Journal of Anxiety Disorders 26 (2012) 197205

Contents lists available at SciVerse ScienceDirect

Journal of Anxiety Disorders

Seeking safety on the internet: Relationship between social anxiety and problematic internet use
Bianca W. Lee a , Lexine A. Stapinski b,
a b

Department of Psychology, Macquarie University, Sydney, NSW 2109, Australia Centre for Emotional Health, Macquarie University, Sydney, NSW 2109, Australia

a r t i c l e

i n f o

a b s t r a c t
As internet use becomes increasingly integral to modern life, the hazards of excessive use are also becoming apparent. Prior research suggests that socially anxious individuals are particularly susceptible to problematic internet use. This vulnerability may relate to the perception of online communication as a safer means of interacting, due to greater control over self-presentation, decreased risk of negative evaluation, and improved relationship quality. To investigate these hypotheses, a general sample of 338 completed an online survey. Social anxiety was conrmed as a signicant predictor of problematic internet use when controlling for depression and general anxiety. Social anxiety was associated with perceptions of greater control and decreased risk of negative evaluation when communicating online, however perceived relationship quality did not differ. Negative expectations during face-to-face interactions partially accounted for the relationship between social anxiety and problematic internet use. There was also preliminary evidence that preference for online communication exacerbates face-to-face avoidance. 2011 Elsevier Ltd. All rights reserved.

Article history: Received 27 April 2011 Received in revised form 13 September 2011 Accepted 6 November 2011 Keywords: Social anxiety Social phobia Internet Addiction Threat Social avoidance

1. Introduction 1.1. Problematic internet use In the last decade there has been an increase in the number of studies investigating how the internet inuences the lives of those who use it (Amichai-Hamburger & Furnham, 2007; Chak & Leung, 2004; McKenna & Bargh, 2000; Shaw & Gant, 2002). At the beginning of 2010, over 1.9 billion individuals were using the internet, of whom 613% are thought to be addicted to its use (ETforecasts, 2010; Morahan-Martin, 2001). Although researchers continue to debate the exact nature and denition of internet addiction (Morahan-Martin, 2005; Young, 1996), the most widely accepted is Davis (2001) cognitive-behavioral model of problematic internet use. According to this model, using the internet to regulate unpleasant moods, becoming attached to the social benets the internet provides, and perceiving more interpersonal control online than ofine, leads to excessive use and the development of compulsions, withdrawal symptoms and negative social, psychological, and/or occupational consequences in the users ofine life (Caplan, 2002; Davis, 2001). Collectively, these factors

Corresponding author at: Department of Psychology, Macquarie University, Sydney, NSW 2109, Australia. Tel.: +61 2 9850 1801. E-mail address: lexine.stapinski@mq.edu.au (L.A. Stapinski). 0887-6185/$ see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.janxdis.2011.11.001

are thought to be symptomatic of problematic internet use (Caplan, 2002). One of the primary functions of the internet is online communication, with almost 10 million Australians using the internet to communicate with family and friends (The Nielsen Company, 2010). Online communication has been shown to have a positive impact on users as a convenient way to maintain and improve current relationships, which may enhance the self-esteem and well-being of those involved (Gross, Juvonen, & Gable, 2002; Selfhout, Branje, Delsing, ter Bogt, & Meeus, 2009). However, if a network of online relationships or a single online partnership becomes all-consuming, then such relationships can negatively impact their ofine lives on a social and/or occupational level (Campbell, Cumming, & Hughes, 2006; Parks & Roberts, 1998). Therefore, it is important to determine which factors predispose individuals to less adaptive use of the internet. Researchers investigating this issue have found that those who partake in instant messaging and online gaming are more likely to be problematic users compared to those who used the internet for other purposes (Caplan, Williams, & Yee, 2009; Chak & Leung, 2004). However, Caplan et al. (2009) found that while only 2% of the variance in problematic internet use was explained by online activities, 36% of the variance was accounted for by the individuals psychological prole. There are several psychological vulnerabilities associated with problematic internet use, including loneliness, depression, substance addictions, shyness, and aggression (Caplan,

198

B.W. Lee, L.A. Stapinski / Journal of Anxiety Disorders 26 (2012) 197205

2007; Caplan et al., 2009; Chak & Leung, 2004; Ebeling-Witte, Frank, & Lester, 2007; Scealy, Phillips, & Stevenson, 2002; Selfhout et al., 2009). It is thought that these psychological factors predispose individuals to social isolation in their ofine lives, so they seek to fulll their interpersonal needs online, and are thus vulnerable to problematic internet use (McKenna & Bargh, 2000). Consistent with this idea, Davis model (2001) proposes that problematic internet use is a consequence of a pre-existing psychopathology, which arises when the internet provides rewards, such as developing better quality relationships online (AmichaiHamburger & Furnham, 2007; McKenna, Green, & Gleason, 2002). Users may come to perceive their online lives as better, both socially and psychologically, and account for this by assuming that their social abilities are better during online compared to face-toface interactions. Once problematic internet use has developed, it may be maintained by a negative feedback loop, where attempts to reduce use produce stressful physiological responses, such as compulsive or withdrawal symptoms (Caplan, 2002; Davis, 2001). Davis model provides a way of understanding the relationship between psychological problems, such as depression and social anxiety, and problematic internet use. 1.2. The relationship between social anxiety and problematic internet use Social anxiety is extremely debilitating due to the negative impact on social networks, and consequent isolation (Gross et al., 2002; La Greca & Lopez, 1998). The disorder is characterized by inated threat expectancies in social-evaluative contexts, and corresponding avoidance of these situations (Rapee & Heimberg, 1997). Avoidance also takes the form of subtle safety behaviors, such as over preparation or speaking quickly (Clark & Wells, 1995). While these behaviors function to temporarily reduce anxiety, social successes such as positive feedback may be attributed to the safety behavior rather than personal ability. Consequently, safety behaviors prevent these individuals from learning that they are overestimating the likelihood of negative evaluation and underestimating their social ability (Alden & Bieling, 1998; Rapee & Heimberg, 1997). Communicating online may be one such safety behavior that allows those with social anxiety to communicate with others while minimizing potential threat and associated anxiety (Erwin et al., 2004; Shepherd & Edelmann, 2005). A growing body of evidence has identied a positive correlation between social anxiety and problematic internet use (Campbell et al., 2006; Caplan, 2007; Chak & Leung, 2004; Ebeling-Witte et al., 2007; Erwin, Turk, Heimberg, Fresco, & Hantula, 2004). Social anxiety has been conceptualized as a risk factor for the development of problematic internet use (Caplan, 2007; Erwin et al., 2004; McKenna & Bargh, 2000), however, few studies have examined the factors that may account for this relationship. One explanation for this relationship may relate to differences in the expectation of negative evaluation from an online compared to ofine audience. Researchers have speculated that the text-based nature of the internet, and the lack of visual cues when communicating online, allows those with social anxiety to conceal, and therefore control, the aspects of their appearance they perceive as leading to negative evaluation, such as sweating and stammering (Alden & Bieling, 1998; Caplan, 2007; Erwin et al., 2004; McKenna & Bargh, 2000). While the internet may be benecial in terms of facilitating engagement with previously avoided social interactions and activities, preference for online communication may lead to the development of problematic internet use (Caplan, 2003; McKenna & Bargh, 2000; Scealy et al., 2002). Preliminary support for this hypothesis comes from recent studies showing that both social shyness and preference for online communication were positively correlated with problematic internet use (Caplan, 2003; Ebeling-Witte et al., 2007).

