Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/257253035

StructureandfunctionofmaladaptivecognitionsinPathologicalInternetUseamongChineseadolescents

ARTICLEinCOMPUTERSINHUMANBEHAVIOR·NOVEMBER2012

ImpactFactor:2.69·DOI:10.1016/j.chb.2012.07.009

CITATIONS

12

READS

87

6AUTHORS,INCLUDING:

YujiaoMai

UniversityofNotreDame

1PUBLICATION12CITATIONS

SEEPROFILE

WeiZhang

FudanUniversity

1,001PUBLICATIONS12,826CITATIONS

SEEPROFILE

Allin-textreferencesunderlinedinbluearelinkedtopublicationsonResearchGate,

lettingyouaccessandreadthemimmediately.

Availablefrom:YujiaoMai

Retrievedon:25March2016

Computers in Human Behavior 28 (2012) 2376–2386

Contents lists available at SciVerse ScienceDirect

Computers in Human Behavior

journal homepage: www.elsevier .com/locate /comphumbeh

Structure and function of maladaptive cognitions in Pathological Internet Useamong Chinese adolescents

Yujiao Mai a, Jianping Hu a, Zheng Yan b, Shuangju Zhen a, Shujun Wang a, Wei Zhang a,⇑a School of Psychology, South China Normal University, Chinab Department of Educational and Counseling Psychology, School of Education, State University of New York at Albany, USA

a r t i c l e i n f o

Article history:Available online 11 August 2012

Keywords:Pathological Internet UseMaladaptive cognitionsSelf-realizationMediatorAdolescents

0747-5632/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.chb.2012.07.009

⇑ Corresponding author. Address: School of PsychUniversity, Shipai, Guangzhou 510631, China.

E-mail address: zhangwei@scnu.edu.cn (W. Zhang

a b s t r a c t

This study empirically investigated the structure and function of maladaptive cognitions related to Path-ological Internet Use (PIU) among Chinese adolescents. To explore the structure of maladaptive cogni-tions, this study validated a Chinese Adolescents’ Maladaptive Cognitions Scale (CAMCS) with twosamples of adolescents (n1 = 293 and n2 = 609). The results of the exploratory factor analysis and confir-matory factor analysis revealed that CAMCS included three distinct factors, namely, ‘‘social comfort,’’‘‘distraction,’’ and ‘‘self-realization.’’ To examine the function of maladaptive cognitions, this study testedan updated cognitive-behavioral model in the third sample of 1059 adolescents. The results of structuralequation model analyses verified both the direct effect of maladaptive cognitions on PIU and their medi-ating role in the relationships between distal factors (social anxiety and stressful life events) and PIUamong Chinese adolescents. Theoretical and practical implications of these findings were discussed.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Extensive research has provided evidence that PathologicalInternet Use (PIU) may have negative influences on developmentof adolescents in areas such as academic achievement, physicalhealth and interpersonal relationships (Chou, Condron, & Belland,2005; Flisher, 2010; Greenfield & Yan, 2006; Shaw & Black,2008). Researchers have proposed theories (Davis, 2001; Krautet al., 1998; Suler, 1999; Young, 1999) to explain how PIU occuramong adolescents. Among these theories, the most representativeone is the cognitive-behavioral model (Davis, 2001). In this model,Davis emphasized the key role that PIU-related maladaptive cogni-tions play in the onset and development of PIU. Davis’ initial workhas inspired and motivated subsequent theory-driven empiricalresearch on PIU (e.g., Caplan, 2002, 2003, 2005, 2010; Davis, Flett,& Besser, 2002; Liu & Peng, 2009). For example, Caplan (2010) re-cently validated a PIU measurement instrument, the GeneralizedProblematic Internet Use Scale 2 (GPIUS2), and empirically testedan updated cognitive-behavioral model of general PIU. These stud-ies have substantially advanced the existing knowledge regardingPIU.

Most of the recent studies on the characteristics of maladaptivecognitions were conducted using subjects with Western culturalbackgrounds. Consequently, new and important research questions

ll rights reserved.

ology, South China Normal

).

have emerged. First, would the structure of maladaptive cognitionsbe the same across different cultures? Second, if there were cul-tural differences in the structure of maladaptive cognitions, thenwould their function be equivalent across cultures? To answerthese questions, we conducted the present study in Chinese ado-lescents, a group that is considered to be at high risk for PIU(e.g., Hechanova & Czincz, 2008; Yen, Yen, & Ko, 2010; Zhang,Amos, & McDowell, 2008). The goals of the study were to (1) ex-plore the structure of maladaptive cognitions among Chinese ado-lescents and (2) examine the function of maladaptive cognitions inthe development of PIU in this group. We aimed to explain themechanism by which PIU develops in Chinese adolescents and toprovide evidence that existing theories of PIU can be generalizedacross cultures.

1.1. The structure of maladaptive cognitions

Davis (2001) introduced a cognitive-behavioral model of PIU.One contribution of this model was that it distinguished betweentwo distinct forms of PIU: specific and generalized PIU. SpecificPIU refers to the pathological use of content-specific functions ofthe Internet (e.g. gambling, stock trading, viewing sexual materi-als), which would be likely manifested in some alternative waysif the individual was unable to access the Internet; in contrast, gen-eralized PIU describes a multidimensional overuse of the Internet,which is associated with the unique social context available onInternet. Davis argued that individuals with generalized PIU wereconsiderably more problematic; thus, generalized PIU should be

Y. Mai et al. / Computers in Human Behavior 28 (2012) 2376–2386 2377

especially interesting to researchers (Caplan, 2002). Anotherimportant contribution of the cognitive-behavioral model of PIUwas that it focused on the maladaptive cognitions associated withPIU. This was the appealing feature of this model. In this model,Davis (2001) proposed that the presence of maladaptive cognitionswas critical for the development and maintaining of PIU. In the fol-lowing part, we will introduce the definition of maladaptive cogni-tions and will discuss the structure of maladaptive cognitions.

1.1.1. Definition of maladaptive cognitionsMaladaptive cognitions refer to cognitive biases that individuals

form toward themselves and the world after they start using theInternet (Davis, 2001). The cognitive biases are extreme thoughtsthat the real ‘‘self’’ is never good enough and only the online ‘‘self’’is satisfying, and these biases are based on self-doubt, a low senseof self-efficacy and negative self-value in the real world as well aspositive feelings about self-worth in the virtual world. An exampleof a maladaptive cognition that could arise from PIU might be, ‘‘Iam worthless offline, but online I am someone.’’ Individuals hold-ing such cognitive biases may simply believe that the online worldis better than the real world. They might believe that ‘‘the Internetis the only place where I am respected,’’ ‘‘nobody loves me offline,’’or ‘‘the Internet is my only friend.’’

Maladaptive cognitions emerge only after individuals begin touse the Internet (Davis, 2001), and they involve comparativethoughts about which is more satisfying, the offline world or theonline world. Thus, the development of maladaptive cognitionsmight be similar to the ‘‘needs-satisfying’’ process that Suler(1999) suggested to explain PIU. Initially, certain needs of an indi-vidual that cannot be satisfied in real life are satisfied by using theInternet. Then, as this ‘‘needs-satisfying’’ process is repeated, theperception that the online world is better than real life would bestrengthened, ultimately becoming an automatic pattern ofthought. Therefore, it could be inferred that the development ofmaladaptive cognitions might be related to specific needs thatare difficult to satisfy in real life. Such a hypothesis has been indi-rectly supported by some studies (Caplan, 2002, 2003, 2005; Daviset al., 2002; Liu & Peng, 2009). For example, Caplan (2002, 2003,2005) found that individuals lacking social skills in real life weremore likely to use the Internet for social interactions to satisfy theirneed for interpersonal relationships. Thus, these individuals mightdevelop the maladaptive cognition of Preference for Online SocialInteraction (POSI).

1.1.2. The structure of maladaptive cognitions in the western contextDavis et al. (2002) first described maladaptive cognitions as a

construct of two-dimensions, social comfort and distraction, anddeveloped the Online Cognition Scale (OCS) based on this notion.Social comfort refers to an individual’s perception of the onlineworld as a social context that is safer than the real world becauserejection can be avoided. Distraction refers to the idea that an indi-vidual can use the Internet to escape from the pressures, tasks andtroubles of the real world. Caplan (2003, 2010) considered generalPIU to be mainly associated with interpersonal Internet usage. Hepresented a type of maladaptive cognition termed POSI and devel-oped the GPIUS as a measuring tool. POSI refers to the belief thatone will perform better and feel better about oneself in online so-cial interactions and relationships than in similar offline activities.POSI is composed of two sub-dimensions (Caplan, 2002): socialbenefit, the perceived social benefits of Internet use, and socialcontrol, the perception of having increased control when interact-ing with others online. Liu and Peng (2009) used a single dimen-sion of ‘‘Preference for a Virtual Life (PVL)’’ to describemaladaptive cognitions. Individuals with PVL consider themselvesto behave better, feel better and be treated better in the virtual

world than the real world. PVL was assessed using combined itemsfrom both OCS (Davis et al., 2002) and GPIUS (Caplan, 2003).

