This article appeared in a journal published by Elsevier.

The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
 Author's personal copy

Computers in Human Behavior 25 (2009) 490–500

Contents lists available at ScienceDirect

Computers in Human Behavior

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

Toward a comprehensive theory of problematic Internet use: Evaluating the role

of self-esteem, anxiety, flow, and the self-rated importance of Internet activities
Ho-Kyung Kim *, Keith E. Davis
School of Journalism and Mass Communications, University of South Carolina, Columbia, SC 29208, USA

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

Article history: Problematic Internet use (PIU) has long been a matter of public concern. Study 1 examined the correlates
Available online 4 December 2008 of a well-validated measure of Internet addiction [Young, K. S. (1998). Caught in the net. New York: John
Wiley and Sons] with 315 students (39% male). Hierarchical regression analyses indicated that low self-
Keywords: esteem, unrealistic optimism, and the importance of 7-positive activities (I7PA) all contributed to the pre-
Anxiety diction of PIU. The inclusion of controls for aspects of Internet uses and demographic characters showed a
Flow clear distinction between PIU and frequent Internet use. Subjects who anticipated that negative outcomes
Problematic Internet use
were more likely to happen to them than to peers were none-the-less more likely to engage in PIU. The
Self-esteem
Sensation-seeking
more important positive activities were, the greater the PIU. SEM model suggested that unrealistic opti-
Theory mism was best treated as part of PIU and that low self-esteem and the I7PA were the powerful predictors
of PIU. Study 2 tested SEM models of psychosocial characteristics with 279 students (48% male). Anxiety
and flow were added as predictors and the felt safety for social contacts was added to Global PIU follow-
ing Caplan [Caplan, S. E. (2002). Problematic Internet use and psychosocial well-being: Development of a
theory-based cognitive-behavioral measurement instrument. Computers in Human Behavior, 18, 553–
575]. The more complex model provided excellent fit and showed that low self-esteem, anxiety, and
the I7PA were positively related to PIU. The relationship both of flow and sensation-seeking was medi-
ated by the 17PA. Research on the topic has been relatively atheoretical. We propose a theoretical tem-
plate to guide future research on PIU.
Published by Elsevier Ltd.

1. Introduction home life or work life begin to suffer; (c) the pleasure of participa-
tion has an exaggerated value; and (d) attempts to reduce or re-
This present study critically examines the concept of problem- strict the activity tend to fail (relapse). These four characteristics
atic Internet use (PIU) and develops a theory of personal character- are part of the definition of an ‘‘addiction” but they omit one key
istics likely to contribute to PIU by young adults. In doing so, this element that cannot be assessed by purely self-report instruments.
study builds on Young’s (1998) ground breaking work on Internet That is the phenomena of a growing tolerance on the respective
addiction and the extensions made by Caplan (2003, 2007) who substances (drugs) so that the addict needs more and more of
conceptualized the phenomena as PIU. Beyond these contributions, the drug to achieve the desired effects (Samet, Waxman, Hatzen-
we identify a theoretical template from Descriptive Psychology buehler, & Hasin, 2007). Griffiths (2001) has argued that one can
(Ossorio, 2006) likely to be useful in development of a comprehen- index tolerance from self-reports of an increasing amount of the
sive model of PIU. activity as being required for one’s satisfaction. Caplan, correctly
Although Young and Caplan use different words for their mea- in our view, restricts the concept being measured to that of ‘‘prob-
sures, we believe that our research will show that their measures lematic Internet use.” There is no serious question that a behavior
are interchangeable but that Caplan has the more nuance concep- such as Internet use is problematic if the activities become com-
tualization of the phenomena. Young (1998, 2004) has developed a pulsive, if they interfere with normal activities of daily living,
measure that taps these characteristics: (a) the activity taking on and if the person can no longer control his/her own use. Second,
an obsessive/compulsive character such that one has trouble plac- it is the activities that one can access through the Internet that
ing appropriate limits on it; (b) participation in the activity tends can become problematic – not the Internet per se. Problematic
to interfere with other important social practices such that one’s behaviors of all sorts are indicative of problems in self-regulation
(Baumeister, Heatherton, & Tice, 1994; Gaillot & Baumeister,
* Corresponding author. Tel.: +1 803 673 8417; fax: +1 803 777 4103. 2007). Potentially, many intrinsically rewarding activities can be-
E-mail address: kim65@mailbox.sc.edu (H.-K. Kim). come problematic in the sense above, but research has shown

0747-5632/$ - see front matter Published by Elsevier Ltd.

doi:10.1016/j.chb.2008.11.001
 Author's personal copy

H.-K. Kim, K.E. Davis / Computers in Human Behavior 25 (2009) 490–500 491

some intrinsically rewarding activities have a much greater poten- 1.2. Current research on Internet usage
tial for the development of problematic characteristics – among
which are games of chance, vicarious violence, pornography, and Recent research about Internet uses shows that among approxi-
the associated masturbation, vicarious access to the lives of the mately 301 million Americans, about 215 million people (71.4%)
rich and famous (Griffiths, 2001). uses the Internet, and the growth percentage from 2000 to 2007 is
Baumeister’s ‘‘energy” or psychological resource model of self- 125.6 (Internet World Stats, 2007). A nationwide study by the Med-
regulation posits that any environmental demands that require icine School (N = 2513) found 13.7% of respondents having difficulty
the use of self-control have the potential to deplete a person’s re- to stay offline for short periods and 12.4% staying online longer than
sources for self-control. In effect, each of us has only so much planned (Aboujaoude, Koran, Gamel, Large, & Serpe, 2006). These
capacity for self-control and when it is stressed by demanding cir- survey figures suggest that between 10% and 15% of Internet users
cumstances, all of us are prone to lapses in self-control. Recently a experience some of the problems associated with PIU.
body of research has related diminished self-control to diminished
blood glucose levels, providing a partial physiological mechanism 1.3. A psychosocial model of problematic Internet use: Conceptual
for the observed behavior effects (Gaillot & Baumeister, 2007). template
How is this general model of self-control deficits relevant to a
study of individual differences that may underlie PIU? 1.3.1. The intentional action paradigm as a template for the analysis of
Some chronic conditions such as social rejection and the problematic behaviors
resulting low self-esteem appear to limit self-control resources Ossorio (2006) has argued and demonstrated that underlying
and/or to reduce the weight placed on them (Finkenauer, Engels, folk psychology is a universal paradigm of persons as agents of
& Baumeister, 2005; Tangney, Baumeister, & Boone, 2004). Recent deliberate and intentional action. He has shown that a parametric
longitudinal research has shown that lower self-esteem prospec- analysis with eight parameters is sufficient to represent any and all
tively predicts health problems two months later and that much cases of the behavior of persons. These include the Individual per-
of the effect is mediated by poor quality social bonds associated son whose behavior it is; the Cognitive parameter, namely the dis-
with low self-esteem (Stinson et al., 2008). Because of the strong tinctions about the world and the self that the behavior require;
conceptual basis for seeing self-esteem as related to the broader the Motivational parameter, namely the state of affairs wanted, de-
psychological concept of self-regulation and the evidence that sired, or that one has reason enough to try to get; the Skill param-
self-regulation is intrinsically involved in addictions, we prefer eter, namely the know how and ability that one has to engage in
to focus on self-esteem rather than upon variables such as lone- the relevant behavior; the Performance parameter, namely the
liness selected by some other researchers. Loneliness is undoubt- overt actions required to produce the behavior, and the Outcome
edly an empirical correlated of PIU but not we think as central to or Achievement parameter, namely the changed state of affairs cre-
its development as low self-esteem (Caplan, 2007; Douglas et al., ated by the behavior. Two additional parameters reflect (1) the
2008). information that the behavior reflects about the Personal Character-
Another characteristic that has the potential to be relevant to istics revealed behavior such as his/her abilities, values, disposi-
PIU is sensation-seeking – which is a personal trait associated with tions, and conditions and (2) about the Significance of the
lower value on self-control vs. a higher value on thrills and imme- behavior for the further social practices that the person is engaging
diate gratification (Slater, 2003). Problems in self-regulation are in. We have drawn on this template to identify gaps in current the-
also related with a perceptual tendency of optimistic bias effect orizing and to identify promising variables for consideration.
in which individuals are more likely to discount risks for them- Research on heavy Internet users contains an implicit contradiction:
selves and overestimate vulnerabilities for others (Weinstein, On one hand, heavy users are reported to be socially isolated and to
1980). This study looks at the relationships between PIU and these have few social skills (Brenner, 1997; Engelberg & Sjöberg, 2004). On
personal characteristics and gives a more detailed rationale for the other hand, as an interactive medium, the Internet is a means to
including these characteristics in a comprehensive model of Global social communication with friends, family, source of information,
PIU below. and potential new associates. Thus it could fulfill needs for social rec-
ognition, social support, and a sense of belonging not currently being
1.1. Evidence of problematic Internet use met in a person’s immediate social world (Beard, 2005; Caplan,
2003; Scherer, 1997). Following Caplan(2003, 2007), we want to sug-
Young (1998) first empirically analyzed problematic Internet gest that it is important to distinguish the types of functions being
behaviors in 1996, and she examined whether Internet use can fulfilled by Internet uses if we are to get a fuller understanding of
be addictive. Recently Caplan (2003, 2007) has developed an inter- PIU. In the case of persons who use the net a lot but primarily to
esting psychosocial model for the study of PIU and devised his own maintain contact with family and friends, they may be at a much
instrument, the Generalized Problematic Internet Use Scale (GPI- lower risk of PIU because the net is merely another tool for maintain-
US; Caplan, 2002). This present study may be seen as an extension ing their important social ties. In contrast, for individuals with much
of Caplan’s psychosocial model of PIU, but with some distinctive smaller social networks and some feelings of alienation, rejection, or
additions that seem responsive to his suggestions in recent papers social isolation, the Internet may then have a much greater potential
(Caplan & High, 2007) and those of La Rose, Lin, and Eastin (2003). for problematic patterns of use.
The conceptions of Young (1998, 2004) and Caplan (2002, 2003,
2007) all embody the key conceptual features of problems in self- 1.3.2. Importance of positive Internet functions
regulation – that it seemed safe to treat the measures as equivalent These functions reflect specific motives for engaging in Internet
– an assumption that we test empirical in a subsample of Study 1 use and thus can reveal the values of the participants. We have devel-
and in Study 2. oped our own measure of 10 potential positive functions that could
A second way in which this study differs from those studying be satisfied by Internet use and asked participants to rate the impor-
problematic users is that we are interested in the entire range of tance of each to them. We further anticipated that not all positive
behaviors that may occur, from no PIU at all to very disturbing functions would be related to PIU. Specifically we anticipated that
PIU. And because there is no normative data with cut-offs to indi- the importance of its use to maintain contact with family and friends,
cate a specific score for being labeled as PI users, we examine the and to research information for work or school would not be predic-
entire range of PIU scores. tive of PIU. But the remaining 7-positive activities, such as joining
 Author's personal copy

