Computers in Human Behavior 72 (2017) 178e188

Contents lists available at ScienceDirect

Computers in Human Behavior

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

Full length article

Associations between psychosocial factors and generalized

pathological internet use in Chinese university students:
A longitudinal cross-lagged analysis
Yu Tian, PhD a, Yulong Bian, PhD b, Piguo Han, PhD a, Fengqiang Gao a, *, Peng Wang a, **
a
Department of Psychology, Shandong Normal University, China
b
Department of Computer Science and Technology, Shandong University, China

a r t i c l e i n f o a b s t r a c t

Article history: This short-term longitudinal study examined the reciprocal associations among preexisting psychosocial
Received 22 November 2016 variables (shyness, loneliness, and interpersonal relationships), maladaptive cognitions, and generalized
Received in revised form pathological Internet use (GPIU) in a Chinese sample. A total of 361 university freshmen (138 males;
16 February 2017
mean age ¼ 18.57 years) participated in the study. Cross-lagged panel analysis indicated that (i) the
Accepted 20 February 2017
Available online 23 February 2017
associations among shyness, maladaptive cognitions, and GPIU were dynamic and bidirectional, and the
increased maladaptive cognitions bidirectionally mediated the relation between shyness and GPIU across
time; (ii) the association between interpersonal relationships and GPIU was dynamic; and (iii) the as-
Keywords:
GPIU
sociation between loneliness and GPIU was dynamic and bidirectional. Additionally, (iv) shyness and
Shyness GPIU had higher stability than interpersonal relationships, loneliness, and maladaptive cognitions did
Loneliness and (v) gender differences were identified in GPIU (T1 and T2). Interventions for GPIU and implications
Interpersonal relationships for future studies were discussed.
Maladaptive cognitions © 2017 Elsevier Ltd. All rights reserved.
Cross-lagged analysis

1. Introduction
Using the Internet has become one of the most popular leisure-
time activities among adolescents aged 10e21 years in China (Liu,
Fang, Deng, & Zhang, 2012). More than 287 million adolescents
are Internet users, making this age cohort the largest population of
Internet users (China Internet Network Information Center
[CNNIC], 2016). Previous research in China has reported patholog-
ical Internet use and dependency in approximately 8% of adoles-
cents (Fan et al., 2008; Lei & Yang, 2007; Liu et al., 2012; Luo & Peng,
2008). Concern about the large number of adolescents using the
Internet has led to increased research into the possible adverse
effects of Internet use on adolescent well-being, academic
achievement, and socialization (Chen, 2012; Fan et al., 2008; Lei &
Yang, 2007; Liu et al., 2012; Lu & Yeo, 2015; Luo & Peng, 2008;
* Corresponding author. Department of Psychology, Shandong Normal University,
No. 88, East Wenhua Road, Ji'nan, Shandong, 250014, China.
** Corresponding author. Department of Psychology, Shandong Normal University,
No. 88, East Wenhua Road, Ji'nan, Shandong, 250014, China.
E-mail addresses: 1427329793@qq.com (Y. Tian), bianyulong_007@hotmail.com
(Y. Bian), hanpiguo1215@163.com (P. Han), gaofengqiang@sdnu.edu.cn (F. Gao),
pengsdnu@163.com (P. Wang).

Niculovi
c, Zivkovi

c, Manasijevi
c, & Strbac, 2014).
The anonymity and absence of nonverbal and demographic cues
provided by the Internet can be beneficial to adolescents' well-
being by offering relief from emotional distress (Caplan & Turner,
2007), as well as enhancing perceptions of social support and
self-esteem (Kraut et al., 2002) and expanding the ranges of
interpersonal relationships (Cotten, 2008). However, pathological
Internet use (PIU), which is characterized by excessive or compul-
sive Internet use and a preoccupation with and loss of control over
this use, results in negative personal and professional conse-
quences (Caplan, 2002; Davis, 2001) and may be detrimental
(Greenfield & Yan, 2006). The number of adolescent Internet users
in China has increased rapidly from 120 million in 2002 to 287
million as of the time of publication (CNNIC, 2016), placing growing
numbers at the risk of developing PIU. This large population and the
increasing trend indicate that investigating how adolescents ac-
quire PIU and developing effective intervention techniques for PIU
are imperative. A number of studies have demonstrated that ado-
lescents with PIU usually suffer from loneliness (Engelberg &
lu,
Sjoberg, 2004; Turkle, 2011; Young, 1998), shyness (Erog
Pamuk, & Pamuk, 2013), poor interpersonal relationships (Odaci
& Çikrikçi, 2014; Sanders, Field, Diego, & Kaplan, 2000), cognitive

http://dx.doi.org/10.1016/j.chb.2017.02.048
0747-5632/© 2017 Elsevier Ltd. All rights reserved.
 Y. Tian et al. / Computers in Human Behavior 72 (2017) 178e188 179

distortion (Lu & Yeo, 2015) and other decreases in well-being (Liu Although numerous studies have shown a significant relation-
et al., 2012; Lu & Yeo, 2015). Therefore, investigating the detailed ship between preexisting psychosocial variables (shyness, loneli-
relationships of these variables and PIU is necessary. ness, and interpersonal relationships), maladaptive cognitions, and
GPIU, few of them have directly tested the directionality of this
association. For example, most studies examining the links be-
2. Literature review
tween shyness and GPIU have looked unidirectionally from either
shyness to GPIU (Caplan, 2002; Casale & Fioravanti, 2011), GPIU to
In identifying the etiology of PIU, the cognitive-behavioral lu et al., 2013), or from shyness to GPIU through
shyness (Erog
model offers theoretical explanations of the origins and patho-
maladaptive cognitions (Tian, Bian, Han, Wang, & Gao, 2015).
genesis of GPIU (Davis, 2001, Fig. 1). Davis (2001) suggested that
However, it may be that the associations among shyness, mal-
psychopathology (such as depression, social anxiety, and substance
adaptive cognitions, and GPIU are bidirectional, with each charac-
dependence) is a distal necessary cause of symptoms of GPIU.
teristic mutually influencing another. An understanding of the
Notably, he argued that underlying psychopathology does not in
directionality of the associations among shyness, maladaptive
itself result in symptoms of GPIU, but is a necessary element in its
cognitions, and GPIU has important practical implications for
etiology. The key factors of GPIU are maladaptive cognitions, which
intervention efforts. Intervention based on incorrect assumptions
are proximal sufficient causes. Furthermore, some studies have
of the nature of the directionality of associations may prove to be
indicated that distal psychopathology renders an individual
ineffective. For example, if the goal of an intervention is to decrease
vulnerable to GPIU through maladaptive cognitions (Caplan, 2005;
GPIU and shyness is not the antecedent but the effect of decreasing
Kalkan, 2012; Li & Wang, 2013; Liu & Peng, 2009; Lu & Yeo, 2015;
maladaptive cognitions, an intervention aimed at decreasing
Mai et al., 2012). Additionally, the social context of the individual
shyness would be unlikely to decrease GPIU. If shyness and mal-
contributes to the causal pathway of GPIUdspecifically, through
adaptive cognitions influence each other over time, interventions
social isolation and a lack of social support from family and friends
can benefit from targeting both shyness and maladaptive
(Davis, 2001; Liu et al., 2012; Lu & Yeo, 2015).
cognitions.
On the basis of the cognitive-behavioral theory described by
To our knowledge, little research has addressed the direction-
Davis (2001) and previous studies (Caplan, 2005; Kalkan, 2012; Li &
ality of associations among preexisting psychosocial variables
Wang, 2013; Liu & Peng, 2009; Lu & Yeo, 2015; Mai et al., 2012),
(shyness, loneliness, and interpersonal relationships), maladaptive
preexisting psychosocial variables (shyness, loneliness, and inter-
cognitions, and GPIU. One study examined causal priority in the
personal relationships) related to GPIU were selected. Specifically,
observed empirical relationship between Internet addiction and
shyness tends to be a distal necessary cause of PIU symptoms,
loneliness. The results indicated that excessive and unhealthy
because the core trait of shyness is social anxiety (Mccabe, 2015;
Internet use increases feelings of loneliness over time (Yao &
Snyder, Smith, Augelli, & Ingram, 1985), which renders an indi-
Zhong, 2014). This study enhanced previous research by using
vidual vulnerable to GPIU through maladaptive cognitions. Addi-
Chinese samples. However, it had some limitations. First, the time
tionally, social support is a significant predictor of interpersonal
span between the two waves of surveys might have been too short
relationships and loneliness (Solomon, Bensimon, Greene, Horesh,
to detect a stable change in Internet addiction and loneliness over
& Eindor, 2015; Zhang, Gao, Fokkema, Alterman, & Liu, 2015),
time. Furthermore, because only two time points were applied in
which tend to be the causal pathway of GPIU rather than its distal
the cross-lagged model, examining the dynamic association be-
necessary cause. Therefore, this study identified maladaptive cog-
tween Internet addiction and loneliness across time was not
nitions as important predictors and proximal causes of PIU,
possible. Excessive and unhealthy Internet use may increase feel-
whereas shyness was identified as the distal necessary cause of
ings of loneliness, and increased loneliness may further increase
GPIU and interpersonal relationships and loneliness as its causal
excessive and unhealthy Internet use. Second, the observed
pathway.

