Abstract

There have been many debates on whether problematic internet use (PIU) and problematic video gaming (PVG) are overlapping concepts. The present study compared the common and unique predictors of PIU and PVG. School children (N = 1,080) completed a set of questionnaires on their internet use, basic psychological needs satisfaction, sensation seeking, impulsivity, internalizing symptoms, and parental factors (Wave 1). Respondents’ PIU and PVG were assessed 2 years later (Wave 3). Results indicate that PIU and PVG were predicted by impulsiveness, online social comfort, internalizing symptoms, parental attachment, and child’s perceived warmth at home. Age at Wave 1, age of first exposure to the internet, sensation seeking, online self-regulation, and parental communication apprehension were all significant longitudinal predictors of PIU, but not PVG. Parental involvement in media use predicted a decreased likelihood of PVG but not PIU.

Keywords: problematic internet use, internet gaming disorder, problematic gaming, behavioral addiction, risk factor

Funding: This study was funded by the Inter-Ministry Cyber Wellness Steering Committee, cochaired by the Ministry of Information, Communications, and the Arts and the Ministry of Education of Singapore.

Disclosures: The authors have no conflicts of interest. Angeline Khoo is currently retired from the National Institute of Education, Nanyang Technological University, Singapore. Albert Kienfie Liau, Dongdong Li, and Angeline Khoo were at the National Institute of Education, Nanyang Technological University, Singapore, during the period of the study. The data set used in the study is part of a larger data set. Other parts of the larger data set have been used in the following publications: Chng et al. (2015) and Choo et al. (2021).

Data Availability: The data set used for this study is available upon reasonable request from Douglas A. Gentile. The data set is not publicly available due to privacy and ethical restrictions.

Correspondence concerning this article should be addressed to El-Lim Kim, Department of Psychology, Iowa State University, 901 Stange Road,Ames, IA 50011, United States. Email: [email protected]

Problematic internet use (PIU) can be defined as maladaptive patterns of technology use that include symptoms such as preoccupation, tolerance, loss of control over internet use, withdrawal, and overuse that often result in impairment of the individual’s essential functioning (Choo et al., 2021). The idea that people can be addicted to the internet was first proposed in the 1990s (Griffiths, 1996; Young, 1998). Since then, many researchers have worked to conceptualize and provide valid measurement for PIU. During this process, the field soon realized that when people are said to be “addicted to the internet,” they could be addicted to different online activities—such as online gaming, social media, online shopping, pornography, and the like. Then the next question arose: Is internet itself addictive, or is internet a medium for other online activities that are addictive? In other words, the question focused on whether problematic use of different online activities should be understood as a single generalized “internet addiction” or be distinguished into specific internet-based addictions, such as internet gaming addiction, social media addiction, and online shopping addiction (Montag et al., 2021; Petry & O’Brien, 2013).

The goal of the present study is to expand the existing literature on problematic technology use by examining different correlates of PIU and problematic video gaming (PVG) to provide further evidence on whether they are two different patterns of “addictive” behaviors. Specifically, the present article focuses on characteristics that are shared by both PIU and PVG, as well as characteristics that are unique to either PIU or PVG.

Problematic Internet Use and Problematic Gaming

Among different types of activities, gaming was one of the different subtypes of online activities that received substantial research attention (as did gaming in general). Based on a large number of accumulated research findings on the disorder as well as the severity of its consequences (Petry et al., 2014), internet gaming disorder was included in the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5; American Psychiatric Association, 2013) as a condition for further study. At the time, the DSM committee found the evidence for gaming disorder to be much stronger than that for internet use disorder. Furthermore, the 11th edition of the International Classification of Diseases included gaming disorder as an official diagnosis (World Health Organization, 2019), giving additional weight to the importance of gaming in contrast to the internet. It is important to note that the diagnosis of internet gaming disorder is not based on the amount of time spent on playing video games. That is, contrary to the popular belief, a person who plays video games for many hours is not necessarily a problematic gamer. The DSM-5 lists symptoms of internet gaming disorder as follows: (a) preoccupation with gaming; (b) withdrawal symptoms when unable to play video games; (c) tolerance, or the need for more intensive gaming experience to be satisfied; (d) unsuccessful attempt to reduce the amount of gaming; (e) loss of interest in other hobbies; (f) continuing to game despite negative consequences; (g) deceiving others (e.g., family members, friends, partners) about the amount of time spent on gaming; (h) using video games to relieve or escape from negative moods; and (i) risking or jeopardizing important academic/occupational opportunities or significant relationships due to gaming.