492 H.-K. Kim, K.E. Davis / Computers in Human Behavior 25 (2009) 490–500

the same interest groups, playing online games, shopping/auction- tion-seeking scores than those who were not dependent. Slater
ing, downloading music, and/or movies, meeting new people, (2003) questioned whether sensation-seeking could be a predictor
becoming famous, and flirting, would have problematic potential. to explain adolescent use of violent Internet messages. He found
There is evidence in the literature (Young, 1998) that each of these that high sensation-seekers were inclined to be alienated from
can become obsessive activities for some persons. This expectation peers, family, and school, and those younger students were more
was tested and supported the creation of a single measure of these likely to visit Websites showing aggressive contents. Thus, while
7, which we label the I7PA, for importance of these 7-positive we considered that PIU could be positively correlated with sensa-
activities. tion-seeking, the mixed findings to date suggest that the relation-
ship is more complex. We will explore correlates of sensation-
1.3.3. Likelihood of negative outcomes seeking to see if they mediate the relationship between sensa-
Weinstein (1980) explained the optimistic bias effect in the fol- tion-seeking and PIU.
lowing contexts; (1) a tendency to think that negative events are
unlikely to occur to them, (2) a tendency to think that positive fu- 1.5. Control variables
ture events are more likely to happen to them compared to their
peers, and (3) a tendency to view others as more likely to lose con- It is especially important to distinguish between problematic
trol and cause accidents than for themselves to do so. use and merely frequent or heavy Internet use. One may use it
Following Campbell and his colleagues (2007) and Weinstein quite heavily for work, for family financial activities, for essential
(1980) examined the unrealistic optimism effect for positive and communication with others, for important information search.
negative consequences from Internet use. As expected, in this Widyanto and Griffiths (2006) emphasized that Internet use
study participants perceived that they were more likely than their should be interpreted in various contexts, but earlier research
peers to experience positive online events (e.g., finding a sought tended to overlook routine Internet use. We think the concept of
item, winning a prize or contest) and less likely to experience neg- PIU requires that the use be compulsive, that it interfere with other
ative events (e.g., auction fraud, e-mail being read). More specifi- important aspects of one’s life (work and family) and that attempts
cally, participants with more Internet use experiences appeared to control it often fail. Thus we ask several questions about Internet
significantly more optimistic than did those with less experience. use so that we could distinguish routine Internet use from PIU. We
PI users would be more concerned about their uncontrollable anticipated that heavy use could be distinguished empirically from
Internet use and results, but they also hope that negative things PIU by means of hierarchical regression techniques, which would
do not happen to their lives. Thus we expect that optimistic per- allow us to distinguish the variance in PIU due to frequent use from
ceptual tendency about possible negative outcomes would be neg- that associated with other variables.
atively related to PIU. Previous studies on PIU have identified three demographic vari-
ables as relevant: (1) age (younger users (<25) tend to be PI users
(e.g., Brenner (1997); (2) gender (men are more likely to be PI
1.4. Personal characteristic predictive of problematic Internet use
users) (i.e., it is consistent with findings by Kubey et al. (2001),
Morahan-Martin and Schumacher (2000), and Scherer (1997), but
1.4.1. Self-esteem
inconsistent with research by Brenner (1997); and (3) year in col-
Because low self-esteem is conceptually and empirically a func-
lege (e.g., Kubey et al. (2001) showed first year college students
tion of perceived rejection, abandonment, or indifference by signif-
tend to be PI users). We collected this information and other demo-
icant others (Leary & MacDonald, 2003; MacDonald & Leary, 2005),
graphic variables so that we could control these influences by
the degree of self-esteem should be a predictor of PIU. Earlier stud-
using residualized PIU scores.
ies showed a link between PIU and levels of self-esteem (Arm-
The literature review thus far suggests the following three
strong, Phillips, & Saling, 2000; Ko, Yen, Chen, Chen, & Yen, 2007;
hypotheses which we propose to test in this study and raise one
Young & Rogers, 1998). Thus, we use low self-esteem as a funda-
interesting question which we will pursue.
mental characteristic that we anticipate to be predictive of suscep-
tibility to PIU. Although low self-esteem is conceptually distinct
from social anxiety, depression, or loneliness, evidence exists for 1.6. Summary of predictions
moderate correlations among these variables (Leary & MacDonald,
2003). In this study we give a priority to self-esteem as a poten- Hypothesis 1a. The greater the importance of 7-positive activities,
tially causal variable because self-esteem is implicated in lapses the greater the likelihood of the person having a pattern of
of self-control and we see the problematic aspects of heavy Inter- problematic Internet use.
net use as directly expressive of such lapses (La Rose et al., 2003;
Stinson et al., 2008; Tangney et al., 2004). Supportive data can be Hypothesis 1b. The greater the importance of maintaining pre-
found in the studies of lonely and depressed individuals who are existing social relationship with one’s family and friends, the lower
more likely to engage in synchronous functions of Internet (e.g., the tendency toward problematic Internet use.
chatting, instant messengers, etc.) (Caplan, 2003; Kubey, Lavin, &
Barrows, 2001; Young & Rogers, 1998). Hypothesis 1c. The greater importance of researching information
for school or work, the lower the tendency of problematic Internet
use.
1.4.2. Sensation-seeking
Findings of previous studies examined the link between PIU
Hypothesis 2. The greater likelihood of negative outcomes, the
and sensation-seeking are not consistent. High sensation-seekers
lower the rate of problematic Internet uses.
were not heavier Internet users and impulsive risk takers (Arm-
strong et al., 2000). Lavin and his colleagues (1999) found that
Hypothesis 3. The higher the person’s self-esteem, the lower his/
PI users were more likely to stay on the net, use synchronous
her problematic Internet use.
functions like multi-playing games or chatting, but their levels
of sensation-seeking was lower than those of non-problematic Research question 1: Is sensation-seeking positively associated
users. In contrast, Lin and Tsai (2002) showed that Internet with problematic Internet use or is its connection to such use more
dependent adolescents had higher disinhibition and overall sensa- indirect and mediated by other factors?
 Author's personal copy

