Factors Predicting Internet Gaming Disorder: A Two-Wave Panel Study Sophie F. Waterloo
Graduate School of Communication University of Amsterdam
Research Master Thesis 27-06-2014
Semester 2
Supervisor: Dr. J.S. Lemmens
Student Number: 5883873 sophie.waterloo@student.uva.nl
Abstract
The adoption of Internet Gaming Disorder (IGD) in the latest edition of the DSM-5 marks an important new shift within research on pathological gaming. The main aim of the current study was to explain the antecedent structure of IGD. In doing so, this study is the first to longitudinally examine the causal relationship between a set of personality and situational factors, and the newly classified IGD. A two-wave panel study was conducted among 643 Dutch adolescent games over the course of six months. Using autoregressive structural equation models, the causal relationship between each identified factor and IGD over time was examined in more detail. The analyses yield valuable insights for the understanding of the pathological involvement with video games among adolescents, in particular with respect to the prevalence, progressiveness and antecedent structure of pathological gaming.
Keywords: pathological gaming, Internet Gaming Addiction, DSM-5, predictors, two-wave panel data, adolescents
Factors Predicting Internet Gaming Disorder
Included in the latest version of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), the issue of pathological involvement with video games, it seems, is on its way as being widely recognized as a clinical disorder termed Internet Gaming Disorder (APA, 2013). The topic of pathological gaming, often described as persistent, recurrent and excessive involvement with computer or video games despite associated problems (Lemmens, Valkenburg & Peter, 2009), has gained widespread attention over the years. Here, scholars have mainly focused on possible risk factors, consequences, prevalence rates and comorbidity, in particular among adolescents (Kuss & Griffiths, 2012). The rationale for studying adolescents in relation to pathological gaming is mainly guided by the perspective that adolescents are most vulnerable to the development of pathological behaviour in comparison to other age groups (Griffiths & Wood, 2000). A closer examination of the literature reveals inconsistencies in both the uses of terminology and the measurement criteria used, indicating the need for an agreed measuring instrument. The specific diagnostic criteria introduced in the DSM-5 for the diagnosis of Internet Gaming Disorder (IGD) responds to this need, additionally opening up new research perspectives. The aim of the current study is to examine the causal relation between predicting variables and the newly classified Internet Gaming Disorder among adolescents in a longitudinal design.
A variety of elements within games can invoke a sense of immersion that attracts, but moreover engages individuals to keep playing (McGonigal, 2011), reflected in different motivations identified in previous research. These include the need for competition, challenge, social interaction, enjoyment, excitement, role-play, and to learn new things, among others (Vorderer, Hartmann & Klimmt, 2003; Yee,
2006; Jansz & Tanis, 2007; Hoffman & Nadelson, 2010). In explaining how gamers become pathologically involved with video games, a large body of research have demonstrated links between personality traits and the pathological involvement of video games, ranging from levels of self-esteem (Ko, Yen, Chen, Chen, & Yen, 2005; Lemmens, Valkenburg & Peter, 2011) and perceived feelings of loneliness (Seay & Kraut, 2007; Parsons, 2005; Kim, Namkoong, Ku & Kim, 2008; Lemmens et al., 2011), to impulsive behaviour and ADHD (Gentile et al., 2011; Weinstein & Weizman, 2012; Johnstone, 2013). Specifically, four different factors related to psychosocial well-being (i.e., social competence, self-esteem, life satisfaction and loneliness) and two stable factors of personality/temperament (i.e., impulsivity and ADHD) are considered as predictors in the current study.
Particular little attention has been directed to the role of situational factors in relation to the pathological involvement with video games among adolescents. Family is however of particular importance, since much of the media use among adolescents takes place within the home environment. From a problem behaviour perspective, family expectations may contribute to an increase in exhibiting problem behaviours of pathological nature (Donovan, Jessor & Costa, 1991). The current study additionally aims to address this gap in the literature by including the factor of family conflict as a possible predictor of pathological gaming.
Determining whether the above mentioned factors act as predictors to the development of IGD establishes a better, and more exhaustive, understanding of the antecedent structure of pathological gaming to date. As such, the central research question is: How are personality and situational factors causally related to the
Theory Defining Pathological Gaming
Games provide a particular immersive digital environment, where players become captivated in the narrative of a virtual game world (McMahan, 2003). In explaining why people play video games, McGonigal (2011) relates this idea of immersion to Csíkszentmihályi’s (1975) concept of ‘flow’, defined as a state of complete absorption in an intrinsically enjoyable activity. Csíkszentmihályi (1975) notes that games in particular provide the possibility to capture people in a state of flow, and play, as he writes, “is the flow experience par excellence” (p. 37, emphasis in original). Here, the combination of self-chosen goals, personally optimized
obstacles, increasing challenges and continuous feedback within games add to this state of flow (McGonigal, 2011). Studies on motivations and video games have found a variety of motivational variables that relate to these flow-inducing characteristics of games, which include the need for competition, challenge, social interaction,
enjoyment, excitement, role-play, and to learn new things (Vorderer et al., 2003; Yee, 2006; Jansz & Tanis, 2007; Hoffman & Nadelson, 2010).
While a number of studies have focused on the positive outcomes of gaming (e.g. learning, collaboration, hand-eye coordination), most studies within media effect research have directed attention towards the potential problematic uses and ill-effects, one of which is pathological involvement with video games (e.g. Fisher, 1994;
Grüsser, Thalemann & Griffiths, 2006; Lemmens et al., 2009; Wood, 2008; Van Rooij, Schoenmakers, Vermulst, Van Den Eijnden & Van de Mheen, 2011). Different terms have been used within research on this topic, including videogame dependence (e.g. Griffiths & Hunt, 1998), problematic game playing (e.g. Seay & Kraut, 2007), gaming disorder, pathological gaming (e.g. Lemmens et al., 2009), and the more
popular term game addiction (e.g. Charlton & Danforth, 2007; Chiu, Lee & Huang, 2004). Pathological gaming, considered a more accurate terminology, is generally defined as the persistent, recurrent and excessive involvement with computer or video games that cannot be controlled despite associated social and emotional problems (Lemmens et al., 2009).
While most studies adopt a similar interpretation of the pathological
involvement with video games, two different approaches to measuring pathological involvement with video games can be distinguished within the literature. First, in addition to frequency of game play, pathological gaming has been assessed using a variety of scales for internet addiction, each measuring a different set of underlying factors of pathological behaviour (e.g. Compulsive Internet Use Scale, Internet Addiction Scale Taiwan). Here, scale items were either rephrased to fit the topic of gaming (Ng & Wiemer-Hastings, 2005; Wan & Chiou, 2006), or used in combination with hours spent on online gaming as a measure for online pathological gaming (Van Rooij et al., 2011). Second, the diagnostic criteria for pathological gambling (DSM-4) have often been adapted and applied in measuring pathological involvement with games (Grüsser, Thalemann, Albrecht & Thalemann, 2005; Salguero & Morán, 2002). The use of this scale in measuring pathological gaming has gained considerable critique, in that the underlying behavioural symptoms may not be directly applicable to gaming (Wood, 2008; Olsen, 2010). These inconsistencies within measuring the concept of pathological gaming question the validity of results regarding prevalence, predictors and its consequences found so far. The DMS-5 sheds new light on both the definition and measurement of pathological gaming, framing it as a pathological disorder with its own set of criteria.