Since the addition of internet gaming disorder in the DSM-5, there has been a debate about whether PIU and PVG should be viewed as two separate nosological entities. Studies suggest that online-connected video games, especially Massively Multiplayer Online Role-Playing Games, are more “addictive” than offline games (Balakrishnan & Griffiths, 2018; Bodi et al., 2020; Chen et al., 2020; Kuss & Griffiths, 2012; Mancini et al., 2019). Given these considerations, there is some merit to provide empirical evidence to differentiate PIU and PVG.

Common Correlates of Problematic Internet Use and Problematic Gaming

From the theoretical perspective, addictive use of technology such as PIU and PVG could be explained using the Interaction of Person–Affect–Cognition–Execution (I-PACE) model (Brand et al., 2019). The I-PACE model of addictive behaviors postulates that the person factors and the “inner circle” (e.g., affect, cognition, and execution-related factors) interact with one another to influence the developmental trajectory of problematic technology use. If two technology-related problematic behaviors show similar correlates, risk and protective factors, and outcomes, then we could argue that they share common psychological mechanisms. These correlates could include person factors such as predisposing variables (e.g., age, impulsivity, sensation seeking), psychopathological factors (e.g., depression, anxiety), and childhood experiences from parent–child interactions (e.g., attachment, parent–child communication style, perceived warmth at home). There could also be affective factors such as internalizing problems (e.g., loneliness), cognitive factors such as reward expectancy (e.g., basic psychological needs satisfaction—autonomy, competence, relatedness), and habitual behaviors (e.g., feeling more comfortable with online presence than offline presence, reduced ability to self-regulate the amount of internet activity).

Past studies demonstrate that individuals with PIU and those with PVG show several similarities in their correlational predictors, such as internalizing problems, psychological needs, and academic achievement. First, both PIU and PVG are associated with numerous internalizing problems (e.g., loneliness, depression, anxiety). For example, PIU and PVG are associated with more stress, anxiety, depressive symptoms, emotional distress, and low self-esteem (Elhai et al., 2017; Király et al., 2014; Moreno et al., 2022; Pontes, 2017; Strittmatter et al., 2016; Wang et al., 2018; Yau et al., 2013). People with PIU and those with PVG both reported more psychological distress compared to the normal population (Moreno et al., 2021; Wong et al., 2020). The two groups were also more likely to be socially withdrawn and were less likely to have close social connections for emotional support (Kato et al., 2020; Oka et al., 2021; Pontes et al., 2015; Wartberg et al., 2017). Second, both PIU and PVG are correlated with lower basic psychological needs satisfaction, such as competence, autonomy, and relatedness (Allen & Anderson, 2018; Li et al., 2016; Scerri et al., 2019). Third, PIU and PVG were associated with lower academic performance (Brunborg et al., 2014; Truzoli et al., 2019). At last, many studies indicate that parental involvement could reduce the risk of PIU and PVG (e.g., Cuong et al., 2021; Gentile et al., 2017; Martins et al., 2020). A warm, stable parent–child relationship is also associated with a lower risk of PIU and PVG (Bender et al., 2020; Karaer & Akdemir, 2019; Liau, Choo, et al., 2015). These similarities in correlates of PIU and PVG may point to shared psychological mechanisms of addictive technology-related behaviors.

Unique Correlates of Problematic Internet Use and Problematic Video Gaming

In contrast, however, some scholars pointed out that the classification issues should focus on the different functionalities of the internet (Tokunaga, 2016). That is, the addiction of the internet should be distinguished from the addiction on the internet. People’s motivation to use the internet depends on the specific internet activities. Not all internet activities involve gaming, and spending more time online is not necessarily related to more time playing online games (Király et al., 2014). People who play excessive amounts of video games are qualitatively different from those who spend excessive amounts of time shopping online. There are different cognitive networks, basic psychological needs satisfactions, and goal-attaining mechanisms involved in the two activities. From this point of view, it is inadequate to use a broader concept of problematic technology use to understand these two groups of individuals. Because of the conceptual ambiguity, some scholars recommended that the term “internet addiction” should not be used (e.g., Montag et al., 2021; Starcevic, 2013).