Based on similar reasoning, some researchers have proposed that online interactions are perceived as safer, in terms of the probability and consequences of negative evaluation, than face-to-face interactions (Amichai-Hamburger & Furnham, 2007; Caplan, 2002; King & Poulos, 1998; McKenna & Bargh, 2000). Consistent with this idea, socially anxious samples report experiencing greater ease interacting on the internet as compared to face-to-face (Erwin et al., 2004). While online communication appears to reduce or regulate social anxiety in the short term (Campbell et al., 2006; Shepherd & Edelmann, 2005), in the longer term condence to communicate with others beyond the online context may be undermined if successful online interactions are attributed to the unique aspects of the internet, such as anonymity, rather than personal attributes (Clark & Wells, 1995; Erwin et al., 2004; McKenna et al., 2002). This may result in a cycle of avoidance that is perpetuated by online communication, and ultimately leads to the development of problematic internet use and the maintenance of interpersonal anxiety (Andersson, 2009; Erwin et al., 2004; Sheeks & Birchmeier, 2007; Shepherd & Edelmann, 2005). Although this hypothesis has not yet been tested directly, preliminary support is derived from investigations of the motives for internet use amongst socially anxious individuals. These participants reported a reliance on the internet as a social outlet that enabled them to avoid face-to-face interactions (Erwin et al., 2004) and other unpleasant situations (Ebeling-Witte et al., 2007; Whang, Lee, & Chang, 2003). This tendency to avoid face-to-face interactions may preclude the benecial effect of learning, through experience, that their fear of negative evaluation is exaggerated (Rapee & Heimberg, 1997). Alternative explanations for the relationship between social anxiety and problematic internet use have focused on potential differences in the quality of online compared to ofine relationships. Physical features, such as attractiveness, have been found to determine how much individuals like one another in face-to-face settings (McKenna et al., 2002). Consequently, signs of visible anxiety may be perceived by others as unattractive, which could dictate how successful those with social anxiety will be at forming face-toface relationships (McKenna & Bargh, 2000; McKenna et al., 2002). Given that concerns about physical appearance are reduced online, socially anxious individuals may be able to develop more meaningful friendships (Bargh, McKenna, & Fitzsimons, 2002; McKenna & Bargh, 2000). Along similar lines, the social compensation hypothesis suggests that socially anxious individuals compensate for poor ofine relationships by seeking attachments in an online environment, where the constraints that usually make them interact poorly are reduced (Campbell et al., 2006; Selfhout et al., 2009; Walther, 1996). Self-disclosure is thought to be the foundation of these less supercial online relationships (McKenna & Bargh, 2000; Parks & Floyd, 1996; Valkenburg & Peter, 2007). Although disclosure can lead to risks such as social rejection and emotional vulnerability (Pennebaker, 1989), the anonymous nature of online interactions is thought to minimize these risks (Amichai-Hamburger & Furnham, 2007; Gross et al., 2002; McKenna & Bargh, 2000). The Real Me model of problematic internet use proposes that some individuals have difculty portraying their true self (p. 34) in their face-toface relationships due to fear of rejection (Bargh et al., 2002). This fear causes them to seek anonymous online relationships, where they can more easily be themselves. This model has been successfully applied to socially anxious participants, who reported they were more likely to portray their true self online (McKenna et al., 2002). One possible explanation for this nding is that the intense fear of negative evaluation experienced by those with social anxiety inhibits self-disclosure in face-to-face relationships. Thus, the comparably less threatening online context may provide the opportunity to fulll the social need to belong and to be understood (Baumeister & Leary, 1995; McKenna & Bargh, 2000). However, the

B.W. Lee, L.A. Stapinski / Journal of Anxiety Disorders 26 (2012) 197205

199

perception of better relationship quality online compared to ofine may produce a reliance on internet-based relationships and ultimately, problematic internet use (Bargh et al., 2002; Erwin et al., 2004; King & Poulos, 1998; McKenna et al., 2002; Peters & Malesky, 2008; Valkenburg & Peter, 2007). The perceived quality of online and ofine relationships will be examined further in the current paper. 1.3. The current study While a number of promising explanations have been proposed to account for the relationship between social anxiety and problematic internet use, few studies have examined these hypotheses empirically. This study administered an online survey to address several as yet unanswered questions. Firstly, this study assessed the relationship between symptoms of social anxiety and problematic internet use while controlling for comorbid depression, general anxiety and stress, which have been implicated as predictors in previous research (Caplan et al., 2009; Selfhout et al., 2009; Shaw & Gant, 2002). Secondly, the study explored whether perceptions of relationship quality partially account for the association between social anxiety and problematic internet use. Considering the work of McKenna and colleagues (McKenna et al., 2002), it was predicted that individuals with social anxiety may be susceptible to problematic internet use due to a perception of greater interdependence, commitment, predictability, breadth and depth of conversation in online as compared to ofine relationships. Thirdly, this study examined whether socially anxious individuals perceive greater control over their self-presentation when communicating online and perceive these interactions to be associated with decreased risk of negative evaluation. Again, it was hypothesized that these factors may partially account for the relationship between social anxiety and problematic internet use. Finally, the current study explored the conceptualization of online communication as a safety behavior that reduces discomfort but perpetuates the social avoidance associated with this disorder. In line with this conceptualization, it was predicted that (1) the relationship between social anxiety and preference for online social interactions would be partially explained by the self-reported use of safety behaviors, and (2) preference for online social interaction would be associated with avoidance of face-to-face interactions while controlling for the level of avoidance attributable to fear of negative evaluation. 2. Methods 2.1. Participants Adults living in Australia who identied that they use the internet for communicative purposes on a regular basis were invited to participate in the study. Participants were recruited via posters placed around Macquarie University, advertisements sent to the Psychology Department mailing lists, and posts on a range of online social, gaming and mental health forums. To facilitate broader distribution, recipients were asked to forward the to others who may be interested to participate. A $50 prize draw was offered as an incentive for participating. A total of 460 individuals began the online survey and of these, 345 participants completed it. Seven participants were excluded from further analyses because they were under 18 years of age, leaving a sample size of 338 with 134 men and 204 women. 2.2. Procedure The online survey was managed using the survey creator, Qualtrics. The survey was accessed via a URL link, and responses were transmitted to a secure server. All procedures were approved