Fig. 1. The cognitiveebehavioral model of pathological Internet use (PIU). Adapted from Davis (2001).
180 Y. Tian et al. / Computers in Human Behavior 72 (2017) 178e188
empirical relationship between Internet addiction and loneliness
was tested. This method cannot test the measurement invariance
across time, which may introduce measurement error in later
analysis. Therefore, more than two waves of surveys and latent
cross-lagged analysis are necessary.
2.1. Cross lagged analysis
Numerous questions in psychological research concern how two
or more variables influence each other over time. An example of
such research questions is “How do rumination and anxiety influ-
ence each other?” Because of ethical limitations, many of these
questions cannot be investigated experimentally. However, cross-
lagged panel design can effectively solve these questions. In this
approach, two or more variables are measured on two or more
occasions. The cross-lagged associations between the variables
over time can be examined while controlling for the effect that the
variables have on themselves (i.e., the autoregression; Rogosa,
1980; Schuurman, Ferrer, De, & Hamaker, 2016). An important
goal of cross-lagged panel studies is to establish causal effects using
cross-lagged regression coefficients, and then compare these as-
sociations with respect to their strength. The strongest association
is judged to provide the most important causal influence that drives
the system, which is also referred to as being “causally dominant.”
This causal influence can assist determining the bidirectional as-
sociations among variables. Additionally, the autoregressive effects
of the cross-lagged panel model indicate how stable the individual
differences in the scores are over time (Schuurman et al., 2016;
Talbot et al., 2012).
2.2. Freshmen of university
The beginning of university life is a developmental period for
students. Several theorists (e.g., Collins & Steinberg, 2006;
Woodhouse, Dykas, & Jude, 2012) have proposed that university
students shift from a relatively greater reliance on parents for
support and interaction to a reliance on interpersonal relationships
(not only family relationships but also peer and other social re-
lationships). Interpersonal relationships tend to be a central pre-
occupation for university students (Brown & Klute, 2003).
Additionally, past studies have demonstrated that heavy Internet
use is related to a decline in family communication and supervision
(Liu et al., 2012; Park, Kim, & Cho, 2008; Van den Eijnden,
Spijkerman, Vermulst, Van Rooij, & Engles, 2010). Therefore, the
shift in interpersonal relationships tends to place university stu-
dents at risk of developing PIU. Furthermore, owing to abundant
leisure time and unlimited Internet access through a range of
wireless tools, contemporary university students tend to spend
large amounts of time online, and are therefore very likely to
experience symptoms of PIU (Chen, 2012; Morahan-Martin &
Schumacher, 2000). Hence, we believe that university freshmen are
more likely to develop GPIU than other type school students.
2.3. Gender difference
Morahan-Martin and Schumacher (2000) suggested that path-
ological Internet users use the Internet for various purposes such as
social interaction with online-only contacts (sharing ideas or fan-
tasies, meeting new people, talking to others with shared interests,
seeking emotional support) and recreational Internet usage
(relaxing, gambling, using adult-only resources, playing games,
using virtual reality, wasting time). However, gender differences
also exist in Internet use purposes. For example, females engage
more with Web blogs and personal home pages, which can be used
as outlets for self-advertisement and self-expression, whereas
males exhibit more interest in the latest information technology
and playing games (Niculovi
c et al., 2014). These different purposes
for Internet use may yield gender differences in specific-field PIU
(SPIU). However, no research has comprehensively described
gender differences in SPIU such as daily routines, social behavior,
academic and occupational productivity, sleep patterns, feelings,
and online activities (e.g., time spent on games, chatting, and
shopping) for a Chinese sample. Hence, additional studies are
necessary to extend previous research by examining gender dif-
ferences in these specific fields.
2.4. The present study
To investigate the formation mechanisms of GPIU, and provide
evidence for the prevention of GPIU in university freshmen, pre-
existing psychosocial variables (shyness, loneliness, and interper-
sonal relationships), maladaptive cognitions, and GPIU were
combined in a cognitive-behavioral model. Furthermore, a latent
cross-lagged panel model design was used to investigate the as-
sociations among preexisting psychosocial variables (shyness,
loneliness, and interpersonal relationships), maladaptive cogni-
tions, and GPIU across time in a Chinese sample. We hypothesized
that (i) shyness, maladaptive cognitions, and GPIU influence each
other over time and that shyness influences GPIU through mal-
adaptive cognitions; (ii) interpersonal relationships and GPIU in-
fluence each other over time; and (iii) loneliness and GPIU
influence each other over time. Additionally, the self-reported
scores for GPIU and SPIU were tested for gender difference, as
well as shyness, loneliness, interpersonal relationships, and mal-
adaptive cognitions.
3. Method
3.1. Participants
The initial number of young people available for participation
was 500 participants (279 females and 221 males). All were uni-
versity students from two educational institutions in Eastern China.
The questionnaire measurements were taken at the beginning of
the university year (T1), 3 months later (T2), and 6 months later
(T3). Of the initial sample, 139 adolescents did not complete the
third wave of measurements (participation rate: 72.20%). The lack
of participation was almost entirely owing to sickness or absence.
Consequently, the final sample comprised 361 adolescents (223
females and 138 males), with a mean age at the beginning of the
study of 18.57 years (standard deviation [SD] ¼ 0.88). A series of
independent sample t tests was conducted to determine whether
participating adolescents differed from nonparticipating adoles-
cents regarding any of the variables applied in this study. None of
these analyses was significant.
A team of researchers from another city came to the university
to administer a series of self-report questionnaires. No university-
related adults were present during the questionnaire administra-
tion, which was conducted by a researcher during one full class
period. Students were informed that no one in their university
would see their reports and that the researchers would not know
who the students were when processing their collective reports. At
the end of the session, the students were briefed about the purpose
of the research and the absolute anonymity of their identities.
Because there were no student names included on the surveys, the
school staff chose to collect passive consent from parents. Thus,
parents were notified and given the option of refusing to allow their
child's participation. Parental consent forms were distributed to all
the students. Close to 96% of the students' parents returned the
consent forms to allow their children's participation.
 Y. Tian et al. / Computers in Human Behavior 72 (2017) 178e188
3.2. Measures
3.2.1. Shyness
Shyness was measured using the Chinese adaptation (Wang,
Wang, Han, Gong, & Gao, 2009) of the Cheek and Buss Shyness
Scale (Cheek & Buss, 1981). The questionnaire measures shyness as
a deficit of social skills. The Chinese version is a 17-item (instead of
14) self-reporting scale. Students rated their agreement with the
items on a 5-point Likert scale ranging from 1 (strongly disagree) to
5 (strongly agree). A sample item for shyness is “I do not find it
difficult to ask other people for information.” The Cronbach's a
values for shyness across three time points were 0.79, 0.88, and
0.83.
3.2.2. Interpersonal relationships
The 10-item scale developed by Fang, Wo, and Lin (2005) was
used to measure the quantity of interpersonal relationships (five
items) and the quality of interpersonal relationships (five items).
The participants responded on a 5-point Likert scale ranging from 1
(strongly disagree) to 5 (strongly agree). A sample item for the
quantity of interpersonal relationships is “has trouble making
friends” and a sample item for the quality of interpersonal re-
lationships is “most of my best friends do not care about my feel-
ings or welfare.” Cronbach's a for the quantity of interpersonal
relationships and quality of interpersonal relationships for the
three time points were 0.89e0.94 and 0.85e0.95, respectively,
demonstrating the reliability of these two subscales in the present
sample.
3.2.3. Loneliness
Loneliness was measured using the Chinese adaptation (Wang,
1995) of the Cheek and Russell Loneliness Scale (Russell, Peplau,
& Ferguson, 1978). This scale consists of 20 self-statements con-
cerning people's satisfaction with their interpersonal relationships.
The statements are answered on a 4-point scale for how often
(1 ¼ never, 4 ¼ often) subjects feel the statements apply to them-
selves, with high scores indicating greater perceived loneliness. A
sample item for loneliness is “I feel lonely even when I am around
other people.” Cronbach's a values for loneliness for the three time
points were 0.88, 0.87, and 0.92.
3.2.4. Maladaptive cognitions
The 14-item scale developed by Li, Zhang, Li, Zhen, and Wang
(2010) was used to measure maladaptive cognitions regarding so-
cial convenience and coping with stress. The participants respon-
ded on a 5-point Likert scale ranging from 1 (strongly disagree) to 5
(strongly agree). A sample item for maladaptive cognition regarding
social convenience was “friends in this online network are more
trustworthy than my real-world friends” and for coping with stress
was “people can temporarily relieve pressure by playing online
games.” Cronbach's a values for social convenience and coping with
stress for the three time points were 0.78e0.88 and 0.85e0.91.
3.2.5. GPIU
GPIU was measured using the Chinese adaptation (Bai & Fan,
2005) of the Internet Addiction Test (Young, 1998), which as-
sesses the degree to which Internet use affects people's daily rou-
tines, social life, academic and occupational productivity, sleep
patterns, feelings, and online activities (time spent on games,
chatting, and shopping) by asking 20 questions (e.g., “How often do
others in your life complain to you about the amount of time you
spend online?”). Student rated their agreement with the items on a
6-point Likert scale ranging from 1 (never) to 6 (always). Cronbach's
a values for GPIU across three time points were 0.82, 0.88, and 0.86.
181
3.3. Analytical approach
Three cross-lagged panel designs were employed. The first
model included measures of shyness, maladaptive cognitions, and
GPIU, the second model included measures of interpersonal re-
lationships and GPIU, and the third model included measures of
loneliness and GPIU. Structural equation modeling was used to test
the three hypotheses of the study. All of the models were tested
through maximum likelihood estimation using MPLUS 7.0. In cross-
lagged models, change in each variable over time is modeled using
the stability coefficients between time-adjacent measures of each
variable (e.g., T1 GPIU predicts T2 GPIU, and T2 GPIU in turn pre-
dicts T3 GPIU), and the cross-lagged relations among the three
models are captured by the cross-lagged effects between two var-
iables (e.g., T1 shyness predicts T2 GPIU, and T1 GPIU in turn pre-
dicts T2 shyness). We also tested the configural invariant and
metric equivalence of each measure across the three occasions (e.g.,
factor loadings involving the same indicator were specified to be
the same over time, and the errors of the same indicator were also
specified to covary over time).
Additionally, the parceling strategy of the factorial algorithm
was used to improve the quality of the indicators and model fit.
Following Rogers and Schmitt (2004), factor analysis was first
performed, and items of each latent variable were ranked from
highest to lowest according to the factor loading size. Each parcel
was sequentially assigned the remaining items with the highest
and lowest rankings, alternating direction through the parcels,
until all items were assigned. For example, in the case of 16 items
assigned to four parcels, Parcel #1 ¼ items ranked 1, 8, 9, and 16;
Parcel #2 ¼ items ranked 2, 7, 10, and 15; Parcel #3 ¼ items ranked
3, 6, 11, and 14; and Parcel #4 ¼ items ranked 4, 5, 12, and 13.
The fit of all the cross-lagged models was assessed using the chi-
squared (c2) statistic, Tucker-Lewis Index (TLI), comparative fit in-
dex (CFI), and root-mean-square error of approximation (RMSEA).
Because the chi-squared statistic is sensitive to large samples,
model fit indices were used as the primary criteria to evaluate
model fit. The TLI and the CFI range between 0 and 1, with values
above 0.90 indicating adequate model fit (Hu & Bentler, 1999). A
rule of thumb for the RMSEA is that values  0.10 indicate poor fit,
values between 0.05 and 0.08 indicate reasonable error of
approximation, and values  0.05 indicate close approximation
(Browne & Cudeck, 1993). The chi-squared test of difference (Dc2)
was used to compare the fit of the nested models. A nonsignificant
Dc2 test indicates that the two models provide an equal fit to the
data, whereas a significant Dc2 suggests that the less constrained
model should be retained.
4. Results
4.1. Descriptive statistics and difference test
The means and standard deviations of shyness, interpersonal
relationships and loneliness, maladaptive cognitions, and GPIU are
presented in Table 1. Independent sample t tests were conducted to
compare males and females through the measures of shyness,
interpersonal relationships and loneliness, maladaptive cognitions,
and GPIU. The results showed that the self-reported scores of
shyness, interpersonal relationships, and loneliness exhibited no
differences with respect to gender at the three times. However,
males had a higher GPIU score than females did at T1 and T2
(Table 1).
The correlations among all variables are shown in Table 2. Cor-
relations among T1, T2, and T3 measures of shyness were in the
0.61e0.71 range. Correlations among T1, T2, and T3 measures of
interpersonal relationships were in the 0.43e0.51 range.
182 Y. Tian et al. / Computers in Human Behavior 72 (2017) 178e188