The dimensions of the maladaptive cognitions presented abovecan be divided into two categories. The first category, termed ‘‘so-cial comfort,’’ involves cognitive biases that the interpersonal selfand the environment of the virtual world may be better than thosein real life. The social comfort (Davis et al., 2002), POSI (Caplan,2003; Caplan, 2010), and PVL (Liu & Peng, 2009) measures men-tioned above can be classified into this category. These measuresare all characterized by a preference for using the Internet’s socialfunctions to satisfy a need for interpersonal relationships (Casale &Fioravanti, 2011; Lee & Stapinski, 2012; Morahan-Martin & Schum-acher, 2003; Shek, Tang, & Lo, 2008; Tekinarslan & Gürer, 2011).Additionally, this category of maladaptive cognitions relates tothe negative characteristics surrounding forming interpersonalrelationships (e.g., loneliness, rejection sensitivity, and lack of so-cial skills). The second category, ‘‘Distraction,’’ involves the mal-adaptive belief that hiding oneself in the Internet world mayhelp one avoid real-life pressures (e.g., Davis et al., 2002). This cat-egory of maladaptive cognitions is characterized by aimless usageof the Internet to satisfy a need to escape from the real world (Da-vis et al., 2002).

1.1.3. The structure of maladaptive cognitions in Chinese cultureChinese adolescents can be considered a high-risk group of PIU,

as PIU was found to be more prevalent among adolescents in Chinathan in their counterparts in Western countries (Block, 2008;Hechanova & Czincz, 2008; Yen et al., 2010; Zhang et al., 2008).The incidence rates of PIU in the USA and Europe ranged between1.5% and 8.2% (Johansson & Götestam, 2004; Siomos, Dafouli, Bra-imiotis, Mouzas, & Angelopoulos, 2008; Villella et al., 2011; Wein-stein & Lejoyeux, 2010; Zboralski et al., 2009), whereas in China,the incidence rate was found to be as high as 13.7% (Block,2008). Thus, it is important to determine the structure of Inter-net-related maladaptive cognitions that Chinese adolescents maydevelop.

It could be considered that the maladaptive cognitions of Chi-nese adolescents should include the two dimensions identified instudies within Western cultures. Similar to adolescents in Westerncountries, Chinese adolescents have a strong need for interpersonalrelationships. Thus, the maladaptive cognitions of Chinese adoles-cents may include the dimension ‘‘social comfort,’’ as the virtualworld might best meet certain social needs that are unattainablein real life. Furthermore, compared with adolescents in Westerncountries, most Chinese adolescents are shouldering greater levelsof academic stress from their parents, schools and society. Hence,the maladaptive cognitions of Chinese adolescents may likely in-clude the dimension ‘‘distraction,’’ as the virtual world may be-come a convenient place to escape from these significant real-lifepressures.

In addition, maladaptive cognitions among Chinese adolescentsmay include other novel dimensions. As mentioned above, adoles-cents in China experience much greater academic stress than ado-lescents in Western countries (Yen et al., 2010). As academicranking is usually an important indicator of a student’s success,it is hard for adolescents who are only at the average level to feela sense of achievement in such an environment. Thus, they are ea-ger to find another means to satisfy their needs for attaining asense of achievement or self-realization. Several studies havefound that adolescents with PIU in China are particularly addictedto playing online games (Chou & Hsiao, 2000; Huang et al., 2009; Li& Chung, 2006; Lo, Wang, & Fang, 2005; Wan & Chiou, 2006). On-line games may happen to meet their needs in this regard, as onemay be able to attain high scores and rankings with sufficient prac-tice and dedication. This may provide a much easier means for aca-demically poor students, who may be neglected by their parents,

2378 Y. Mai et al. / Computers in Human Behavior 28 (2012) 2376–2386

teachers and classmates, to gain a sense of success. They mayspend substantial amounts of time surfing online and may try hardto become superior in online games by scoring more points andwinning more awards, thus gaining a sense of self-value thatwould be difficult to attain in the real world. Leung (2003) alsofound that individuals with PIU tended to try out new roles inthe virtual world and to participate in activities in which they ex-celled to experience a high level of self-esteem and imaginary self-value. Based on these studies, we propose that the structure ofmaladaptive cognitions in Chinese adolescents may include a thirddimension termed ‘‘self-realization.’’

1.2. The function of maladaptive cognitions

1.2.1. The direct and mediating effects of maladaptive cognitionsAccording to Davis’ cognitive-behavioral model (2001), mal-

adaptive cognitions act as proximal causes that directly result inPIU symptoms, which in turn trigger various PIU-related negativeoutcomes. On the other hand, the emotional vulnerability or pre-disposition of individuals to anxiety, depression and loneliness,along with external stressors such as negative life events, mayact as distal causes that influence PIU indirectly and may be med-iated by maladaptive cognitions.

The hypothesis that maladaptive cognitions have direct effectson PIU has been supported by strong evidence (Caplan, 2002,2003, 2005, 2010; Li, Zhang, Li, Zhen, & Wang, 2010; Liu & Peng,2009; Peng & Liu, 2010). Empirical studies (Caplan, 2003, 2005;Li et al., 2010) also support the hypothesis that maladaptive cogni-tions may mediate the relationship between distal causes and PIU.For instance, Caplan (2003, 2005) found that in the Western con-text the maladaptive cognitions POSI played a full mediatory rolein the relationship between the individual vulnerability dimension(termed ‘‘lack of social skills’’) and PIU.

The findings of Li et al. (2010) described the major effects ofmaladaptive cognitions on the development of PIU among Chineseadolescents. However, the tool by which maladaptive cognitionswere measured in their study was developed under a Westernbackground. As mentioned above, the maladaptive cognitions ofChinese adolescents may have a structure different from that ofadolescents in Western countries. Hence, to explore the commonfunction of maladaptive cognitions across cultures, it is necessaryto assess the function of maladaptive cognitions in the develop-ment of PIU using a specially designed measurement suitable forChinese adolescents within the context of Chinese culture.

1.2.2. The function of maladaptive cognitions among Chineseadolescents

The function of maladaptive cognitions in the development ofPIU among Chinese adolescents may be similar to that in adoles-cents in Western countries. Maladaptive cognitions may act asproximal causes directly resulting in PIU symptoms and relatednegative outcomes, or mediate the effects of distal causes, includ-ing individual vulnerability and external stress, on PIU (Davis,2001). Concerning individual vulnerability such as depression, so-cial anxiety, and substance dependence (Davis, 2001), social anxi-ety was found to be an important factor in the development of PIU(Caplan, 2007; High & Caplan, 2009; Lee & Stapinski, 2012; Lo et al.,2005; Weinstein & Lejoyeux, 2010). Social anxiety refers to emo-tional responses in social situations that may lead to avoidancebehaviors, such as strong apprehension, worry or fear. Individualswith social anxiety may prefer to use the Internet to avoid face-to-face social interactions with others (Shepherd & Edelmann, 2005).Regarding external stressors such as some life events (Davis, 2001),stressful life events were found to be important antecedents of PIU(Leung, 2007; Li et al., 2010; Li, Wang, & Wang, 2009). Stressful lifeevents may include stress from family, school, peers or other social

networks in daily life (Liu, Liu, Yang, & Zhao, 1997). When facingsignificant life stress, individuals may utilize new media such asthe Internet in an attempt to escape real life (Whang, Lee, & Chang,2003). Including these factors in a model of PIU may help betterpredict and explain the function of maladaptive cognitions.

The present study was conducted in two steps. In Study 1, a Chi-nese Adolescents Maladaptive Cognitions Scale (CAMCS) wasdeveloped to examine the structure of maladaptive cognitions inChinese adolescents. In Study 2, a model was constructed to ex-plore the positive predictive effects of maladaptive cognitions onPIU symptoms and related negative outcomes in Chinese adoles-cents, as well as their mediating effects on the relationship be-tween distal causes and PIU.

2. Study 1

2.1. Method

2.1.1. Sample and procedureWhen analyzing the structure of a scale, both exploratory factor

analysis (EFA) and confirmatory factor analysis (CFA) are necessary(Floyd & Widaman, 1995; Hau, Wen, & Cheng, 2004; Reise, Waller,& Comrey, 2000). Thus, in the present study, data for these analyseswere based on two independent samples of middle school students(n1 = 293; n2 = 609).