H.-K. Kim, K.E. Davis / Computers in Human Behavior 25 (2009) 490–500 493

2. Study 1: Method the redundant items and rewording one item, 10 items, such as
shopping/auctioning fraud, someone hacking email and/or files,
2.1. Participants getting a virus/worm, being stalked by an individual met online,
receiving excessive junk emails (‘‘spam”), becoming addicted to
Participants (N = 315) were enrolled in a variety of undergradu- Internet use, Identity theft, online harassment (‘‘flaming”), having
ate programs, with the majority in Journalism and Mass Communi- username/password stolen, employer secretly monitoring individ-
cations (27%), followed by History (16%), Sociology (16%), Sport ual online use, were selected to measure the negative outcomes
and Entertainment Management (13%), Economics (11%), Educa- variable. Responses were scored on a 7-point scale (e.g., 1 = less
tion (9%), and Psychology (8%) at a large southeastern university. likely to happen to me, 4 = average (same amount), 7 = more likely
Prior to conducting the survey, participants were informed that to happen to me), with the higher scores showing a greater likeli-
the purpose of this study was to learn more about individual Inter- hood of negative outcomes to happen to the participants. The mean
net use, in specific, social networking Web sites like MySpace or was 2.91 (SD = 0.58) and Cronbach’s a was .86.
Facebook, and outcomes. Their informed consent was solicited
and questionnaire packets were completed by participants in class- 2.2.1.3. Self-esteem. Participants’ self-esteem was assessed using
room settings. The procedures were approved by the University the 10-item Rosenberg Self-Esteem Scale (1965). Each item was an-
Institutional Review Board (IRB) and participation was voluntary. swered on a 5-point Likert-type scale with higher scores indicating
Among the 315 college students, 192 (61%) were female and higher self-esteem. Sample items included, ‘‘I feel that I have a
123 (39%) were male, with a mean age of 22.3 years (SD = 5.75 number of good qualities” and ‘‘I am able to do things as well as
years). The majority (90.2%) of the participants fell in the 18–25 most other people.” Average self-esteem was 4.25 (SD = 0.58),
age range. The sample included 2.9% freshmen, 11.1% sophomores, and the reliability was .86.
24.1% juniors, 52.4% seniors, and 9.5% others. The sample was
mainly Caucasian (72.4%) and Windows users (86.3%) as their com- 2.2.1.4. Sensation-seeking (SS). Participants completed the 11-item
puter operating system. sensation-seeking scale, used in Kalichman and Rompa’s Study
(1995), and one more item, suggested by Slater (2003) that showed
reliability and validity to predict risky behaviors. Sample items
2.2. Measures
were: (1) I get bored and restless if I have nothing to do; (2) I
can see myself seeking pleasures around the world with ‘‘exciting”
2.2.1. Predictor variables
people. The response format was from 1 to 5 (where 1 = not at all
2.2.1.1. Importance of positive activities. Participants were asked to
like me, 3 = neutral (mixed), 5 = very much like me). The mean was
rate the importance of the positive functions for using social
3.01 (SD = 0.83), and the scale was internally consistent (a = .85).
networking sites compared to other students (of the same gen-
der, at the same college). The current item pool was constructed
2.2.2. Dependent variable
first by asking a focus group of students to identify positive
2.2.2.1. Problematic Internet use-1 (PIU-1). In the US, there are two
aspects of the Internet. Next the 18 activities and events elicited
well-established measures – Young’s Internet Addiction Test
were reviewed, and only those with the 10 endorsements with
(IAT) (1998) and Caplan’s Generalized Problematic Internet Use
higher mean scores were retained. All items ranged from 1 to
Scale (GPIUS) (2002) – that could be used to assess the behaviors
7 (1 = less important to me, 4 = about the same amount, 7 = more
indicative of PIU.
important to me). The higher scores indicated that positive activ-
We tested Young’s IAT with a small number of students. The
ities were more important to the participants than to other
analysis of the item pool suggested that we could eliminate four
students.
items from the IAT because of their redundancy with other items
Because we anticipated that the importance of Internet use to
in the pool and not lose any significant internal consistency or con-
maintain contact with family and friends, and to research informa-
tent coverage. Responses ranged from 1 (strongly disagree) to 5
tion for work or school would not be predictive of PIU, these were
(strongly agree), and the average was 1.90 (SD = 0.66, a = .92).
retain for separate analysis from the seven items that we expected
We also examined the relationship between Young’s IAT (1998)
to hold the potential for problematic use. These 7-positive func-
and Caplan’s GPIUS (2002) in a separate subsample of 70 under-
tions were joining the same interest groups, playing online games,
graduates (39 male and 31 female) enrolled Introductory Psychol-
shopping/auctioning, downloading music, and/or movies, meeting
ogy 101 course. In order to avoid order effects, we combined the
new people, trying to become famous, and flirting and/or locating a
items alternating items from each measure to make them appear
romantic partner (boyfriend, girlfriend). Using the full sample, we
to be one scale. The results are presented below.
did a linear trend ANOVA for each item in relation to the total PIU
score to verify that all seven items showed the expected linear
3. Study 1: Results
trend. They did; the greater the importance of the activity to indi-
viduals, compared to others of the same gender/age, the higher
3.1. Analysis of Young’s IAT (PIU-1) and Caplan’s GPIUS (PIU-2)
their PIU scores. Because the trends were so consistent, these seven
items were combined into a new importance of 7-positive activi-
A total, combined 30 items with Young’s 16 IAT items (PIU-1)
ties measure (I7PA). The average I7PA was 3.14 (SD = 1.16) and
and Caplan’s 14 GPIUS items (PIU-2) that parallel the IAT were
Cronbach’s a was .74. Additionally, the mean score of importance
tested to clarify the relationship of these two scales. Among the to-
of maintaining pre-existing relationships (IMPR) was 5.52
tal 29 items of Caplan’s measure, we thought only 14 items that di-
(SD = 1.32, a = .40), and the average importance of researching
rectly assessed the behavioral manifestations of PIU. Table 1
was 4.36 (SD = 1.75).
contains the exact wording, item means, SDs for all 16 IAT items
and 14 GPIUS items used, with active and declarative sentences.
2.2.1.2. Likelihood of negative outcomes (LNO). Participants were The mean PIU-1 was 2.04 (SD = 0.81, a = .94) and the average
asked to compare the likelihood of each negative event happening PIU-2 was 2.08 (SD = 0.82, a = .93).
to them vs. other students (of the same gender, at the same col- The zero-order Pearson correlation analysis showed that PIU-1
lege) as a result of using social networking sites. A small group of and PIU-2 were getting at the same underlying dimension
students generated 17 potentially negative events. After removing (r = .93, p < .001) and had the same relationships with other predic-
 Author's personal copy

494 H.-K. Kim, K.E. Davis / Computers in Human Behavior 25 (2009) 490–500

Table 1
Descriptive statistics of PIU-1 items of Young’s IAT (1998) and PIU-2 Items of Caplan’s GPIUS (2002) assessed in Study 1.

PIU-1 Items M SD PIU-2 items M SD

(1) I usually stay online longer than originally intended 3.96 0.94 (1) I have been unsuccessful attempts to control Internet 1.70 1.08
use
(2) I often neglect household chores to spend more time online 2.29 1.32 (2) I have lost track of time online 3.03 1.36
(3) I prefer the excitement of the Internet to offline relationships 1.51 0.96 (3) I have gone online for longer than I intended 3.31 1.23
(4) I frequently form new relationships with fellow online users 1.81 1.05 (4) I spend more time online than I expect to 3.00 1.34
(5) Others in my life complain about the amount of time I spend online 1.89 1.06 (5) I miss being online if I can’t go on it 2.13 1.35
(6) My grades and/or school work suffers because of the amount of time I spend online 1.91 1.13 (6) I have gotten into trouble at work/school because of my 1.63 1.04
online activity
(7) I prefer to check e-mail before doing something else that needs to be done 2.84 1.36 (7) I am preoccupied with the Internet if I can not log on 1.91 1.13
for some period of time
(8) My job performance and/or productivity suffers because of the Internet 1.73 1.08 (8) It is hard to stop thinking about what is waiting for 1.61 0.97
me online
(9) I become defensive and/or secretive when anyone asks me what I do online 1.67 0.96 (9) I have missed classes or work because of online 1.64 1.14
activities
(10) I often daydream about things I do online 1.53 1.02 (10) I have been guilty about time online 1.89 1.16
(11) I’m constantly anticipating when I go online again 1.73 1.08 (11) I feel lost if I can’t go online 1.61 1.01
(12) Life without the Internet would be boring, empty, and/or joyless 2.14 1.32 (12) When not online, I wonder what is happening online 1.78 1.19
(13) I tend to snap, yell, and/or act annoyed if someone bothers me while I am online 1.66 1.05 (13) I have tried to stop using Internet for long period of 1.74 1.11
time
(14) I frequently lost sleep due to late-night Internet and/or email use 2.11 1.45 (14) I have been unable to reduce time online 2.14 1.12
(15) I often find myself saying, ‘‘just a few more minutes” when online 2.23 1.34
(16) I’ve tried and failed to cut down the amount of time spent online 1.60 1.01