by the Macquarie University Human Research Ethics Committee. Respondents completed measures assessing internet use, psychological well-being, the quality of a chosen online and ofine relationship, and their perceptions of threat during an online and ofine social interaction. In order to test the relevant hypotheses, online and ofine relationship quality was assessed by matching relationships according to their duration. Measures assessing perception of social threat online compared to ofine were counterbalanced to control for order effects. 2.3. Measures 2.3.1. Internet usage A modied version of Erwin et al. (2004) internet usage survey was used to determine the frequency and breadth of online applications used in an average week. The original survey was modied to incorporate online pornography, online gaming, internet dating, and more recently developed social applications including Facebook, Myspace, Twitter, Messenger (MSN). 2.3.2. Depression, Anxiety and Stress Scale The Depression, Anxiety and Stress Scale-21-item version (DASS-21; Lovibond & Lovibond, 1995) was used to measure symptoms of depression, anxiety, and stress. Respondents were asked to rate on a 4 point scale how much each statement applied to them in the past week ranging from 0 to 3, where higher scores reect more severe psychopathology. Good internal consistency for the depression ( = .88), anxiety ( = .82), and stress ( = .90) subscales have been demonstrated (Henry & Crawford, 2005), and comparable internal consistencies were found in the current study ( = .91, = .81 and = .87, respectively). The DASS-21 has high convergent and discriminant validity with other validated measures of depression and anxiety (Henry & Crawford, 2005). 2.3.3. Social anxiety The self-report version of the Liebowitz Social Anxiety Scale (LSAS-SR; Fresco et al., 2001) was used to measure social anxiety across two different domains: fear/anxiety and avoidance. Participants were given 24 statements describing various social situations (e.g., meeting strangers) and asked to rate their level of fear on a 4 point scale from 0 (none) to 3 (severe), and avoidance in the last week from 0 (never avoided this activity) to 3 (usually avoided this activity). The LSAS-SR has good construct validity for all subscales (r = .62.66) and total score (r = .83; Fresco et al., 2001). Good internal consistencies for all subscales ( = .73.84) and total score ( = .94) in a non clinical sample have been found (Fresco et al., 2001). Comparable internal consistency was demonstrated in the current sample for the subscales ( = .88.92) and total score ( = .97). 2.3.4. Fear of negative evaluation The Brief Fear of Negative Evaluation scale II (bFNEII) is the 8-item shortened version of the original scale used to measure fear of negative evaluation in face-to-face environments (Carleton, McCreary, Norton, & Asmundson, 2006). Participants rated the extent to which statements were characteristic of them (e.g., I am afraid others will not approve of me) on a 5 point scale ranging from 0 to 4. Carleton et al. (2006) reported high convergent validity of the bFNEII, and good internal consistency ( = .96), which was replicated in the present study ( = .96). 2.3.5. Relationship quality The Levels of Development in Online Relationships survey (LoD; Parks & Floyd, 1996) was used to measure various factors that reect relationship quality. Subscales assessing breadth (e.g., our communication ranges over a wide variety of topics), depth (e.g.,

200

B.W. Lee, L.A. Stapinski / Journal of Anxiety Disorders 26 (2012) 197205

I usually tell this person exactly how I feel), commitment (e.g., I am very committed to maintaining this relationship), interdependence (e.g., the two of us depend on each other), and predictability (e.g., I can accurately predict how this person will respond to me in most situations) were used in the current study. The extent to which the respondents agreed or disagreed with various statements about their chosen relationships were ranked on a scale from 1 (strongly disagree) to 7 (strongly agree). To assess relationship quality separately for online and ofine relationships, participants were rst instructed: Please think of the people with whom you primarily communicate with (more than 90% of the time) over the internet. Choose the person with whom you are closest and answer the following questions in relation to that person. Participants were then asked to think of a face-to-face friend they have known for roughly the same amount of time as the online friend they chose. This allowed the online and ofine relationship to be matched according to the duration of the chosen online friendship. Content validity of the LoD scale has been established, along with good internal consistency across these subscales ( = .85.88; Parks & Floyd, 1996). Good internal consistency was also demonstrated in the current study for both online ( = .81.90) and ofine ( = .8.90) versions. 2.3.6. Generalized Problematic Internet Use Scale The Generalized Problematic Internet Use Scale (GPIUS; Caplan, 2002) was used to measure the cognitions, behaviors, and outcomes of problematic internet use according to Davis model (2001). It consists of seven subscales: mood alteration (e.g., I go online to make myself feel better when I am down), social benets (e.g., I feel more condent socializing online than ofine), negative consequences (e.g., I have missed class or work because I was online), compulsivity (e.g., I have made unsuccessful attempts to control my internet use), excessive use (e.g., I spend a good deal of time online), withdrawal (e.g., I nd it hard to stop thinking about what is waiting for me online), and interpersonal control (e.g., I have control over how others perceive me online) with a total of 28 questions. Respondents rated their agreement or disagreement with each statement using a 5 point scale, where higher scores indicated more severe problematic internet use. Prior research has established good content validity with respect to Davis denition of problematic internet use (Caplan, 2002). The subscales show good internal consistency ( = .78.85; Caplan, 2002), and this was also demonstrated in the current sample for the total score ( = .93) and its subscales ( = .62.90). 2.3.7. Preference for Online Social Interaction The Preference for Online Social Interaction scale (POSI; Caplan, 2003) was used to measure to what extent the respondent preferred online over face-to-face communication. Responses were rated on a 5 point scale where higher scores indicated stronger preference for online interactions. The POSI showed good internal consistency ( = .86) in a normal sample (Caplan, 2003), and also in the current sample ( = .87). 2.3.8. Safety behaviors The Subtle Avoidance Frequency Examination (SAFE) was utilized to assess the use of safety behaviors commonly used by socially anxious individuals. Respondents were asked to rate how often they would engage in 32 behaviors during a face-to-face social situation (e.g., avoid eye contact) ranging from 0 (never) to 5 (always). In a sample with social phobia, the SAFE demonstrated good construct validity and strong internal consistency ( = .89; Cuming et al., 2009). This was replicated in the present sample ( = .95).