2.1.1.1. Sample 1. This sample consisted of 293 students (49.02% fe-males) from general secondary schools in Guangdong, southernChina. The mean age of the participants was 15.41 years(SD = 2.08) with a range of 12–19 years. All participants had morethan 1 year of Internet experience, and 10.24% (40.00% of whichwere females) were diagnosed with PIU according to the diagnosticcriteria described by Ko et al. (2005). That is, individuals with a to-tal score of more than 63 in the Revised Chinese Internet AddictionScale (CIAS-R; Chen, Weng, Su, Wu, & Yang, 2003) would be diag-nosed as PIU. More information about CIAS-R and this criteriacould be found in the following Section 2.1.2.

2.1.1.2. Sample 2. This sample consisted of 609 students (47.03% fe-males) from general secondary schools in Guangdong, southernChina. The mean age of the participants was 15.23 years(SD = 1.73) with a range of 12–19 years. All participants had morethan 1 year of Internet experience, and 11.99% (30.14% of whichwere females) were diagnosed with PIU according to the diagnosticcriteria described by Ko et al. (2005).

The data collection procedure was the same for both samples:after informed consent was obtained from both the school andthe participants, the assessment was conducted in the classroomwithin 10 min, and the questionnaires were collected immediatelyafter completion. After excluding invalid responses and data fromparticipants without any Internet experience, the percentages ofvalid data from both samples exceeded 90% (91.56% and 90.90%,respectively).

2.1.2. Measures2.1.2.1. Maladaptive cognitions. CAMCS was used to measure mal-adaptive cognitions. The initial items on the scale originated fromthree procedures. (1) 15 items were obtained from interviews with50 students who were diagnosed with PIU according to the diag-nostic criteria reported by Ko et al. (2005). Based on the definitionof maladaptive cognitions given by the cognitive-behavioral mod-el, the interview questions mainly involved two aspects: Internet-related all-or-none cognitions about the world and self, focusing on‘‘things existing only in the Internet world, but not in real life’’ andincluding words expressing absolute meanings such as ‘‘only’’ or

Y. Mai et al. / Computers in Human Behavior 28 (2012) 2376–2386 2379

‘‘never.’’ Examples of the items included ‘‘I don’t have to thinkabout the difficult homework when I am surfing online,’’ ‘‘I feelmyself more powerful when surfing online,’’ ‘‘I can talk bravelyonly online,’’ and ‘‘Only friends online will tell you more sincerewords.’’ (2) Based on a sample of 300 general secondary school stu-dents from Guangdong, an analysis was conducted on the 20-itemmaladaptive cognitions subscale (including social comfort and dis-traction) from the OCS (Davis et al., 2002). After eliminating itemsfor which scores were too high (those with means greater than 4when scored on a 5-point scale) and the discrimination of whichwas too low, a factor analysis was conducted on the remainingitems. Items that could not be clearly classified into a certaindimension were excluded from further analyses. In total, 17 itemswere obtained. (3) Two items were borrowed from the GPIUS(Caplan, 2002), as they were obviously different in meaning fromitems in the OCS.

The items selected from the OCS and GPIUS were translated bi-directionally. A Psychology doctoral student first translated all ofthe items into Chinese, and then two graduate students majoringin English translated them back into English. Another two graduatestudents then compared the two versions of the items to confirmthat they expressed the same meaning. After this procedure wasrepeated twice, the final Chinese items were considered to havethe same meaning as the original English versions.

Twenty secondary school students were invited to assess theitems. According to the assessment, items that were difficult tounderstand, items with multiple meanings and items overlappedby other items were eliminated. Finally, 30 items were includedin the pilot survey. They were scored on a 5-point scale rangingfrom ‘‘totally disagree’’ to ‘‘totally agree,’’ with higher scores indi-cating greater levels of maladaptive cognitions.

2.1.2.2. Pathological Internet Use (PIU). The Revised Chinese InternetAddiction Scale (CIAS-R; Chen et al., 2003) was used to assess thePIU tendencies of adolescents. The scale was composed of two sub-scales: Core Symptoms of Internet Addiction (CSIA) and RelatedProblems of Internet Addiction (RPIA). The former included threedimensions: compulsive use (e.g. I cannot control my impulse forusing the internet.), withdrawal (e.g. I would feel uncomfortableif I have not been using the internet for some time.) and tolerance(e.g. I need to spend more time online now than before until I feelsatisfied.). The latter included two dimensions: interpersonal andhealth-related problems (e.g. My interactions with families andfriends decrease because of using the internet.) and time manage-ment problems (e.g. I would stay up late using the internet andthus feel listless on the next day.). The entire 26-item scale wasscored on a 4-point scale ranging from ‘‘totally disagree’’ to ‘‘totallyagree.’’ The total scores on the CIAS-R ranged from 26 to 104, withhigher scores indicating a higher severity of PIU. Individuals with atotal score of more than 63 in CIAS-R would be diagnosed as PIUaccording to the criteria proposed by Ko et al. (2005). CIAS-R andits diagnostic criteria (Ko et al., 2005) appeared to be more com-prehensive than other existing measurements used to assess PIU(Hechanova & Czincz, 2008). The internal reliability of the scaleand the two subscales in the original study ranged from .79 to.93. The CIAS-R had good reliability and validity (Ko et al., 2009).The Cronbach’s a in the present study ranged from .67 to .92.

2.1.2.3. Participant background. The participants were required togive information on gender, age, and duration of Internet experi-ence. There were five options for duration of Internet experience:‘‘less than 3 months,’’ ‘‘3 months to 6 months,’’ ‘‘6 months to1 year,’’ ‘‘1 year to 2 years,’’ or ‘‘more than 2 years.’’

2.1.3. AnalysesSPSS 13.0 and LISREL 8.72 were used to implement the analyses.

Before the analyses, we inspected the data for missing values. Theaverage percentage of missing values per variable was .40% (rang-ing from .00% to 1.70%). All missing values were then replacedusing the expectation maximization (EM) algorithm method inSPSS (Allison, 2002).

Following stringent procedures for scale development (Floyd &Widaman, 1995; Hau et al., 2004; Reise et al., 2000), the data anal-yses comprised three steps. In Step 1, a primitive item analysis wasconducted with Sample 1 to screen out low-quality items, and anEFA with Sample 1 was executed to explore the initial frameworkof adolescents’ maladaptive cognitions. In Step 2, a CFA was per-formed with Sample 2 to confirm the factor structure of the mea-surement as revealed in Step 1. Finally, in Step 3, reliability andvalidity analyses for the scale were conducted on the merged dataof Sample 1 and Sample 2, including correlation and regressionanalyses.

2.2. Results and discussion

2.2.1. Exploring the factor structure of CAMCS using EFATo screen out low-quality items and gain initial items for EFA, a

primitive item analysis was conducted on Sample 1. Items that metthe following criteria were considered low-quality items: with thestandard deviation lower than 1, the skewness of distribution high-er than 1, the item-test correlation lower than .40, or the item dis-crimination lower than .60. Finally, four candidate items wereexcluded. With the remaining 26 items, the Bartlett’s test was sta-tistically significant, v2 = 2873.30, df = 325, p < .001, and Aiser–Meyer–Olkin statistics (KMO = .92) was greater than its thresholdvalue of .50. Thus, the data was suitable for EFA.

To explore the initial framework of maladaptive cognitions, anEFA was executed with the remaining 26 items on Sample 1. Theanalysis followed the principal axis factoring method with Promax(Kappa = 4) Oblique Rotation. The initial run produced a six-factorunforced solution according to Kaiser’s criterion. However, the re-sult of Cattell’s screen test on the sedimentation graph suggested athree-factor solution. We then deleted 14 items that either failed toload substantially on any factor (i.e., with factor loadings less than.40), loaded almost equally on more than one factor, or belonged tofactors that explained less than 3.0% of the variance or includedless than three items. With the remaining 12 items (KMO = .87;Bartlett’s test v2 = 970.78, df = 66, p < .001), the second run of EFAproduced a clear three-factor solution based on Cattell’s screen teston the sedimentation graph and preliminary eigenvalues. Each fac-tor included four items with factor loadings ranging from .40 to .78(see Table 1); the eigenvalues of the three factors were 4.33, 1.50and 1.04, respectively. The three factors accounted for 36.11%,12.16% and 8.70% of the explainable variance, respectively, andthe overall scale explained 57.27% of the variance. All correlationcoefficients between the factors were all statistically significant,r = .64, .49, .42, ps < .001.

Based on the existing literature and examining the meaning ofitems in each group, the three factors were initially labeled as ‘‘so-cial comfort,’’ ‘‘distraction’’ and ‘‘self-realization.’’