Table 1
Means, and standard deviations of shyness, interpersonal relationship and loneliness, maladaptive cognitions, GPIU and gender differences.

Variable Total (n ¼ 361) Male (n ¼ 138) Female (n ¼ 223) t df p

M SD M SD M SD

1. T1 Shyness 45.43 10.05 45.25 10.10 45.55 10.04 0.73 359 0.78
2. T1 IR 31.92 4.73 32.01 4.98 31.86 4.57 0.29 359 0.77
3. T1 Loneliness 39.51 8.51 38.91 8.53 39.89 8.50 1.01 359 0.29
4. T1 MC 33.90 9.35 33.94 9.67 33.88 9.00 2.85 359 0.52
5. T1 GPIU 34.95 10.08 36.17 10.40 34.20 9.82 2.10 359 <0.05
6. T2 Shyness 47.90 10.35 47.23 11.46 48.32 9.61 0.97 359 0.33
7. T2 IR 32.92 4.68 32.74 5.26 33.03 4.29 0.56 359 0.58
8. T2 Loneliness 41.51 18.88 40.71 10.24 42.01 22.64 0.63 359 0.53
9. T2 MC 34.79 10.43 35.10 11.06 34.61 10.92 2.22 359 0.87
10. T2 GPIU 36.86 11.85 38.14 13.08 36.06 11.70 1.00 359 <0.01
11. T3 Shyness 46.85 10.75 45.68 11.23 47.57 10.40 1.62 359 0.11
12. T3 IR 33.08 4.73 33.18 5.05 33.01 4.53 0.34 359 0.73
13. T3 Loneliness 41.51 9.65 40.87 10.37 41.91 9.17 0.08 359 0.32
14. T3 MC 33.83 10.67 34.70 11.35 33.30 10.21 1.22 359 0.22
15. T3 GPIU 37.68 11.78 37.62 11.57 37.71 11.93 0.71 359 0.94

Note: IR ¼ interpersonal relationship, MC ¼ maladaptive cognitions.

Table 2
Correlations of shyness, interpersonal relationship, loneliness, maladaptive cognitions, and GPIU over a three-wave test.

Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

1. T1 Shyness 1
2. T1 IR 0.21** 1
3. T1 Loneliness 0.54** 0.55** 1
4. T1 MC 0.43** 0.08 0.33** 1
5. T1 GPIU 0.39** 0.14** 0.32** 0.71** 1
6. T2 Shyness 0.71** 0.22** 0.50** 0.36** 0.36** 1
7. T2 IR 0.23** 0.47** 0.37** 0.09 0.11* 0.29** 1
8. T2 Loneliness 0.21** 0.24** 0.41** 0.14** 0.17** 0.23** 0.23** 1
9. T2 MC 0.35** 0.15** 0.33** 0.67** 0.63** 0.39** 0.09 0.19** 1
10. T2 GPIU 0.35** 0.17** 0.32** 0.57** 0.74** 0.35** 0.15** 0.22** 0.71 1
11. T3 Shyness 0.61** 0.21** 0.48** 0.31** 0.32** 0.68** 0.26** 0.17** 0.38** 0.35** 1
12. T3 IR 0.21** 0.43** 0.37** 0.08 0.06 0.25** 0.51** 0.23** 0.11* 0.06 0.25** 1
13.T3 Loneliness 0.40** 0.52** 0.67** 0.30** 0.31** 0.44** 0.38** 0.41** 0.34** 0.37** 0.44** 0.43** 1
14. T3 MC 0.30** 0.06 0.23** 0.60** 0.57** 0.34** 0.14** 0.12* 0.69** 0.57** 0.35** 0.12* 0.35** 1
15. T3 GPIU 0.29** 0.14** 0.31** 0.51** 0.61** 0.31** 0.15** 0.24** 0.58** 0.71** 0.40** 0.18** 0.44** 0.69** 1
*
Correlation is significant at the 0.05 level (2-tailed).
**
Correlation is significant at the 0.01 level (2-tailed).

Correlations among T1, T2, and T3 measures of loneliness were in
the 0.41e0.67 range. Correlations among T1, T2, and T3 measures of
maladaptive cognitions were in the 0.57e0.67 range. Correlations
among T1, T2, and T3 measures of GPIU were in the 0.61e0.74
range. The associations among the measures of shyness, interper-
sonal relationships, loneliness, maladaptive cognitions, and GPIU
were in the expected direction. That is, shyness, loneliness, mal-
adaptive cognitions, and GPIU were positively correlated with each
other, and interpersonal relationships was negatively correlated
with shyness, loneliness, maladaptive cognitions, and GPIU
(Table 2).
Additionally, pairwise comparisons indicated that the values for
interpersonal relationships, loneliness, maladaptive cognitions, and
GPIU at T2 were equal to those at T3. However, the values were
significantly greater than those at T1. Additionally, shyness at T3
was significantly greater than at T2 and shyness at T2 was signifi-
cantly greater than at T1 (see Table 3).
4.2. The cross-lagged associations among shyness, maladaptive
cognitions and GPIU
Establishing Measurement Invariance Over Time. To test hypoth-
esis (i), the measurement model's fit for the latent variables of the
study was examined. Item-parcels were used as indicators of the
latent variables. Four parcels were used as indicators for shyness,
two for maladaptive cognitions, and four for GPIU. The latent var-
iables of shyness, maladaptive cognitions, and GPIU were specified
to covary over time. Errors of the same indicator were also specified
to covary over time. Additionally, factor loadings involving the
same indicator were specified to be the same over time. To deter-
mine whether the measurement model could be considered
invariant or equivalent over time, the model fit of two alternative
models (differing in number of parameter restrictions) was tested
and compared (Cole & Maxwell, 2003) using the corrected chi-
squared difference test (Satorra & Bentler, 2001).
First, a longitudinal confirmatory factor analysis was under-
taken, which included all observed and latent variables from each
time with freely estimated parameters (all the observed indicators
of latent variables and the errors of each indicator were specified to
covary over time). The measurement model consisted of nine latent
variables corresponding to the three variables of the study being
measured three times. This measurement model (unconstrained
model) exhibited good fit indexes (see M1 in Table 4). This model
was compared with a more restrictive model (metric invariance
model) in which all the indicators of latent variable factors were
specified as being equal across time. Although the data model fit
was good (see M2 in Table 4), the chi-squared test of difference
indicated that the model without metric invariance fit the data
more closely (Dc2 ¼ 30, Ddf ¼ 18, p < 0.05). Partial metric invari-
ance was then tested by freely estimating the first factor indicator
 Y. Tian et al. / Computers in Human Behavior 72 (2017) 178e188 183

Table 3
Pairwise comparisons of shyness, interpersonal relationship and loneliness, maladaptive cognitions, GPIU over a three-wave test.

Items T1 (M, SD) T2 (M, SD) T3 (M, SD) F (1, 360) Pairwise comparisons

Shyness 45.43 (10.05) 47.90 (10.35) 49.85 (10.75) 8.40** T1 < T2 < T3
IR 31.92 (4.73) 32.92 (4.68) 33.08 (4.73) 18.99** T1 < T2 ¼ T3
Loneliness 39.51 (8.51) 41.51 (18.88) 41.51 (9.65) 26.03** T1 < T2 ¼ T3
MC 32.17 (9.35) 33.56 (10.43) 33.83 (10.67) 12.21** T1 < T2 ¼ T3
GPIU 34.95 (10.10) 37.35 (11.85) 37.68 (11.78) 28.31** T1 < T2 ¼ T3
*
Correlation is significant at the 0.05 level (2-tailed).
**
Correlation is significant at the 0.01 level (2-tailed).

Table 4 maladaptive cognitions and GPIU were bidirectional: GPIU and

Fit statistics for model comparison of cross-lagged associations among shyness, maladaptive cognitions positively predicted each other across time;
maladaptive cognitions and GPIU. the associations between shyness and maladaptive cognitions were
Model M1 M2 M3 M4 M5 dynamic and bidirectional: shyness positively predicted maladap-
tive cognitions across time; and increased T2 maladaptive cogni-
Chi-square (c2) 1557 1587 1585 1756 1843
df 459 477 475 497 494 tions (T1 < T2) positively predicted T3 shyness (T3 > T2). The
RMSEA 0.05 0.05 0.05 0.06 0.07 association between shyness and GPIU was dynamic: initial GPIU
TLI 0.93 0.98 0.98 0.94 0.93 positively predicted increased shyness at T2 (T1 < T2) (Fig. 2).
CFI 0.94 0.98 0.98 0.93 0.94
Therefore, hypothesis (i) was verified.
Note: M1 ¼ unconstrained model, M2 ¼ metric invariance model, M3 ¼ partial The above paths were indicative of some mediational mecha-
metric invariance model, M4 ¼ autoregressive model, M5 ¼ cross-lagged model. nisms. The link between T1 shyness and T3 GPIU was mediated
through the increased maladaptive cognitions at T2, and the link
between T1 GPIU and T3 shyness was mediated through the
of shyness at T1 (pairwise comparisons indicated the first in- increased maladaptive cognitions at T2. The significance of these
dicator's effect of shyness at T1 ¼ 0.41 vs. T3 ¼ 0.61, exhibiting a mediational paths was tested using bootstrapping. With this aim,
maximum invariance). A good model fit was achieved for the par- 1000 bootstrap samples were created from the original data set by
tial metric invariance model (see M3 in Table 4) and an equally good random sampling with replacement. According to the proposal of
fit to the data was achieved with the unconstrained model Shrout and Bolger (2002), a mediational effect is significant at the
(Dc2 ¼ 28, Ddf ¼ 16, p ¼ 0.11). Thus, the partial metric invariance 0.05 level if the 95% confidence level does not include zero. The
model was used in the subsequent analysis. results indicate that all mediational paths were statistically signif-
Autoregressive Model. The next step was to test an autoregressive icant: the indirect effect of T1 shyness on T3 GPIU through T2
model, which included autoregressive paths between the T1, T2, maladaptive cognitions (95% CI: 0.12e0.15) and the indirect effect
and T3 measures of shyness, interpersonal relationships, and GPIU. of T1 GPIU on T3 shyness through T2 maladaptive cognitions
All autoregressive paths were significant and are displayed in Fig. 2. (0.15e0.19). Thus, maladaptive cognitions played a bidirectional
The fit indices for the autoregressive model were excellent (see M4 mediating role between shyness and GPIU across time.
in Table 4).
Hypothesized Cross-Lagged Model. Building on the above autor- 4.3. The cross-lagged associations between interpersonal
egressive model, the fit of the hypothesized model was tested. It relationship and GPIU
included cross-lagged predictive paths from shyness and mal-
adaptive cognitions to GPIU, from maladaptive cognitions and GPIU Establishing Measurement Invariance Over Time. The measure-
to shyness, and from shyness and GPIU to maladaptive cognitions ment model consisted of six latent variables corresponding to the
across time. The fit indices for this model were excellent (see M5 in two variables of the study measured three times; two parcels were
Table 4). The results indicated that the associations between used as indicators for interpersonal relationships, and four parcels