Note: N = 70 college students. PIU-1 = problematic Internet use-1 measured by Young’s Internet Addiction Test (IAT) (1998); PIU-2 = problematic Internet use-2 measured by
Caplan’s Generalized Problematic Internet Use Scale (GPIUS) (2002). Response scale: 1 = strongly disagree; 5 = strongly agree.

tor variables. PIU-1 and PIU-2 were negatively correlated with self- used them to stay informed (e.g., downloading, searching informa-
esteem (r = .35, p < .01; r = .27, p < .05, respectively), whereas tion, etc.).
these measurements were positively correlated with the amount
of time spent online (r = .49, p < .001; r = .51, p < .001, respectively), 3.3. Gender differences in Internet uses
the longest time spent online (r = .30, p < .01; r = .36, p < .01,
respectively), the time on social networking sites (r = .41, Study 1 showed that the values of Internet functions were dif-
p < .001; r = .39, p < .001, respectively), LNO (r = .57, p < .001; ferent to men and to women. Women (M = 5.31, SD = 1.90) were
r = .51, p < .001, respectively), and I7PA (r = .35, p < .001; r = .35, more likely to maintain their real-life ties with family than men
p < .001, respectively). (M = 4.81, SD = 2.02), F(1, 313) = 4.94, p = .03, g2 = .02. On the other
Next, we performed the multiple regression with only demo- hand, Internet uses for playing games, downloading music or mov-
graphic and general Internet uses variables and saved each residual ies, becoming famous, and flirting or locating a romantic partner
score of PIU-1 and of PIU-2, then run the regression with six pre- were more important to men, F(1, 313) = 21.84, p < .001, g2 = .07;
dictor variables (e.g., self-esteem, SS, IMPR, research, I7PA, and F(1, 313) = 18.16, p < .001, g2 = .06; F(1, 313) = 23.98, p < .001,
LNO), because of a small number of participants (N = 70). The pre- g2 = .07; F(1, 313) = 25.43, p < .001, g2 = .08, respectively. Interest-
dictive powers of each measurement, PIU-1 and PIU-2 were similar ingly, men (M = 3.45, SD = 2.01) were more concerned about
(R2 = 33.4, p < .001, R2 = 25.7, p < .001, respectively), and only LNO becoming addicted to Internet use than women (M = 2.70, SD = 1.
contributes positively to predict PIU-1 (b = .45, p < .001) and PIU- 69), F(1, 313) = 12.70, p < .001, g2 = .04. Men’s PIU scores
2 (b = .40, p < .01). (M = 2.06, SD = .74) were higher than women’s scores (M = 1.80,
SD = .58), F(1, 313) = 11.49, p < .001, g2 = .04.
3.2. Descriptive findings
3.4. Hypotheses testing
Among the 315 college students, an overwhelming percentage
91.4% (N = 288) had used the Internet more than 5 years. 40% Before testing hypotheses, a correlation analysis was checked to
(N = 126) of the students spend more than 2 h per day on the Inter- ensure that PIU and other variables, in fact, significantly correlated
net, and the majority (82.5%, N = 260) had experiences of spending with one another (Table 2). The zero-order Pearson correlation
more than 3 h in one day on the net. analysis showed that PIU was negatively correlated with gender
Two hundred and eighty five participants (90.5%) use social net- (r = .19, p < .001), academic year (r = .22, p < .001), and self-es-
working sites like MySpace or Facebook and 289 participants teem (r = .30, p < .001). On the other hand, it was positively cor-
(91.7%) have experiences of creating their own profile on these related with the amount of time spent online (r = .37, p < .001),
sites more than once. 72.1% (N = 227) of the participants used Face- the longest time spent online (r = .21, p < .001), hours spent on so-
book most often while 16.5% (N = 52) of the participants used My- cial networking sites (r = .32, p < .001), visiting times to networking
Space. The majority (68.9%, N = 217) spent less than 1 h per day on sites (r = .24, p < .05), SS (r = .18, p < .01), I7PA (r = .37, p < .001), and
these sites, and 48.6% (N = 153) of the surveys said they visited LNO (r = .30, p < .001).
these sites 1–5 times daily. When the participants were asked to The hierarchical multiple regression provided a test of relation-
select one, primary reason for using social interacting sites, among ships between PIU and predictors (e.g., I7PA, IMPR, research, LNO,
9 specific reasons, 70.8% (N = 223) said that they used these sites to self-esteem, and SS), after considering the contribution of estab-
keep in touch with family and friends. 14.6 (N = 46) used them to lished demographic and general Internet uses. These relationships,
kill time while a small percentage of the students (3.8%, N = 12) along with the control variables, total to explain 39.9% of variance
in PIU.
 Author's personal copy

H.-K. Kim, K.E. Davis / Computers in Human Behavior 25 (2009) 490–500 495

Table 2
Means and standard deviations of, and correlations among, variables assessed in Study 1 (zero-order Pearson correlation coefficients).

Variables M SD 1 2 3 4 5 6 7 8 9 10 11 12 13 14
* * *** ** ** ***
(1) Gender 1.61 0.49 – .14 .002 .10 .13 .06 . 05 .26 .15 .15 .003 .24 .05 .19***
(2) Academic year 3.55 0.91 – .06 .01 .08 .13* .14* .13* .15** .03 .06 .20*** .03 .22***
(3) Years online 4.90 0.37 – .09 .06 .02 .10 .01 .07 .07 .09 .06 .06 .004
(4) Hours per day online 3.26 1.11 – .37*** .42*** .22*** .001 .09 .06 .08 .14* .14* .37***
(5) Longest hours online 4.17 0.81 – .21*** .10 .10 .02 .01 .13* .09 .08 .21***
(6) Hours social network 1.82 1.10 – .70*** .01 .10 .21*** .02 .29*** .08 .32***
(7) Visiting times network 3.24 1.36 – .06 .01 .23*** .03 .22*** .16** .24***
(8) Sensation-seeking 3.01 0.83 – .04 .002 . 08 .25*** .11 .18**
(9) Self-esteem 4.25 0.58 – .15** .04 .08 .26*** .30***
(10) IMPA 5.52 1.32 – .27*** .37*** .01 .01
(11) Research 4.36 1.75 – .37*** .11 .05
(12) I7PA 3.14 1.16 – .30*** .37***
(13) LNO 2.91 1.08 – .30***
(14) PIU-1 1.90 0.66 –

Note: N = 315 college students. For gender, 1 = male, 2 = female. IMPA, importance of maintaining pre-existing relationships; I7PA, importance of 7-positive activities; LNO,
likelihood of negative outcomes; PIU-1, problematic Internet use-1 measured by Young’s Internet Addiction Test (IAT) (1998).
*
p < .05.
**
p < .01.
***
p < .001.

The first hypothesis (H1) checked the relationship between the and that is just part of being PIU. Thus, we treated LNO as the
subjective values of online activities. As predicted, I7PA (b = .24, DV and tested its relationship.
p < .001) contributed positively to the model predicting PIU, but
IMPR (b = .06, p = .252) and researching information (b = .08, 3.6. SEM model analysis
p = .134) were not related to PIU. The second hypothesis (H2) exam-
ined the optimism bias effect, but LNO (b = .14, p < .01) contributed SEM model design to integrate the expected and unexpected
positively when entered into the model. The third hypothesis (H3) findings was conducted and the resulting analysis was both very
looked at the connection between self-esteem and PIU, and as pre- good fit to the current data (v2 (5, N = 315) = 3.60, p = .61,
dicted, self-esteem (b = .25, p < .001) was negatively related to PIU. CFI = 1.00, RMSEA = .00). Fig. 1 showed the importance of our re-
Overall, H1 and H3 were supported, but H2 was not. vised conceptualization of LNO as an aspect of PIU and of how sen-
The research question (RQ1) concerning sensation-seeking was sation-seeking operates through the importance of positive
also examined. After controlling for general Internet use patterns, activities to Internet participants.
sensation-seeking (b = .13, p < .05) had a significant positive effect,
but it failed to retain significance after entry of other psychological 4. Study 1: Discussion
variables (b = .07, p = .135). Thus this result requires additional
analyses. What do these findings do for the understanding of PIU? They
show that a clear distinction can be made between heavy Internet
3.5. Further analysis use and problematic use. They help resolve the inconsistent find-
ings about the role of sensation-seeking in the development of
The current findings raise two questions; (1) why was sensa- PIU. The model suggests that sensation-seeking has a positive ef-
tion-seeking not positively associated with PIU? and (2) what is fect (or relation to) problematic use only when persons with high
the direction of the relationship between perceived likelihood of sensation-seeking scores find some of the Internet activities highly
negative outcomes and PIU? important. Otherwise, sensation-seekers may not be prone to prob-
For the first question, structural equation modeling (SEM)
was performed to clarify direct and indirect relationships. We
used the comparative fix index (CFI) and the root mean square
error of approximation (RMSEA) for indicators of fit as Fan and
his colleagues (1999) recommended because these two indices
are particularly insensitive to sample size. Browne and Cudeck
(1993) suggested values of RMSEA between .05 and .08 indi-
cate an acceptable fit of the model and CFI should be equal
to or greater than .90 to accept the model. The data showed
suitable model fit (v2 (1, N = 315) = 2.71, p = .10, CFI = .98,
RMSEA = .07). SS was a significant direct predictor of I7PA
(b = .25, p < .001) and I7PA was a direct predictor of PIU
(b = .37, p < .001). This path model showed that the SS–PIU
relationship was entirely mediated by degree of endorsements
of positive functions.
The current data indicated that LNO and PIU were positively re-
lated (b = .14, p < .01). This finding is consistent with Caplan’s re-
sults for conceptually similar variables in his 2003 study. But one Fig. 1. Structural equation modeling 1: direct and indirect effects of degrees of
may still ask just why does this relationship exist? We see it as a sensation-seeking, self-esteem, importance of 7-positive activities on Global PIU in
part-whole relationship, not a predictive-outcome relationship. Study 1 (N = 315). (v2 (5, N = 315) = 3.60, p = .609, CFI = 1.000, RMSEA = .000).
*
p < .05; **p < .01; ***p < .001. Notes: (1) Arrows indicate path coefficients, which are
The idea is that when one is addicted, one goes ahead taking standardized regression weights (b), the coefficients value is shown above the
known risks without giving the consequences a heavy weight, relevant arrows.
 Author's personal copy