Apart from the conceptual aspect, people with PIU and PVG often show different behavioral patterns (Baggio et al., 2018; Rozgonjuk et al., 2023). As expected, those with PVG used internet predominantly for online gaming, whereas those with PIU used internet for a wide range of activities, such as gaming, chatting, and social networking (Balhara et al., 2021). That is, problematic internet users were not necessarily problematic gamers, and problematic gamers were not necessarily problematic internet users (Balhara et al., 2021; Király et al., 2014). The correlation between PIU and PVG ranged between r = .13 and r = .63 (Montag et al., 2015; van Rooij et al., 2012), demonstrating that the associations between PIU and PVG ranged from weak to moderately strong, depending on the samples.

Other studies have also found that the populations for PIU and PVG do not overlap greatly. For example, studies reported that PIU affects the general population, whereas PVG affects more males than females (Balhara et al., 2021; Holdoš, 2017; Kuss & Griffiths, 2012; Rehbein et al., 2010; Teng et al., 2021; Wartberg et al., 2021; Young, 1996). Some studies also reported that females are more likely to report PIU than males, but it is possible that the observed sex differences were due to females’ inclination to use more social media (Lei et al., 2018; Su et al., 2020). In other words, when we use “problematic internet use” as an umbrella term to assess general preoccupation with the internet use regardless of specific activities, then the sex differences in specific internet use addictions may become less prominent.

Nonetheless, a lack of overlap in populations affected does not in itself demonstrate different disorders. To demonstrate that PVG and PIU are distinct disorders requires stronger evidence, such as having distinctly different risk and protective factors. There is not as much research examining this question, but there is some. Compared to PIU, PVG was more strongly associated with negative interpersonal relationships, such as poorer social skills (King et al., 2013). People with PVG also reported poor parental attachment (Bonnaire & Phan, 2017; Jeong et al., 2020; Teng et al., 2020).

The Present Study

Past literature indicates that there are some similarities and differences in the correlates of PIU and PVG. For example, PIU and PVG seem to be related to different populations (e.g., PIU for general population and PVG for males). Nonoverlapping populations, however, do not suggest that the two are distinct taxa. The people who are addicted to playing slot machines are not the same people who are addicted to playing blackjack. Yet, these are not two different disorders—they are simply different morphologies of gambling disorder. One way to determine whether PIU and PVG are distinct technology disorders or different morphologies of the same underlying technology disorder is to assess whether they are predicted by different risk and protective factors. That is, if one explanatory variable predicts PVG but not PIU, then this may suggest that they are different disorders. In addition, it is useful to examine whether PIU and PVG share comorbidities. For example, if problematic internet users were depressed and anxious, but problematic gamers were socially withdrawn (e.g., high on loneliness), then this may suggest distinct disorders. These questions require large-scale longitudinal data to answer.

The goal of the present study is to examine the common and unique correlational predictors of PIU and PVG. Specifically, we examined commonly identified risk and protective factors on the subsequent development of PIU and PVG, such as described by the I-PACE model (Brand et al., 2019), including person factors (e.g., impulsivity, sensation seeking, attachment, parent–child communication style, perceived warmth at home, depression, anxiety), affective factors (e.g., loneliness), and cognitive factors (e.g., basic psychological needs—autonomy, competence, relatedness, online social comfort, self-regulation of online activities). Given the exploratory nature of the study, no specific hypotheses were formed. However, we predicted that different risk factors would be associated with PIU and PVG, respectively.

Method

Respondents and Procedure

The data were collected from 2010 (Wave 1) to 2012 (Wave 3). This study was a part of a broader research project that examined the effectiveness of cyber-wellness programs. The cyber-wellness programs commonly aimed to help students to develop healthy habits in using internet. The programs were educational and preventive in nature, not treatment services for any digital technology-related disorders.1 The study was approved by the Institutional Review Board of Nanyang Technological University, Singapore.

Schools were sampled based on the following procedure. The Ministry of Education (MOE) and cyber-wellness programs provided lists of schools that have used their services. Schools were categorized based on whether they received programs conducted: (a) by both MOE and cyber-wellness providers, (b) by MOE only, (c) by cyber-wellness providers only, or (d) by neither. The number of schools randomly selected from each group was determined according to calculated percentages to obtain a stratified representative sample. Students from the randomly selected schools were invited to participate in the study. The grade level of the students ranged from Primary 3 to Secondary 5 (equivalent to Grades 3–11 in the United States). Students from lower primary and secondary grades were oversampled so that they could be followed up for another 2 years.