Fig. 2. Cross-lagged associations among shyness, maladaptive cognitions and GPIU. Note: MC ¼ maladaptive cognitions. Only significant associations are displayed. Coefficients
represent standardized values. *Coefficient is significant at the 0.05 level. **Coefficient is significant at the 0.01 level.
184 Y. Tian et al. / Computers in Human Behavior 72 (2017) 178e188
Table 5
Fit statistics for model comparison of cross-lagged associations between interper-
sonal relationship and GPIU.
Model M1 M2 M3
Chi-square (c2) 517 528 509
df 120 133 134
RMSEA 0.05 0.05 0.05
TLI 0.94 0.98 0.95
CFI 0.95 0.98 0.96
Note: M1 ¼ unconstrained model, M2 ¼ metric invariance
M3 ¼ autoregressive model, M4 ¼ cross-lagged model.
for GPIU. This measurement model showed good fit indexes (see
M1 in Table 5). Additionally, a good model fit was also achieved for
the metric invariance model (see M2 in Table 5), fitting the data
equally as well as the unconstrained model did (Dc2 ¼ 11, Ddf ¼ 13,
p ¼ 0.31).
Autoregressive Model. The next step was to test an autoregressive
model, which included autoregressive paths between the T1, T2,
and T3 measures of interpersonal relationships and GPIU. All
autoregressive paths were significant and are displayed in Fig. 3. Fit
indices were excellent for the autoregressive model (see M3 in
Table 5).
Hypothesized Cross-Lagged Model. Building on the above autor-
egressive model, we tested the fit of the hypothesized model. It
included cross-lagged predictive paths from interpersonal re-
lationships to GPIU, and from GPIU to interpersonal relationships
across time. The fit indices for this model were excellent (see M4 in
Table 5). The results indicate that the associations between inter-
personal relationships and GPIU were dynamic: initial interper-
sonal relationships negatively predicted later increased GPIU
(T1 < T2); however, increased GPIU negatively predicted final
interpersonal relationships (Fig. 3). Therefore, hypothesis (ii) was
verified.
4.4. The cross-lagged associations between loneliness and GPIU
Establishing Measurement Invariance Over Time. The measure-
ment model consisted of six latent variables corresponding to the
two variables of the study measured three times; two parcels were
used as indicators for loneliness, four parcels for GPIU. This mea-
surement model showed good fit indices (see M1 in Table 6).
Additionally, a good model fit was achieved for the metric invari-
ance model (see M2 in Table 6), fitting the data equally as well as
the unconstrained model did (Dc2 ¼ 18, Ddf ¼ 11, p ¼ 0.07). Thus,
the unconstrained model was then used for the subsequent
analysis.
Autoregressive Model. The next step was to test an autoregressive
model, which included autoregressive paths between T1, T2, and T3
Fig. 3. Cross-lagged associations between interpersonal relationship and GPIU. Note: IR ¼ interpersonal relationship. Only significant associations are displayed. Coefficients
represent standardized values. *Coefficient is significant at the 0.05 level. **Coefficient is significant at the 0.01 level.
measures of loneliness and GPIU. All autoregressive paths were
significant and are displayed in Fig. 4. Fit indices were excellent for
the autoregressive model (see M3 in Table 6).
M4 Hypothesized Cross-Lagged Model. Building on the above autor-
505 egressive model, the fit of the hypothesized model was tested. It
130 included cross-lagged predictive paths from loneliness to GPIU, and
0.06 from GPIU to loneliness across time. Fit indices were excellent for
0.95
this model (see M4 in Table 6). The results indicate that the asso-
0.94
ciation between loneliness and GPIU was dynamic and bidirec-
model,
tional: loneliness positively predicted GPIU across time, whereas
increased GPIU at T2 negatively predicted later loneliness (Fig. 4).
Thus, hypothesis (iii) was verified.
5. Discussion
5.1. The reciprocal associations among shyness, maladaptive
cognitions and GPIU
The present study revealed that maladaptive cognitions played a
bidirectional mediating role between shyness and GPIU, which
indicates that shyness and GPIU influence each other through
maladaptive cognitions across time. In other words, individuals
with severe maladaptive cognitions may suffer from more severe
shyness and GPIU than others. According to the model described by
Davis (2001), the cognitive dysfunctions are a sufficient determi-
nant of subsequent symptoms of GPIU. These symptoms, in turn,
feed back into the subsequent cognitive dysfunctions and shyness.
Generally, this finding was consistent with most previous research
findings that maladaptive cognitions play a crucial role in devel-
oping PIU (Caplan, 2005; Kalkan, 2012; Li & Wang, 2013; Liu &
Peng, 2009; Lu & Yeo, 2015; Mai et al., 2012).
Importantly, GPIU was found to render an individual vulnerable
to shyness through maladaptive cognitions across time. This is
because individuals with PIU were more satisfied than others with
their Internet usage because they felt that the Internet is more
entertaining, fun, and interactive, helping them to escape from
their real-world responsibilities and identity (Chou & Hsiao, 2000).
However, maladaptive cognitions increased later shyness and GPIU.
Additionally, shyness was also found to renders an individual
vulnerable to GPIU through maladaptive cognitions across time.
Shy people may find it difficult to show warmth and empathy with
other people in face-to face-conversation owing to their having
poorer interpersonal skills. They may also believe that they are less
liked and accepted by their peers. However, the Internet may also
provide them with anonymity and an absence of nonverbal and
demographic cues that can facilitate their communication,
relieving them the anxiety of face-to-face interaction, increasing
their sense of comfort and social control (Caplan, 2002; Tian et al.,
2015). However, the apparent benefits may be misleading (i.e.,
maladaptive cognitions), further increasing later shyness and GPIU.
 Y. Tian et al. / Computers in Human Behavior 72 (2017) 178e188
Table 6
Fit statistics for model comparison of cross-lagged associations between loneliness
and GPIU.
Model M1 M2 M3
Chi-square (c2) 401 419 403
df 120 131 123
RMSEA 0.04 0.04 0.05
TLI 0.97 0.97 0.95
CFI 0.96 0.96 0.96
Note: M1 ¼ unconstrained model, M2 ¼ metric invariance
M3 ¼ autoregressive model, M4 ¼ cross-lagged model.
Additionally, the positive direct effect of GPIU on later increased
shyness was also verified. The attraction of the Internet for shy
individuals may be that it enables them to disguise their identities
easily and avoid experiencing anxiety in social interactions (Erog lu
et al., 2013). However, The more time they spend online, the less
time they engage in offline social interactions. This could eventually
lead to a strong dependency on the Internet, which may lead to
their becoming alienated from normal socialization, thereby
perpetuating reduced social and communication skills (Durkee
et al., 2012). Consequently, they may experience rejection in so-
cial activities and suffer increased social anxiety (Ehrensaft, 2005;
Wentzel & Mcnamara, 1999), which in turn exacerbates their
shyness.
5.2. The reciprocal associations between interpersonal relationships
and GPIU, loneliness and GPIU
According to cognitive-behavioral model, interpersonal re-
lationships and loneliness tend to be the causal pathway of GPIU
(Davis, 2001). However, this study does not conclusively demon-
strate that interpersonal relationships and loneliness are causal
pathways of GPIU, owing to their associations being dynamic and
bidirectional. Some studies have identified that the cycle of inter-
personal relationships, loneliness, and PIU involves individual with
poor interpersonal relationships and loneliness using the Internet
to compensate for offline social interaction thereby developing PIU.
PIU and lack of social interactions in real life then exacerbate their
feelings of loneliness and poor interpersonal relationships (Lu &
Yeo, 2015; Yao & Zhong, 2014). All these mechanisms may
contribute to the creation of a “snowball” or “cascade” effect at the
beginning of university life, which is a developmental stage at
which the occurrence of poor interpersonal relationships and
loneliness predicts increased GPIU. GPIU in turn increases the
likelihood of experiencing further poor interpersonal relationships
and loneliness. Consequently, GPIU stabilizes into trait-like risks
(0.67 / 0.69), whereas interpersonal relationships (0.46 / 0.43)
and loneliness tend toward unstable “state-like” risks
(0.39 / 0.35).
Heinrich and Gullone (2006) suggested that loneliness is
Fig. 4. Cross-lagged associations between loneliness and GPIU. Note: Only significant associations are displayed. Coefficients represent standardized values. *Coefficient is significant
at the 0.05 level. **Coefficient is significant at the 0.01 level.
185
assigned a special importance in adolescence, which continues to
be regarded as a period when loneliness is particularly prevalent. In
particular, for freshmen who are shifting from a stage wherein they
M4 relatively heavily rely on their parents for support to a stage
