N = 76). Levene's test for equality did not allow us to
2
The average sum score of the BDI appears in addition to the mean score from Table 1 in light of the conventional
standardized norms and cut-off points of depression scores that are usually considered in their sum scores.

14 The Digital Footprints of Adolescent Distress
assume equality of variances between the two groups (F = 12.31, p = .001). The results of the t-
test comparison (equal variances not assumed) demonstrated a significant difference in BDI
scores between users that published posts with explicit references to personal distress (M =
19.60, SD = 14.69, N = 10) and users that did not (M = 9.00, SD = 7.19, N = 76), t = 2.25, p =
.050. A calculation of the standardized differences between the two means indicated a large
effect size (Cohen's d = .92). These results lend support to our first hypothesis that explicit
references to distress imply higher levels of depression among adolescents.
In light of our general dichotomous goal to use Facebook as a detection tool that determines
whether or not the adolescent suffer from depression (i.e., a 'yes or no' question), we continued
and examined the predictive value of posts with explicit references, specifically to depressive
symptoms, using a non-parametric, Chi-square test. A dichotomous variable was created to
differentiate between depressed and non-depressed participants. The standard cut-off score for
mild depression is 14. However, we used a cut-off score of 13 because two items from the
original BDI scale (sexual and suicidal behaviors) were omitted in the current study and the
adjusted cut-off score was re-calculated to 12.667. Based on this cut-off score, 25 participants
(29.1%) were found to experience at least mild depression. A Chi square test of independence
demonstrated a significant difference , χ2 (1, 86) = 9.19, p = .002. Remarkably, seven out of ten
(70%) participants who published posts with explicit references to depressive symptoms
experienced at least mild depression, whereas 18 out of 76 (24%) participants who did not
publish depressive posts fell in that same category. This finding is specifically informative
because it suggests that explicit references to depression significantly increase the chances that
the adolescent is indeed suffering from depression (hypothesis #1).

15 The Digital Footprints of Adolescent Distress
Explicit references to distress and social rejection. A t-test was conducted to compare social
rejection scores for adolescents with and without explicit references to any type of distress.
Levene's test for equality was significant (F = 9.80, p = .002). The difference in social rejection
scores for students who expressed distress (M = 2.15, SD = 1.21, N = 10) and those who did not
(M = 1.70, SD = .69, N = 76) was not significant, t = -1.14, p = .280 (equal variances not
assumed). A Chi square test of independence, using a dichotomous variable for social rejection
based on the upper quartile of the scale (2.25), did not demonstrate significant differences in
social rejection, with and without explicit posts of psychosocial distress, χ2 (1, 86) = .14, p =
.705.
Explicit references to distress and victimization of bullying. A t-test was conducted to
compare victimization scores for adolescents with and without explicit references to distress.
Levene's test for equality was significant (F = 24.51, p < .001). The test did not demonstrate a
significant difference in victimization scores for adolescent with explicit references to distress
(M = 1.95, SD = 1.06, N = 10) and without references to distress (M = 1.45, SD = .40, N = 76), t
= -1.48, p = .171 (equal variances not assumed). The following Chi square test, which used a
dichotomous variable of victimization based on the upper quartile of the scale (1.67), did not
demonstrate significant differences in victimization of bullying with and without references to
psychosocial distress, χ2 (1, 86) = .92, p = .336. Thus, it cannot be concluded that explicit
manifestations of any type of distress or psychosocial distress are significantly linked to
experiences of social rejection or victimization of bullying.
Discussion
The main goal of Study 1 was to examine the predictive validity of explicit online references
to personal distress. The results of Study 1 confirmed our first hypothesis that explicit references

16 The Digital Footprints of Adolescent Distress
to distress predict higher levels of depression among adolescents. Most of the documented
explicit references to distress were to depressive symptoms (58% of the distressed posts) and,
remarkably, seven out of the ten (70%) participants who published posts with explicit references
to depressive symptoms experienced at least mild depression. These findings extend previous
research that examined the validity of explicit references to depression (Holleran, 2010; Moreno
et al., 2012) to include younger SNS users (i.e., adolescents rather than adult).
Nevertheless, even though the findings in Study 1 showed that explicit references to personal
distress in Facebook postings are predictive of self-reported symptoms of depression, distress
references could not predict psychosocial distress (i.e., social rejection and victimization of
bullying). Moreover, even the most frequent explicit distress references (to depression) are rare:
Only 2% of adolescent postings collected contained explicit references to personal distress. In
addition, only 12% of the number of adolescents posted Facebook status updates that included
explicit depression references. Vice versa, most of the Facebook timelines of the adolescents
who, according to the self-report measures, were suffering from depressive symptoms (18 out of
25), did not include explicit reference to any type of personal distress. This false negative
identification error, whereby the test (explicit distress references in Facebook postings) failed to
detect a depressive individual, requires further research that examines whether additional, less
explicit Facebook activity features, may serve as online indicators for distress.
Study 2
The first goal of Study 2 is to replicate theStudy 1 findings in a new and larger sample of
Israeli adolescents. The second goal is to broaden the scope of Facebook activity features that
could potentially add to the prediction of psychological distress. As documented in recent
research, both general measures of psychological well-being (Burke & Kraut, 2016) as well as