2.2.2. Confirming the factor structure of CAMCS using CFATo confirm the measurement model of CAMCS and formally as-

sess its construct validity, a CFA was performed with Sample 2. Acovariance matrix and the maximum likelihood fitting methodwere used for the analysis.

To evaluate the model, we followed the guidelines proposed byHau et al. (2004). First, we examined the parameter estimates foran improper solution. All were found to be statistically significant,and no abnormal values were found. Next, we examined the fit

M2

M4

M6

M12

M9

M15

M16

M17

M22

M26

M30

M5

Factor 1

Factor 2

.38* * *

.44* * *

.47* * *

.57* * *

.47* * *

.71* * *

.62* * *

.51* * *

.66* * *

.51* * *

67* * *

.63* * * .65* * *

2χ = 211.59, df = 53, df/2χ = 4.02, NNFI = .93, CFI = .94, RMSEA

= .070, SRMR = .049.

Note. *** p < .001. All coefficient estimates are completely standardized.

Fig. 2. Fitness indexes and standardized estimates for the two-factor model.

Table 1Exploratory factor analysis of Chinese adolescents maladaptive cognitions scale.

Items Factorloadings

1 2 3

I can get to know a person better on the Internet than inperson.

.58

When surfing online, I can say things that could never besaid in real life.

.68

I always feel embarrassed when talking with others unless Italk through the Internet.

.65

Only friends online will tell you more sincere words. .55Sometimes I use the Internet to put off something. .63I feel the most comfortable online because I don’t have to

worry about the exams..68

I don’t have to think about my responsibilities when I amsurfing online.

.62

I don’t have to think about the difficult homework when Iam surfing online.

.40

I am more respected online than in real life. .78I am at my best online. .50People accept me for who I am online. .58I feel myself more powerful when surfing online. .42

2380 Y. Mai et al. / Computers in Human Behavior 28 (2012) 2376–2386

indices of the model to determine whether the model fit the datawell. A model satisfying the following four criteria would be con-sidered acceptable: v2/df < 5, Non-Normed Fit Index (NNFI) > .90,Comparative Fit Index (CFI) > .90, Root Mean Square Error ofApproximation (RMSEA) < .05, and Standardized Root Mean SquareResidual (SRMR) < .08 (Hau et al., 2004; Marsh, Hau, & Wen, 2004).Results showed that the three-factor model fit the data well,v2 = 184.60, df = 51, v2/df = 3.62, NNFI = .94, CFI = .95, RMSEA =.066, SRMR = .045 (see Fig. 1). Third, we examined the modificationindices produced by LISREL, finding that no new path that couldstatistically significantly improve the model’s fit was suggested.Fourth, we examined the parameter estimates again to determinewhether the factor structure was proper. The factor loadings wereat least moderately large in magnitude, ranging from .42 to .71 (.57average), indicating that all the items converged meaningfully ontothe scale.

However, as the correlation coefficient between ‘‘socialcomfort’’ and ‘‘self-realization’’ was .81 and it is suggested that

M2

M4

M6

M12

M9

M15

M16

M17

M22

M26

M30

M5

Social Comfort

Distraction

Self-Realization

.42* * *

.50* * *

.54* * *

.62* * *

.47* * *

.71* * *

.62* * *

.51* * *

.68* * *

.53* * *

.69* * *

.63* * *

.57* * *

.64* * *

.81* * *

2χ = 184.60, df = 51, df/2χ = 3.62, NNFI = .94, CFI =.95, RMSEA

= .066, SRMR = .045.

Note. *** p < .001. All coefficient estimates are completely standardized.

Fig. 1. Fitness indexes and standardized estimates for the three-factor model.

two factors might overlap if their correlation coefficient was higherthan .70 (Jia & Jia, 2009; MacKenzie, Podsakoff, & Jarvis, 2005), wetried to merge the items within ‘‘social comfort’’ and ‘‘self-realiza-tion’’ and developed a two-factor model (see Fig. 2). To comparethe fitness of the two-factor model with the three-factor model,another CFA was conducted on the two-factor model,v2 = 211.59, df = 53, v2/df = 4.02, NNFI = .93, CFI = .94,RMSEA = .070, SRMR = .049. This two-factor model showed a sta-tistically significantly larger chi-squares, Dv2 = 26.98, Ddf = 2,p < .0001. Its NNFI and CFI were reduced by .01, whereas RMSEAand SRMR increased by .004. These results indicated that thethree-factor model was better fit than the two-factor model, usingthe rules for model comparison (Hau et al., 2004; Marsh et al.,2004).

2.2.3. Reliability and validity analysesThe reliability and validity of the CAMCS were analyzed using

the merged data of Sample 1 and Sample 2 (n = 902).

2.2.3.1. Reliability. The coefficient of internal consistency for theoverall scale was good (a = .81). For each of the three subscales,the coefficients of internal consistency ranged from moderate tohigh: ‘‘social comfort’’ (a = .66), ‘‘distraction’’ (a = .67), and ‘‘self-realization’’ (a = .74).

2.2.3.2. Convergent and discriminant validity. An initial correlationanalysis was executed among the 12 items. The results indicatedthat all items were statistically significantly correlated with eachother (Pearson’s r = .10–.46, ps < .01), indicating that the scalehad good convergent validity. Second, an independent t-test wasconducted to compare the means between the two sets of correla-tion coefficients (r values) among inner-factor items (items sharinga common factor) and those among inter-factor items (itemsbelonging to different factors). Before the t-test, the r values ofthe two sets of correlation coefficients were transformed into theirFisher’s Zr values. Results indicated that the mean coefficientsamong the inner-factor items (mean Zr = .38) were statistically sig-nificantly larger than those among the inter-factor items (meanZr = .23), t (64) = 6.01, p < .001, d = 1.66, 95% CI [.09, .19], indicatingthat the scale had good discriminant validity.

Social Anxiety

Stressful Life Events

Maladaptive Cognitions

PIU Symptoms

Negative Outcomes

+

+

Note. PIU = Pathological Internet Use.

+

+

Fig. 3. Model M1.

Y. Mai et al. / Computers in Human Behavior 28 (2012) 2376–2386 2381

2.2.3.3. Construct validity. A hierarchical regression analysis withthe Enter method was conducted to investigate the effects of eachtype of maladaptive cognition on PIU. The results indicated thatwith the effects of gender and age considered, all three types ofmaladaptive cognitions could positively predict PIU in adolescents(b = .12, b = .28, b = .24, p < .001). These three types of maladaptivecognitions in total explained extra 25% of the variation in PIU(R2 = .03 for Step 1, p < .001; DR2 = .25 for Step 2, p < .001). Theseresults were consistent with the theoretical hypothesis of mal-adaptive cognitions proposed by Davis (2001), indicating that theconcept validity of CAMCS was good. In addition, the standardizedeffect sizes of the ‘‘distraction’’ and ‘‘self-realization’’ dimensionswere larger than that of the ‘‘social comfort’’ dimension; this willbe discussed in the general discussion.

Social Anxiety

Stressful Life Events

Maladaptive Cognitions

PIU Symptoms

Negative Outcomes

+

+

+

+

Note. PIU = Pathological Internet Use.

+

+

Fig. 5. Model M3.

Social Anxiety

Stressful Life Events

Maladaptive Cognitions

PIU Symptoms

Negative Outcomes

+

+

+

+

Note. PIU = Pathological Internet Use.

Fig. 4. Model M2.

2.2.4. Interpretations of the three factorsThe results of the EFA and CFA, as well as the reliability and

validity analyses, provided support for the hypothesis that thescale of maladaptive cognitions among Chinese adolescents con-tained three factors: ‘‘social comfort,’’ ‘‘distraction,’’ and ‘‘self-realization.’’

The first factor is similar to the social comfort factor proposedby Davis et al. (2002) and the POSI factor suggested by Caplan(2003), Caplan (2010), which refers to an individual’s preferencefor engaging in social interactions online rather than face-to-facecommunication and his/her feelings of safety and security in beingpart of virtual social network.

The second factor is consistent with the maladaptive cognitiondimension of distraction introduced by Davis et al. (2002), whichinvolves using the Internet for the purpose of escape. An individualmay use the Internet to be distracted from a stressful event, task, orthought.

The third factor is a new category of maladaptive cognitionsintroduced in the present study. It refers to an individual’s desireto gain respect from an imaginary audience or to enhance one’s selfby accomplishing difficult tasks online to gain a sense of self-worththat may be lacking in real life.

All three factors are analogous to the important reasons Davisdescribed for why individuals may be drawn to spending time on-line. They reflect the experienced sentiment of a preference for avirtual context rather than the real world.