2.3.9. Perceived probability and consequences of threat The Probability and Consequences of Threat survey was used to measure respondents perception of social threat in terms of the likelihood of negative evaluation from a hypothetical conversation partner (probability), and their estimate of how bad this would be should it occur (consequences; Rapee & Abbott, 2007). To apply this scale separately to online and ofine interactions, two vignettes were developed. The online vignette read: Imagine that you begin to chat with a new acquaintance over the internet. You are put in contact through a mutual friend. Although you have never met them face-to-face, you are likely to come across them again on the internet. Think about how you would handle this situation, i.e. introducing yourself, carrying on a conversation, etc. Below are a number of judgments your new acquaintance may make about you. Use the following scale to indicate how likely it is they will make these judgments. The ofine vignette was identical to the rst, however participants were asked to Imagine that you are meeting a new acquaintance for the rst time. You are introduced at a social function by a mutual friend. Although you have never met them before, you are likely to see them again. A small pilot study was conducted (n = 4), to ensure that the online and ofine vignettes could be correctly differentiated. For each negative judgment, (e.g., he/she thinks that your conversation is boring) participants were asked to rate how likely and how bad it would be on a scale of 0 (not at all likely/bad) to 4 (extremely likely/bad). The total online and ofine threat scores were derived by multiplying the probability and consequence score for each item, and then summing these together (c.f., Carr, 1974). Good internal consistency for this measure has been established in previous research ( = .93; Rapee & Abbott, 2007), and in the current sample for the online ( = .90) and the ofine ( = .92) measure. 2.4. Data coding and analysis Categorical variables with options endorsed by small participant numbers (n < 30) were combined with a conceptually related group. Difference scores were not used to evaluate differences between online and ofine relationship quality and threat perception due to concerns that the interpretation of such analyses can be misleading (see Edwards, 2002; Zuckerman et al., 2002). Instead, hypotheses were tested using a repeated measures procedure where variables were treated as within-subject predictors with two levels (online and ofine). Mediation analyses were used to explore the mechanisms thought to contribute to the direct relationship between variables of interest (see Fig. 1). The current study followed Baron and Kennys recommended steps (1986; as shown in Fig. 1) in conjunction with 1000 bootstrap resamples to determine the 95% condence interval for the indirect effect. Given the cross-sectional nature of the data, ndings derived from these mediation analyses only suggest possible causal relationships. As multiple tests were conducted per hypothesis, a Benjamini and Hochberg (1995) correction was applied to each set to control the Type-1 error rate. All reported p-values and condence intervals have been adjusted for this error rate correction. 3. Results 3.1. Testing assumptions The assumptions of normality and sphericity were met for all variables. The equal variance assumption was met for all variables with the exception of two: social avoidance and preference for online social interaction. A log10 data transformation was

B.W. Lee, L.A. Stapinski / Journal of Anxiety Disorders 26 (2012) 197205

201

A. The Direct effect Independent Variable Dependent Variable

Path c

B. The Indirect effect Path a

Mediator Variable

Path b

Independent Variable Path c

Dependent Variable

Fig. 1. Diagram of the general mediation model including direct and indirect effects.

considered to reduce this violation. However, Vittinghoff et al. (2005) assert that addressing these violations is not necessary unless results differ qualitatively before and after transformation (p. 127). The pattern of results were the same, therefore a transformation was not implemented.

3.2. Sample characteristics The mean age of the sample was 29.75 years (SD = 12.88) where men (M = 33.55, SD = 15.11) were signicantly older than women (M = 27.25, SD = 10.49; t(336) = 4.21, p < .001). The age range was between 18 and 74 years. Sixty-eight percent of participants were not living with a spouse or partner, 74% were born in Australia, 76% were Caucasian, 67% had education beyond secondary school and 34% were full-time students. The sample had a mean social anxiety score of 37.86 (SD = 26.27, range: 0141), with 52% (n = 176) of the sample scoring at level indicative of probable social anxiety disorder according to the Liebowitz Social Anxiety Scale (Stein, 2004). Thirty-two percent of the sample had higher levels of stress (M = 12.53, SD = 8.79, range: 042), 54% had higher levels of anxiety (M = 7.68, SD = 7.41, range: 034), and 43% had higher levels of depression (M = 10.24, SD = 9.28, range: 040) than the normative data obtained from a non-clinical sample (Henry & Crawford, 2005). Therefore, roughly half of the current sample appears to be psychologically healthy. According to the theoretical cut-off scores proposed by Caplan (2002), the mean score of problematic internet use (M = 70.24, SD = 21.08) is within the non-pathological range (2884). However, the current sample is more disordered than Caplans normative sample (M = 53.17, SD = 23.15) with 24.85% exhibiting problematic levels of internet use (a score above 84). The descriptive statistics for perceptions of threat and relationship quality are presented in Table 1.

Seven participants were unable to complete the relationship quality measures for a friend they primarily communicate with online, and 2 participants did not complete the Probability and Consequences of Threat survey. Data from these participants were excluded from the relevant section of the analysis. Excluded participants did not signicantly differ from the rest of the sample in terms of social anxiety, F(1, 336) = .32, p = .58. There was a trend towards lower levels of problematic internet use in the excluded group, F(1, 336) = 3.67, p = .056, which is unsurprising given the majority were excluded because they could not identify an online friendship for the relationship questionnaire. However, this result should be treated with caution given the small sample size (n = 9) in the excluded group. 3.3. Demographic variables associated with problematic internet use In the current sample, younger participants, F(1, 336) = 19.5, p < .001, 2 = .06, individuals not living with a partner, F(1, 336) = 11.11, p = .003, 2 = .03, and those with lower income, F(1, 335) = 3.99, p = .03, 2 = .02 were more likely to have problematic internet use. Respondents born in Australia, F(1, 336) = 6.91, p = .018, 2 = .02, and those who were not of Caucasian race, F(1, 336) = 13.34, p < .001, 2 = .04, were also more likely to develop problematic internet use. Given these ndings, age, marital status, income, race, and country of birth may be extraneous variables, so were controlled for in all analyses where problematic internet use was the dependent variable. 3.4. Testing hypotheses 3.4.1. Social anxiety and aspects of problematic internet use A multivariate analysis indicated that social anxiety was associated with all aspects of problematic internet use (all ps < .05, all 2 s > .02). Thus, a univariate approach was employed by using the total score derived from the 7 subscales. An exploratory factor analysis supported a one factor solution, which gave an eigenvalue of 3.79 and accounted for 54% of the variance. Higher social anxiety was associated with more problematic internet use as assessed by this total score, F(1, 330) = 101.94, p < .001, 2 = .24, 95% CI [.31, .45]. When controlling for the effect of anxiety, F(1, 327) = 7.77, p = .01, 2 = .02; stress, F(1, 327) = 2.15, p = .14, 2 = .01, and depression, F(1, 327) = 3.53, p= .08, 2 = .01, social anxiety remained a signicant predictor of problematic internet use, F(1, 327) = 28.71, p < .001, 2 = .08, 95% CI [.17, .36].