17 The Digital Footprints of Adolescent Distress
specific ones, such as depression (De Choudhury et al., 2013) have been found to leave digital
footprints on a person’s social network activity. We therefore complement the top-down search
for explicit distress references with a bottom-up search for Facebook activity features that are not
rooted in traditional clinical assessment tools and definitions.
Method
Participants and procedure. A few modifications were made to the data collection
procedure, which both facilitated the data collection itself and expanded the scope of data
collected (e.g., by including 'Likes', 'Groups', and other people's reactions to the user postings).
All changes are provided in the supplementary material. A total of 162 adolescents (51.3% girls)
participated in Study 2. The collected Facebook data included the 'About' section (i.e., basic
personal details and hobbies as well as number of Facebook friends) and the users' 'Timeline'
activity. Timelines included Facebook posts, created ('Owner-Created'), or shared ('Owner-
Shared') by the user, as well as posts that included a reference to the user ('Others-Tagged') or
that were attached to the user 'Timeline' ('Others-Created'). The Facebook data included both
quantitative data and qualitative verbal content. A quantification process of the verbal content
(Chi, 1997) was applied: First, all the collected data was carefully read several times to identify
recurrent themes. Following, a coding scheme was developed, while taking into consideration
previous work that yielded non-trivial categories that may serve as non-explicit digital footprints
of distress (Ophir, Asterhan, & Schwarz, 2017).
The About section. The content analysis process yielded six distinct content categories: (a)
Fun & news, (b) Commercials, (c) Belonging, (d) Values & involvement, (e) Gothic & dark, and
(f) Extreme & offensive. The research coders were then instructed to judge which category is
best suited to a given item from this section. We calculated the frequencies of each content

18 The Digital Footprints of Adolescent Distress
category in order to construct the idiosyncratic compound of each one of the collected About
sections. For example, the About section of a particular teenager may consist of 70% 'Fun and
news' and 30% 'Commercial content'.
Timeline section. The quantitative data from the 'Timeline' section included the overall
number of posts on a participant's timeline during three months of activity, as well as the
proportions of each type of post: 'Owner-Created', 'Owner-Shared', 'Other-Tagged', and 'Other-
Created'. Each participant was also given a total score for the number of posts that appear on
his/her timeline because of actions undertaken by other people ("Others-Total"), that is: the sum
of "Other-Tagged" and "Other-Created". The relative frequency of "Others-Total" out of the
overall number of posts on one's timeline reflects the pivot point between the user own postings
and the postings published by others. We have also calculated the total number of Likes and
Comments a teenager received for his/her own Timeline postings (i.e., 'Owner-Created' or
'Owner-Shared') in the last three months.
In addition to these numerical data, qualitative features were also extracted from the
Timeline section, by focusing on the content of posts, which were published or shared by the
users (Owner-Created and Owner-Shared). Similar to Study 1, posts were first examined for
explicit references to personal distress. Nevertheless, even though, self-reported depression rates
were approximately the same as in Study 1 (see Results), only four participants in this sample
showed explicit references of distress in their Facebook postings (three of whom also received
high self-reported depression scores). Based on previous findings on Facebook postings of
distress (Ophir et al., 2017), we coded the content along six categories: (a) number of first person
single pronouns, (b) number of first person plural pronouns, (c) number of other people in
pictures, (d) valence of content, (e) attitudes towards others, and (f) poetic-concrete content.

19 The Digital Footprints of Adolescent Distress
Good interrater reliability was achieved in all categories, except for poetic-concrete content,
which was found more subjective and amorphous. The poetic-dramatic ratings were excluded
from further analyses. The complete content analysis procedure is provided in the supplementary
material along with detailed descriptions of the scales of measurements and inter-rater reliability
statistics for each coding category.
Results
Self-report measures of distress. Corresponding with the findings in Study 1 and the
literature on psychological distress, significant correlations were documented between the three
self-report measures of distress (Table 1). Similar to Study 1, the average sum score of the BDI
was 9.81 (SD = 8.64). Thirty-one percent of the participants reported at least mild depression.
Self-reported distress and features of the Facebook About section. Table 2 presents zero
order correlations between the different quantitative measures from the Facebook About section
and the three measures of distress. The total number of Check-Ins (adolescents sharing their
geographic location at a particular moment with Facebook contacts) was found to be negatively
correlated with social rejection (r = -.17, p = .028) as well as with depression (r = -.16, p = .047).
The number of Apps a person is subscribed to was negatively correlated with Depression (r = -
.16, p = .047). However, the overall pattern that emerges from the data shows that most of the
quantitative "About" categories and self-reported distress measures are not associated (Table 2).
Table 3 presents the zero-order correlations between self-report distress measures and the six
content categories, which were based on classifications of the types of Likes and Groups in the
Facebook About section. The data in Table 3 shows that the frequency of 'Gothic and dark'
content was positively correlated with social rejection (r = .29, p <.001) and with victimization