Internet Gaming Disorder
With the specific criteria for the diagnosis of pathological involvement with games introduced in the fifth edition of the DSM, a more consistent methodology concerning pathological gaming can be achieved. Specifically, nine different criteria have been established to determine whether an individual can be diagnosed with gaming disorder (APA, 2013). Of these nine criteria, six appear to overlap with the criteria for pathological gambling that have been used in previous studies on pathological gaming, which are: Preoccupation, tolerance, withdrawal, persistence, escape, and problems (Lemmens, Valkenburg & Gentile, in press). The other three criteria are identified as measuring deceit, displacement, and severe consequences. This indicates a more exhaustive measure of underlying pathological behaviours relating to gaming than has previously been employed. Additionally, the DSM-5 guidelines note that an individual should meet at least five or more out of the nine criteria to be diagnosed with IGD (APA, 2013).
Based on the specific criteria introduced in the DSM-5, an Internet Gaming Disorder (IGD) scale and index was tested by Lemmens et al. (in press), which proved to be valid and reliable. Preliminary results further indicated that the
prevalence of IGD is higher among young adults than among adolescents (Lemmens et al., in press). However, it has generally been found that adolescents are most prone to developing pathological involvement with games (Griffiths & Wood, 2000), which in turn could lead to problems regarding general well-being, school performances and daily routines (Ko et al., 2005). It is therefore of importance to gain a better
Predicting problematic media behaviour
In media effect research, the notion that the effect of media on cognitions, emotions, attitudes, psychology and behaviour is dependent on individual differences and social-context variables is shared across a number of conditional media effect models (Valkenburg & Peter, 2013). As Valkenburg and Peter (2013) conceptualize in the Differential Susceptibility to Media Effects Model (DSMM), variables of dispositional, developmental and social susceptibility can act both as predictors of media use and as moderators of the effects of media use on media response states. Dispositional variables are internal characteristics, such as gender,
personality/temperament, cognitions, attitudes, motivations, identity and moods (Valkenburg & Peter, 2013). Here, a distinction can be made between transient, and more stable dimensions of personality over time. Developmental variables refer to the cognitive and emotional developmental levels related to age as predictors of media use and preferences. Similarly, social contexts such as family, siblings, friends or schools can encourage or discourage particular media use (Valkenburg & Peter, 2013).
Previous studies on pathological gaming have found associations with factors of psychosocial well-being, reflecting more transient dimensions of personality, and the more stable personality/temperament traits such as impulsivity and ADHD. These variables fit the conceptualization of dispositional susceptibility as identified by Valkenburg and Peter (2013). Of additional interest is the conceptualization of social contexts as predictor of particular media use. Within research on pathological gaming, the inclusion of social context variables is limited. From a problem behaviour
perspective, similar factors are considered as contributors to the development of problematic behaviours such as delinquent behaviour, problem drinking and
substance abuse (Donovan et al., 1991). The Problem Behaviour Theory, a social psychological framework focused on explaining emergent problematic behaviours among adolescents (Donovan et al., 1991), posits that psychosocial personality factors reflect proneness to problem behaviours. Additionally, social contexts such as family and school can be of influence on the development of problem behaviour, as much of the socializing concerning healthy behaviours among adolescents is carried out within these contexts (Donovan et al., 1991).
In predicting both media use and problem behaviour, thus, personality factors and situational factors such as family and school appear to act as predictors
(Valkenburg & Peter, 2013; Donovan, Jessor & Costa, 1991). In light of this, different personality and situational factors can be considered to act as predictors of IGD. In line with existing empirical research on pathological gaming, different factors related to personality and social context are considered. For personality factors, a distinction is made between more transient factors of psychosocial well-being and the more stable factors of personality/temperament. Additionally, the situational factor of family conflict is considered, for reasons that will be discussed next.
Psychosocial well-being and pathological gaming
Research on pathological gaming has extensively studied associations and causal relations with a variety of factors relating to psychosocial well-being
(Lemmens et al., 2011; Gentile et al, 2011). Psychosocial well-being is a composite term mixing the concept of wellness or well-being, defined as the notion of health as a dynamic state or process, with the concept of psychosocial, which refers to the
importance of both intrapersonal and interpersonal functioning (Lent, 2004). This encompasses factors at both ends of the well-being spectrum, such as life satisfaction, social competence, self-esteem, introversion, anxiety, depression and loneliness. As
the literature suggests, individuals who suffer from poor psychosocial well-being are more prone to develop problematic media behaviours such as pathological gaming (Davis, 2001).
The factors of psychosocial well-being considered in this study are social competence, self-esteem, life-satisfaction and loneliness. In particular social competence, self-esteem and loneliness have been found to causally relate to the pathological involvement with games within longitudinal designs. First, lower social competence has been found to act as significant risk factors for developing
pathological gaming behaviour (Gentile et al., 2011; Lemmens et al., 2011). Social competence refers to the behaviour that reflects successful social functioning, with a relative tendency to be sociable or associate with one’s peers (Howes, 1987;
Inderbitzen & Foster, 1992). Lower social competence as a predictor of pathological gaming can be explained in a variety of ways. First, online games in particular allow players with lower social competence to fulfil unmet emotional and psychological needs in real life, offering an environment with safe and secure environments for social interaction (Young, 1998; Chak & Leung, 2004). Second, individuals might spend more time playing games as a way to avoid real life social situations (Peters & Malesky, 2008). Therefore, it is expected that lower social competence acts as a predictor to IGD (H1).
Self-esteem refers to the evaluation that individuals make about themselves that expresses a self-judgment of approval, disapproval and personal worth (Demo & Savin-Williams, 1992; Zimmerman, Copeland, Shope & Dielman, 1996), which are formed by reflected appraisals, social comparison, and self-attribution (Rosenberg, Schooler & Schoenbach, 1989). Playing games has been found to relate to lower levels of self-esteem among adolescents, mainly because the possibility to be
successful in games allow for the compensation of a weak self-image (Roe & Muijs, 1998). In the current study, lower levels of self-esteem are expected to predict IGD (H2).
Life satisfaction is defined as a general cognitive assessment of a person’s subjective well-being (Diener, Emmons, Larsen & Griffin, 1985). While no
significant causal effect between life satisfaction and pathological gaming has been found as of yet, research have indicated strong associations. In particular a lower perceived life satisfaction has been found to correlate with online game addiction (Ko et al., 2005). It has been argued to predict pathological gaming as online games provide a means to counterbalance dissatisfaction in real life (Wan & Chiou, 2006). Additionally, studies have shown that symptoms of depression, a related construct to lower levels of perceived life satisfaction, also correlate with the pathological
involvement of games (Seo, Kang & Yom, 2009). As such, lower life satisfaction is expected to predict IGD (H3).
Research on the relation between loneliness and pathological gaming shows contradictory results, as a longitudinal study by Seay and Kraut (2007) found loneliness to be a significant predictor of problematic gaming, while the study by Lemmens et al. (2011) indicates that loneliness acts as both a cause and a
consequence of pathological gaming. Loneliness, referring to an unpleasant
experience that derives from important deficiencies in an individual’s social network and heavily dependent on peer influences (Peplau & Perlman, 1982; Bauminger & Kasari, 2000), may be alleviated by playing (online) games due to the inexpensive and effortless means of getting into contact with others and avoiding feelings of loneliness (Chappell, Eatough, Davies & Griffiths, 2006). In turn, it has been argued that playing games might displace and limit the possibilities to connect and socialize
with other individuals in real life, leading to increased feelings of loneliness (Caplan, 2003, Lemmens et al., 2011). As such, loneliness is hypothesized as being both a cause and consequence of IGD in this study (H4).