Table 1 Descriptive statistics for perception of threat and relationship quality. Variable Online Mean Probability of threat Consequences of threat Interdependence Breadth Depth Predictability Commitment 4.81 10.28 17.99 15.96 26.61 22.90 23.54 SD 4.40 7.71 5.84 4.78 7.49 4.77 5.30 Range 025 028 428 321 535 528 528 Face-to-face Mean 6.20 13.37 20.50 17.33 27.54 23.44 24.62 SD 5.42 7.53 5.73 4.12 7.11 4.55 4.48 Range 028 028 428 521 535 828 928

202

B.W. Lee, L.A. Stapinski / Journal of Anxiety Disorders 26 (2012) 197205 Note. SA: social anxiety; bFNE: Brief Fear of Negative Evaluation scale; ftf: face-to-face; GPIUS: Generalized Problematic Internet Use Scale. Avoidance: avoidance of face-to-face interactions in the past week. POSI: preference for online social interactions. ** p < .01. *** p < .001.

Table 2 Summary of mediation results (1000 bootstrap resamples).

Dependent variable (DV)

GPIUS

3.4.4. Online communication as a safety behavior A mediation analysis revealed a signicant indirect effect where the relationship between social anxiety and preference for online social interaction was partially mediated by the tendency to use safety behaviors with a 95% CI [.02, .1] (see Table 2). It was also of interest to explore whether preference for online communication exacerbates avoidance of face-to-face interactions. Therefore, another mediation analysis evaluated whether preference for online social interaction predicts avoidance of face-to-face interactions (as assessed by the Liebowitz Social Anxiety Scale) above and beyond avoidance attributable to fear of negative evaluation (as assessed by the Brief Fear of Negative Evaluation scale).

Mediator variable (MV)

Independent variable (IV)

bFNE SA

SA

SA

Probability of online threat Probability of ftf threat POSI SAFE

GPIUS

Avoidance POSI

3.4.3. Perception of threat online versus ofine As predicted, those with higher levels of social anxiety perceived more interpersonal control during online than ofine interactions, F(1, 330) = 18.47, p < .001, 2 = .05, 95% CI [.02, .04]. Furthermore, repeated measures analyses showed that those with higher social anxiety perceived less threat during online than face-to-face interactions, F(1, 334) = 71.05, p < .001. Upon further exploration, it was found that those with higher social anxiety perceived the probability of threat occurring to be less likely during online than ofine interactions, F(1, 334) = 60.88, p < .001, but no signicant difference was perceived between the consequences of such threat in an online and ofine environment, F(1, 334) = 3.02, p = .083. Since social anxiety was associated with a difference in the perceived probability of threat occurring online compared to ofine, these variables were examined as factors that may account for the relationship between social anxiety and problematic internet use. This possibility was tested using two mediation analyses (see Table 2). The relationship between social anxiety and problematic internet use was not signicantly mediated by perceived probability of threat during online interactions 95% CI [.004, .046]. However, perceived probability of threat during face-to-face interactions did signicantly mediate the relationship between social anxiety and problematic internet use, where the indirect effect was signicant at 95% CI [.02, .14].

Direct effect (c)

3.4.2. Quality of online relationships A repeated measures analysis indicated that those with higher levels of social anxiety reported communicating with more people online than ofine, F(1, 329) = 6.65, p = .026. Contrary to expectations, those with higher levels of social anxiety did not perceive better quality online than ofine relationships in terms of interdependence, F(1, 329) = 0.83, p = .45; breadth, F(1, 329) = .35, p = .61; depth, F(1, 329) = 0.21, p = .68; predictability, F(1, 329) = 4.24, p = .08, or commitment F(1, 329) = 1.09, p = 39. However, main effect analyses conrmed that social anxiety was associated with decreased breadth, depth and predictability of relationships irrespective of communication mode (ps < .05), while interdependence and commitment in relationships did not differ (ps > .05). To further explore the relationship between problematic internet use and relationship quality, a separate univariate analysis was conducted for each aspect of relationship quality, irrespective of social anxiety level. This revealed that reports of high interdependence F(1, 323) = 6.66, p = .02, 2 = .02, 95% CI [.12, 87], and predictability F(1, 323) = 6.03, p = .03, 2 = .02, 95% CI [1.05, .12] in online relationships were associated with problematic internet use. For face-to-face relationships, low levels of breadth, F(1, 323) = 24.39, p = .001, 2 = .07, 95% CI [1.84, .79]; depth, F(1, 323) = 13.18, p = .001, 2 = .04, 95% CI [.87, .26]; predictability, F(1, 323) = 8.17, p = .015, 2 = .03, 95% CI [1.18, .22], and commitment, F(1, 323) = 10.44, p = .004, 2 = .03, 95% CI [1.29, .31] were associated with more problematic internet use. No other relationship quality variables predicted problematic internet use (all ps > .05, all 2 s < .02).

Adjusted R2 (%)

29.64

31.26
***

Total effects (c)

.38***

95% CI indirect effect

[.00, .05]

[.02, .14]
**

Indirect effect (a b)

.36***

Effect of MV on DV (b)

Effect of IV on MV (a)

***

.32*** .54***

.05***

.12

.29*** .11***

.31

.67

**

.44*** .08***

.3

***

.09** .06**

.02

.08

[.05, .19] [.02, .10]