496 H.-K. Kim, K.E. Davis / Computers in Human Behavior 25 (2009) 490–500

lematic use at all. The model supports the implications of Baumei- characteristics have potential as predictors of Global PIU. These
ster’s model of self-regulation problems as arising from depletion are anxiety (excessive worry) (Spada, Lanston, Nikčević, & Moneta,
of self-control resources (Baumeister et al., 1994; Gaillot & Bau- 2008) and flow (Thatcher, Wretshko, & Fridjhon, 2008). While anx-
meister, 2007). Chronic low self-esteem is one indicator of such iety would be expected to have a negative relationship to self-es-
low self-control resources. Low self-esteem was strongly positively teem, we anticipate that it will make an independent and
associated both with PIU to various online activities and with significant negative contribution to the prediction of Global PIU.
biased thinking about potential negative consequences (e.g., per- The more anxious the person is, the greater the likelihood of him
sonal information sold, being misled, etc.). finding the Internet a safe and secure place for making social con-
In general this model is consistent with Caplan’s (2003) emphasis tacts, and for becoming overly dependent upon Internet usage.
on the role of psychosocial variables in the prediction and under- Flow, in contrast, is conceptually a positive factor in Internet usage
standing of PIU. We found, consistent with the studies by Valken- because it represents the tendency to become totally absorbed in
burg and Peter (2007) that for those participants who used the the activity at hand. To the extent that a person finds the opportu-
Internet to communicate with family and friends, heavy usage had nities and activities on the net important, there is a greater chance
little negative implications for PIU. In contrast, those who used the that a sense of flow will develop and that she/he will become
net to make new friends were much more likely to have high PIU overly involved in the Internet. In the case of flow, we anticipate
scores. that its relationship to global PIU can either be direct or that it
The finding about those with high scores on PIU having high can be mediate by the importance of Internet activities. Thus, the
scores on the possibilities of negative outcomes happening to them following three hypotheses are evaluated via SEM modeling to test
invites the question, ‘‘Why?” We propose that problematic Internet the proposed model and compare it to alternative models.
users keep using the Internet even though they perceive possible
Hypothesis 1. High self-esteem will be negatively associated with
vulnerability because the threats to self are not ‘‘real.” Rather, they
the latent construct of Global PIU.
are ‘‘true” in the sense of being known and thus recognized but not
seen as relevant to the self. ‘‘Real” is what one acts on, ‘‘true” is
Hypothesis 2a. High anxiety will be positively associated with the
merely what one believes. Other examples of such gaps between
latent construct of Global PIU.
the ‘‘real” and the ‘‘true” exist in the context of most problematic
behaviors. The warning labels on cigarette packs are one example.
Hypothesis 2b. High self-esteem will be negatively associated
The anti-binge drinking efforts on US college campuses are an-
with anxiety.
other. People can tell you the risks, but somehow these do not
count in their action tendencies.
Hypothesis 3a. The positive relationship between sensation-seek-
ing and the latent construct of Global PIU will be positively medi-
4.1. PIU Study 2
ated by the importance of 7-positive Internet activities.
Based on the SEM model from Study 1, we have argued that PIU
Hypothesis 3b. Flow is positively associated with the latent con-
and the unrealistic evaluation of negative outcomes should be con-
struct of Global PIU and this connection mediated by the impor-
ceived as aspects of a global latent construct (which we call Global
tance of 7-positive Internet activities.
PIU) which incorporates problematic Internet use and the distorted
evaluation of potential negative outcomes, wherein the user under-
estimates their likelihood of happening to him or her. Recognizing 5. Study 2: Method
that negative outcomes are likely and yet not having that constrain
one’s use fit the same pattern of addictive behaviors in which ad- 5.1. Participants
dicted smokers downplay the reality of the threats of cancer and
other illness or methamphetamine uses treat the threats to life and The Study 2 sample were 279 college students (145 females and
health as not a deterrent (Richmond and Anderson,1994; Stead, 134 males) with different academic background, Management
Perera, & Lancaster, 2008). A second aspect of the SEM model is that (34.8%), History (23.7%), Chemistry (18%), Sport and Entertainment
it suggests the way in which the personal trait of sensation-seeking Management (9.7%), Journalism and Mass Communications (8%),
is involved in PIU but shows it to be mediated by the importance of and Psychology (5.8%). The mean age of the sample was 21.4 years
the activities available on the net. Only for persons who value one or (SD = 3.2 years) and the age range was 16–42 years. The majority
more of the seven types of activities does sensation-seeking relate to (94.2%) of the sample fell in the 16–25 age range. The respondents
degree of PIU. contained 2.9% freshmen, 13.6% sophomores, 30.8% juniors, 49.8%
Our purpose in the second study was to develop a more com- seniors and 2.9% other. The sample was 76.3% Caucasian, 12.2% Afri-
plex model both of the constellation of PIU characteristics and of can American, 5.7% Asian/Asian-American, and 5.8% described
their potential antecedents. We were fortunate to have several the- themselves as ‘‘other.” Informed consent and debriefing were pro-
oretical resources for the development of both ideas. First, Caplan’s vided and the procedures were approved by the University IRB.
(2002, 2003) conception of PIU included an important dimension The previous questionnaires used in Study 1 (i.e., general Internet
that seemed theoretically distinct to us, but of great interest. In uses, sensation-seeking, self-esteem, importance of positive activi-
his measure, a set of 14 items reflecting the feeling that the Inter- ties, likelihood of negative outcomes, Young’s IAT, Caplan’s GPIUS,
net was a safe place for making interpersonal contacts were included. and biographical questions) including the two new scales (i.e., anx-
We may give these items the brief label, ‘‘Safe Social Contacts iety and flow) were distributed to each participant in classroom with
(SSC)” with the understanding that the full descriptive label is im- permission of the instructors.
plied in our usage. His research suggested a link between this view
of the Internet and engaging in problematic usage. In this study, we 5.2. Measures
separately assess the dimension of Internet interpersonal safety as
a factor that is part of the broader latent construct of Global PIU. 5.2.1. Predictor variables
In addition, we have examined the research literature and the In this sample, the mean sensation-seeking score was 3.19
recent meta-analysis of qualitative studies (Douglas et al., 2008) (SD = 0.73, a = .82), and the average self-esteem was 4.21 (SD =
and have come to the conclusion that two additional personal 0.61, a = .88). Anxiety as assessed by Spada et al. (2008) involved
 Author's personal copy