3. Study 2

To examine the effects of maladaptive cognitions on PIU amongChinese adolescents, the present study specified a model M1 (seeFig. 3) based on the existing literature (Caplan, 2002, 2003, 2005,2010; Davis, 2001; Li et al., 2010). In this model, maladaptive cog-nitions positively predict PIU symptoms, which in turn result in re-lated negative outcomes (e.g., interpersonal and health problems),and maladaptive cognitions also fully mediate the positive linksbetween social anxiety and stressful life events and PIU symptoms.

We also proposed an alternative model M2 (see Fig. 4) to exam-ine the mediating roles of maladaptive cognitions in a more rigor-ous manner. In model M2, social anxiety and stressful life eventsinfluenced PIU symptoms directly without the mediating effectsof maladaptive cognitions. In addition, if the mediating role of mal-adaptive cognitions was supported, we proposed another alterna-tive model M3 to examine whether maladaptive cognitions hadfull or partial mediating effects (see Fig. 5). In model M3, maladap-tive cognitions partially mediate the positive links from social anx-iety and stressful life events to PIU symptoms, that is, socialanxiety and stressful life events had direct effects as well as indi-rect effects through maladaptive cognitions on PIU Symptoms.

3.1. Method

3.1.1. Sample and procedureThe participants consisted of 1174 students from general sec-

ondary schools. The main data collection procedure was the sameas that described for Study 1. After eliminating invalid data anddata from participants without any Internet experience, the num-ber of subjects used for the final analyses was 1059 (48.07% fe-males), accounting for 90.20% of the total data collected. Themean age of the participants was 15.25 years (SD = 1.74), with arange of 12–19 years. All subjects had experience using the Inter-net for more than 1 year, and 11.89% (32.54% females) were diag-nosed with PIU according to the diagnostic criteria presented byKo et al. (2005).

3.1.2. MeasuresWhen the participant background and maladaptive cognitions

were assessed by the measurements same as those applied inStudy 1, social anxiety, stressful life events, PIU symptoms, andPIU negative outcomes were assessed using instruments describedin the following text.

Social anxiety was assessed using the Interaction AnxiousnessScale (IAS) developed by Leary (1983) and revised by Ma (1999).

Tabl

e2

Pear

son

corr

elat

ion

coef

fici

ents

amon

gob

serv

able

vari

able

sof

the

mea

sure

men

tm

odel

.

Var

iabl

e1

23

45

67

89

1011

1213

1415

1617

1So

cial

An

xiet

y1

–2

Soci

alA

nxi

ety

2.5

3***

–3

Soci

alA

nxi

ety

3.6

3***

.51*

**

–4

Inte

rper

son

alpr

essu

re.1

3***

.08*

.12*

**

–5

Aca

dem

icpr

essu

re.2

0***

.15*

**

.20*

**

.54*

**

–6

Hea

lth

adap

tati

onpr

oble

ms

.07*

.04

.08*

**

.56*

**

.45*

**

–7

Loss

.15*

**

.13*

**

.17*

**

.54*

**

.48*

**

.48*

**

–8

Pun

ish

men

t.1

1***

.05

.07*

.55*

**

.52*

**

.54*

**

.52*

**

–9

Oth

erev

ents

.13*

**

.12*

**

.15*

**

.49*

**

.46*

**

.45*

**

.55*

**

.47*

**

–10

Soci

alco

mfo

rt.1

9***

.03

.14*

**

.14*

**

.12*

**

.09*

**

.11*

**

.13*

**

.06

–11

Dis

trac

tion

.18*

**

.05

.13*

**

.13*

**

.11*

**

.03

.11*

**

.12*

**

.07*

.57*

**

–12

Self

real

izat

ion

.14*

**

.08*

.11*

**

.16*

**

.21*

**

.11*

**

.09*

*.0

9**

.09*

*.3

3***

.43*

**

–13

Com

puls

ive

use

.15*

**

.08*

.17*

**

.18*

**

.23*

**

.13*

**

.16*

**

.17*

**

.12*

**

.29*

**

.44*

**

.43*

**

–14

Wit

hdr

awal

.20*

**

.16*

**

.20*

**

.17*

**

.23*

**

.15*

**

.15*

**

.15*

**

.13*

**

.27*

**

.38*

**

.37*

**

.63*

**

–15

Tole

ran

ce.2

0***

.10*

**

.20*

**

.19*

**

.25*

**

.15*

**

.14*

**

.17*

**

.13*

**

.29*

**

.41*

**

.40*

**

.74*

**

.67*

**

–16

Inte

rper

son

alre

lati

onsh

ips

.11*

**

.05

.12*

**

.23*

**

.23*

**

.17*

**

.17*

**

.15*

**

.16*

**

.20*

**

.25*

**

.34*

**

.53*

**

.56*

**

.60*

**

–17

Hea

lth

/tim

em

anag

emen

t.2

0***

.13*

**

.20*

**

.17*

**

.20*

**

.17*

**

.14*

**

.15*

**

.18*

**

.24*

**

.31*

**

.29*

**

.54*

**

.59*

**

.68*

**

.62*

**

–M

ean

2.89

3.22

3.02

1.58

2.09

1.15

1.68

.96

1.29

2.72

2.66

2.48

1.98

1.98

1.8

1.74

1.89

9St

d.de

viat

ion

.77

.55

.8.8

5.9

5.8

31.

176

.91

1.28

.81

.84

.85

.62

.56

.58

.59

.563

*p

<.0

5.**

p<

.01.

***

p<

.001

.

2382 Y. Mai et al. / Computers in Human Behavior 28 (2012) 2376–2386

This 15-item measure is scored using a 5-point scale, with higherscores indicating a higher tendency toward social anxiety. Theinternal consistency reliability of the scale was .87, and the retestreliability was .80. Some items were modified to fit the partici-pants of the present study; for example, the term ‘‘the boss’’was replaced with ‘‘the elder.’’ Cronbach’s a for the scale was.81 in the present study.

Stressful life events were assessed with the Adolescent LifeEvents Scale (ASLEC), which was developed by Liu et al. (1997)and revised by Gao (2006). Spanning six dimensions (interper-sonal relationships, academic stress, health adaptation problems,loss, penalties, and other events) and 26 items, this scale is scoredon a 5-point scale, with a higher total score of stressful life eventsindicating a greater impact on the individual. The scale’s internalconsistency reliability was .89. Cronbach’s a of this scale was .89in the present study.

PIU symptoms and PIU related negative outcomes in the ado-lescents were assessed respectively with the two subscales, CSIAand RPIA, of the CIAS-R developed by Chen et al. (2003) (see Sec-tion 2.1.2). Cronbach’s a for these two subscales in the presentstudy was .90 and .88, respectively.

3.1.3. AnalysesBefore the analyses, we inspected the data for missing values,

finding that the average percentage of missing values per variablewas .24% (ranging from .00% to 1.30%). All missing values werethen replaced using the EM algorithm method in SPSS (Allison,2002).

Using SPSS 13.0 and LISREL 8.72, to compare the three hypoth-esized models, the analyses comprised two steps. In Step 1, a CFAwas conducted to fit a measurement model for the next step. InStep 2, SEM analyses were performed to validate the structuralmodel (relations between the constructs) of each of the threemodels M1, M2, and M3, based on the measurement model con-structed in Step 1. The procedure was conducted in accordancewith the two-step method for using structural equitation model-ing to test and develop theories, as proposed by Anderson andGerbing (1988) and recommended by Hau et al. (2004). The max-imum likelihood fitting method and the covariance matrix wereused for CFA and SEM.

3.2. Results and discussion

3.2.1. Fitting a measurement modelIt was suggested that if the sample size is sufficiently large,

then it would be better to adopt the parceling method to avoidan excessive number of parameters being estimated in one model(Bandalos & Finney, 2001; Hau et al., 2004). In addition, after par-celing, the data would more closely approximate a normal distri-bution and would have better quality (Hau et al., 2004).Therefore, we used parcels as indicators (observable variables)to measure the latent variables (social anxiety, stressful lifeevents, maladaptive cognitions, PIU symptoms and PIU-relatednegative outcomes). The parcels used were the means of the sub-sets of items for each latent variable. The parceling strategy in-volved items in the same dimension being grouped into thesame subset. Specifically, the indices of social anxiety were thethree item parcels randomly grouped in the IAS; the indices ofstressful life events were the six dimensions in the ASLEC; theindices of maladaptive cognitions were the three dimensions inthe CAMCS; the indices of PIU symptoms were the three dimen-sions in the CSIA subscale from the CIAS-R; and the indices ofPIU-related negative outcomes were the two dimensions in theRPIA subscale from the CIAS-R. Finally, a measurement modelcomposed of seventeen indicators (observable variables), fourexogenous latent variables and three endogenous latent variables

Table 3Factor loadings of indicators to their latent variables in the measurement model.