20 The Digital Footprints of Adolescent Distress
of bullying (r = .24, p = .002). The frequency of 'fun and news' was positively correlated with
victimization of bullying (r = .16, p = .039).
Self-reported distress and features of the Facebook Timeline. Zero order correlations
between self-reported distress and the different types of timeline posts are presented in Table 4.
Social rejection was found to be negatively correlated with the pivot point score of Others-Total,
r = -.21, p = .009, specifically with the proportions of Other-Tagged, r = -.24, p = .002.
Table 5 presents correlations between self-reported distress and the sum scores of Likes,
Comments and the six content categories. The total number of comments or likes a teenager
received for their own posts was not found to be associated with any of the three self-reported
levels of distress. Social rejection was negatively associated with Other people in pictures, the
number of other individuals that appear in the pictures posted by the teenager him/herself, r = -
.21, p = .012. Interestingly, social rejection was positively correlated with the combined score of
attitudes towards others, r = .18, p = .019.
Similar trends were observed in victimization of bullying: Victimization was also negatively
associated with Other people in pictures, r = -.18, p = .033 and the correlation between
victimization and attitudes towards others approached significance, r = .14, p = .068.
Victimization of bullying was also positively correlated with the number of first person single
pronouns adolescents used in their Facebook postings, r = .24, p = .008.
Although the size of the reported significant correlations was small (.17 =< r = < .29), taken
together, the results suggest that even though explicit teenage references to distress in online
Facebook postings is rare, less explicit features of behavior on the adolescent's personal
Facebook space can predict social rejection and victimization of bullying (hypotheses #2 and
#3). In contrast, self-reported depression demonstrated only two (negative) correlations, namely

21 The Digital Footprints of Adolescent Distress
with the number of Check in's and Apps. In the next section, we apply discriminant analyses to
examine which of the abovementioned Facebook features may distinguish between adolescents
who suffer from psychosocial distress from those who do not. In this way we could consolidate,
and make sense of, the online clinical picture of psychosocial distress.
Discriminant analysis. A discriminant analysis was performed to examine whether and to
what extent Facebook variables can be used to discriminate between adolescents with high levels
of social rejection (upper quartile, SR ≥ 8) and adolescents with low levels of social rejection
(lower quartile, SR ≤ 4). Five Facebook variables, which were significantly correlated with
social rejection served as predictors of membership in the one of the two groups of social
rejection: (1) Check-ins, (2) posts by others on Timeline (Others-Total), (3) Gothic and dark
content themes from the About section, (4) Other people in pictures from the user Timeline
postings, and (5) Attitudes towards others in Timeline postings.
The discriminant analysis transformed the original Facebook variables into one new variable
(i.e., a canonical variable) that combines the five separate variables, appropriately weighted, into
a single index which maximally discriminates between low and high social rejection. The overall
test of function was significant, Wilks λ = .79, χ2 = 19.31, df = 5, Canonical correlation = .46, p =
.002, with 68.2% correct re-classification of originally grouped cases. These results suggest that
the composite of the five Facebook activity features of Posts by others, Check-Ins, Gothic and
dark content, Other people in pictures, and Attitudes towards others can significantly
discriminate between adolescents with high levels of social rejection and low levels of social
rejection.
A second discriminant analysis was performed to examine whether and to what extent
Facebook variables can be used to discriminate between adolescents with high levels of

22 The Digital Footprints of Adolescent Distress
victimization of bullying (upper quartile, Victimization ≥ 10) and adolescents with low levels of
social rejection (lower quartile, Victimization ≤ 7). Four Facebook variables significantly
correlated with victimization of bullying and served as predictors of membership in one of the
two groups of victimization: (1) Fun and news themes from the About section, (2) Gothic and
dark content from the About section, (3) First person single usage in Timeline postings, and (4)
Other people in pictures from the Timeline section.
The overall test of function was significant, Wilks λ = .84, χ2 = 10.44, df = 4, Canonical
correlation = .40, p = .034, with 61.5% correct re-classification of originally grouped cases.
These results suggest that the above four Facebook categories may significantly discriminate
between adolescents who suffer from victimization of bullying and those who do not.
Discussion
The main goal of Study 2 were to explore the predictive value of both explicit distress
references as well as non-explicit, additional Facebook activity features. The results of Study 2
suggest that even though explicit distress references are rare in teenage online Facebook
postings, some forms of distress (i.e., social rejection and victimization) leave less explicit, yet
discernable digital footprints on adolescents' Facebook activities (hypotheses #2 and #3).
Discriminant analyses suggested composites of Facebook features that can serve as indicators
for the existence of social rejection and victimization of bullying: On the one hand, the online
clinical picture of social rejection seems to reflect 'real' social difficulties (i.e., low frequency of
Others-Total, low number of Other people in pictures, and low number of Check-ins). On the
other hand, adolescent that are socially rejected publish more postings of their own (i.e., more
Owner-created and Owner-shared posts, the complementary side of 'Others-Total') and their
communication behavior expresses positive attitudes towards others. Finally, socially rejected

23 The Digital Footprints of Adolescent Distress
adolescents endorse distinct online preferences and participate in marginal Facebook groups and
pages with Gothic and dark content.
Similar to social rejection, the online picture of adolescents who are victims of bullying may
reflect the 'offline' social status of the user (i.e., low number of Other people in pictures) and may
be detected through their interest in Gothic and dark content. Moreover, adolescent who
experience acts of bullying were found to be more active on Facebook (i.e., high number of
'Your posts' and high frequency of the primary content category 'Fun and news'). Finally, they
tend to publish posts in which they refer to themselves (First person single pronouns) more often.
Further theoretical implications and limitations are discussed in the general discussion section.
General discussion
The current research investigated whether and how data logged and stored on online social
network sites can be used as a tool to detect adolescents' psychological distress. We conducted
two consecutive studies to address three research hypotheses according to which: (1) online
explicit references to distress is predictive of adolescents' depression; (2) personal distress might
appear indirectly, in other, non-explicit, online behaviors; and (3) psychosocial distress can be
detected from Facebook activities. Whereas Study 1 focused on explicit references to distress, in
Study 2 we searched for less explicit signs of distress as well. In both studies, features of
adolescents' Facebook activity was compared with their self-reported answers to well-established
measures of depression and of psychosocial distress (i.e., social rejection and victimization of
bullying). An integration and discussion of the main findings from both studies is presented
below, separately for depression and for psychosocial distress.
Depression. A small number of postings in Study 1 contained explicit references to personal
distress and mainly to depressive symptoms. These explicit postings were published by 12% of