.53*** .14***

.38

45.10 29.22

B.W. Lee, L.A. Stapinski / Journal of Anxiety Disorders 26 (2012) 197205

203

The bFNEII was selected for this purpose because it assesses the fears that are central to social anxiety disorder (Rapee & Heimberg, 1997), but omits any assessment of social avoidance. As predicted, preference for online social interaction was found to be signicantly associated with face-to-face avoidance in the past week after controlling for the level of avoidance explained by fear of negative evaluation (see path b in Table 2). Furthermore, the relationship between fear of negative evaluation and social avoidance was found to be partially mediated by having a preference for online social interactions with a 95% CI [.05, .19] (see Table 2). 4. Discussion As expected, the current study showed that social anxiety was strongly associated with greater problematic internet use, producing a large effect size. This converges with previous ndings, which have similarly found an association between social anxiety and internet use (Caplan, 2007; Erwin et al., 2004). Furthermore, this study extended previous research by demonstrating that social anxiety uniquely predicted problematic internet use when controlling for variance accounted for by general psychopathological symptoms. Subscales assessing depression, general anxiety and stress produced modest effect sizes, with only anxiety reaching signicance as a predictor of problematic internet use. This contrasts with previous research demonstrating a moderate relationship between depression and problematic internet use (Caplan et al., 2009; Ko et al., 2008), and suggests that social anxiety should be controlled for in future studies given the potential overlap with depression. The current study also found that those with higher social anxiety reported communicating with more people online as compared to face-to-face. This nding is consistent with the social compensation hypothesis, which proposes that the internet may promote online friendships at the expense of face-to-face social networks (Peters & Malesky, 2008; Selfhout et al., 2009; Valkenburg & Peter, 2008). Therefore, the effect of online relationships on the social well-being of those with social anxiety may be two-fold, in that it may assist the formation of new friendships, but also exacerbate avoidance of face-to-face interactions. 4.1. Relationship quality, social anxiety and problematic internet use According to current models, social anxious individuals perceive their relationships to be impoverished and unsatisfactory (e.g., Erwin et al., 2004). Consistent with this idea, the current study found that social anxiety was associated with the perception of poorer quality relationships, both online and ofine. It was predicted that online communication may be appealing for social anxious individuals because they perceive more interdependence, depth, breadth, predictability, and commitment in their online relationships relative to ofine relationships. This was not supported by the current data as no discrepancy was observed between the quality of online compared to ofine relationships as a function of social anxiety. Thus, the online medium does not appear to improve perceived relationship quality, rather social anxiety is associated with poor relationship quality across both mediums. Although these results are inconsistent with proposed models of social anxiety and problematic internet use (Bargh et al., 2002; Erwin et al., 2004; King & Poulos, 1998; Valkenburg & Peter, 2007), the current study appears to be the rst to directly examine the relationship between social anxiety and the quality of online relationships. There are several possible explanations for these unexpected ndings. Firstly, it may be that aspects other than those measured in the current study are critical to the perception of preferable online

compared to face-to-face relationships. Thus, future researchers may be interested in examining other factors that may make online relationships preferable for individuals with social anxiety, such as feelings of satisfaction and fulllment, or achieving a sense of belonging (King & Poulos, 1998; Sheeks & Birchmeier, 2007). Secondly, in order to match relationship duration across online and ofine contexts, only one relationship of each type was selected, and participants answered questions with this relationship in mind. Therefore, it is possible that the selected relationship may not be representative of participants relationship quality more generally. However, this explanation seems unlikely, as previous studies using general samples suggest that reporting about one online and one ofine relationship is representative of general friendship quality (McKenna et al., 2002; Parks & Roberts, 1998). The association between relationship quality and problematic internet use, irrespective of social anxiety level, was also explored. This revealed that those who reported poorer quality face-to-face relationships, and those with greater dependence on and predictability in their online relationships, were vulnerable to problematic internet use. This nding suggests that high quality face-to-face friendships may provide protection against development of problematic internet use. Clinically, this points to the potential benets of targeting the development and maintenance of healthy face-to-face relationships in interventions for problematic internet use. 4.2. Threat, social anxiety and problematic internet use Researchers have hypothesized that online communication is perceived as less threatening with less risk of negative consequences as compared to face-to-face interactions (AmichaiHamburger & Furnham, 2007; Caplan, 2002; King & Poulos, 1998; McKenna & Bargh, 2000). However, this idea has received little empirical attention. Consistent with previous research (Caplan, 2007; Erwin et al., 2004), this study showed that higher social anxiety was associated with a perception of greater control during online social interactions as compared to face-to-face interactions. As Erwin et al.s (2004) ndings suggest, online interactions may be perceived as more controllable because an online audience is unable to hear mistakes in speech or see visible signs of anxiety. In addition, this study was the rst to examine whether socially anxious individuals perceive online interactions to be less threatening in terms of the probability and consequences of negative evaluation. As expected, those with higher social anxiety perceived the probability of threat to be signicantly less during online than face-to-face interactions. However, the consequences of negative evaluation were perceived as equal regardless of the medium via which they interacted. Contrary to expectations, it therefore seems that aspects of the online context, such as anonymity, are not sufcient to reduce the perceived consequences of negative evaluation. It has been suggested that decreased social threat perception during online interactions could explain why those with social anxiety are more vulnerable to problematic internet use (AmichaiHamburger & Furnham, 2007; Caplan, 2002; King & Poulos, 1998). However, the results of this study indicated that this relationship is not mediated by perceived probability of online threat, but rather by the perceived probability of face-to-face threat. This seems to suggest that socially anxious individuals are not drawn to online interactions because these are perceived as safe, but rather because face-to-face interactions are so threatening. This is consistent with Erwin et al.s (2004) report that despite perceiving greater control during online interactions, socially anxious individuals were nevertheless fearful of these interactions. On the whole, the current ndings suggest that socially anxious individuals are at risk of developing problematic internet use because their inated

204

B.W. Lee, L.A. Stapinski / Journal of Anxiety Disorders 26 (2012) 197205

perceptions of negative evaluation during face-to-face interactions drive them to nd safer means of communication. 4.3. Problematic internet use as a safety behavior It was predicted that online communication may act as a safety behavior for socially anxious individuals, in that it facilitates avoidance of more threatening face-to-face interactions. This hypothesis was supported in the current study by four lines of evidence. As described above, individuals with social anxiety perceived the probability of social threat to be reduced during online as compared to ofine interactions, and tend to communicate with a greater number of people online than face-to-face. Thus, online communication may provide those with social anxiety a means of avoiding threatening face-to-face interactions, leading to excessive internet use. Furthermore, the relationship between social anxiety and having a preference for online social interactions was partially accounted for by the tendency to adopt a range of alternate safety behaviors in social situations. This indicates that for individuals with social anxiety, online communication may be one of many safety behaviors employed to neutralize or avoid threatening social interactions. Finally, the current study used a mediation analysis to assess whether preferring online social interaction was associated with an exacerbation of social avoidance. Safety behaviors are thought to reinforce the maladaptive cognitions, and thus perpetuate avoidance of threatening situations, because the lack of fear is attributed to the safety behavior rather than the individuals ability (Rapee & Heimberg, 1997). The indirect effect was signicant, suggesting that online communication exacerbates avoidance of face-to-face interactions in those with social anxiety even when controlling for the variance directly attributable to fear of negative evaluation. The current ndings complement previous research which found an association between the use of online communication and avoidance of social interactions (Ebeling-Witte et al., 2007; Erwin et al., 2004; King & Poulos, 1998; McKenna & Bargh, 2000; Whang et al., 2003). 4.4. Clinical implications The present ndings have implications for those who work with clients experiencing social anxiety and/or problematic internet use. It is important that therapists are aware of factors that predispose individuals to problematic internet use, such as impoverished face-to-face relationships, and preference or reliance on the internet for social interaction. Interventions to decrease vulnerability to problematic internet use may incorporate development of adaptive anxiety management strategies, and promotion of positive face-toface friendships. Early intervention may be particularly important for young adults and adolescents, given that younger respondents were more likely to have problematic internet use which may be related to increased exposure to the internet at an earlier developmental stage. The current ndings also bear relevance to the use of the internet as a therapeutic device. Evidence is accumulating to support online therapy approaches to the treatment of social anxiety and other disorders, and this mode is becoming increasingly popular amongst clinicians (Andersson, 2009). Given the current ndings, online therapy may be particularly appealing to socially anxious individuals given the greater interpersonal control and increased comfort it provides. However, the present ndings also suggest that online communication may be a safety behavior for those with social anxiety, and thus caution is needed when applying online therapy to those who are susceptible to problematic internet use. In these populations it will important to incorporate regular exposure to face-to-face interactions to provide corrective learning experiences and reduce avoidance in this context (Rapee &