Indices Social anxious Life events Maladaptivecognitions

PIU symptoms PIU-relatednegativeoutcomes

Social Anxiety 1 .81Social Anxiety 2 .65Social Anxiety 3 .78Interpersonal pressure .76Academic pressure .69Health adaptation problems .69Loss .72Punishment .73Other events .67Social comfort .65Distraction .55Self realization .82Compulsive use .81Withdrawal .77Tolerance .89Interpersonal relationships .76Health/time management .82

Table 4Fitness indexes for model comparison.

Model v2 df v2/df NNFI CFI RMSEA SRMR

M1 423.87 114 3.72 .97 .98 .052 .057M2 440.56 114 3.86 .97 .98 .052 .071M3 384.39 112 3.43 .98 .98 .049 .042

Y. Mai et al. / Computers in Human Behavior 28 (2012) 2376–2386 2383

was constructed. Table 2 presents the Pearson correlation coeffi-cients among all the indicators in the measurement model.

A CFA was performed to test this measurement model. The re-sults indicated that all parameter estimates were statistically sig-nificant, and no abnormal values were found. All of the key fitindices of the model exceeded the criteria recommended by Hauet al. (2004) and Marsh et al. (2004), v2 = 367.22, df = 109, v2/df = 3.37, NNFI = .98, CFI = .98, RMSEA = .048, SRMR = .040. Theloadings of all of the indicators on their latent variables were rea-sonably high, ranging from .55 to .89, with a mean of .76 (seeTable 3). Therefore, this model fit the data well.

3.2.2. Fitting the structural modelResults from SEM (see Table 4) showed that all of the three

models fit the data well, moreover, M1 was statistically

Social Anxiety

Stressful Life Events

Maladaptive Cognitions PIU Symptoms

Negative Outcomes

.16* * *

.20* * *

.54* * *

.89* * *

df/2χ = 384.39, df = 112, df/2χ = 3.43, NNFI = .98, CFI =.98,

RMSEA = .049, SRMR = .042.

Note. PIU = Pathological Internet Use.

*** p < .001. All coefficient estimates are completely standardized.

.15* * *

.11* * *

Fig. 6. Fitness indexes and standardized estimates for the final model of M3.

significantly better fit than M2, Dv2(1) = 16.69, p < .0001, whereasM3 was statistically significantly better fit than M1,Dv2(2) = 39.48, p < .0001. Thus, M3 was selected as the final model(see, Fig. 6).

The detailed results of model M3 were discussed as follow. Intotal, the equation of PIU symptoms explained 40.00% of the vari-ation (R2 = .40), and the equation of PIU-related Negative Outcomesexplained 79.50% of the variation (R2 = .80), suggesting that themodel was effective for explaining PIU.

Specifically, the standardized direct effect of maladaptive cogni-tions on PIU symptoms was .54, p < .001, the standardized directeffect of PIU symptoms on PIU-related negative outcomes was.89, p < .001. These results support the hypothesis that maladaptivecognitions positively predict PIU symptoms, which in turn result inrelated negative outcomes. It is consistent with the theoreticalhypothesis of the cognitive-behavioral model proposed by Davis(2001).

Both social anxiety and stressful life events had positive directeffects on PIU symptoms, standardized estimations of these directeffects were .16 (p < .001) and .20 (p < .001) respectively. At thesame time, the effects of both factors were partially mediated bymaladaptive cognitions. The proportion of the mediating effect ofmaladaptive cognitions on the relationship between social anxietyand PIU symptoms was 50.00%, and that on the relationship be-tween stressful life events and PIU symptoms was 37.50%. Theseresults support the hypothesis that maladaptive cognitions par-tially mediate the positive links from social anxiety and stressfullife events to PIU symptoms. These findings confirm the theoreticalhypothesis suggested by Davis (2001) that maladaptive cognitions,as a proximal factor, mediate the effects of distal factors on PIU.However, they were not consistent with the findings of Caplan(2003, 2005) and Li et al. (2010), who reported that maladaptivecognitions fully mediated the effects of distal factors on PIU. Theseresults will be discussed further in the general discussion.

4. General discussion

The present study proposed a new dimension (‘‘self-realiza-tion’’) in the structure of maladaptive cognitions among Chineseadolescents in addition to the two dimensions (‘‘social comfort’’and ‘‘distraction’’) found among adolescents in Western countries.Moreover, the present study also confirmed the decisive role that

2384 Y. Mai et al. / Computers in Human Behavior 28 (2012) 2376–2386

maladaptive cognitions play in the development of PIU among Chi-nese adolescents.

4.1. Chinese adolescents’ maladaptive cognitions and the mechanismof PIU

The present study enriched the conceptual structure of mal-adaptive cognitions to some extent by finding a new dimension(‘‘self-realization’’) that has not yet been considered. This dimen-sion, together with two other dimensions (‘‘social comfort and dis-traction’’) found in Western cultures, was consistent with the threetypes of needs (social interaction, escape and self-realization; Sul-er, 1999) that might be satisfied by using the Internet. However,the question why the new dimension of ‘‘self-realization’’ wouldbe found in Chinese adolescents whereas not in adolescents inWestern countries might lead to further consideration of thecauses of maladaptive cognitions. Although the Internet may sat-isfy different types of needs in different individuals, whether Inter-net use would cause certain maladaptive cognitions depends onwhether individuals lack mechanisms to satisfy these needs in reallife. The result that the structure of maladaptive cognitions in Chi-nese adolescents was different from that of adolescents in Westerncountries indicates that the structure of maladaptive cognitionsmight differ across cultures. Studies using participants from differ-ent cultural backgrounds may be necessary to obtain a more com-plete understanding of maladaptive cognitions.

The results of Study 2 verified both the direct effect of maladap-tive cognitions on PIU and their mediating role in the relationshipsbetween distal factors and PIU among Chinese adolescents. This en-riches and extends the literature on the cognitive-behavioral modelof PIU (e.g., Caplan, 2003, 2005; Davis, 2001; Li et al., 2010) andprovides evidence for the general function of maladaptive cogni-tions in the mechanism of PIU development. Although the specificcontent of maladaptive cognitions may differ in different cultures,studies from both China (e.g., Li et al., 2010) and Western countries(Caplan, 2003, 2005) have demonstrated that maladaptive cogni-tions play a decisive role in the development of PIU.

4.2. Explaining the high prevalence of PIU among Chinese adolescents

The present study confirmed the higher PIU prevalence found inChinese adolescents compared to that in European and Americanadolescents. It is important to consider the underlying reasonsfor this increased prevalence of PIU.

Results from the regression analyses in Study 1 indicated thatall three types of maladaptive cognitions had effects on PIU. Partic-ularly, ‘‘self-realization’’ and ‘‘distraction’’ played more importantroles in the development of PIU than ‘‘social comfort.’’ By contrast,the social function of the Internet has been found to be stronglyassociated with PIU in Western studies (e.g., Caplan, 2007; Casale& Fioravanti, 2011; Lee & Stapinski, 2012), and PIU may be causedby maladaptive cognitions related to social comfort (e.g., POIS byCaplan (2005)). These different findings suggest that comparedwith Western countries, the structure of maladaptive cognitionsmight be more complex among Chinese adolescents and prove tobe a decisive factor that would result in PIU. In other words, Chi-nese adolescents might be at a greater risk for PIU. This may helpexplain the higher prevalence of PIU among Chinese adolescentsthan among adolescents in Western countries.

4.3. Implications for PIU Prevention and Interventions

Consistent with earlier studies (Caplan, 2003, 2005; Li et al.,2010), we confirmed that the presence of maladaptive cognitionsplays a decisive role in the development and maintenance of PIU.PIU was affected to a large extent by psychological factors, suggest-

ing it might be reasonable to use psychotherapy (World HealthOrganization, 1990) as a PIU intervention rather than drug therapyor electric shock and electromagnetic treatments. Numerous stud-ies and practical experience have indicated that removing psycho-logical dependence is a key and effective way to solve addictiveproblems, including both substance abuse and behavioral addic-tions (Aston-Jones & Druhan, 1999; Holden, 2001; Nestler, 2001).Among the commonly used psychotherapy approaches, the ap-proach of displacing irrational beliefs or dysfunctional schemaused in cognitive-behavioral therapy was found to have long-termeffects on PIU (Young, 2007). Because maladaptive cognitions havebeen found to have decisive effects on PIU, it is reasonable to treatindividuals with PIU by helping them change their existing mal-adaptive cognitions.