24 The Digital Footprints of Adolescent Distress
the sample. The results of Study 1 confirmed hypothesis #1 and indicated that explicit references
to distress do predict higher levels of depression among adolescents. Specifically, seven out of
the ten (70%) participants who published posts with explicit references to depressive symptoms
experienced at least mild depression. Study 2 documented only four participants who published
explicit references to distress, three of whom indeed received high depression scores. Based on
these combined findings, we conclude that explicit postings of referring to depression are rare,
but when they do appear on adolescents' Facebook accounts, they are very informative and
substantively increase the chances that the adolescent is indeed experiencing depression. This
finding extends previous research that examined the validity of explicit references to depression
in adult samples (Holleran, 2010; Moreno et al., 2012) to the adolescent population .
However, our results also qualify previous research by warning against over-reliance on
explicit references to depression. Over-reliance on explicit references may result in false
negative errors in identification (type II error), whereby the test (i.e., online explicit references to
depression) fails to detect an existing effect (i.e., depressive symptoms). In both studies, the
majority of depressed adolescents did not publish explicit references to any type of personal
distress. In the absence of explicit references to depression, adolescent depression may be not so
easy to detect via online activities. The analyses presented in Study 2 failed to find any
additional differences in Facebook activity behaviors that could distinguish between depressive
and non-depressive adolescents. It is possible that our failure to detect Facebook features that
associate with depression is a result of the relatively small sample of adolescents who were
depressed at the time of the study. Moreover, it appears that adolescents publish significantly less
(verbal) content than adults' users of SNS. The overall number of postings of adolescents in three
months documented in Study 2 was 11, of which only four posts approximately (39%) contained

25 The Digital Footprints of Adolescent Distress
truly personal content ('Owner created'). In comparison, previous research on adult Twitter users
documented close to1,133 Twitter posts in a three months period on average (De Choudhury et
al., 2013). More research is needed, preferably with larger samples, data sets that include more
SNS activity features and computerized language processing techniques to extract significant
online (non-direct) signs for depression from a larger set of potential Facebook features.
Longitudinal research is also recommended to trace changes in online Facebook activity, which
can be considered a leisure activity, to detect decrease in pleasure or interest over time.
Psychosocial distress. In contrast to depression, social rejection and victimization of
bullying were not significantly associated with explicit references to any type of distress.
Interestingly however, the findings presented in Study 2 revealed a number of less explicit
Facebook activity features that predicted social rejection and victimization of bullying. Study 2,
which included a larger sample and a richer data set, lend initial support to hypothesis #2 and #3
by demonstrating how experiences of social rejection and victimization of bullying may be
expressed in subtler and indirect Facebook behaviors of adolescents. Discriminant analyses
suggested several Facebook features that can serve as indicators for the existence of social
rejection or victimization of bullying.
Our findings contribute to the growing body of research that aims to predict mental health
states from social media. Existing research in this now field, has predominantly focused on more
defined DSM-based constructs of mental disorders, such as depression (De Choudhury et al.,
2013) or post-traumatic stress disorder (e.g., Reece et al., 2016). To the best of our knowledge,
the current research is the first to show that sub-clinical negative psychosocial distress (i.e.,
social rejection and victimization of bullying) can be detected through (non-explicit) features of
SNS behaviors. These digital footprints of psychosocial distress are especially valuable

26 The Digital Footprints of Adolescent Distress
considering the low frequency of explicit references to any type of distress, which was
documented in both studies and the importance of social relationships in adolescence.
From a more theoretical point of view, the findings from Study 2 may also advance the
dialogue between two competing theories on the link between internet usage and social status
and skills (Kraut et al., 2002): The rich-get-richer model posits that extroverted, socially adept
individuals are more likely to benefit from online communication, than introverts. They form
valuable online relationships and their "real-life" social skills are replicated online (Kraut et al.,
2002). The social compensation model argues to the contrary that introverted people with less
social skills benefit more from Internet communication because they are now granted with an
exceptional opportunity to engage in social interactions in less threatening environments
(Sivashanker, 2013). The emergence of the internet also allowed introverts to reveal hidden
facets of their "true self" (Amichai-Hamburger, Wainapel, & Fox, 2002; Bargh, McKenna, &
Fitzsimons, 2002; McKenna, Green, & Gleason, 2002), their inner feelings, hidden fantasies, and
"real" identity (Winnicott, 1960). This theoretical dispute continues to this day and the available
empirical research does not provide conclusive evidence to either hypothesis (Baker & Oswald,
2010; Lee, 2009).
Our findings contribute to this theoretical discussion and suggest that the relationships
between social status and Internet usage are more complex than can be explained simply by
either hypothesis. We did not find a simple relationship between social status and overall number
of Facebook postings (as would be predicted by both theoretical hypotheses). The relationships
that were found addressed different types of online postings (i.e., 'Your posts' versus 'Others-
Total'), thus corroborating with both theoretical hypotheses and perhaps offering a theoretical
solution for this dispute. Even if adolescents who experience social rejection are more likely to