Factors of personality/temperament and pathological gaming
Impulsivity is a personality trait characterized by behavioural inhibition, including a higher propensity for rapid, unplanned reactions to internal or external stimuli without regard to the negative consequences of these reactions (Barratt & Patton, 1983). In particular the failure to self-control has been found to correlate with different types of addictive behaviours, including internet addiction and online game addiction (Young, 1998; Kim et al., 2008). Additionally, a longitudinal study by Gentile et al. (2011) found that greater impulsivity acts as significant risk factor for developing pathological gaming behaviour. The relation between impulsivity and pathological internet use can be explained through the notion that pathological use of these media can be characterized as an impulse control disorder similar to the
symptoms of pathological gambling (Young, 1998; Cao, Su, Liu & Gao, 2007). Higher levels of impulsivity are therefore expected to act as a predictor of IGD (H5).
Related to impulsivity, ADHD has also been found to be associated to pathological involvement with games and other forms of problematic behaviour (Walther, Morgenstern & Hanewinkel, 2012; Chan & Rabinowitz, 2006). ADHD constitutes three primary symptoms, which include poor sustained attention,
impulsiveness and hyperactivity, which typically arise early in childhood and remain persistent over time (Barkley, 1997). According to a longitudinal study by Molina and Pelham (2003), adolescents that engage in substance abuse reported to display
symptoms of ADHD. Further, problematic internet use has also been found to relate to ADHD (Weinstein & Lejoyeux, 2010; Yoo et al. 2004). However, it remains
unclear whether playing video games increases ADHD symptoms, or whether those that show ADHD symptoms are more attracted to videogames. As the symptoms of ADHD among children and adolescents have been ascribed to both genetic and environment factors (Hudziak, Derks, Althoff, Rettew & Boomsma, 2005), it is hypothesized that individuals who show more symptoms of ADHD are more likely to show higher levels of IGD (H6).
Family conflict and media behaviour
In line with media effect theories and problem behaviour theories, social context can contribute to the development of particular media use and problematic behavioural tendencies among adolescents (Valkenburg & Peter, 2013; Donovan et al., 1991). However, research on the explanatory nature of situational factors such as family and school on pathological gaming remains scarce. As adolescents primarily consume media at home, family present a particular interesting situational variable to take into consideration. Media habits have already been found to differ across families with different communication patterns (Lull, 1980). Family conflict can be argued to be associated with problematic use of media, as media is often primarily consumed in the home environment. Additionally, in families with high levels of conflict, low family involvement, and inadequate parental monitoring, adolescent problem behavior has been found to become more likely (Ary et al., 1999). In a study by De Leo and Wulfurt (2013) higher family conflict was also found to coincide with higher levels of internet addiction. Here, internet addiction could also include heightened levels of online game play. As such, increased levels of family conflict are expected to act as a predictor of IGD (H7).
Method Sample
A two-wave longitudinal panel survey study was conducted among Dutch adolescents from 6 randomly selected schools located in different parts of the country. In October 2013, 1431 adolescents participated in the first wave of the study (44.7% male). The second wave was fielded six months later, in which 1056 adolescents participated (43.8% male). This respondent attrition of 375 respondents (26.2%) for the second wave was mainly due to the final examination and graduation period, illness, or unavailability of supervising teachers. Further, 123 respondents could not be matched between waves because of discrepancies in respondents’ names or student numbers. Between the two waves, a total of 933 questionnaires were matched. Within this sample, the age of the respondents ranged from 11 to 17, with a mean age of 13.3 (SD = 1.1) in wave 1 and 13.7 (SD = 1.2) in wave 2. Within the sample of 933
matched respondents, 647 adolescents indicated to have played games throughout the two waves. Three respondents were eliminated for reasons of systematic missing values, and one extreme univariate outlier was detected across the predictor variables. Random missing values within variables in the dataset were imputed by means of regression imputation in Amos 20. The resulting final sample consisted of 643 game-playing adolescents, with an average age of 13.7 (SD = 1.2, 59.6% male). After matching the responses between waves, respondents’ names and student numbers were deleted from the final dataset.
Procedure
Respondents were recruited from various randomly selected high schools throughout the Netherlands. After acquiring active consent from the schools and teachers, and passive consent from respondents’ parents, respondents filled in a
paper-and-pencil survey. All surveys were filled in during school hours. Respondents were assured that their answers were handled anonymously, and that teachers, parents or fellow students would not be informed of any individual results. To match the responses across the two waves, respondents were asked to fill in their name or their personal student number, or both. Respondents who indicated not having played any video game in the last six months were asked to skip the questions on game related topics. The survey was presented in the Dutch language. It took respondents on average 20 minutes to complete the survey, after which they were thanked for their participation and received a small present.
Measures
Internet Gaming Disorder (IGD). The short version of the Internet Gaming Disorder (IGD) index (Lemmens et al., in press) was used to assess the extent to which adolescents displayed disordered gaming behaviour. Note that the index items are not only applicable to online games, but games in general. The index consisted of 9 items, for which respondents were asked whether they had experienced the
situations expressed in statements in the past six months. The nine items corresponded to the nine dimensions proposed by the DSM-5 (APA, 2013): Preoccupation,
tolerance, withdrawal, persistence, escape, problems, deceit, displacement and severe consequences. Sample items include ‘in the past six months, did you lose interest in other hobbies or activities because you wanted to play games?’, and ‘in the past six months, did you keep the amount of time you actually spent on games a secret?’ In compliance with the DSM-5 guidelines, respondents could answer on a dichotomous answer scale with either yes or no. The items were summed to form a continuous scale ranging from 0 to 9, with higher scores indicating higher levels of IGD. The mean score in wave 1 was 1.03 (SD = 1.48), and in wave .90 (SD = 1.49).
Social Competence. Social competence was assessed with four items based on instruments for measuring social skills and interpersonal competence among adolescents by Buhrmester et al. (1988). Respondents were asked to rate how difficult or easy they found the following four situations: 1) ‘starting a conversation with a stranger’, 2) expressing my feelings to someone else’, 3) ‘introducing myself to someone I have not met before’, and 4) ‘talking to someone about something I feel ashamed of’. Responses were provided on a 5-point scale, ranging from very difficult (1) to very easy (5). The items were averaged to create the scale scores, with higher scores indicating greater social competence. The four items were found to form a reliable scale in both the first (M = 3.14, SD = .73, α = .73) and second wave (M = 3.15, SD = .80, α = .78).
Self-esteem. Self-esteem was measured with six items from the Rosenberg Self Esteem Scale (Rosenberg et al., 1989). Respondents were asked to indicate whether they agreed or disagreed with statements such as ‘I feel that I have a number of good qualities’ and ‘I feel I do not have much to be proud of’ (reverse coded), on a Likert scale ranging from 1 (totally disagree) to 5 (totally agree). The items were averaged to create the scale scores, with higher scores reflecting greater self-esteem. After reverse coding two items, Cronbach’s alpha for this scale was .81 (M = 3.89, SD = .64) for the first wave, and .82 (M = 3.82, SD = .66) for the second wave.