Heimberg, 1997; Valkenburg & Peter, 2008). Turkle (1995) provides an alternative perspective on this issue, stating that the internet is no more addictive than therapy when it works as a pathway to psychological growth (p. 12). Andersson (2009) has concurred that while online therapy for social anxiety may reinforce face-to-face avoidance, it enhances the effectiveness of therapy by providing a safe environment for interaction and expression. Therapists need to be aware of the potential benets and disadvantages of online therapy, and use this knowledge to educate clients and plan therapeutic interventions. 4.5. Limitations Several limitations of the current study should be noted. First and foremost, this study is limited by the cross-sectional nature of the data collected. Considering the work by Cole and Maxwell (2003), it is possible that the mediation relationships identied in a cross-sectional study would not be replicated with longitudinal data. Thus, the mediation ndings in the current paper suggest possible causal pathways through which problematic internet use may develop in those with social anxiety. However, longitudinal and experimental data are needed to validate these relationships. Furthermore, the ongoing debate as to whether a uni-directional or bidirectional relationship exists between social anxiety and problematic internet use (Caplan, 2003; Davis, 2001), makes longitudinal research vital to clarify the nature of this relationship. A further limitation of the current study relates to the Generalized Problematic Internet Use Scale (GPIUS), which has not yet been examined for construct validity. Unfortunately, there are currently no instruments available that have demonstrated construct validity for identifying those with internet use problems. Given the paucity of available assessment measures, the GPIUS was selected for use in the current study because it is one of the most commonly used measures in the eld, and prior research has demonstrated good content validity with respect to Davis denition of problematic internet use (Caplan, 2002). As internet access becomes more readily available and problematic internet use becomes more prevalent (Morahan-Martin, 2001), future research to support the construct validity of the GPIUS is vital. Some limitations of the recruitment method and sample warrant comment. Although attempts were made to recruit broadly via online forums, it is likely that a majority of the sample were recruited via and list servers that started from the Macquarie University Department of Psychology. Therefore, the sample may not be representative of the broader population of internet users. In addition, older individuals were not as well represented in the sample as the younger age range, and therefore caution should be taken when generalizing to older individuals. Finally, it should be noted that scores on the Liebowitz Social Anxiety Scale indicated a probable diagnosis of social anxiety disorder for 34% of participants, and a very probable or extremely probable diagnosis for 18% of participants. This representation is high for a general sample, and is likely attributable to the recruitment strategy which included posts on numerous gaming and mental health forums. It will be important for future studies to replicate the current ndings in a treatment-seeking sample. Although several predictors of problematic internet use were identied, the effect size for some predictors was small, indicating only a modest contribution in terms of variance explained. It may be that stronger or different relationships between variables will be observed in a clinical sample. Despite these limitations, the current study provides preliminary evidence that socially anxious individuals use online communication to feel safer in the face of negative evaluation.

B.W. Lee, L.A. Stapinski / Journal of Anxiety Disorders 26 (2012) 197205

205

This provides a theoretical framework for understanding why these individuals are vulnerable to the development of internet use problems, and informs the development of interventions aimed at preventing and treating these issues. Acknowledgement The authors wish to thank Alan Taylor for his guidance regarding the statistical procedures described in the current paper. References
Alden, L. E., & Bieling, P. (1998). Interpersonal consequences of the pursuit of safety. Behavior Research and Therapy, 36, 5364. Amichai-Hamburger, Y., & Furnham, A. (2007). The positive net. Computers in Human Behavior, 23, 10331045. Andersson, G. (2009). Using the internet to provide cognitive behavior therapy. Behavior Research and Therapy, 47, 175180. 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), 3348. Baron, R. M., & Kenny, D. A. (1986). The moderatormediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. Personality and Social Psychology, 51(6), 11731182. Baumeister, R. F., & Leary, M. R. (1995). The need to belong: desire for interpersonal attachments as a fundamental human motivation. Psychological Bulletin, 117, 497529. Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society (Series B), 57, 289300. Campbell, A. J., Cumming, S. R., & Hughes, I. (2006). Internet use by the socially fearful: addiction or therapy? Cyberpsychology & Behavior, 9(1), 6981. Caplan, S. E. (2002). Problematic internet use and psychological well-being: development of a theory-based cognitive-behavioral measurement instrument. Computers in Human Behavior, 18, 553575. Caplan, S. E. (2003). Preference for online social interaction: a theory of problematic internet use and psychosocial well-being. Communication Research, 30(6), 625648. Caplan, S. E. (2007). Relations among loneliness, social anxiety and problematic internet use. Cyberpsychology & Behavior, 10(2), 234242. Caplan, S. E., Williams, D., & Yee, N. (2009). Problematic internet use and psychosocial well-being among MMO players. Computers in Human Behavior, 25, 13121319. Carleton, R. N., McCreary, D. R., Norton, P. J., & Asmundson, G. J. G. (2006). Brief fear of negative evaluation scale-revised. Depression and Anxiety, 23, 297303. Carr, A. T. (1974). Compulsive neurosis: a review of the literature. Psychological Bulletin, 81, 311318. Chak, K., & Leung, L. (2004). Shyness and locus of control as predictors of internet addiction and internet use. Cyberpsychology & Behavior, 7(5), 559570. Clark, D. M., & Wells, A. (1995). A cognitive model of social phobia. In: R. G. Heimberg, M. R. Liebowitz, D. A. Hope, & F. R. Schneier (Eds.), Social phobia: diagnosis, assessment, and treatment (pp. 6993). New York, USA: Guilford Press. Cole, D. A., & Maxwell, S. E. (2003). Testing mediational models with longitudinal data: questions and tips in the use of structural equation modeling. Journal of Abnormal Psychology, 112(4), 558577. Cuming, S., Rapee, R. M., Kemp, N., Abbott, M. J., Peters, L., & Gaston, J. E. (2009). A self-report measure of subtle avoidance and safety behaviors relevant to social anxiety: development and psychometric properties. Journal of Anxiety Disorders, 23, 879883. Davis, R. A. (2001). A cognitive-behavioral model of pathological internet use. Computers in Human Behavior, 17, 187195. Ebeling-Witte, S., Frank, M. L., & Lester, D. (2007). Shyness, internet use and personality. Cyberpsychology & Behavior, 10(5), 713716. Edwards, J. R. (2002). Alternatives to difference scores. Polynomial regression analysis and response surface methodology. In: F. Drasgow, & N. Schmitt (Eds.), Measuring and analyzing behavior in organizations. Advances in measurement and data analysis.. San Francisco, CA, USA: Jossey-Bass Inc. Erwin, B. A., Turk, C. L., Heimberg, R. G., Fresco, D. M., & Hantula, D. A. (2004). The internet: home to a severe population of individuals with social anxiety disorder? Anxiety Disorders, 18, 629646. ETforecasts. (2010). Internet user forecast by country: executive summary. Retrieved from. http://www.etforecasts.com/products/ES intusersv2.htm Fresco, D. M., Coles, M. E., Heimberg, R. G., Liebowitz, M. R., Hami, S., Stein, M. B., et al. (2001). The Liebowitz Social Anxiety Scale: a comparison of the psychometric properties of self-report and clinician-administered formats. Psychological Medicine, 31, 10251035. Gross, E. F., Juvonen, J., & Gable, S. L. (2002). Internet use and well-being in adolescence. Journal of Social Issues, 58(1), 7590.