H.-K. Kim, K.E. Davis / Computers in Human Behavior 25 (2009) 490–500 497

positive beliefs about worry which ranged on a 5-point scale (e.g., (N = 218) of the individuals spent at least 1 h per day on the net.
1 = strongly disagree, 3 = neutral (mixed), 5 = strongly agree) with Interestingly, 88.5% (N = 247) of the students had experiences of
higher scores indicating higher positive perceptual tendencies about spending more than 3 h in one day online.
concern (e.g., ‘‘worrying helps me to solve problems,” ‘‘I need to wor- The majority (91.4%) used social networking sites and nearly all
ry in order to remain organized”). The mean score of anxiety was 2.40 respondents (93.5%, N = 261) had created their own personal Web
(SD = 1.03) and this scale was internally consistent (a = .94). space based on these sites. Facebook was the most popular site and
The literature review showed that flow experiences related to 83.5% (N = 233) of the students updated their profile on this site
PIU (Thatcher et al., 2008). Webster and his colleagues (1993) orig- while only 7.9% frequently used MySpace. 75.2% (N = 210) of the
inally developed the flow scale with 12 items. We replaced the students spent less than 1 h per day on these sites, but approxi-
words ‘‘Lotus 1-2-3” with ‘‘the Internet,” and used the present mately half of the students visited these sites at least 1 time per
tense (e.g., ‘‘when using the Internet, I am totally absorbed in what day (43.7%, N = 98). The following two reasons of using the Internet
I am doing,” ‘‘The Internet is intrinsically interesting to me”). Re- were common; (1) 35.5% (N = 97) used it to stay informed (down-
sponses ranged on a 5-point scale (e.g., 1 = strongly disagree, load, information search), and (2) 34.4% (N = 94) used it to keep in
3 = neutral (mixed), 5 = strongly agree) and the mean was 3.35 touch with family and/or friend.
(SD = 0.45, a = .70).
As the last predictor variable, the same question measuring the 6.2. Gender differences in Internet uses
importance of 7-positive activities (I7PA) was used (e.g., ‘‘shop-
ping/auctioning”, ‘‘joining the same interest groups”). The average In terms of the importances of Internet functions, there were
of the I7PA was 2.95 (SD = 1.00) and a = .71. significant gender differences. Women (M = 5.08, SD = 1.75) would
like to say that online activity for communicating with family was
5.2.2. Dependent variable more essential than men (M = 4.64, SD = 1.90), F(1, 275) = 3.95,
Since we treated the likelihood of negative outcomes (LNO) as p = .05, g2 = .01. However, other activities such as joining groups,
part of PIU, we used the same measurement for LNO used in Study downloading entertainment resources, playing games, meeting
1. The mean was 3.0 (SD = 0.97) and a = .83. new people, becoming famous, and flirting were more important
Next, the separate subsample of 70 students from Study 1 showed to men, F(1, 274) = 5.04, p = .03, g2 = .02; F(1, 275) = 5.45, p = .02,
that the slightly revised version ‘‘PIU-1” of Young’s IAT (1998) and g2 = .02; F(1, 275) = 40.65, p < .001, g2 = .13; F(1, 275) = 9.80,
Caplan’s 14 GPIUS items ‘‘PIU-2” (2002) were essentially the same p = .002, g2 = .03; F(1, 275) = 19.27, p < .001, g2 = .07; F(1, 275) =
(r = .93, p < .001), thus we used the same combined, single scale with 12.44, p < .001, g2 = .04, respectively. Men (M = 2.97, SD = 1.69)
IAT and GPIUS. Caplan’s (2002) original measure for the GPIUS con- were more worried about becoming addicted to Internet use than
tained not just the 14 items that we extracted for their direct reflec- women (M = 2.64, SD = 1.64), F(1, 276) = 2.77, p = .09, g2 = .01.
tion of problematic behaviors but also 15 other items, all but one of Women had a propensity to think hacking into emails or files
which were related to the concept of seeing the Internet as a safe was likely to happen to women (M = 3.25, SD = 1.48) than to men
place for making social contacts (SSC) (see Table 3). The remaining (M = 2.97, SD = 1.29), F(1, 276) = 2.81, p = .09, g2 = .01. Interestingly,
one item, ‘‘I have spent a good deal of time online” was deleted be- men had higher levels of SSC, PIU-1, and PIU-2 than women,
cause it seemed not to be a problematic behavior compared to other F(1, 277) = 15.61, p < .001, g2 = .05; F(1, 277) = 14.17, p < .001,
items. The mean of SSC was 1.76 (SD = 0.66) and Cronbach’s a was g2 = .05; F(1, 277) = 11.72, p < .001, g2 = .04, respectively.
.91. The average of PIU-1was 1.89 (SD = 0.65, a = .91) and the mean
of PIU-2 was 1.90 (SD = .71, a = .91). 6.3. Hypotheses testing

6. Study 2: Results Study 1 showed considerable influences of participants’ differ-

ent Internet uses patterns (i.e., years online, hours per day online,
6.1. Descriptive findings longest hours per day online, hours per day using social network-
ing sites, and numbers of visiting times to network sites) and
Among the 279 respondents, a large proportion 92.8% (N = 259) demographic characters (i.e., gender and academic year in college)
had been using the Internet for longer than 5 years and 75.5% on the levels of PIU. To build a more comprehensive model, we
controlled for these effects by using residualized scores of four
Table 3 dependent variables (i.e., SSC, LNO, PIU-1, and PIU-2). The zero-or-
Descriptive statistics of 14 Safe Social Contacts (SSC) Items of Caplan’s GPIUS in der Pearson correlation analysis explained positive relationships
Study 2. both between SS and I7PA (r = .26, p < .001) and between flow
SSC items M SD and I7PA (r = .23, p < .001) (see Table 4). Self-esteem was nega-
tively correlated with anxiety (r = .28, p < .001). Psychosocial
(1) I am worthless offline, but online I am someone 1.33 0.70
(2) I am more comfortable with computers than with people 1.64 0.95 variables (i.e., self-esteem, anxiety, flow, and I7PA) were positively
(3) I feel safer relating to others online rather than face-to-face 1.64 0.95 correlated with the four dependent variables, but flow and the
relationships residualized LNO was not significantly related. SS was only posi-
(4) I go online to make myself feel better when I’m down 1.65 0.97
tively correlated with the residualized PIU-2. The correlation be-
(5) I don’t worry about how I look when socializing online 2.80 1.45
(6) I use the Internet to talk with others when I feel isolated 2.04 1.15
tween PIU-1 and PIU-2 was .88 (p < .001).
(7) I use the net to make myself feel better when I’m down 1.69 1.00 Structural equation modeling (SEM) performed and the latent
(8) I am treated better online than in face-to-face relationships 1.44 0.77 construct of Global PIU was measured by the four dependent vari-
(9) I am more confident socializing online than offline 1.58 0.94 ables. As shown in Fig. 2, it had good fit (v2 (24, N = 279) = 47.73,
(10) I am treated better by others online than I am offline 1.52 0.85
p = .003, CFI = .97, RMSEA = .06).
(11) I seek others online when I feel isolated 1.73 1.04
(12) I feel safer relating to others online rather than face-to-face 1.52 0.95 The first hypothesis (H1) predicted the negative connection of
(13) I have control over how others perceive me online 2.45 1.32 self-esteem and Global PIU. Fig. 2 showed that individuals with
(14) I do not worry about relationship commitment when I am 1.63 1.02 low self-esteem had higher PIU tendency (b = .39, p < .001). The
socializing online second hypothesis (H2a) posited that anxiety was positively re-
Note: N = 279 college students. GPIUS, Generalized Problematic Internet Use Scale lated to Global PIU and the model indicated significantly positive
by Caplan (2002). Responses ranged from 1 (strongly disagree) to 5 (strongly agree). relationship between anxiety and Global PIU (b = .12, p < .05). As
 Author's personal copy

498 H.-K. Kim, K.E. Davis / Computers in Human Behavior 25 (2009) 490–500

Table 4
Means and standard deviations of, and correlations among, variables assessed in Study 2 (zero-order Pearson correlation coefficients).

Variables M SD 1 2 3 4 5 6 7 8 9
***
(1) Sensation-seeking 3.19 0.73 – .000 .01 .07 .26 .007 .07 .06 .13*
(2) Self-esteem 4.21 0.61 – .28*** .03 .06 .41*** .25*** .42*** .39***
(3) Anxiety 2.40 1.03 – .13* .004 .20*** .18** .21*** .13*
(4) Flow 3.35 0.45 – .23*** .17** .01 .16** .19***
(5) I7PA 2.95 0.99 – .26*** .27*** .25*** .28***
(6) Residualized SSC 0.00 0.99 – .28*** .77*** .73***
(7) Residualized LNO 0.00 0.99 – .36*** .30***
(8) Residualized PIU-1 0.00 0.99 – .88***
(9) Residualized PIU-2 0.00 0.99 –

Note: N = 279 college students. I7PA, importance of 7-positive activities; SSC, safe social contacts; LNO, likelihood of negative outcomes; PIU-1, problematic Internet use-1
measured by Young’s Internet Addiction Test (IAT) (1998); PIU-2, problematic Internet use-2 measured by Caplan’s Generalized Problematic Internet Use Scale (GPIUS)
(2002).
*
p < .05.
**
p < .01.
***
p < .001.