Life satisfaction. Selected items from the Satisfaction with Life Scale were used to assess life satisfaction (Diener et al., 1985). A 5-point Likert scale (1 = totally disagree to 5 = totally agree) measures the overall judgement of one’s life with five items. Sample items include ‘In the past six months I have been satisfied with my life’ and ‘In the past six months my life was in most ways close to my ideal’. The items formed a reliable scale, where higher scores reflect a greater life satisfaction, in both
the first wave (M = 3.73, SD = .74, α = .81) and the second wave (M = 3.66, SD = .81, α = .89).
Loneliness. Respondents’ loneliness was measured with the five items with the highest item-total correlations from the 20-item UCLA loneliness scale (Russell, 1996). Respondents were asked to indicate whether they agreed or disagreed with statements on a 5-point Likert response scale, ranging from totally disagree (1) to totally agree (5). Sample items include ‘In the past six months I felt alone’ and ‘In the past six months I felt that no one really knows me well’. Higher scores on this scale indicate a greater sense of loneliness. The items were averaged to create the scale scores and were found to form a reliable scale in both the first (M = 1.71, SD = .75, α = .86) and second wave (M = 1.79, SD = .83, α = .91).
Impulsivity. Adolescents’ level of impulsivity was measured by means of four selected items obtained from the Barratt Impulsiveness Scale (Barratt, 1959). Respondents were asked to rate on a Likert scale ranging from 1 (totally disagree) to 5 (totally agree), the extent to which the following statements applied to themselves: 1) ‘I do things without thinking’ 2) ‘I act on impulse’ 3) ‘I act on the spur of the moment’ 4) ‘I make decisions quickly’. Higher scores on this scale indicate greater level of impulsivity. The four items formed a reliable scale in the first (M = 2.77, SD = .73, α = .71) and in the second wave (M = 2.84, SD = .78, α = .67).
ADHD. To assess the extent to which respondents displayed symptoms of Attention Deficit Hyperactivity Disorder, nine items from the DSM-4 (APA, 2000) were used. In particular, the items focused on attention deficit. Respondents were asked to indicate how often particular situations were applicable to themselves on a 5 point scale, as proposed by Kessler et al. (2005), ranging from never (1) to very often (5). Example items are ‘I am easily distracted’ and ‘I have difficulties organizing
tasks’. After averaging the items to create scale scores, Cronbach’s alpha was found to be .86 (M = 2.42, SD = .66) for the first wave, and .89 (M = 2.56, SD = .75) for the second wave. Higher scores on this scale indicate that respondents show more
symptoms of ADHD.
Family conflict. Five items of the conflict subscale of the Family Environment Scale were used to measure the degree of family conflict (Moos & Moos, 1994). Respondents were asked to indicate how often a variety of scenarios regarding criticizing each other, hitting each other, arguing, cursing and throwing with things within the family occurred, with response options ranging from never (1) to very often (5). Sample items include ‘How often do people criticize each other at home’ and ‘How often do people curse at each other in your home’. Scores were averaged to create scales, with higher scores reflecting greater family conflict. The scale showed reliability as indicated by a Cronbach’s alpha of .76 (M = 2.14, SD = .63) for the first wave, and .79 (M = 2.19, SD = .69) for the second wave.
Data Analysis
!! To investigate the causal relationship between IGD and the constructs of psychosocial well-being, ADHD and situational constructs, eight separate two-wave autoregressive cross-lagged panel models were analysed in structural equation modelling (Amos 20). In autoregressive structures, the past values of a variable are used to predict future variables of the same variable plus random error, thus
incorporating lagged effects among the indicators (Kline, 2011). This is reflected in the stability coefficients included in the model that, together with cross-lagged coefficients, predict aggregate change over two time points, allowing for a more ‘pure’ influence of each construct of interest (see Schlueter, Davidov & Schmidt, 2006, for more information). The cross-lagged paths included in the model test the
causal-correlational longitudinal relationship between IGD and the considered cause and effect factors, additionally allowing for the exploration of reciprocal effects. The variable for IGD was included as a manifest variable, while the considered factors in the analyses represented latent variables. The indicators for each latent variable at both time points were included in the tested models.
The variables included in the analyses were non-normally distributed, assessed through Shapiro-Francis tests as well as through a visual inspection of histograms and normal probability plots, resulting in the violation of the assumption of multivariate normality. As this assumption is necessary for accurate significance testing, bootstrap analyses were run for each structural equation model using Maximum Likelihood estimation (2000 bootstrap samples). Bootstrapping refers to the method of resampling as developed by Efron (1979), and through repetition simulates the drawing of numerous random samples from a population (Kline, 2011). As such, bootstrap analyses provide an estimation approach without making assumptions about the distribution of variables.
The two-step approach as described by Bollen (1989) was employed for each model, consisting of the identification of a measurement model in the first step and the structural part of the model in the second step (see Kline, 2011, for more information). In the respecification of the measurement models, modifications of measurement error patterns were considered when covariances between error terms of indicators within one time point were suggested. Additionally, error terms of
indicators at time point 1 were allowed to covary with corresponding indicators at time point 2, as is common in longitudinal structural equation models (Little, 2009). For each suggested modification, a chi-square difference test was performed to assess significant improvement of the model. The final models are presented in the results.
To determine model fit, the following measures were collectively considered: chi-square, CFI, RMSEA, and 90% confidence interval. A non-significant chi-chi-square, a CFI over .90, RMSEA below .05, and 90% confidence interval (CI) below .05 for lower bound and below .10 for higher bound, ideally, point towards a good model fit (Kline, 2011; Barrett, 2007).
Results Descriptive results and zero-order correlations
According to the DSM-V, an individual should meet at least five or more out of the nine criteria to be diagnosed with Internet Gaming Disorder (IGD). In the first wave, 24 respondents (2%) fit the diagnosis of IGD, whereas 25 respondents (2.3%) in the second wave met this diagnostic cut-off point. Only 1% of the adolescents (n = 10) reported having experienced five or more criteria across both waves. The zero-order correlations for the considered variables within and between the two waves are presented in Table 1. Except for the variables self-esteem and social competence, all factors were significantly correlated with either IGD in wave 1 or IGD in wave 2. As Table 1 shows, family conflict and ADHD were significantly correlated with IGD both within and between waves. Furthermore, all significant zero-order correlations were in the expected direction; higher impulsivity, lower life satisfaction, higher loneliness, more ADHD symptoms and more family conflict were related to higher scores on IGD in both waves. Between waves, IGD was also significantly correlated (r = .53, p < .001). Because the data is non-normally distributed, the non-significant correlations found for self-esteem and social competence might be due to variance. These variables are therefore still modelled in autoregressive structural equation models, as these eliminate a considerable proportion of potentially confounding
! Fa m il y Co n fl ic t W2 - W1 - *** .62 ADHD W2 - *** .23 *** .31 W1 - *** .66 *** .34 *** .27 Lo n el in es s W2 - .27 *** .3 3 *** .1 9 *** .3 1 *** W1 - *** .53 *** .30 *** .27 *** .24 *** .22 Li fe S at is fa ct io n W2 - -.36 *** -.5 4 *** -.2 8 *** -.3 7 *** -.2 1 *** -.3 2 *** W1 - *** .49 -.4 7 *** -.3 9 *** -.3 4 *** -.2 6 *** -.2 4 *** -.2 5 *** Se lf E st ee m W2 - .31 *** .5 6 *** -.3 7 *** -.5 2 *** -.2 6 *** -.2 8 *** -.1 9 *** -.2 7 *** W1 - .53 *** .4 8 *** .3 9 *** -.5 0 *** -. 43 *** -.3 7 *** -.3 1 *** -.2 3 *** -.2 1 *** Im p u lsi vi ty W2 - -.16 *** -.1 2 ** -.1 6 *** -.1 9 *** .1 8 *** .2 1 *** .3 4 *** .4 9 *** .1 5 *** .2 5 *** W1 - .47 *** -.1 9 *** -.1 1 ** -.2 0 *** -.1 5 *** .1 5 *** .1 9 *** .4 8 *** .3 4 *** .2 3 *** .1 8 *** So ci al Co m p et en ce W2 - -.03 .02 .25 ** .2 7 *** .1 6 *** .3 0 *** -.2 4 *** -.3 4 *** -.1 6 *** -.1 7 *** -.0 5 -.0 8 * W1 - *** .62 .0 4 .0 4 .2 8 *** .1 9 *** .2 5 *** .2 6 *** -.2 7 *** -.2 6 *** -.1 8 *** -.1 3 ** -.0 9 * -.0 6 IG D W2 .03 -.0 2 .0 4 .1 6 *** .0 2 -.0 7 -.0 3 -.1 2 .0 8 .1 2 ** .1 4 *** .2 2 *** .1 1 ** .1 0 * W1 -.02 .06 .1 6 *** .1 2 ** -.1 0 * -.0 5 -.1 3 ** -.0 9 * .1 7 *** .0 4 .2 7 *** .1 6 *** .1 6 *** .0 8 * W1 W2 W1 W2 W1 W2 W1 W2 W1 W2 W1 W2 W1 W2 So ci al Co m p et en ce Im p u lsi vi ty Se lf E st ee m Li fe Sa ti sf ac ti on Lo n el in es s ADHD Fam il y Co n fl ic t N ot e. * p < .05. ** p < .01. *** p < .001 ! T abl e 1. C or re lat ions w it hi n and be tw ee n w av es 1 and 2 (N = 463)
variance (Schlueter et al., 2006), but interpretation should be approached with caution.