Henry, J. D., & Crawford, J. R. (2005). The short-form version of the Depression Anxiety Stress Scales (DASS-21): construct validity and normative data in a large non-clinical sample. British Journal of Clinical Psychology, 44, 227239. King, S. A., & Poulos, S. T. (1998). Using the internet to treat generalized social phobia and avoidant personality disorder. Cyberpsychology & Behavior, 1(1), 2936. Ko, C.-H., Yen, J.-Y., Chen, C.-S., Chen, C.-C., & Yen, C.-F. (2008). Psychiatric comorbidity of internet addiction in college students: an interview study. CNS Spectrum, 13(2), 147153. La Greca, A. M., & Lopez, N. (1998). Social anxiety among adolescents: linkages with peer relations and friendships. Journal of Abnormal Child Psychology, 26(2), 8394. Lovibond, P. F., & Lovibond, S. H. (1995). The structure of negative emotional states: comparison of the depression anxiety stress scales (DASS) with the Beck depression and anxiety inventories. Behavior Research and Therapy, 33(3), 335343. McKenna, K. Y. A., & Bargh, J. A. (2000). Plan 9 from Cyberspace: the implications of the internet for personality and social psychology. Personality and Social Psychology Review, 4, 5775. McKenna, K. Y. A., Green, A. S., & Gleason, M. E. J. (2002). Relationship formation on the internet: whats the big attraction? Journal of Social Issues, 58(1), 931. Morahan-Martin, J. (2001). Impact of Internet abuse for college students. In: C. Wolfe (Ed.), Learning and teaching on the World Wide Web (pp. 191219). San Diego, CA: Academic Press. Morahan-Martin, J. (2005). Internet Abuse: addiction? Disorder? Symptom? Alternative explanations? Social Science Computer Review, 23(1), 3948. Parks, M. R., & Floyd, K. (1996). Making friends in Cyberspace. Journal of Communication, 46(1), 8097. Parks, M. R., & Roberts, L. D. (1998). Making MOOsic: the development of personal relationships on line and a comparison to their off-line counterparts. Journal of Social and Personal Relationships, 15(4), 517537. Pennebaker, J. W. (1989). Confession, inhibition, and disease. In: L. Berkowitz (Ed.), Advances in experimental social psychology (pp. 211244). New York: Academic. Peters, C. S., & Malesky, J. L. A. (2008). Problematic usage among highly-engaged players of massively multiplayer online role playing games. Cyberpsychology & Behavior, 11(4), 481484. Rapee, R. M., & Abbott, M. J. (2007). Modelling relationships between cognitive variables during and following public speaking in participants with social phobia. Behavior Research and Therapy, 45, 29772989. Rapee, R. M., & Heimberg, R. G. (1997). A cognitive-behavioral model of anxiety in social phobia. Behavior Research and Therapy, 35(8), 741756. Scealy, M., Phillips, J. G., & Stevenson, R. (2002). Shyness and anxiety as predictors of patterns of internet usage. Cyberpsychology & Behavior, 5(6), 507515. Selfhout, M. H. W., Branje, S. J. T., Delsing, M., ter Bogt, T. F. M., & Meeus, W. H. J. (2009). Different types of internet use, depression, and social anxiety: the role of perceived friendship quality. Journal of Adolescence, 32, 819833. Shaw, L. H., & Gant, L. M. (2002). In defense of the Internet: the relationship between internet communication and depression, loneliness, self-esteem, and perceived social support. Cyberpsychology & Behavior, 5(2), 157171. Sheeks, M. S., & Birchmeier, Z. P. (2007). Shyness, sociability, and the use of computer-mediated communication in relationship development. Cyberpsychology & Behavior, 10, 6470. Shepherd, R.-M., & Edelmann, R. J. (2005). Reasons for internet use and social anxiety. Personality and Individual Differences, 39, 949958. Stein, D. J. (2004). Clinical manual of anxiety disorders. Virginia, USA: American Psychiatric Publishing. The Nielsen Company. (2010 January 22). Led by Facebook, Twitter, global time spent on social media sites up 82% year over year [Web log post]. Retrieved from. http://blog.nielsen.com/nielsenwire/global/led-by-facebooktwitter-global-time-spent-on-social-media-sites-up-82-year-over-year/ Turkle, S. (1995). Life on the screen: identity in the age of the internet. New York: Simon and Schuster. Valkenburg, P. M., & Peter, J. (2007). Preadolescents and adolescents online communication and their closeness to friends. Developmental Psychology, 43(2), 267277. Valkenburg, P. M., & Peter, J. (2008). Adolescents identity experiments on the internet. Communication Research, 35(2), 208231. Vittinghoff, R., Glidden, D. V., Shiboski, S. C., & McCulloch, C. E. (2005). Regression methods in biostatistics: linear, logistic, survival and repeated measures models. USA: Springer Science + Business Media Inc. Walther, J. B. (1996). Computer-mediated communication: impersonal, interpersonal, and hyperpersonal interaction. Communication Research, 23, 343. Whang, L. S.-M., Lee, S., & Chang, G. (2003). Internet over-users psychological proles: a behavior sampling analysis on internet addiction. Cyberpsychology & Behavior, 6(2), 143150. Young, K. S. (1996). Psychology of computer use: XL Addictive use of the Internet. A case that breaks the stereotype. Psychological Reports, 79, 899902. Zuckerman, M., Gagne, M., Nafshi, I., Knee, C. R., & Kieffer, S. C. (2002). Testing discrepancy effects: a critique, a suggestion, and an illustration. Behavior Research Methods, Instruments, & Computers, 34(3), 291303.