The results from the regression analyses in Study 1 indicated that‘‘self-realization’’ and ‘‘distraction’’ played more important roles inthe development of PIU than ‘‘social comfort,’’ whereas in Westerncountries, ‘‘social comfort’’ was found to play a more important role(e.g., Caplan, 2003, 2005, 2010). These differing results imply thatPIU treatments characterized by correcting maladaptive cognitionsmay need to be adjusted according to different cultures becausethe importance of certain maladaptive cognitions may differ acrosscultures. As Chinese adolescents were more influenced by the mal-adaptive cognition of ‘‘self-realization,’’ parents and educatorsmight help them develop more interests and special skills in real lifeand at the same time educate them according to their aptitudeswhile avoiding an over-emphasis on academic achievement.

4.4. Future research

Although we found the effect of maladaptive cognitions on PIUsymptoms was large in the current study, this does not account forthe total variance in PIU. The results do not support the hypothesisin the cognitive-behavioral model that maladaptive cognitionswere sufficient causes for PIU. The relationship between maladap-tive cognitions and PIU may perhaps be moderated by a third fac-tor. For example, a recent study by Li et al. (2010) indicated thateffort control moderating the effects of maladaptive cognitionson PIU. Future studies should integrate other moderated factorsto further investigate the mechanism of PIU.

Although the results of the current study indicated that themediating effect of maladaptive cognitions was significant, the ef-fects of stressful life events and social anxiety on PIU were not en-tirely mediated by maladaptive cognitions. There might be otherproximal antecedents of PIU in addition to maladaptive cognitions,such as emotional aspects of the behavioral regulation system. Forinstance, some studies have found that emotional intelligence canpredict PIU (Beranuy, Oberst, Carbonell, & Chamarro, 2009). Futureresearch should integrate emotional factors to further investigatethe mechanism of PIU.

When interpreting these results, it is important to rememberthat this research was cross-sectional. Although SEM analysescan assess the plausibility of hypothetical causal processes, strongcausal inferences are not appropriate when this method is usedwith cross-sectional or correlational data. A longitudinal or exper-imental design would provide a stronger framework for identifyingcausal effects.

Acknowledgements

This research was supported by grants from the National Natu-ral Science Foundation of China (31170998) and the Guangdong(China) Natural Science Foundation (10151063101000045) toWei Zhang, and by grant from the National Education SciencesPlanning Projects of China (EBA110324). The views expressed inthis paper are those of the authors’ and do not necessarily reflect

Y. Mai et al. / Computers in Human Behavior 28 (2012) 2376–2386 2385

the views of the granting agencies. We thank the children who par-ticipated in our study and the schools that assisted our study invarious ways.

References

Allison, P. D. (2002). Missing data. Thousand Oaks, CA, US: Sage.Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice. A

review and recommended two-step approach. Psychological Bulletin, 103(3),411–423. http://dx.doi.org/10.1037/0033-2909.103.3.411.

Aston-Jones, G., & Druhan, J. (1999). Behavioural pharmacology: Breaking the chainof addiction. Nature, 400(6742), 317–319. http://dx.doi.org/10.1038/22432.

Bandalos, D. L., & Finney, S. J. (2001). Item parceling issues in structural equationmodeling. In G. A. Marcoulides & R. E. Schumacker (Eds.), New developments andtechniques in structural equation modeling (pp. 269–296). Mahwah, NJ, US:Lawrence Erlbaum Associates, Inc.

Beranuy, M., Oberst, U., Carbonell, X., & Chamarro, A. (2009). Problematic Internetand mobile phone use and clinical symptoms in college students: The role ofemotional intelligence. Computers in Human Behavior, 25(5), 1182–1187. http://dx.doi.org/10.1016/j.chb.2009.03.001.

Block, J. J. (2008). Issues for DSM-V: Internet addiction. American Journal ofPsychiatry, 165(3), 306–307. http://dx.doi.org/10.1176/appi.ajp.2007.07101556.

Caplan, S. E. (2002). Problematic Internet use and psychosocial well-being:Development of a theory-based cognitive-behavioral measurementinstrument. Computers in Human Behavior, 18(5), 553–575. http://dx.doi.org/10.1016/S0747-5632(02)00004-3.

Caplan, S. E. (2003). Preference for online social interaction: A theory of problematicInternet use and psychosocial well-being. Communication Research, 30(6),625–648. http://dx.doi.org/10.1177/0093650203257842.

Caplan, S. E. (2005). A social skill account of problematic Internet use. Journal ofCommunication, 55(4), 721–736. http://dx.doi.org/10.1111/j.1460-2466.2005.tb03019.x.

Caplan, S. E. (2007). Relations among loneliness, social anxiety, and problematicInternet use. CyberPsychology & Behavior, 10(2), 234–242. http://dx.doi.org/10.1089/cpb.2006.9963.

Caplan, S. E. (2010). Theory and measurement of generalized problematic Internetuse: A two-step approach. Computers in Human Behavior, 26(5), 1089–1097.http://dx.doi.org/10.1016/j.chb.2010.03.012.

Casale, S., & Fioravanti, G. (2011). Psychosocial correlates of Internet use amongItalian students. International Journal of Psychology, 46(4), 288–298. http://dx.doi.org/10(1080/00207594), 2010, 541256.

Chen, S., Weng, L., Su, Y., Wu, H., & Yang, P. (2003). Development of a ChineseInternet addiction scale and its psychometric study. Chinese Journal ofPsychology, 45(3), 279–294.

Chou, C., Condron, L., & Belland, J. C. (2005). A review of the research on Internetaddiction. Educational Psychology Review, 17(4), 363–388. http://dx.doi.org/10.1007/s10648-005-8138-1.

Chou, C., & Hsiao, M. C. (2000). Internet addiction, usage, gratification, and pleasureexperience. The Taiwan college students’ case. Computers & Education, 35(1),65–80. http://dx.doi.org/10.1016/S0360-1315(00)00019-1.

Davis, R. A. (2001). A cognitive-behavioral model of pathological Internet use.Computers in Human Behavior, 17(2), 187–195. http://dx.doi.org/10.1016/S0747-5632(00)00041-8.

Davis, R. A., Flett, G. L., & Besser, A. (2002). Validation of a new scale for measuringproblematic Internet use: Implications for pre-employment screening.CyberPsychology & Behavior, 5(4), 331–345. http://dx.doi.org/10.1089/109493102760275581.

Flisher, C. (2010). Getting plugged in: An overview of Internet addiction. Journal ofPaediatrics and Child Health, 46(10), 557–559. http://dx.doi.org/10.1111/j.1440-1754.2010.01879.x.

Floyd, F. J., & Widaman, K. F. (1995). Factor analysis in the development andrefinement of clinical assessment instruments. Psychological Assessment, 7(3),286. http://dx.doi.org/10.1037/1040-3590.7.3.286.

Gao, Y. M. (2006). The confirmatory factor analysis of Adolescent Self-rating LifeEvents Check-List among the students in medical vocational college. ChineseJournal of Medical Education, 26(4), 36–58.

Greenfield, P., & Yan, Z. (2006). Children, adolescents and the Internet: A new fieldof inquiry in developmental psychology. Developmental psychology, 42(3),391–394. http://dx.doi.org/10.1037/0012-1649.42.3.000.

Hau, K. T., Wen, Z. L., & Cheng, Z. J. (2004). Structural equation model and itsapplications. Beijing, China: Education Science Press.

Hechanova, M. R. M., & Czincz, J. (2008). Internet addiction in Asia: Reality or myth?<http://www.crdi.ca/uploads/user-S/12531993781rad4E1AD.DOC> Retrieved16.06.12.

High, A. C., & Caplan, S. E. (2009). Social anxiety and computer-mediatedcommunication during initial interactions: Implications for the hyperpersonalperspective. Computers in Human Behavior, 25(2), 475–482. http://dx.doi.org/10.1016/j.chb.2008.10.011.

Holden, C. (2001). ‘Behavioral’ addictions: Do they exist? Science, 294(5544),980–982. http://dx.doi.org/10.1126/science.294.5544.980.

Huang, R. L., Lu, Z., Liu, J. J., You, Y. M., Pan, Z. Q., Wei, Z., et al. (2009). Features andpredictors of problematic Internet use in Chinese college students. Behaviour &Information Technology, 28(5), 485–490. http://dx.doi.org/10.1080/01449290701485801.

Jia, R., & Jia, H. H. (2009). Factorial validity of problematic Internet use scales.Computers in Human Behavior, 25(6), 1335–1342. http://dx.doi.org/10.1016/j.chb.2009.06.004.

Johansson, A., & Götestam, K. G. (2004). Internet addiction: Characteristics of aquestionnaire and prevalence in Norwegian youth (12–18 years). ScandinavianJournal of Psychology, 45(3), 223–229. http://dx.doi.org/10.1111/j.1467-9450.2004.00398.x.