Cross-lagged effects between IGD and factors of psychosocial well-being To examine whether the underlying constructs of psychosocial well-being, identified as life satisfaction, loneliness, social competence and self-esteem, act as predictors for IGD, or involve a reciprocal relationship as expected for loneliness, four autoregressive structural equation models were tested for each of these latent construct.
The first hypothesis included the construct of social competence, where lower levels of social competence at time point 1 were expected to predict higher levels of IGD at time point 2. Estimation for this model converged to an acceptable model fit as reflected in the fit indices, however, chi-square remained significant, χ2 (26, N = 643) = 69.90, p < .001, CFI = .98, RMSEA = .05 (90% CI: .037, .066). The model shows a significant positive effect from IGD at time point 1 to social competence, predicting an increase in social competence at time point 2 (β = .08, B = .02, p < .05; see Figure 1 for an overview). The model revealed a non-significant positive effect of social competence at time point 1 on IGD at time point 2 (β = .05, B = .21, p = .154). Lower levels of social competence, thus, do not predict higher levels of IGD over the course of six months. Hence, the first hypothesis is not supported. Interpretation of this model should be met with caution, since social competence was uncorrelated with IGD both within and between waves.
Note. * p < .05. ** p < .01. *** p < .001.
Figure 1. Standardized regression weights of IGD and social competence between waves 1 and 2
The second hypothesis focused on esteem, for which lower levels of self-esteem at time point 1 were expected to predict higher levels of IGD at time point 2. The model revealed acceptable fit indices, but a significant chi-square, χ2 (63, N = 643) = 135.44, p < .001, CFI = .98, RMSEA = .04 (90% CI: .032, .052). Self-esteem at time point 1 appears to significantly and positively predict IGD at time point 2 (β = .08, B = .21, p < .05), indicating that higher levels of self-esteem predict higher levels of IGD over the course of six months (see Figure 2 for an overview). Higher levels of IGD at time point 1 did not predict change in levels of self-esteem at time point 2 (β = .02, B = .01, p = .578). These results indicate that self-esteem does act as a predictor of IGD over time, but not in the expected direction. As such, the second hypothesis is not supported. Again, interpretation of this model should be met with caution, since self-esteem was not correlated with IGD within each of the two waves.
Note. * p < .05. ** p < .01. *** p < .001.
Figure 2. Standardized regression weights of IGD and self-esteem between waves 1 and 2
For the third hypothesis, regarding the construct of life satisfaction, lower levels of life satisfaction at time point 1 were expected to predict higher levels of IGD at time point 2. An acceptable model fit is obtained for the final model, reflected in acceptable fit indices and chi-square approaching non-significance, χ2 (44, N = 643) = 62.30, p = .036, CFI = 1.0, RMSEA = .03 (90% CI: .007, .039). Cross-lagged effects revealed a non-significant negative effect from IGD at time point 1 on life satisfaction at time point 2 (β = -.02, B = -.01, p = .634), and a non-significant positive effect from life satisfaction at time point 1 to IGD at time point 2 (β = .04, B = .11, p = .286). These findings indicate that lower levels of life satisfaction do not significantly predict higher levels of IGD six months later; the third hypothesis is not supported.
The fourth hypothesis concerns the construct of loneliness, hypothesized as being both a cause and consequence of IGD. Higher levels of loneliness at time point 1 were expected to predict higher levels IGD at time point 2, whereas higher levels of
IGD at time point 1 were expected to increase feelings of loneliness. Model fit indices proved acceptable, yet chi-square remained significant, χ2 (44, N = 643) = 76.20, p < .01, CFI = 1.0, RMSEA = .03 (90% CI: .020, .046). The model revealed
nonsignificant negative effects of loneliness at time point 1 on IGD at time point 2 (β = -.06, B = -.03, p = .099), and a non-significant effect of IGD at time point 1 on loneliness at time point 2 (β = -.01, B = -.03, p = .714). These findings indicate that loneliness does not act as a predictor and consequence of IGD, and higher levels of loneliness acts neither as a predictor nor as a consequence of IGD over the course of six months. Hence, the fourth hypothesis is not supported.
Cross-lagged effects between IGD and factors of personality/temperament To assess whether the factors of personality/temperament identified in this study act as predictors to the development of IGD, two separate autoregressive structural equation models for impulsivity and ADHD were tested. For the fifth hypothesis, a higher level of impulsivity at time point 1was expected to result in higher levels of IGD at time point 2. The final model revealed acceptable fit indices yet a significant chi-square, χ2 (26, N = 643) = 52.15, p = .002, CFI = .98, RMSEA = .04 (90% CI: .024, .055). The cross-lagged paths revealed no significant effects between IGD at time point 1 and impulsivity at time point 2 (β = .02, B = .01, p = .620), nor between impulsivity at time point 1 and IGD at time point 2 (β = .07, B = -.16, p = .087). As such, higher levels of impulsivity do not predict a significant increase in levels of IGD six months later. Therefore, the fifth hypothesis is not supported.
The sixth hypothesis evaluates whether higher reports of ADHD symptoms at time point 1 predicts higher levels of IGD at time point 2. The model revealed
chi-square, χ2 (140, N = 643) = 224.83, p < .001, CFI = .99, RMSEA = .03 (90% CI: .023, .038). The cross-lagged paths showed non-significant negative effects from IGD at time point 1 to symptoms of ADHD at time point 2 (β = -.05, B = -.02, p = .116). Similarly, it showed non-significant negative effects of symptoms of ADHD at time point 1 to IGD at time point 2 (β = -.01, B = -.01, p = .885). These results indicate that symptoms of ADHD do not significantly predict higher levels of IGD six months later. As such, the sixth hypothesis is not supported.