417 wherein they generally rely on peer relationships and other social
125 relationships, the decline in communication with their parents may
0.05 cause them to suffer from more loneliness and reduced interper-
0.95
sonal interaction than they do during other periods of their
0.97
adolescence. However, the Internet significantly enhances their
model,
sense of identity, enables closer relationships with friends, and
increases their bonding with people worldwide, which makes them
more likely to use the Internet to develop feelings of belonging and
companionship (Chou, 2001). Gradually, they become more drawn
to their online activity, using it as a means to diminish or escape
from their struggles with loneliness. This further isolates them
from interpersonal interactions and thereby causes them to suffer
more loneliness. Furthermore, the more time they spend online,
the less time they engage in offline social interactions, and they
may miss out on opportunities to learn interpersonal skills. This
could eventually lead to them experiencing rejection and being
disliked by other people, thereby perpetuating their poor inter-
personal relationships.
However, the association between loneliness and GPIU was not
consistent with the findings of previous studies. Yao and Zhong
(2014) found that excessive and unhealthy Internet use can in-
crease feelings of loneliness over time. This finding may have been
caused by their adopting a different concept of Internet use. In our
study, Internet use was considered as a social phenomenon,
whereas Yao and Zhong focused on people's mental and emotional
states, which are adversely affected by the overuse of the Internet.
In addition, a different sample was used. University freshmen with
a mean age of 18.57 years were followed in our study, whereas the
mean age of Yao and Zhongs' sample was 21.63. The associations
between loneliness and GPIU are dynamic, and therefore, different
ages may yield different associations. Third, different analysis
methods were used. In the present study, the measurement
invariance of Internet addiction and loneliness was tested at first in
a latent variable model, which can ensure invariance in the mea-
surement of Internet addiction and loneliness across time. By
contrast, the observed Internet addiction and loneliness were
applied directly in Yao and Zhongs' study. This method cannot
ensure the invariance of the measurement of Internet addiction and
loneliness across time, and may thereby introduce an error in later
cross-lagged analysis.
5.3. Gender difference of GPIU
Gender differences were found in GPIU (T1 and T2). To further
investigate the gender differences, a series of independent t tests
were conducted on daily routines, social life, academic and occu-
pational productivity, sleep patterns, feelings, and online activities
186 Y. Tian et al. / Computers in Human Behavior 72 (2017) 178e188
(time spent on games, chatting, and shopping). The results reveal
that females spend more time on chatting, whereas males spend
more time on games; no gender difference were identified in other
fields. These results are consistent with previous research findings
(McAndrew & Jeong, 2012; Niculovi
c et al., 2014; Tan & Yang, 2014;
Tifferet & Vilnai-Yavetz, 2014). This because could be interpreted
according to gender stereotypes, in which typical male character-
istics are considered to consist of independence, autonomy and
self-sufficiency in contrary to typical female characteristics such as
interdependence, nurture, and concern for others (e.g., Chodorow,
1978; Cross & Markus, 1993). In this view, female interpersonal
dependency is higher than that of males (Sanathara, Gardner,
Prescott, & Kendler, 2003). In particular, university students shift
from a relatively greater reliance on parents for support and
interaction to a reliance on not only family relationships, but also
on peer relationships and social relationships; some freshman may
suffer from poor interpersonal relationships and loneliness. In this
situation, females are more likely to communicate online with
family and peers for social support, whereas males are more willing
to spend large amounts of time on games for entertainment
(Hupfer & Detlor, 2007; Niculovi
c et al., 2014).
6. Implications for prevention of GPIU
These findings have implications for the treatment and pre-
vention of GPIU in university freshmen. The results show a trans-
actional, dynamic set of associations among shyness, interpersonal
relationships and loneliness which combine to produce a feedback
loop that perpetuates GPIU. This finding suggests that interventions
(including those aimed at maladaptive cognitions) can break this
chain in various locations. Cognitive behavioral therapy indicates
that cognitive restructuring can be applied as therapeutic strategies
regarding Internet applications (Davis, 2001; Oikawa et al., 2012).
For example, Young and Case (2004) found that education and
training about the risks of excessive or compulsive Internet use can
alleviate many symptoms of GPIU. Additionally, interventions
should target initial interpersonal relationships and loneliness.
Previous studies have found that decreased social interaction is
related to poor interpersonal relationships and loneliness. Thus,
periodic collective counseling and collective activities have been
advocated, particularly for university freshmen experiencing a
decline in family communication and supervision. These activities
can assist students to engage in positive interactions such as
respecting, encouraging, and helping each other, thereby facili-
tating positive interpersonal relationships and reducing loneliness.
Additionally, shyness (0.61e0.66) and GPIU (0.66e0.69) had a
higher level of stability than interpersonal relationships
(0.43e0.46), loneliness (0.35e0.39) and maladaptive cognitions
(0.44e0.57) did. These results indicated shyness, loneliness, inter-
personal relationships, maladaptive cognitions, and GPIU were
somewhat stable at the beginning of university life, suggesting that
emerging individual differences persist over time (Guo, Sun,
Breitsmith, Morrison, & Connor, 2014). It is also noteworthy that
shyness and GPIU had a comparatively higher level of stability than
loneliness, interpersonal relationships, and maladaptive cognitions
did. A construct with a higher level of stability is considered to be
more “trait-like,” whereas a construct with a lower level of stability
is more “state-like” (Kenny & Zautra, 2001, p. 243). Therefore, the
treatment and prevention of GPIU should be carried out as early as
possible at the beginning of university life.
7. Limitations and future directions
Although the current study elucidated a mechanism for the
development of GPIU in China, it has a number of limitations that
should be addressed in future research. First, longitudinal studies
that span a longer study period, especially beginning earlier in
university adolescence, are needed in order to clarify long-term
stabilities, as well as developmental changes in the associations
among preexisting psychosocial variables (shyness, loneliness, or
interpersonal relationships), maladaptive cognitions, and GPIU.
Participants in the current study were followed over the course of
half a year. Longer studies with longer intervals may further
elucidate transactional development among preexisting psycho-
social variables (shyness, loneliness, or interpersonal relationships),
maladaptive cognitions, and GPIU.
Second, although self-reporting is useful in longitudinal
research, we cannot rule out the possibility that the results may be
influenced by the issue of shared-method variance because all the
study variables were obtained from self-reports. Future research
that incorporates data from other informants or sources, such as
teacher, parental, or peer reports are needed to verify the validity of
the present findings.
Finally, this study was conducted in a Chinese cultural context.
Although it extends our understanding and sheds light on cultural
differences, its generalizability must be further explored. For
example, a recent study found that Koreans are more likely to use
social networking sites to increase their social contacts but tended
to spend less time browsing contacts' profiles than Malaysians and
Chinese Internet users do (Lee, Kim, Golden, Kim, & Park, 2016).
Thus, different cultures may encourage different types of GPIU and
exhibit different GPIU formation mechanisms.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.chb.2017.02.048.
References
Bai, X., & Fan, F. M. (2005). A study on the internet dependence of college students:
The revising and applying of a measurement. Psychological Development and
Education, 4, 99e104. http://dx.doi.org/1001-4918(2005)04-0099-104.
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In
K. A. Bollen, & J. S. Long (Eds.), Testing structural equation models (pp. 136e162).
Newbury Park, CA: Sage. http://dx.doi.org/10.1177/0049124192021002005.
Brown, B. B., & Klute, C. (2003). Friendships, cliques, and crowds. In G. R. Adams, &
M. D. Berzonsky (Eds.), Blackwell handbook of adolescence (pp. 330e348).