27 The Digital Footprints of Adolescent Distress
actively use social media as a less threatening social substitute (social compensation hypothesis),
as documented in their tendency to publish more postings of their own, and even if they are using
internet communication as an opportunity to reveal hidden facets of their "true self" (Amichai-
Hamburger et al., 2002) by joining marginal online groups with gothic and dark content, their
efforts of communication do not always bear fruits. Despite their wishes to use social media as a
means to develop social relations (as documented in the higher scores of Positive attitudes
towards others), their social difficulties continue and extend into their online relationships, as
predicted by the rich-get-richer hypothesis, and they do not manage to attract the social support
they may have hoped for. They receive fewer tags and fewer postings by others, their pictures
include fewer friends, and they hesitate to announce their physical presence using the Check-in
option.
This integrative perception of these two allegedly competing hypotheses is repeated in
victimization of bullying. Victimization showed a similar pattern to social rejection with fewer
friends in pictures and more ‘likes’ of gothic and dark content. Additionally, victims of bullying
may prefer the SNS sphere, which only includes those individuals that they have allowed in. This
is reflected in their excessive usage of Facebook activities (high number of 'Owner-created' and
'Owner-shared' postings as well as their high frequency of Likes and Group in the primary
content category of 'Fun and news'). It is also possible that they feel more comfortable to discuss
their own affairs online as is manifested in their increased usage of First person single pronouns.
In this way, the current research suggests that the two theoretical hypotheses do not necessarily
exclude each other. Adolescents who struggle with social situations may indeed try to
compensate for their underdeveloped social skills with increased usage of SNS, yet at the same
time their efforts may not yield the social benefits they expected.

28 The Digital Footprints of Adolescent Distress
Limitations and future research
In addition to the above-mentioned contributions of the present work, we would also like to
point out several limitations that have to be taken into account and should direct future research
efforts. First, the reliance on self-report surveys rather than on clinical interviews to assess
psychological and psychosocial distress, though validated and well-established, limits our ability
to establish the participants' full mental health status. We made significant efforts to identify
uncompleted questionnaires or negligent respondents, but it is possible that some adolescents
were not devoted to the research and chose to be less straightforward.
Second, even though the present methodology allows for the establishment of predictions and
associations between variables, it does not allow us to determine the direction of the
relationships between them. It is commonly assumed that personal distress is manifested or
implicitly reflected in social media usages, but the present results cannot rule out the opposite,
namely that social media usages create or contribute to personal feelings of distress (e.g., that
posting excessively or engaging in Dark and Gothic online content leads to social rejection or
victimization in the real world). Third, some of the coding categories of the verbal content relied
on high-inference, human coding. Though good inter-rater reliabilities were documented in most
categories included in the final analyses, the coding process is very time-consuming and can
address small samples only. Future studies may leverage automated, unsupervised data mining
and natural language processing methods to examine larger data sets and to include a larger set
of Facebook features as potential predictors of psychological distress.
In conclusion
The take-home message of the research presented here can be concluded as follows: (a)
adolescent postings of explicit distress in Facebook are rare; (b) when they do appear, they

29 The Digital Footprints of Adolescent Distress
usually include references to depressive symptoms and they are indeed predictive of depression
among adolescents; and (c) psychosocial distress (i.e., negative interpersonal experiences of
social rejection and victimization of bullying) leaves less explicit, digital footprints that can be
discerned and detected. We contend that these results should be seen as first illustrations for the
new opportunities generated with the rise of social media. Further research is required,
preferably in larger samples and with automated language processing tools to prove the proposed
direction of the research and reveal a more comprehensive picture of how distress is expressed in
online behavior. Scholars that have already used such computerized tools to predict personal
features such as age, gender, or introversion (Schwartz et al., 2013) are predicting that early
screening of mental health conditions from social media activity logs is right around the corner
(Csepeli & Nagyfi, 2017; Park et al., 2014). The findings presented here show this seems a
viable and realistic expectation, and the next step in this line of research.