Cross-lagged effects between IGD and family conflict
To assess whether the situational factor of family conflict predicts higher levels of IGD over time, two additional autoregressive structural equation models were tested. Here, more family conflict at time point 1 was expected to predict higher levels of IGD at time point 2. An acceptable model fit is obtained for the final model reflected in acceptable fit indices, yet chi-square remains significant, χ2 (42, N = 643) = 91.46, p < .001, CFI = .98, RMSEA = .04 (90% CI: .031, .055). The crossed-lagged effect from family conflict at time point 1 to IGD at time point 2 is non-significant (β = .01, B = .04, p = .741). Further, the model reveals a non-significant negative effect of IGD at time point 1 on family conflict at time point 2, (β = -.04, B = -.02, p = .247). These findings indicate that higher levels of family conflict are not a significant predictor of IGD over the course of six months; the seventh hypothesis is therefore not supported.
Discussion
With the inclusion of Internet Gaming Disorder (IGD) in the Diagnostic and
Statistical Manual of Mental Disorders (DSM-5), problematic gaming is now being recognized as an official clinical disorder. Pathological gaming, often also termed ‘game addiction’, ‘problematic gaming’ or gaming dependence’, has long been a topic
of interest within media effect research. Due to the different definitions, and the variety of measurement criteria that have been used to research this concept, some inconsistencies remain with regard to the prevalence, predictors and consequences. The current study focused on determining which factors predict the development of IGD over time by including those factors that have previously been found to
significantly predict constructs theoretically close to IGD (e.g. game addiction, problematic gaming, or pathological gaming). Additionally, this study also examined whether the situational factor of family conflict could be of influence in the
development of IGD. The main research aim was therefore to understand how, over the course of six months, the set of factors psychosocial well-being,
personality/temperament and family conflict causally related to the newly classified Internet Gaming Disorder. Using autoregressive cross-lagged panel models for each latent construct, the causal relations between these factor and IGD could be examined in more detail. This way, both cause and effect relations with IGD were modelled, which allowed for a proper assessment of whether the examined factors indeed solely act as predictors, effects, or whether a reciprocal relationship was in order.
The causal analyses yielded two important findings. First, in contrast to the expected effect, social competence appears to be predicted by IGD, rather than predict pathological involvement with games. Specifically, the findings indicate that higher levels of IGD predict a small increase in perceived social competence over time. This diverges from previous longitudinal research on pathological gaming that found social competence to be a risk factor for the development of pathological gaming (Gentile et al., 2011; Lemmens et al., 2011). It has often been suggested that the world of
(online) games offers a safe and secure environment for people who are socially unskilled, allowing for an anonymous setting where their unmet social needs in real
life can be compensated for, and where real-life social situations can be avoided (Young, 1998; Chak & Leung, 2004; Peters & Malesky, 2008). Taking this even further, Chiu et al. (2004) argue that playing games may replace the development of social relationships and could consequently result in the destruction of social skills. The finding in this study point towards the exact opposite, suggesting that increased levels of pathological gaming can aid the development of social skills over time. The second important finding relates to the effect of self-esteem, with higher levels of self-esteem predicting an increase in levels of IGD. This finding again is incongruent with previous research, where lower levels of self-esteem have been found to act as a predictor of pathological gaming (Lemmens et al., 2011; Roe & Muijs, 1998). This finding is generally interpreted as an indication that being successful in games allows for compensation, or serves as a comforter against negative feelings associated with a weak self-image (Roe & Muijs, 1998). However, the finding in the current study reverses this perspective, suggesting that individuals who are more confident are more likely to engage in pathological gaming over time. It seems, thus, that both the effect of self-esteem and social competence combined point towards a more positive perspective on pathological gaming. While this seems counterintuitive, the motivations to engage in video game play provide an interesting framework. Elements of competition and challenge within games have been found to serve as motivations along with the social characteristic mainly found in massively multiplayer online role playing games (Ducheneaut & Moore, 2005; Ryan, Rigby & Przybylski, 2006; Hoffman & Nadelson, 2010). Here, Vorderer et al. (2003) relate the competitive process of gaming to self-esteem, in particular within a social context where players compete with an opponent. This evokes social
self-esteem, additionally striving for positive moods (Vorderer et al., 2003). A heightened sense of confidence is needed, where individuals perceive themselves as being capable to achieve victory in social competition. In turn, Ducheneaut and Moore (2005) argue that the social nature of multiplayer online games might improve social skills. Players continually confront situations such as meeting new people, leading a small group of players, coordinate and work in teams, participate in sociable interaction with other players for mutual advancement. As such, the positive effects found for self-esteem and social competence might be explained by the particular preference for social competition in multiplayer online games. It is those players who have a heightened sense of self-esteem that seek social competition, thereby
enhancing particular social skills that can be used in offline settings as well and maintaining his or her self-esteem in the process.
Particularly important to the above-discussed findings, however, is the sample of gamers used in this study. Based on the diagnosis criteria of the DSM-5, only 2.3% of the adolescents in wave 2 fit the diagnosis of IGD, reporting five or more criteria that apply to their situation. Moreover, only 1% met five or more criteria of IGD over the course of both waves. Therefore, the variances within the measure of IGD largely stem from those having selected one to four criteria, and thus mostly represent regular gamers. The positive effects, as such, are not surprising when taking into account that the adolescents mainly display regular gaming behaviour as opposed to problematic gaming behaviour. This would additionally in part explain why life satisfaction, loneliness, impulsivity, ADHD and family conflict did not yield any significant effects. In particular impulsivity and the construct of loneliness underlying
psychosocial well-being have been found to predict pathological gaming in previous longitudinal studies (Gentile et al., 2011; Lemmens et al., 2011). These constructs did
however significantly correlate with IGD within waves, except for life satisfaction in the second wave, indicating linear relationships. The situational factor of family conflict, while no significant effects were found, was significantly and positively correlated with IGD both within and between waves. This is in line with previous research on different problematic media use, which found family conflict to be associated with problematic internet behaviour (De Leo & Wulfert, 2013).
A different reason for the non-significant effects of the constructs considered in this study concerns the drawbacks of a two-wave panel design. Kenski and Romer (2006) mentioned the importance of the time period between waves, and how this can possibly have an effect on the analysis, in terms of both the ability to observe change and the potential predictors of that change. When a selected time period between waves is too long, the causal effect of interest might have already diminished and no longer evident, resulting in unjust dismissal of causal relationships (Kenski & Romer, 2006). Similarly, a too limited time window may not permit enough measurable change to occur (Kenski & Romer, 2006). Within media effect research, longitudinal analysis is often conducted with half yearly or yearly time lags (e.g. Aubrey, 2006; Gentile et al., 2011; Peter & Valkenburg, 2011). The selected time period of six months between waves within this study could therefore have been ineffective in capturing the measurable change in IGD and its considered predictors. Pathological behaviour in general can be understood as a progressive process (Nakken, 2013). As such, the measure of IGD in particular, offering a more exhaustive measure of pathological gaming, indicates a development over a longer period of time and could possibly not have actualized in considerable change over the course of six months. An additional drawback of a two-wave panel design is the problem of
of the content of the questions, and the subsequent different responses due to the recognition of the same questions in the second wave. Respondents may give special attention to particular issues due to the prompting of thought processes after the initial survey (Reis & Judd, 2000). Additionally, respondents may have the desire to appear consistent in their answers across waves (Reis & Judd, 2000). Consequently, this problematizes the interpretation of change, as it is uncertain whether the effects are in fact due to actual change of behaviour or the process of repeated questioning. In light of these drawbacks, the causal findings within this study should be interpreted with caution.