6.3.1. Gender
Although gender differences were not a primary focus of our
studies, together the two studies yield interesting findings. Col-
ley and Maltby (2008) found that men and women used the
Internet differently and their Internet uses had brought signifi-
cant changes into their real lives (e.g., women found great values
of online activities for searching general information, academic
purpose, reconnecting with old friends, seeking a romantic part-
ner, and shopping, while Internet uses for making connections
with people all over the world and searching for a job were
more important to men). The present study also showed gender
differences in Internet uses and the following patterns appeared
in both studies. Through the net, women wanted to keep in
touch with family, whereas men were more likely to play games,
download music or movies, become famous, and flirt. Only Study
2 showed considerable importances of Internet uses for joining
Fig. 2. Structural equation modeling 2: direct and indirect effects of self-esteem,
the same interests groups and meeting new people on men.
anxiety, flow, sensation-seeking, and importance of 7-positive activities on Global Interestingly, in both studies, men participants would say that
PIU in Study 2 (N = 279). (v2 (24, N = 279) = 47.73, p = .003, CFI = .973, RMSEA = their chances of becoming addicted to Internet use were much
.060). *p < .05; **p < .01; ***p < .001. Notes: (1) Arrows indicate path coefficients, higher than female participants. This is consistent of the findings
which are standardized regression weights (b), the coefficients value is shown
of Colley and Maltby (2008) (i.e., men were more concerned
above the relevant arrows. (2) The model examined all possible direct and indirect
effects. Insignificant paths are omitted for clarity of presentation. about negative outcomes such as Internet addiction, sexual video
clips, and unauthorized online information). This perceptual
anticipated (H2b), self-esteem was negatively related to anxiety inclination is related with the results that men had higher levels
(b = .28, p < .001). The third hypothesis considered the indirect ef- of PIU-1 than women in both studies. Men also had higher PIU-2
fects both of SS (H3a) and of flow (H3b) to Global PIU, and I7PA and SSC scores than women in Study 2.
mediated these relationships. As predicted, SS was positively con- Additional analyses showed that unrealistic optimism was pos-
nected to I7PA (b = .23, p < .001) and there was positive link be- itively related to PIU. As our comprehensive model demonstrated,
tween flow and I7PA (b = .22, p < .001). Finally, I7PA was PI users know the higher possibilities of negative happenings re-
positively related to Global PIU (b = .23, p < .001). Overall, all the sulted from their uncontrollable, compulsive online behaviors.
hypotheses were supported. More seriously, men PI users keep staying on the net and enjoy
In order to achieve a more compelling test of the model pre- becoming committed with new people even though their Internet
sented in Fig. 2, we examined several alternatives that involved uses could interfere with their daily activities. It triggers the same
single changes typically deleting one of the pathways specified in question again, ‘‘why?”.
the model above. In each case the indices of goodness of fit were
compared to those of model 2. In no cases were they superior or
as good as those of model 2. 7. Study 2: Discussion
The only one, labeled model 3, that came close to being as good
a fit as that presented above in Fig. 2, is one in which both anxiety The more complex models tested in Study 2 largely confirmed
and flow were given direct rather than mediated relationships to our expectations and built toward a more complete picture of
Global PIU. The goodness of fit indices for model 3 were v2 (24, the psychological processes underlying problematic Internet use
N = 279) = 57.70, p = .000, CFI = .962, RMSEA = .072. The direct link (PIU). First PIU is itself a complex latent variable reflecting both
between anxiety and Global PIU in model 3 was not significant alternative measures of problems with the Internet (Caplan’s,
(b = .10, p = .076) but the direct connection between flow and Glo- 2002; Young’s, 1998) as well as unrealistic views of one’s risk for
bal PIU was significant (b = .11, p = .044). We conclude that, on Internet problems other than addiction, and one’s sense of the
both statistical and conceptual grounds, model 2 seems to best re- Internet as a safe place for making social contacts (SSC). Thus our
flect the psychological process ongoing in PIU. final latent model is similar to, but not identical to Caplan’s
 Author's personal copy

H.-K. Kim, K.E. Davis / Computers in Human Behavior 25 (2009) 490–500 499

(2002). We see evidence from both studies that the unrealistic this set of variables. Caplan’s items assessing the perceived safety
views of one risk is a facet of PIU. of the Internet for making social contacts also seems to be an
Among predictors, self-esteem was a powerful negative predic- intrinsic part of PIU.
tor of use and its contribution was supplemented by trait anxiety None-the-less, the study has several limitations. The data are
or worry. The importance of the activities to the person also con- entirely cross-sectional, thus inferences about causal directions
tributed strongly positively to PIU, and interestingly the value are not warranted. The samples, while moderately large, was made
placed these activities mediated the relation both of sensation- up entirely of relative young college students at one large state
seeking tendencies and flow. The more a value placed on the activ- university. The SEM model, developed in Study 1 has been repli-
ities, the great the PIU and also the greater the sensation-seeking cated and extended to include more variables. Furthermore, it
and .the greater the likelihood of a sense of flow. would be preferable to gather data over time so that the plausibil-
Given the broad conceptual template that has guided our re- ity of causal directions from sensation-seeking, low-self-esteem,
search, we are pleased that an important aspect of knowledge – and the importance of Internet functions could be assessed at an
namely the perceived riskiness of Internet use has been shown to early point in time and their long-term effects on PIU and the dis-
function as part of the latent concept of PIU. As we anticipated regard of negative Internet outcomes determined.
the value placed on specific activities that were easily available The integration of variables from Caplan’s model and our model
on the Internet also play a major role in the potential for PIU. Three could provide a more comprehensive overall model. Caplan and
personal characteristics – self-esteem, anxiety, and the tendency to Turner (2007) emphasized emotional support in computer-medi-
get into a flow – all contributed to the prediction of PIU either di- ated communication (CMC) by engaging with diverse online activ-
rectly or via a mediated path through the importance of positive ities. For future research, they suggested looking at the ‘‘efficacy of
Internet activities. The one area that we did not investigate was online emotional support” rather than merely pointing out positive
that of skills and related variables such as self-efficacy or a sense vs. negative consequences from using the Internet. Also, because
of personal control. Caplan and High’s (2007) research suggests there is reason to believe that the Internet plays a greater role in
that this is a fruitful avenue for variables that will complete a com- the lives of a younger cohort (ages 11–18) and in different cultural
prehensive model of PIU. groups, such as Asians or Eastern Europeans, it will be important to
replicate and extend these findings to these samples.
7.1. Practical implications