Ko, C. H., Yen, J. Y., Chen, S. H., Yang, M. J., Lin, H. C., & Yen, C. F. (2009). Proposeddiagnostic criteria and the screening and diagnosing tool of Internet addictionin college students. Comprehensive Psychiatry, 50(4), 378–384. http://dx.doi.org/10.1016/j.comppsych.2007.05.019.

Ko, C. H., Yen, C. F., Yen, C. N., Yen, J. Y., Chen, C. C., & Chen, S. H. (2005). Screening forInternet addiction: An empirical study on cut-off points for the Chen InternetAddiction Scale. The Kaohsiung Journal of Medical Sciences, 21(12), 545–551.http://dx.doi.org/10.1016/S1607-551X(09)70206-2.

Kraut, R., Patterson, M., Lundmark, V., Kiesler, S., Mukophadhyay, T., & Scherlis, W.(1998). Internet paradox: A social technology that reduces social involvementand psychological well-being? American Psychologist, 53(9), 1017–1031. http://dx.doi.org/10.1037/0003-066X.53.9.1017.

Leary, M. R. (1983). Understanding social anxiety: Social, personality and clinicalperspectives. Newbury Park, CA, US: Sage.

Lee, B. W., & Stapinski, L. A. (2012). Seeking safety on the Internet: Relationshipbetween social anxiety and problematic Internet use. Journal of AnxietyDisorders, 26(1), 197–205. http://dx.doi.org/10.1016/j.janxdis.2011.11.001.

Leung, L. (2003). Impacts of Net-generation attributes, seductive properties of theInternet, and gratifications-obtained on Internet use. Telematics and Informatics,20(2), 107–129. http://dx.doi.org/10.1016/S0736-5853(02)00019-9.

Leung, L. (2007). Stressful life events, motives for Internet use, and social supportamong digital kids. CyberPsychology & Behavior, 10(2), 204–214. http://dx.doi.org/10.1089/cpb.2006.9967.

Li, S. M., & Chung, T. M. (2006). Internet function and Internet addictive behavior.Computers in Human Behavior, 22(6), 1067–1071. http://dx.doi.org/10.1016/j.chb.2004.03.030.

Li, H., Wang, J., & Wang, L. (2009). A survey on the generalized problematic Internetuse in Chinese college students and its relations to stressful life events andcoping style. International Journal of Mental Health and Addiction, 7(2), 333–346.http://dx.doi.org/10.1007/s11469-008-9162-4.

Li, D., Zhang, W., Li, X., Zhen, S., & Wang, Y. (2010). Stressful life events andproblematic Internet use by adolescent females and males: A mediatedmoderation model. Computers in Human Behavior, 26(5), 1199–1207. http://dx.doi.org/10.1016/j.chb.2010.03.031.

Liu, X., Liu, L., Yang, J., & Zhao, G. (1997). Reliability and validity of the adolescentself-rating life events checklist. Chinese Journal of Clinical Psychology, 5(1),34–36.

Liu, M., & Peng, W. (2009). Cognitive and psychological predictors of the negativeoutcomes associated with playing MMOGs (massively multiplayer onlinegames). Computers in Human Behavior, 25(6), 1306–1311. http://dx.doi.org/10.1016/j.chb.2009.06.002.

Lo, S. K., Wang, C. C., & Fang, W. (2005). Physical interpersonal relationships andsocial anxiety among online game players. CyberPsychology & Behavior, 8(1),15–20. http://dx.doi.org/10.1089/cpb.2005.8.15.

Ma, H. (1999). Interaction Anxiousness Scale (IAS). In X. D. Wang (Ed.), Rating scalesfor mental health (pp. 230–232). Beijing, China: Chinese Mental Health Journal.

MacKenzie, S. B., Podsakoff, P. M., & Jarvis, C. B. (2005). The problem ofmeasurement model misspecification in behavioral and organizationalresearch and some recommended solutions. Journal of Applied Psychology,90(4), 710–730. http://dx.doi.org/10.1037/0021-9010.90.4.710.

Marsh, H. W., Hau, K. T., & Wen, Z. (2004). In search of golden rules: Comment onhypothesis-testing approaches to setting cutoff values for fit indexes anddangers in overgeneralizing Hu and Bentler’s (1999) Findings. StructuralEquation Modeling, 11(3), 320–341. http://dx.doi.org/10.1016/S0747-5632(03)00040-2.

Morahan-Martin, J., & Schumacher, P. (2003). Loneliness and social uses of theInternet. Computers in Human Behavior, 19(6), 659–671.

Nestler, E. J. (2001). Total recall–the memory of addiction. Science, 292(5525),2266–2267.

Peng, W., & Liu, M. (2010). Online gaming dependency: A preliminary study inChina. CyberPsychology, Behavior, and Social Networking, 13(3), 329–333. http://dx.doi.org/10.1089/cyber.2009.0082.

Reise, S. P., Waller, N. G., & Comrey, A. L. (2000). Factor analysis and scale revision.Psychological Assessment, 12(3), 287–297. http://dx.doi.org/10.1037//1040-3590.12.3.287.

Shaw, M., & Black, D. W. (2008). Internet addiction: Definition, assessment,epidemiology, and clinical management. CNS Drugs, 22(5), 353–365. http://dx.doi.org/10.2165/00023210-200822050-00001.

Shek, D. T. L., Tang, V. M. Y., & Lo, C. Y. (2008). Internet addiction in Chineseadolescents in Hong Kong: Assessment, profiles, and psychosocial correlates.The Scientific World JOURNAL: Child Health and Human Development, 8, 776–787.http://dx.doi.org/10.1100/tsw.2008.104.

Shepherd, R. M., & Edelmann, R. J. (2005). Reasons for Internet use and socialanxiety. Personality and Individual Differences, 39(5), 949–958. http://dx.doi.org/10.1016/j.paid.2005.04.001.

Siomos, K. E., Dafouli, E. D., Braimiotis, D. A., Mouzas, O. D., & Angelopoulos, N. V.(2008). Internet addiction among Greek adolescent students. CyberPsychology &Behavior, 11(6), 653–657. http://dx.doi.org/10.1089/cpb.2008.0088.

2386 Y. Mai et al. / Computers in Human Behavior 28 (2012) 2376–2386

Suler, J. R. (1999). To get what you need: Healthy and pathological Internet use.CyberPsychology & Behavior, 2(5), 385–393. http://dx.doi.org/10.1089/cpb.1999.2.385.

Tekinarslan, E., & Gürer, M. D. (2011). Problematic Internet use among Turkishuniversity students: A multidimensional investigation based on demographicsand Internet activities. International Journal of Human Sciences, 8(1), 1028–1051.

Villella, C., Martinotti, G., Di Nicola, M., Cassano, M., La Torre, G., Gliubizzi, M. D.,et al. (2011). Behavioural addictions in adolescents and young adults: Resultsfrom a prevalence study. Journal of Gambling Studies, 27(2), 203–214. http://dx.doi.org/10.1007/s10899-010-9206-0.

Wan, C. S., & Chiou, W. B. (2006). Why are adolescents addicted to online gaming?An interview study in Taiwan. CyberPsychology & Behavior, 9(6), 762–766.http://dx.doi.org/10.1089/cpb.2006.9.762.

Weinstein, A., & Lejoyeux, M. (2010). Internet addiction or excessive Internet use.The American Journal of Drug and Alcohol Abuse, 36(5), 277–283. http://dx.doi.org/10(3109/00952990), 2010, 4918.

Whang, L. S. M., Lee, S., & Chang, G. (2003). Internet over-users’ psychologicalprofiles: A behavior sampling analysis on Internet addiction. CyberPsychology &Behavior, 6(2), 143–150. http://dx.doi.org/10.1089/109493103321640338.

World Health Organization (1990). International classification of diseases and relatedhealth problems (10th ed.). Geneva, Switzerland: Author.

Yen, C. F., Yen, J. U. Y. U., & Ko, C. H. (2010). Internet addiction: Ongoing research inAsia. World Psychiatry, 9(2), 97.

Young, K. S. (2007). Cognitive behavior therapy with Internet addicts: Treatmentoutcomes and implications. CyberPsychology & Behavior, 10(5), 671–679. http://dx.doi.org/10.1089/cpb.2007.9971.

Young, K. S. (1999). Net compulsions: The latest trends in the area of internetaddiction. <http://www.netaddiction.com/net_compulsions.html> Retrieved07.05.07.

Zboralski, K., Orzechowska, A., Talarowska, M., Darmosz, A., Janiak, A., Janiak, M.,et al. (2009). The prevalence of computer and Internet addiction among pupils.Postepy Hig Med Dosw, 63, 8–12.

Zhang, L., Amos, C., & McDowell, W. C. (2008). A comparative study of Internetaddiction between the United States and China. CyberPsychology & Behavior,11(6), 727–729. http://dx.doi.org/10.1089/cpb.2008.0026.