To effectively determine the complex nature of IGD and its set of predictors, further research is needed with a longitudinal design that is able to capture the possible slower progressive nature of IGD, in particular when examining younger adolescents. Carefully selected time periods between waves could yield more detailed results, as well as the inclusion of three or more waves to establish a more discerning baseline. Further, while this study chose to adopt the IGD index in compliance with DSM-5 guidelines, the use of the IGD scale (Lemmens et al., in press) as opposed to the index might have allowed for more variance. The use of the dichotomous answer scale might be too strict, capturing solely the extremes of problematic gaming behaviour, and none of the smaller effects in change over time.
Other limitations of this study that possibly affected the generalizability of the results include the reliance on self-reports, and the sampling procedure. While the schools were selected at random, the selection of adolescents within each school was based on convenience. That is, the inclusion of participating adolescents depended on which classes the main contact person of each school could provide. Although the sample size was large enough to represent a wide range of characteristics, thus
allowing for substantial inferences about the larger population, certain over- or underrepresentation of particular peer group cultures should be accounted for.
Additionally, an interesting approach for further research would be to include gender as a moderator, and verify whether the predictors for pathological involvement in gaming for adolescent boys are different from adolescent girls. While Lemmens et al. (2011) did not find a moderating effect of gender for the psychosocial causes and consequences of pathological gaming, there is mounting evidence of gender
differences in the prevalence and development patterns of pathological gaming (Chiu et al., 2004; Ko et al., 2005; Gentile, 2009), which might be further understood through the diagnostic criteria of IGD.
Conclusion
Despite the presented limitations and the failure to find support for the expected antecedent structure including the considered personality and situational factors on the development of Internet Gaming Disorder (IGD), this study does offer valuable insights in problematic gaming behaviour among adolescents. This study is the first to longitudinally examine how factors of psychosocial well-being,
personality/temperament and family conflict causally relate to IGD among a large sample of Dutch adolescent gamers. Self-esteem has been identified as a small positive predictor of IGD. The addition of modelling situational factors in relation to IGD, addressing a gap in the literature on pathological gaming, revealed a significant association with IGD both within and across different time points. The understanding of IGD is further enriched by the low prevalence of IGD among adolescents found in this study, and the implication that IGD likely involves a slow gradual process of change, which are factors to be accounted for in future research.
These findings inconsistent with previous research have practical implications for the understanding and clinical diagnosis of IGD. The prevalence of IGD among adolescents appears to be much lower than previous measurements of pathological gaming have indicated, suggesting that problematic gaming behaviour is less of a common problem among this young age group than previously thought. However, this does not mean that parents should not be attentive to signs of pathological gaming, as most initial pathological patterns identified in adulthood developed during
adolescence (Wagner & Anthony, 2002).
The nine dimensions captured in the measurement of IGD constitute an
important new shift within research on pathological gaming. Replication studies using the measurement of IGD can help further validate the prevalence and progressive nature of pathological gaming, additionally allowing for a solid differentiation between different age groups, and determining the role of extraneous variables. The criteria specifically designed to effectively diagnose Internet Gaming Disorder, as described in the DSM-5, can serve as a foundation for future research on the topic, generating a new line of consistent measurement for the notion of pathological gaming within media effect research and beyond.
References
APA (2000). Diagnostic and Statistical Manual of Mental Disorders, 4th edition, Text Revision. Washington, DC: American Psychiatric Association. APA (2013). Diagnostic and Statistical Manual of Mental Disorders, 5th edition.
Washington, DC: American Psychiatric Association.
Ary, D. V., Duncan, T. E., Biglan, A., Metzler, C. W., Noell, J. W., & Smolkowski, K. (1999). Development of adolescent problem behavior. Journal of Abnormal Child Psychology, 27(2), 141-150.
Aubrey, J. S. (2006). Effects of sexually objectifying media on self-objectification and body surveillance in undergraduates: Results of a 2-year panel study. Journal of Communication, 56(2), 366-386.
Barkley, R. A. (1997). Behavioral inhibition, sustained attention, and executive functions: Constructing a unifying theory of ADHD. Psychological Bulletin, 121(1), 65.
Barrett, P. (2007). Structural equation modelling: Adjudging model fit. Personality
and Individual Differences, 42(5), 815-824.
Barratt, E. S. (1959). Anxiety and impulsiveness related to psychomotor efficiency. Perceptual and Motor Skills, 9(3), 191-198.
Bauminger, N., & Kasari, C. (2000). Loneliness and friendship in high-functioning children with autism. Child Development, 71(2), 447-456.
Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley. Chiu, S. I., Lee, J. Z., & Huang, D. H. (2004). Video game addiction in children and
teenagers in Taiwan. CyberPsychology & Behavior, 7(5), 571-581.
internet addiction in a sample of Chinese adolescents. European Psychiatry, 22(7), 466-471.
Caplan, S. E. (2003). Preference for online social interaction A theory of problematic internet use and psychosocial well-Being. Communication Research, 30(6), 625-648.
Chak, K., & Leung, L. (2004). Shyness and locus of control as predictors of internet addiction and internet use. CyberPsychology & Behavior, 7(5), 559-570. Chan, P. A., & Rabinowitz, T. (2006). A cross-sectional analysis of video games and
attention deficit hyperactivity disorder symptoms in adolescents. Annals of General Psychiatry, 5(16), 1-10.
Chappell, D., Eatough, V., Davies, M. N., & Griffiths, M. (2006). EverQuest—It’s just a computer game right? An interpretative phenomenological analysis of online gaming addiction. International Journal of Mental Health and Addiction, 4(3), 205-216.
Charlton, J. P., & Danforth, I. D. (2007). Distinguishing addiction and high engagement in the context of online game playing. Computers in Human Behavior, 23(3), 1531-1548.
Csikszentmihalyi, M. (1975). Beyond boredom and anxiety. San Francisco: Jossey- Bass Publishers.
Davis, R. A. (2001). A cognitive-behavioral model of pathological Internet use. Computers in Human Behavior, 17(2), 187-195.
De Leo, J. A., & Wulfert, E. (2013). Problematic Internet use and other risky behaviors in college students: An application of problem-behavior theory. Psychology of Addictive Behaviors, 27(1), 133.
Demo, D. H., & Savin-Williams, R. C. (1992). Self-concept stability and change during adolescence. Self-perspectives Across the Life Span, 4, 116-150. Desai, R. A., Krishnan-Sarin, S., Cavallo, D., & Potenza, M. N. (2010). Video-
gaming among high school students: health correlates, gender differences, and problematic gaming. Pediatrics, 126(6), e1414-e1424.
Diener, E. D., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The satisfaction with life scale. Journal of Personality Assessment, 49(1), 71-75.
Donovan, J. E., Jessor, R., & Costa, F. M. (1991). Adolescent health behavior and conventionality-unconventionality: An extension of problem-behavior therapy. Health Psychology, 10(1), 52.
Ducheneaut, N., & Moore, R. J. (2005). More than just ‘XP’: learning social skills in massively multiplayer online games. Interactive Technology and Smart Education, 2(2), 89-100.