30 The Digital Footprints of Adolescent Distress
Bibliography
Asterhan, C. S. C. & Rosenberg, H. (2015). The promise, reality and dilemmas of secondary
school teacher-student interactions in Facebook: The teacher perspective. Computers &
Education, 85, 134–148.
Achenbach , T. M. (1991). Manual for the Child Behaviour Checklist/4-18 and 1991 Profile. In:
Burlington, VT: University of Vermont, Department of Psychiatry.
American Psychiatric, A. (2013). Diagnostic and Statistical Manual of Mental Disorders (DSM-
5®): American Psychiatric Pub.
Amichai-Hamburger, Y., Wainapel, G., & Fox, S. (2002). "On the Internet No One Knows I'm
an Introvert": Extroversion, Neuroticism, and Internet Interaction. CyberPsychology &
Behavior, 5(2), 125-128. doi:10.1089/109493102753770507
Baker, L. R., & Oswald, D. L. (2010). Shyness and online social networking services. Journal of
Social and Personal Relationships, 27(7), 873-889. doi:10.1177/0265407510375261
Bargh, J. A., McKenna, K. Y. A., & Fitzsimons, G. M. (2002). Can you see the real me?
Activation and expression of the "true self" on the Internet. Journal of Social Issues,
58(1), 33-48. doi:10.1111/1540-4560.00247
Bazarova, N. N., Choi, Y. H., Schwanda Sosik, V., Cosley, D., & Whitlock, J. (2015). Social
sharing of emotions on Facebook: Channel differences, satisfaction, and replies.
Beck, A. T. (1974). The development of depression: A cognitive model.
Beck, A. T., Steer, R. A., Ball, R., & Ranieri, W. F. (1996). Comparison of Beck Depression
Inventories–IA and –II in psychiatric outpatients. Journal of Personality Assessment,
67(3), 588-597. doi:10.1207/s15327752jpa6703_13
Brown, B. B., & Larson, J. (2009). Peer relationships in adolescence. In R. M. L. L. Steinberg
(Ed.), Handbook of adolescent psychology, Vol 2: Contextual influences on adolescent
development (3rd ed.) (pp. 74-103). Hoboken, NJ, US: John Wiley & Sons Inc.
Burke, M., & Kraut, R. E. (2016). The relationship between Facebook use and well‐being
depends on communication type and tie strength. Journal of Computer‐Mediated
Communication, 21(4), 265-281.
Cavazos-Rehg, P. A., Krauss, M. J., Sowles, S. J., Connolly, S., Rosas, C., Bharadwaj, M., . . .
Bierut, L. J. (2017). An analysis of depression, self-harm, and suicidal ideation content
on tumblr. Crisis.
Christofides, E., Muise, A., & Desmarais, S. (2009). Information Disclosure and Control on
Facebook: Are They Two Sides of the Same Coin or Two Different Processes?
Cyberpsychology & Behavior, 12(3), 341-345. doi:10.1089/cpb.2008.0226
Coppersmith, G., Dredze, M., Harman, C., Hollingshead, K., & Mitchell, M. (2015). CLPsych
2015 Shared Task: Depression and PTSD on Twitter.
Coppersmith, G., Dredze, M., Harman, C., Hollingshead, K., & Mitchell, M. (2015, 2015).
CLPsych 2015 Shared Task: Depression and PTSD on Twitter.
Csepeli, G., & Nagyfi, R. (2017). Facebook-Diagnostics: Detection of Mental Hygiene Problems
Based on Online Traces. In Leisure, Health and Well-Being (pp. 171-178): Springer.

31 The Digital Footprints of Adolescent Distress
De Choudhury, M., Gamon, M., Counts, S., & Horvitz, E. (2013). Predicting Depression via
Social Media. ICWSM, 13, 1-10.
Hargittai, E., & Litt, E. (2011). The tweet smell of celebrity success: Explaining variation in
Twitter adoption among a diverse group of young adults. New media & society, 13(5),
824-842.
Hew, K. F. (2011). Students’ and teachers’ use of Facebook. Computers in Human Behavior,
27(2), 662-676. doi:10.1016/j.chb.2010.11.020
Hew, K. F., & Cheung, W. S. (2012). Use of Facebook: A case study of Singapore students'
experience. Asia Pacific Journal of Education, 32(2), 181-196.
doi:10.1080/02188791.2012.685560
Hobbs, W. R., Burke, M., Christakis, N. A., & Fowler, J. H. (2016). Online social integration is
associated with reduced mortality risk. Proceedings of the National Academy of Sciences,
113(46), 12980-12984.
Holleran, S. E. (2010). The early detection of depression from social networking sites. (71),
ProQuest Information & Learning, US.
Kim, Y. S., & Leventhal, B. (2008). Bullying and suicide. A review. International Journal of
Adolescent Medicine and Health, 20(2), 133-154. doi:10.1515/IJAMH.2008.20.2.133
Kosinski, M., Stillwell, D., & Graepel, T. (2013). Private traits and attributes are predictable
from digital records of human behavior. Proceedings of the National Academy of
Sciences, 110(15), 5802-5805.
Kraut, R., Kiesler, S., Boneva, B., Cummings, J. N., Helgeson, V., & Crawford, A. M. (2002).
Internet paradox revisited (Version 1) [doi:10.1111/1540-4560.00248]. United Kingdom:
Blackwell Publishing.
Lambiotte, R., & Kosinski, M. (2014). Tracking the digital footprints of personality. Proceedings
of the IEEE, 102(12), 1934-1939.
Lee, S. J. (2009). Online communication and adolescent social ties: Who benefits more from
Internet use? (Version 3) [doi:10.1111/j.1083-6101.2009.01451.x]. United Kingdom:
Wiley-Blackwell Publishing Ltd.
Lenhart, A., Purcell, K., Smith, A., & Zickur, K. (2010). Social Media and Young Adults. In:
Washington, DC: Pew Research Center. Available at:
http://pewinternet.org/Reports/2010/Social-Media-and-Young-Adults.aspx. Accessed
May 16, 2013.
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.
Developmental Psychology, 48(2), 369-380. doi:10.1037/a0026338
McDougall, P., Hymel, S., Vaillancourt, T., & Mercer, L. (2001). The consequences of
childhood peer rejection. In Interpersonal rejection (pp. 213-247). New York, NY, US:
Oxford University Press.
McKenna, K. Y. A., Green, A. S., & Gleason, M. E. J. (2002). Relationship formation on the
Internet: What's the big attraction? Journal of Social Issues, 58(1), 9-31.