Malden, MA: Blackwell Publishing.
Caplan, S. E. (2002). Problematic Internet use and psychosocial well-being: Devel-
opment of a theory-based cognitiveebehavioral measurement instrument.
Computers in Human Behavior, 18, 553e575. http://dx.doi.org/10.1016/S0747-
5632(02)00004-3.
Caplan, S. E. (2005). A social skill account of problematic internet use. Journal of
Communication, 55(4), 721e736. http://dx.doi.org/10.1111/j.1460-2466.2005.
tb03019.x.
Caplan, S. E., & Turner, J. S. (2007). Bringing theory to research on computer-
mediated comforting communication. Computers in Human Behavior, 23,
985e998. http://dx.doi.org/10.1016/j.chb.2005.08.003.
Casale, S., & Fioravanti, G. (2011). Psychosocial correlates of internet use among
italian students. International Journal of Psychology, 46(4), 288e298. http://dx.
doi.org/10.1080/00207594.2010.541256.
Cheek, J. M., & Buss, A. H. (1981). Shyness and sociability. Journal of Personality and
Social Psychology, 41, 330e339. http://dx.doi.org/10.1037/0022-3514.41.2.330.
Chen, S. K. (2012). Internet use and psychological well-being among college stu-
dents: A latent profile approach. Computers in Human Behavior, 28(6),
2219e2226. http://dx.doi.org/10.1016/j.chb.2012.06.029.
China Internet Network Information Center (CNNIC). (2016). Statistical report on the
development of internet in China (p. 8). http://www.cnnic.net.cn/hlwfzyj/
hlwxzbg/qsnbg/ (in Chinese).
Chodorow, N. J. (1978). The reproduction of mothering: Psychoanalysis and the soci-
ology of gender. Berkeley: University of California Press.
Chou, C. (2001). Internet heavy use and addiction among taiwanese college stu-
dents: An online interview study. Cyberpsychology & Behavior, 4(5), 573e585.
http://dx.doi.org/10.1089/109493101753235160.
Chou, C., & Hsiao, M. C. (2000). Internet addiction, usage, gratification, and pleasure
experience: The Taiwan college students' case. Computers & Education, 35,
65e80. http://dx.doi.org/10.1016/S0360-1315(00)00019-1.
Cole, D. A., & Maxwell, S. E. (2003). Testing meditational models with longitudinal
 Y. Tian et al. / Computers in Human Behavior 72 (2017) 178e188
data: Questions and tips in the use of structural equation modeling. Journal of
Abnormal Psychology, 112(4), 558e577. http://dx.doi.org/10.1037/0021-843X.
112.4.558.
Collins, W. A., & Steinberg, L. (2006). Adolescent development in interpersonal
context. In N. Eisenberg, W. Damon, & R. Lerner (Eds.) (6th ed.,Social, emotional
and personality developmentHandbook of child psychology:Vol. 3 (pp.
1003e1068). New York: Wiley http://dx.doi.org/10.1002/9780470147658.
chpsy0316.
Cotten, S. R. (2008). Students' technology use and the impacts on well-being. New
Directions for Student Services, 124, 55e70. http://dx.doi.org/10.1002/ss.295.
Cross, S. E., & Markus, H. R. (1993). Gender in thought, belief, and action: A cognitive
approach. In A. E. Beall, & R. J. Sternberg (Eds.), The psychology of gender (pp.
55e98). New York: Guilford Press.
Davis, R. A. (2001). A cognitiveeBehavioral model of pathological internet use.
Computers in Human Behavior, 17(2), 187e195. http://dx.doi.org/10.1016/S0747-
5632(00)00041-8.
Durkee, T., Kaess, M., Carli, V., Parzer, P., Wasserman, C., Floderus, B., et al. (2012).
Prevalence of pathological internet use among adolescents in europe: De-
mographic and social factors. Addiction, 107(12), 2210e2222. http://dx.doi.org/
10.1111/j.1360-0443.2012.03946.x.
Ehrensaft, M. K. (2005). Interpersonal relationships and sex differences in the
development of conduct problems. Clinical Child & Family Psychology Review,
8(1), 39e63. http://dx.doi.org/10.1007s00249-005-0021-y.
Engelberg, E., & Sjoberg, L. (2004). Internet use, social skills, and adjustment. Cyber
Psychology & Behavior, 7(1), 41e47. http://dx.doi.org/10.1089/
109493104322820101.
Eroglu, M., Pamuk, M., & Pamuk, K. (2013). Investigation of problematic internet
usage of university students with psychosocial levels at different levels. Pro-
cedia e Social and Behavioral Sciences, 103(26), 551e557. http://dx.doi.org/10.
1016/j.sbspro.2013.10.372.
Fang, X. Y., Wo, J. Z., & Lin, X. Y. (2005). Development of chinese college student
adjustment scale. Studies of Psychology and Behavior, 3(2), 95e101.
Fan, F., Su, L., Cao, F., Gao, X., Jin, Y., Xiao, H., et al. (2008). Development and psy-
chometric analysis of internet addiction predict test for adolescents. Chinese
Journal of Clinical Psychology, 16(1), 1e4 (in Chinese).
Greenfield, P., & Yan, Z. (2006). Children, adolescents, and the internet: A new field
of inquiry in developmental psychology. Developmental Psychology, 42,
391e394. http://dx.doi.org/10.1037/0012-1649.42.3.391.
Guo, Y., Sun, S., Breitsmith, A., Morrison, F. J., & Connor, M. D. (2014). Behavioral
engagement and reading achievement in elementary-school-age children: A
longitudinal cross-lagged analysis. Journal of Educational Psychology, 107(2),
332e347. http://dx.doi.org/10.1037/a0037638.
Heinrich, L. A., & Gullone, E. (2006). The clinical significance of loneliness: A liter-
ature review. Clinical Psychology Review, 26, 695e718. http://dx.doi.org/10.1016/
j.cpr.2006.04.002.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure
analysis: Conventional criteria versus new alternatives. Structural Equation, 6(1),
1e55. http://dx.doi.org/10.1080/10705519909540118.
Hupfer, M. E., & Detlor, B. (2007). Beyond gender differences: Self-concept orien-
tation and relationship-building applications on the internet. Journal of Business
Research, 60, 613e619. http://dx.doi.org/10.1016/j.jbusres.2006.06.010.
Kalkan, M. (2012). Productiveness of interpersonal cognitive distortions on uni-
versity students' problematic internet use. Children and Youth Services Review,
34(7), 1305e1308. http://dx.doi.org/10.1016/j.childyouth.2012.03.003.
Kenny, D. A., & Zautra, A. (2001). Trait-state models for longitudinal data. In
L. M. Collins, & A. G. Sayer (Eds.), New methods for the analysis of change (pp.
243e263). Washington, DC: American Psychological Association. http://dx.doi.
org/10.1037/10409-008.
Kraut, R., Kiesler, S., Boneva, B., Cummings, J., Helgeson, V., & Crawford, A. (2002).
Internet paradox revisited. Journal of Social Issues, 58, 49e74. http://dx.doi.org/
10.1111/1540-4560.00248.
Lee, S. L., Kim, J. A., Golden, K. J., Kim, J. H., & Park, S. A. (2016). A cross-cultural
examination of SNS usage intensity and managing interpersonal relationships
online: The role of culture and the autonomous-related self-construal. Frontiers
in Psychology, 7(376), 1e12. http://dx.doi.org/10.3389/fpsyg.2016.00376.
Lei, L., & Yang, Y. (2007). The development and validation of adolescent pathological
Internet use scale. Acta Psychologica Sinica, 39(4), 688e696 (in Chinese).
Liu, Q. X., Fang, X. Y., Deng, L. Y., & Zhang, J. T. (2012). Parenteadolescent commu-
nication, parental internet use and internet-specific norms and pathological
internet use among chinese adolescents. Computers in Human Behavior, 28(4),
1269e1275. http://dx.doi.org/10.1016/j.chb.2012.02.010.
Liu, M., & Peng, W. (2009). Cognitive and psychological predictors of the negative
outcomes associated with playing MMOGs (massively multiplayer online
games). Computer in Human Behavior, 25, 1306e1311. http://dx.doi.org/10.1016/
j.chb.2009.06.002.
Li, H., & Wang, S. (2013). The role of cognitive distortion in online game addiction
among Chinese adolescents. Children and Youth Services Review, 35(9),
1468e1475. http://dx.doi.org/10.1016/j.childyouth.2013.05.021.
Li, D., Zhang, W., Li, X., Zhen, S., & Wang, Y. (2010). Stressful life events and prob-
lematic internet use by adolescent females and males: A mediated moderation
model. Computers in Human Behavior, 26(5), 1199e1207. http://dx.doi.org/10.
1016/j.chb.2010.03.031.
Luo, H., & Peng, Y. (2008). Relationship of Internet addiction and family environ-
ment and attachment of the youth. Chinese Journal of Clinical Psychology, 16(3),
319e321 (in Chinese).
187
Lu, X., & Yeo, K. J. (2015). Pathological internet use among Malaysia university
students: Risk factors and the role of cognitive distortion. Computers in Human
Behavior, 45, 235e242. http://dx.doi.org/10.1016/j.chb.2014.12.021.
Mai, Y., Hu, J., Yan, Z., Zhen, S., Wang, S., & Zhang, W. (2012). Structure and function
of maladaptive cognitions in pathological internet use among Chinese adoles-
cents. Computers in Human Behavior, 28(6), 2376e2386. http://dx.doi.org/10.
1016/j.chb.2012.07.009.
McAndrew, F. T., & Jeong, H. S. (2012). Who does what on Facebook? Age, sex, and
relationship status as predictors of Facebook use. Computers in Human Behavior,
28(6), 2359e2365. http://dx.doi.org/10.1016/j.chb.2012.07.007.
Mccabe, R. E. (2015). Review of the mindfulness and acceptance workbook for social
anxiety and shyness: Using acceptance and commitment therapy to free
yourself from fear and reclaim your life. Canadian Psychology, 56(1), 152e153.
http://dx.doi.org/10.1037/a0038090.
Morahan-Martin, J., & Schumacher, P. (2000). Incidence and correlates of patho-
logical internet use among college students. Computers in Human Behavior,
16(1), 13e29. http://dx.doi.org/10.1016/S0747-5632(99)00049-7.
Niculovi