References
Changing intrinsically rewarding behaviors is notoriously hard
to do. Just providing information about the potential dangers has Aboujaoude, E., Koran, L. M., Gamel, N., Large, M. D., & Serpe, R. T. (2006). Potential
proved over and over again not to be a deterrent. Among the markers for problematic Internet use: A telephone survey of 2513 adults.
important ingredients for prevention and treatment are these: (a) International Journal of Neuropsychiatric Medicine, 11, 750–755.
Armstrong, L., Phillips, J. G., & Saling, L. L. (2000). Potential determinants of heavier
Do not allow the person to return to the same social setting in internet usage. International Journal of Human Computer Studies, 53, 537–550.
which the problem behavior developed; (b) Use cognitive-behav- Bandura, A. (1986). Social functions of thought and action: A social cognitive theory.
ioral therapeutic techniques that changes negative self-concepts, Englewood Cliffs: Prentice-Hall.
Baumeister, R. F., Heatherton, T., & Tice, D. (1994). Losing control: How and why
anxiety, and low self-esteem (Beck, 1976; Bergner, 2007; Bergner
people fail at self-regulation. San Diego, CA: Academic Press.
& Holmes, 2000); (c) Use self-efficacy enhancement procedures Beard, K. W. (2005). Internet addiction: A review of current assessment techniques
such as those developed by Bandura (1986) and Marlatt and Dono- and potential assessment questions. Cyberpsychology and Behavior, 8, 7–14.
van (2005) to enhance attempting alternative behaviors and to Beck, A. (1976). Cognitive therapy and the emotional disorders. New York: Meridian.
Bergner, R. M. (2007). Status dynamics: Creating new paths to therapeutic change. Ann
persist in them in the face of relapses. Young (1998) has pioneered Arbor, MI: BurnsPark.
treatment efforts in the United States and the South Korean gov- Bergner, R. M., & Holmes, J. (2000). Self concepts and self concept change: A status
ernment has invested in treatment programs and built a national dynamic formulation. Psychotherapy, 37, 36–44.
Brenner, V. (1997). Psychology of computer use: XLVII. Parameters of Internet use,
network of 140 Internet addiction counseling centers (New York abuse and addiction: The first 90 days of the Internet usage survey.
Times, 2007). Korean researchers have developed checklists to Psychological Reports, 80, 879–882.
diagnose the degree of addiction since 2002, and the measure- 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. 136–162).
ments are targeted to two different age groups, adolescents (K- Newbury Park, CA: Sage.
Scale) and adults (A-Scale). Each measurement has two different Campbell, J., Greenauer, N., Macaluso, K., & End, C. (2007). Unrealistic optimism in
forms, self-report and observer-report (Korea Agency for Digital internet events. Computers in Human Behavior, 23, 1273–1284.
Caplan, S. E. (2002). Problematic Internet use and psychosocial well-being:
Opportunity, 2008). Development of a theory-based cognitive-behavioral measurement instrument.
Computers in Human Behavior, 18, 553–575.
7.2. Strengths and limitations of the studies Caplan, S. E. (2003). Preference for online social interaction: A theory of problematic
Internet use and psychosocial well-being. Communication Research, 30, 625–648.
Caplan, S. E. (2007). Relations among loneliness, social anxiety, and problematic
The present studies have brought together previously unexam- Internet use. Cyberpsychology and Behavior, 10, 234–242.
ined variables and shown how they work together to give a more Caplan, S. E., & High, A. C. (2007). Beyond excessive use: The interaction between
comprehensive model of PIU. The inclusion of careful controls for cognitive and behavioral symptoms of problematic Internet use. Communication
Research Report, 23, 265–271.
aspects of Internet uses and demographic characteristics in our Caplan, S. E., & Turner, J. S. (2007). Bringing theory to research on computer-mediated
regression analyses allowed us to distinguish heavy use from prob- comforting communication. Computers in Human Behavior, 23, 985–998.
lematic use. A new variable, similar in some respects to Caplan’s Colley, A., & Maltby, J. (2008). Impact of the Internet on our lives: Male and female
personal perspectives. Computers in Human Behavior, 24, 2005–2013.
(2003) preference for online social interaction but considerably Douglas, A. C., Mills, J. E., Niang, M., Stepchenkova, S., Byun, S., Ruffini, C., et al.
broader in item content, called the importance of positive Internet (2008). Internet addiction: Meta-synthesis of qualitative research for the
functions, showed promise both as a predictor of PIU and as a decade 1996–2006. Computers in Human Behavior, 24, 3027–3044.
Engelberg, E., & Sjöberg, L. (2004). Internet use, social skills, and adjustment.
mediator of the relationship between sensation-seeking, flow, Cyberpsychology and Behavior, 7, 41–47.
and PIU. The measure of PIU was a slightly modified version of a Fackler, M. (2007). In Korea, a boot camp cure for Web obsession. New York Times.
well-established scale-IAT (Young, 1998), and it showed higher Available from: <http://www.nytimes.com/2007/11/18/technology/18rehab.
html?pagewanted=all> Retrieved 19.05.08.
internal consistency than earlier studies did (a = .92). A small scale Fan, X., Thompson, B., & Wang, L. (1999). Effects of sample size, estimation methods,
comparative study of the IAT measure (Young, 1998) and Caplan’s and model specification on structural equation modeling fit indexes. Structural
GPIUS (2002) revealed them to function as essential identical with Equation Modeling, 6, 56–83.
 Author's personal copy

500 H.-K. Kim, K.E. Davis / Computers in Human Behavior 25 (2009) 490–500

Finkenauer, C., Engels, R. C. M. E., & Baumeister, R. F. (2005). Parenting behavior and Rosenberg, M. (1965). Society and the adolescent self-image. NJ: Princeton University
adolescent behavioural and emotional problems: The role of self-control. Press.
International Journal of Behavioural Development, 29, 58–69. Ossorio, P. G. (2006). The behavior of persons. Ann Arbor, MI: Descriptive Psychology
Gaillot, M. T., & Baumeister, R. F. (2007). The physiology of willpower: Linking blood Press.
glucose to self-control. Personality and Social Psychology Review, 11, 303–327. Samet, S., Waxman, R., Hatzenbuehler, M., & Hasin, D. S. (2007). Assessing
Griffiths, M. (2001). Sex on the Internet: Observations and implications for Internet addiction: Concepts and instruments. NIDA Science and Practice Perspectives, 4,
sex addiction. Journal of Sex Research, 35, 333–342. 19–31.
Internet World Stats (2007). Internet usage and population statistics for North Scherer, K. (1997). College life online: Healthy and unhealthy Internet use. Journal of
America. Available from: <http://www.internetworldstats.com/stats14.htm> College Students Development, 38, 655–665.
Retrieved 22.04.08. Slater, M. D. (2003). Alienation, aggression, and sensation-seeking as predictors of
Kalichman, S. C., & Rompa, D. (1995). Sexual sensation seeking and sexual adolescent use of violent film, computer, and Website content. Journal of
compulsivity scales: Reliability, validity, and predicting HIV risk behavior. Communication, 53, 105–121.
Journal of Personality Assessment, 65, 586–601. Spada, M. M., Lanston, B., Nikčević, A. V., & Moneta, G. B. (2008). The role of
Ko, C. H., Yen, J. Y., Chen, C. C., Chen, S. H., & Yen, C. F. (2007). Factors predictive for metacognitions in problematic Internet use. Computers in Human Behavior, 24,
incidence and remission of Internet addiction in young adolescents: A 2325–2335.
prospective study. Cyberpsychology and Behavior, 10, 545–551. Stead, L. F., Perera, R., & Lancaster, T. (2008). Telephone counseling for smoking
Korea Agency for Digital Opportunity and Promotion (2008). Available from: cessation (Cochrane Review). In The Cochrane Library (Issue 3). Chichester,
<https://www.kado.or.kr/> Retrieved 05.11.08. England: Wiley.
Kubey, R. W., Lavin, M. J., & Barrows, J. R. (2001). Internet use and collegiate Stinson, D. A., Logel, C., Zanna, M. P., Holmes, J. G., Cameron, J. J., Wood, J. V., et al.
academic performance decrements: Early findings. Journal of Communication, (2008). The cost of lower self-esteem: Testing a self- and social-bonds model of
51, 366–382. health. Journal of Personality and Social Psychology, 94, 412–428.
La Rose, R., Lin, C. A., & Eastin, M. S. (2003). Unregulated Internet usage: Addiction, Tangney, J. P., Baumeister, R. F., & Boone, A. L. (2004). High self-control predicts
habit, or deficient self-regulation? Media Psychology, 5, 225–253. good adjustment, less pathology, better grades, and interpersonal success.
Lavin, M., Marvin, K., McLarney, A., Nola, V., & Scott, L. (1999). Sensation seeking and Journal of Personality, 72, 271–324.
collegiate vulnerability to Internet dependence. Cyberpsychology and Behavior, 2, Thatcher, A., Wretshko, G., & Fridjhon, P. (2008). Online flow experiences,
425–430. problematic Internet use and Internet procrastination. Computers in Human
Leary, M. R., & MacDonald, G. (2003). Individual differences in trait self-esteem: A Behavior, 24, 2236–2254.
theoretical integration. In M. R. Leary & J. Tangney (Eds.), Handbook of self and Valkenburg, P. M., & Peter, J. (2007). Internet communication and its relation to
identity (pp. 401–418). New York: Guilford Press. well-being: Identifying some underlying mechanisms. Media Psychology, 9,
Lin, S. S. J., & Tsai, C. C. (2002). Sensation seeking and Internet dependence of 43–58.
Taiwanese high school adolescents. Computers in Human Behavior, 18, 411–426. Webster, J., Trevino, L. K., & Ryan, L. (1993). The dimensionality and correlates of
MacDonald, G., & Leary, M. R. (2005). Why does social exclusion hurt? The flow in human–computer interactions. Computers in Human Behavior, 9,
relationship between social and physical pain. Psychological Bulletin, 131, 411–426.
202–223. Weinstein, N. (1980). Unrealistic optimism about future life events. Journal of
Marlatt, G. A., & Donovan, D. M. (2005). Relapse prevention maintenance strategies in Personality and Social Psychology, 39, 806–820.
the treatment of addiction behaviors (2nd ed.). New York: Guilford. Widyanto, L., & Griffiths, M. (2006). Internet addiction: A critical review.
Morahan-Martin, J. M., & Schumacher, P. (2000). Incidents and correlates of International Journal of Mental Health Addiction, 4, 31–51.
pathological Internet use among college students. Computers in Human Young, K. S. (1998). Caught in the net. New York: John Wiley and Sons.
Behavior, 16, 13–29. Young, K. S. (2004). Internet addiction: A new clinical phenomenon and its
Richmond, R. L., & Anderson, P. (1994). Research in general practice for smokers and consequences. American Behavioral Scientist, 48, 402–415.
excessive drinkers in Australia and the UK: III. Dissemination of interventions. Young, K. S., & Rogers, R. C. (1998). The relationship between depression and
Addictions, 89, 49–62. Internet addiction. Cyberpsychology and Behavior, 1, 25–28.