32 The Digital Footprints of Adolescent Distress
doi:10.1111/1540-4560.00246
Moreno, M. A., Christakis, D. A., Egan, K. G., Jelenchick, L. A., Cox, E., Young, H., . . .
Becker, T. (2012). A Pilot Evaluation of Associations Between Displayed Depression
References on Facebook and Self-reported Depression Using a Clinical Scale. Journal of
Behavioral Health Services & Research, 39(3), 295-304. doi:10.1007/s11414-011-9258-7
Moreno, M. A., Jelenchick, L. A., Egan, K. G., Cox, E., Young, H., Gannon, K. E., & Becker, T.
(2011). Feeling bad on Facebook: depression disclosures by college students on a social
networking site. Depression and Anxiety, 28(6), 447-455. doi:10.1002/da.20805
Nolen-Hoeksema, S., Wisco, B. E., & Lyubomirsky, S. (2008). Rethinking rumination.
Perspectives on Psychological Science, 3(5), 400-424. doi:10.1111/j.1745-
6924.2008.00088.x
Obar, J. A., & Wildman, S. S. (2015). Social media definition and the governance challenge: An
introduction to the special issue.
Ophir, Y. (2017). SOS on SNS: Adolescent distress on social network sites. Computers in
Human Behavior, 68, 51-55.
Ophir, Y., Asterhan, C. S. C., & Schwarz, B. B. (2017). Unfolding the notes from the walls:
Adolescents’ depression manifestations on Facebook. Computers in Human Behavior, 72,
96-107. doi:http://dx.doi.org/10.1016/j.chb.2017.02.013
Park, G., Schwartz, H. A., Eichstaedt, J. C., Kern, M. L., Kosinski, M., Stillwell, D. J., . . .
Seligman, M. E. P. (2014). Automatic Personality Assessment Through Social Media
Language. Journal of Personality and Social Psychology, No Pagination Specified.
doi:10.1037/pspp0000020
Park, S., Lee, S. W., Kwak, J., Cha, M., & Jeong, B. (2013). Activities on Facebook reveal the
depressive state of users. Journal of medical Internet research, 15(10).
Pennebaker, J. W., & King, L. A. (1999). Linguistic styles: Language use as an individual
difference. Journal of Personality and Social Psychology, 77(6), 1296-1312.
doi:10.1037/0022-3514.77.6.1296
Phua, J., Jin, S. V., & Kim, J. (2017). Uses and gratifications of social networking sites for
bridging and bonding social capital: A comparison of Facebook, Twitter, Instagram, and
Snapchat. Computers in Human Behavior, 72(Supplement C), 115-122.
doi:https://doi.org/10.1016/j.chb.2017.02.041
Pickett, C. L., Gardner, W. L., & Knowles, M. (2004). Getting a Cue: The Need to Belong and
Enhanced Sensitivity to Social Cues. Personality and Social Psychology Bulletin, 30(9),
1095-1107. doi:10.1177/0146167203262085
Reece, A. G., Reagan, A. J., Lix, K. L. M., Dodds, P. S., Danforth, C. M., & Langer, E. J.
(2016). Forecasting the onset and course of mental illness with Twitter data. arXiv
preprint arXiv:1608.07740.
Rey, J. M., & Bird, K. D. (1991). Sex differences in suicidal behaviour of referred adolescents.
The British Journal of Psychiatry, 158, 776-781. doi:10.1192/bjp.158.6.776
Rigby, K. (1998). Manual for the Peer Relations Questionnaire (PRQ). Point Lonsdale, Victoria,

33 The Digital Footprints of Adolescent Distress
Australia: The Professional Reading Guide.
Rosenberg, H., Ophir, Y., & Asterhan, C. S. C. (2018). Virtual safe zone: Teacher psycho-
pedagogical support of teenage students through social media in times of war. Teaching
and Teacher Education, 73, 35-42.
Rude, S. S., Gortner, E.-M., & Pennebaker, J. W. (2004). Language use of depressed and
depression-vulnerable college students [Press release]
Saleem, S., Prasad, R., Vitaladevuni, S. N. P., Pacula, M., Crystal, M., Marx, B., . . . Speroff, T.
(2012, 2012). Automatic Detection of Psychological Distress Indicators and Severity
Assessment from Online Forum Posts.
Schwartz, H. A., Eichstaedt, J. C., Kern, M. L., Dziurzynski, L., Ramones, S. M., Agrawal, M., .
. . Seligman, M. E. (2013). Personality, gender, and age in the language of social media:
The open-vocabulary approach. PloS one, 8(9), e73791.
Sivashanker, K. (2013). Cyberbullying and the digital self. Journal of the American Academy of
Child & Adolescent Psychiatry, 52(2), 113-115. doi:10.1016/j.jaac.2012.11.008
Sosik, V. S., & Bazarova, N. N. (2014). Relational maintenance on social network sites: How
Facebook communication predicts relational escalation. Computers in Human Behavior,
35, 124-131.
Tsugawa, S., Kikuchi, Y., Kishino, F., Nakajima, K., Itoh, Y., & Ohsaki, H. (2015). Recognizing
depression from twitter activity.
Tsugawa, S., Kikuchi, Y., Kishino, F., Nakajima, K., Itoh, Y., & Ohsaki, H. (2015, 2015).
Recognizing depression from twitter activity.
Varnik, P. (2012). Suicide in the World. International Journal of Environmental Research and
Public Health, 9(3), 760-771. doi:10.3390/ijerph9030760
Velting, D. M., Shaffer, D., Gould, M. S., Garfinkel, R., Fisher, P., & Davies, M. (1998). Parent–
victim agreement in adolescent suicide research. Journal of the American Academy of
Child & Adolescent Psychiatry, 37(11), 1161-1166. doi:10.1097/00004583-199811000-
00015
Wang, J., Iannotti, R. J., & Nansel, T. R. (2009). School bullying among adolescents in the
United States: Physical, verbal, relational, and cyber. Journal of Adolescent health, 45(4),
368-375.