Efron, B. (1979). Bootstrap methods: Another look at the jackknife. The Annals of Statistics, 7(1), 1-26.
Fisher, S. (1994). Identifying video game addiction in children and adolescents. Addictive Behaviors, 19(5), 545-553.
Gentile, D. A. (2009). Pathological video game use among youth 8 to 18: A national study. Psychological Science, 20(5), 594–602.
Gentile, D. A., Choo, H., Liau, A., Sim, T., Li, D., Fung, D., & Khoo, A. (2011). Pathological video game use among youths: A two-year longitudinal study. Pediatrics, 127(2), e319-e329.
Griffiths, M. D., Davies, M. N., & Chappell, D. (2004). Online computer gaming: A comparison of adolescent and adult gamers. Journal of Adolescence, 27(1), 87-96.
Griffiths, M. D., & Hunt, N. (1998). Dependence on computer games by adolescents. Psychological reports, 82(2), 475-480.
Griffiths, M., & Wood, R. T. (2000). Risk factors in adolescence: The case of gambling, videogame playing, and the Internet. Journal of Gambling Studies,16(2-3), 199-225.
Grüsser, S. M., Thalemann, R., & Griffiths, M. D. (2006). Excessive computer game playing: Evidence for addiction and aggression?. CyberPsychology &
Behavior, 10(2), 290-292.
Grüsser, S. M., Thalemann, R., Albrecht, U., & Thalemann, C. N. (2005). Excessive computer usage in adolescents: A psychometric evaluation. Wiener Klinische Wochenschrift, 117(5-6), 188-195.
Hoffman, B., & Nadelson, L. (2010). Motivational engagement and video gaming: A mixed methods study. Educational Technology Research and
Development, 58(3), 245-270.
Howes, C. (1987). Social competence with peers in young children: Developmental sequences. Developmental Review, 7(3), 252-272.
Hudziak, J. J., Derks, E. M., Althoff, R. R., Rettew, D. C., & Boomsma, D. I. (2005). The genetic and environmental contributions to attention deficit hyperactivity disorder as measured by the Conners’ Rating Scales. American Journal of Psychiatry, 162(9), 1614-1620.
Inderbitzen, H. M., & Foster, S. l. (1992). The teenage inventory of social skills: Development, reliability, and validity. Psychological Assessment, 4(4), 451-459.
Jansz, J., & Tanis, M. (2007). Appeal of playing online first person shooter games. CyberPsychology & Behavior, 10(1), 133-136.
Johnstone, S. (2013). Computer gaming and ADHD: Potential positive influences on behavior. Technology and Society Magazine, IEEE, 32(1), 20-22.
Kenski, K., & Romer, D. (2006). Analysis of panel data. In D. Romer, K. Kenski, K. Winneg, C. Adasiewicz & K. H. Jamieson (Eds.), Capturing campaign dynamics 2000 & 2004. The national Annenberg Election Survey (pp. 151-164). Philadelphia: University of Pennsylvania Press.
Kessler, R. C., Adler, L., Ames, M., Demler, O., Faraone, S., Hiripi, E. V. A., ... & Walters, E. E. (2005). The World Health Organization Adult ADHD Self-Report Scale (ASRS): a short screening scale for use in the general population. Psychological Medicine, 35(02), 245-256.
Kim, E. J., Namkoong, K., Ku, T., & Kim, S. J. (2008). The relationship between online game addiction and aggression, self-control and narcissistic personality traits. European Psychiatry, 23(3), 212-218.
Kline, R. B. (2011). Principles and practice of structural equation modeling. New York: Guilford press.
Ko, C., Yen, J., Chen, C., Chen, S., & Yen, C. (2005). Gender differences and related factors affecting online gaming addiction among Taiwanese adolescents. The Journal of Nervous and Mental Disease, 193(4), 273–277.
Ko, C. H., Yen, J. Y., Liu, S. C., Huang, C. F., & Yen, C. F. (2009). The associations between aggressive behaviors and Internet addiction and online activities in adolescents. Journal of Adolescent Health, 44(6), 598-605.
Lemmens, J. S., Valkenburg, P., & Peter, J. (2009). Development and validation of a game addiction scale for adolescents. Media Psychology, 12, 77-95.
Lemmens, J. S., Valkenburg, P. M., & Peter, J. (2011). Psychosocial causes and consequences of pathological gaming. Computers in Human Behavior, 27,
144-152.
Lemmens, J. S., Valkenburg, P. M., & Gentile, D. A. (in press). The Internet Gaming Disorder scale and index.
Lent, R. W. (2004). Toward a unifying theoretical and practical perspective on well- being and psychosocial adjustment. Journal of Counseling Psychology,51(4), 482.
Little, T. D. (2009). Longitudinal structural equation modeling. New York: Guilford Press.
Lo, S. K., Wang, C. C., & Fang, W. (2005). Physical interpersonal relationships and social anxiety among online game players. CyberPsychology & Behavior, 8(1), 15-20.
McGonigal, J. (2011). Reality is broken: Why games make us better and how they can change the world. Penguin.
McMahan, A. (2003). Immersion, engagement and presence. The video game theory reader, 67-86.
Molina, B. S., & Pelham Jr, W. E. (2003). Childhood predictors of adolescent substance use in a longitudinal study of children with ADHD. Journal of Abnormal Psychology, 112(3), 497.
Nakken, C. (2013). The addictive personality: Understanding the addictive process and compulsive behavior. Minnesota: Hazelden Publishing.
Ng, B. D., & Wiemer-Hastings, P. (2005). Addiction to the internet and online gaming. CyberPsychology & Behavior, 8(2), 110-113.
Olson, C. K. (2010). Children's motivations for video game play in the context of normal development. Review of General Psychology, 14(2), 180.
games as facilitator of internet addiction. Dissertation Abstracts International, 66, 8-B.
Peplau, L. A., & Perlman, D. (1982). Perspectives on loneliness. Loneliness: A sourcebook of current theory, research and therapy, 1-18.
Peters, C. S., & Malesky, L. A. (2008). Problematic usage among highly-engaged players of massively multiplayer online role playing games. CyberPsychology & Behavior, 11(4), 481-484.
Reis, H. T., & Judd, C. M. (Eds.). (2000). Handbook of research methods in social and personality psychology. Cambridge: Cambridge University Press. Roe, K., & Muijs, D. (1998). Children and computer games: A profile of the heavy
user. European Journal of Communication, 13(2), 181-200.
Rosenberg, M., Schooler, C., & Schoenbach, C. (1989). Self-esteem and adolescent problems: Modeling reciprocal effects. American Sociological Review, 54(6), 1004-1018.
Ryan, R. M., Rigby, C. S., & Przybylski, A. (2006). The motivational pull of video games: A self-determination theory approach. Motivation and emotion,30(4), 344-360.
Salguero, R. A., & Morán, R. M. B. (2002). Measuring problem video game playing in adolescents. Addiction, 97(12), 1601-1606.
Schlueter, E., Davidov, E., & Schmidt, P. (2006). The dynamics of authoritarianism and anomia: Applying autoregressive cross-lagged and latent growth models to a three-wave panel study. In K. Montfort, H. Oud, & A. Satorra (Eds.), Longitudinal Models in the Behavioral and Related Sciences. Mahwah, New Jersey: Lawrence Erlbaum Publishers, EAM Book Series.