c, M., Zivkovi

c, D., Manasijevi

c, D., & Strbac, N. (2014). Study of pathological
internet use, behavior and attitudes among students population at technical
faculty bor, university of belgrade. Computers in Human Behavior, 39(39), 78e87.
http://dx.doi.org/10.1016/j.chb.2014.06.020.
Odaci, H., & Çikrikçi, O. € (2014). Problematic internet use in terms of gender,
attachment styles and subjective well-being in university students. Computers
in Human Behavior, 32(1), 61e66. http://dx.doi.org/10.1016/j.chb.2013.11.019.
Oikawa, H., Kamagata, K., Arai, M., Fukasawa, A., Yokota, H., Ide, T., et al. (2012).
Modelling excessive internet use:s revision of R. Davis's cognitive-behavioural
model of pathological internet use. Polish Psychological Bulletin, 42(3),
129e139. http://dx.doi.org/10.2478/v10059-011-0018-6.
Park, S. K., Kim, J. Y., & Cho, C. B. (2008). Prevalence of internet addiction and
correlations with family factors among South Korean adolescents. Adolescence,
172(43), 895e909.
Rogers, W. M., & Schmitt, N. (2004). Parameter recovery and model fit using
multidimensional composites: A comparison of four empirical parceling algo-
rithms. Multivariate Behavioral Research, 39(3), 379e412. http://dx.doi.org/10.
1207/S15327906MBR3903_1.
Rogosa, D. (1980). A critique of cross-lagged correlation. Psychological Bulletin, 88,
245e258. http://dx.doi.org/10.1037/0033-2909.88.2.245.
Russell, D., Peplau, L. A., & Ferguson, M. L. (1978). Developing a measure of lone-
liness. Journal of Personality Assessment, 42(3), 290e294. http://dx.doi.org/10.
1207/s15327752jpa4203_11.
Sanathara, V. A., Gardner, C. O., Prescott, C. A., & Kendler, K. S. (2003). Interpersonal
dependence and major depression: A etiological inter-relationship and gender
differences. Psychological Medicine, 33(5), 927e931. http://dx.doi.org/10.1017/
S0033291703007542.
Sanders, C. E., Field, T. M., Diego, M., & Kaplan, M. (2000). The relationship of
internet use to depression and social isolation among adolescents. Adolescence,
35(138), 237e242.
Satorra, A., & Bentler, P. M. (2001). A scaled difference chi-square test statistic for
moment structure analysis. Psychometrika, 66(4), 507e514. http://dx.doi.org/10.
1007/BF02296192.
Schuurman, N. K., Ferrer, E., De, B. M., & Hamaker, E. L. (2016). How to compare
cross-lagged associations in a multilevel autoregressive model. Psychological
Methods, 21(2), 206e221. http://dx.doi.org/10.1037/met0000062.
Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and nonexperimental
studies: New procedures and recommendations. Psychological Methods, 7(4),
422e445. http://dx.doi.org/10.1037//1082-989X.7.4.422.
Snyder, C. R., Smith, T. W., Augelli, R. W., & Ingram, R. E. (1985). On the self-serving
function of social anxiety: Shyness as a self-handicapping strategy. Journal of
Personality & Social Psychology, 48(4), 970e800 http://dx.doi.org/10.1037/0022-
3514.48.4.970.
Solomon, Z., Bensimon, M., Greene, T., Horesh, D., & Eindor, T. (2015). Loneliness
trajectories: The role of posttraumatic symptoms and social support. Journal of
Loss & Trauma, 20(20), 1e21. http://dx.doi.org/10.1080/15325024.2013.815055.
Talbot, L., Stone, S., Gruber, J., Hairston, I., Eidelman, P., & Harvey, A. (2012). A test of
the bidirectional association between sleep and mood in bipolar disorder and
insomnia. Journal of Abnormal Psychology, 121(1), 39e50. http://dx.doi.org/10.
1037/a0024946.
Tan, W. K., & Yang, C. Y. (2014). Internet applications use and personality. Telematics
and Informatics, 31(1), 27e38. http://dx.doi.org/10.1016/j.tele.2013.02.006.
Tian, Y., Bian, Y. L., Han, P. G., Wang, P., & Gao, F. Q. (2015). The effect of shyness on
Internet addiction: The mediating effects of immersion tendency and network-
related maladaptive cognition. Chinese Journal of Special Education, 12, 83e89.
Tifferet, S., & Vilnai-Yavetz, I. (2014). Gender differences in Facebook self-
presentation: An international randomized study. Computers in Human
Behavior, 35(6), 388e399. http://dx.doi.org/10.1016/j.chb.2014.03.016.
Turkle, S. (2011). Alone together: Why we expect more from technology and less from
each other. New York: Basic Books.
Van den Eijnden, R. J. J. M., Spijkerman, R., Vermulst, A. A., Van Rooij, T. J., &
Engles, R. C. M. E. (2010). Compulsive internet use among adolescents: Bidi-
rectional parentechild relationships. Journal of Abnormal Child Psychology, 38,
77e89. http://dx.doi.org/10.1007/s10802-009-9347-8.
Wang, D. F. (1995). A study on the reliability and validity of the Russel Loneliness
Scale. Chinese Journal of Clinical Psychology, 3(1), 23e25. http://dx.doi.org/
10.16128/j.cnki.1005-3611.1995.01.006.
Wang, Q. Q., Wang, P., Han, L., Gong, R. Y., & Gao, F. Q. (2009). The revision of college
188 Y. Tian et al. / Computers in Human Behavior 72 (2017) 178e188
students' shyness scale and its relevant study. Psychological Science, 32(1),
204e206. http://dx.doi.org/10.1679/j.cnki.1671-6981.2009.01.048.
Wentzel, K. R., & Mcnamara, C. C. (1999). Interpersonal relationships, emotional
distress, and prosocial behavior in middle school. Journal of Early Adolescence,
19(1), 114e125. http://dx.doi.org/10.1177/0272431699019001006.
Woodhouse, S. S., Dykas, M. J., & Jude, C. (2012). Loneliness and peer relations in
adolescence. Review of Social Development, 21(2), 273e293. http://dx.doi.org/10.
1111/j.1467-9507.2011.00611.x.
Yao, M. Z., & Zhong, Z. (2014). Loneliness, social contacts and internet addiction: A
cross-lagged panel study. Computers in Human Behavior, 30(30), 164e170.
http://dx.doi.org/10.1016/j.chb.2013.08.007.
Young, K. S. (1998). Internet addiction: The emergence of a new clinical disorder.
Cyberpsychology & Behavior, 1(3), 237e244. http://dx.doi.org/10.1089/cpb.1998.
1.237.
Young, K. S., & Case, C. J. (2004). Internet abuse in the workplace: New trends in risk
management. Cyberpsychology and Behavior, 7(1), 105e111. http://dx.doi.org/10.
1089/109493104322820174.
Zhang, B., Gao, Q., Fokkema, M., Alterman, V., & Liu, Q. (2015). Adolescent inter-
personal relationships, social support and loneliness in high schools: Mediation
effect and gender differences. Social Science Research, 53, 104e117. http://dx.doi.
org/10.1016/j.ssresearch.2015.05.003.