34 The Digital Footprints of Adolescent Distress
Table 1
Descriptive statistics and zero order correlations of the self-report measures
Pearson correlation with
M (SD)
Social rejection
Victimization of bullying
Study 1 (N = 86)
Social Rejection
1.75 (.78)
Victimization of bullying
1.51 (.54)
.52**
Depression (BDI)
.55 (.48)
.54**
.40**
Study 2 (N =
162)
Social Rejection
1.59 (.66)
Victimization of bullying
1.56 (.60)
.51**
Depression (BDI)
.52 (.45)
.54**
.48**
Note: ** = Correlation is significant at the 0.01 level (2-tailed). * = Correlation is significant at the 0.05 level (2-tailed).
Table 2
Descriptive statistics of the 'About' section quantitative features and their correlations with Self-reported
distress (N = 162)
Pearson correlation with
M (SD)
Social
rejection
Victimization
of bullying
Depression
Friends
534.26 (536.91)
.05
.13
.14
Check- In's
14.09 (16.44)
-.17*
-.13
-.16*
Sports
9.93 (25.29)
-.15~
.02
-.08
Music
35.44 (93.92)
-.03
-.01
-.06
Movies
12.75 (26.30)
.08
.03
.01
TV
18.42 (26.38)
-.01
.07
-.08
Books
2.52 (7.18)
.11
.12
.05
Apps
3.47 (6.78)
.04
-.03
-.16*
Likes
384.89 (558.18)
-.07
.05
-.05
Groups
8.08 (29.21)
-.01
-.05
-.09
Note: ** = Correlation is significant at the 0.01 level. * = Correlation is significant at the 0.05 level. ~ = Correlation approached
significance (p = .06).

35 The Digital Footprints of Adolescent Distress
Table 3
Descriptive statistics of the 'About' section Content categories and their correlations with Self-reported
distress (N = 162)
Pearson correlation with
M (SD) in %
Social
Rejection
Victimization
of Bullying
Depression
Fun & news
65 (16.67)
.11
.16*
.08
Commercials
18.92 (13.38)
-.10
-.14
-.03
Belonging
7.57 (7.77)
-.08
-.11
-.08
Values & social engagement
7.06 (7.42)
-.09
-.08
-.09
Gothic & dark
0.70 (2.48)
.29***
.24**
.11
Extreme & offensive
0.74 (1.64)
.01
.02
.03
Note: Correlations are significant at a two-tailed p-level of * > .05, ** > .01 level or *** > .001.
Table 4
Descriptive statistics of the 'Timeline' Types of posts and correlations with Self-reported distress (N =
162)
Person correlation with
M (SD)
Social
Rejection
Victimization of
Bullying
Depression
Overall Postings
10.62 (15.94)
.17*
.11
.04
Owner-Created
38.80% (28.04)
.10
.00
.05
Owner-Shared
16.84% (25.67)
.13
.09
.01
Others-Tagged
28.41% (26.60)
-.24**
-.12
-.05
Others-Created
12.19% (19.88)
.01
.06
-.06
Others-Total:
Tagged & Created
40.61% (30.01)
-.21**
-.07
-.08
Note: * = Correlation is significant at the 0.05 level. ** = Correlation is significant at the 0.01 level. Overall Postings = The
overall number of posts, including all types of posts. Owner-Created = posts, which were created and published by the participant
her/himself. Owner-Shared = posts (usually, created by others), which were shared by the participant her/himself. Others-Tagged
= posts (usually, created by others), in which other people "tagged" (attached) the participant name. Others-Created = post, which
were attached by other people to the participant timeline.
Table 5
Descriptive statistics of the 'Timeline' posting characteristics and their correlations with self-reported
distress (N = 162)
Pearson correlation with
Range
M (SD)
Social
rejection
Victimization
of Bullying
Depression
Likes on posts
Sum
212.40 (503.64)
.08
.10
.14
Comments on posts
Sum
26.89 (60.94)
.01
.00
.09
First person single
Sum
6.24 (13.86)
.11
.24**
.08
First person plural
Sum
1.49 (5.44)
.07
.06
.06
Other people in pictures
Sum (1-4)
17.85 (22.55)
-.21*
-.18*
-.04
Valence of content
Sum (Positive –
Negative)
2.44 (4.05)
.13
.12
.05
Attitudes towards others
Sum (Positive –
Negative)
1.06 (2.27)
.18*
.14
.12
Note: * = Correlation is significant at the 0.05 level. ** = Correlation is significant at the 0.01 level. The Correlation with 'Other

36 The Digital Footprints of Adolescent Distress
people in pictures' was calculated on 131 participants who had at least one picture of themselves in their timeline. Sum (Positive
– Negative) = a total, combined score, which subtracted the number of negative (valance/attitudes) posts from the number of
positive (valance/attitudes) posts.