All selected schools were mainstream schools following the Singaporean educational system (e.g., no private international schools). None of the selected schools offered integrated program (a scheme that allows high-performing students to skip the General Certificate of Education Ordinary Level exam and directly proceed to General Certificate of Education Advanced Level or other equivalent exams), which is expected given that a relatively small number of schools offer integrated program. All schools were similar in that they offered full-day classes, and students could choose to participate in extracurricular activities or cocurricular activities after class based on their choice. At the beginning of the study (Wave 1), there were 3,079 respondents (50.4% male) who agreed to participate in the study. The average age for the group was 13.01 (SD = 2.40) at Wave 1. Most respondents self-identified as ethnic Chinese (70.6%), followed by Malay (17.7%), Indian (7.3%), Eurasian (0.6%), and other races (3.8%). This ethnic profile is fairly representative of Singapore. The type of housing respondents lived in was used as a proxy measure of their socioeconomic status. Majority of the respondents (85.0%) lived in Housing & Development Board flats, followed by private condominiums (9.1%), terrace/semidetached houses (4.1%), and others (1.8%)—this housing profile is also representative of Singapore. All consent procedures required by the Singaporean MOE, the Singaporean Ministry of Information, Communications and the Arts, and participating schools were followed. Respondents completed a set of online questionnaires during class time in their schools. Questionnaires were administered in class by teachers who received detailed instructions from research personnel.

Two years later, respondents were contacted again to complete another set of questionnaires, which included a measure of PIU and PVG. Of the initial 3,079 student respondents (Wave 1), 1,080 of them completed the measure of PIU (Wave 3), and 976 completed the measure of PVG (Wave 3). Students who had graduated from Primary or Secondary schools between Waves 1 and 3 were not followed up at Wave 3.

Measures

Average Grade

At Wave 1, respondents reported their last exam scores in English, math, and second language. Exam scores were measured using an ordinal scale, where 1 = scores below 50 (out of 100), 2 = scores between 50 and 59, 3 = scores between 60 and 69, 4 = scores between 70 and 79, 5 = scores between 80 and 89, and 6 = scores above 90. The mean of the three exam scores was used as the average grade.

Age of First Exposure to the Internet

At Wave 1, respondents reported their approximate age when they have first used the internet. An ordinal scale was used to measure the age of first exposure to the internet, such that score of 1 = below age of 4, score of 2 = age of 4, score of 3 = age of 5, and so forth.

Basic Psychological Needs Satisfaction

At Wave 1, Basic Psychological Needs Scale (Deci & Ryan, 2000; Ilardi et al., 1993) was used to assess respondents’ basic psychological needs satisfaction in real life. The scale has 21 items and three subscales of seven items each, measuring autonomy, competence, and relatedness. Respondents rated the items on a 4-point scale, from 1 = not true at all, 4 = really true. Cronbach’s α values for autonomy, competence, and relatedness were .63, .64, and .68, respectively.

Sensation Seeking

At Wave 1, the Brief Sensation Seeking Scale (Hoyle et al., 2002) was used to measure sensation seeking. Respondents rated eight items on a 4-point Likert scale, ranging from 1 = not true at all to 4 = really true. Sample item includes “I like exciting parties.” High scores indicate high sensation-seeking tendencies. Cronbach’s α for the scale was .81.

Impulsivity

At Wave 1, the students’ level of impulsivity was measured using 10 items adapted from the Barratt Impulsiveness Scale used by Patton et al. (1995). The 10 items were selected based on analyses done in a related but different longitudinal project (Gentile et al., 2011; Liau, Neo, et al., 2015). Students rated each item on a 4-point scale ranging from 1 = strongly disagree to 4 = strongly agree. A higher score indicated a higher level of impulsivity. Cronbach’s α was .76.

Online Social Comfort

At Wave 1, online social comfort was assessed using the social comfort subscale (13 items) of the Online Cognition Scale (Davis et al., 2002). Respondents rated to what extent they agree with different statements describing online presence (e.g., “I am most comfortable online,” “Online relationships can be more fulfilling than offline ones”) on a 4-point Likert-type scale, 1 = not true at all, 4 = really true. Cronbach’s α for the scale was .95.

Online Self-Regulation

At Wave 1, online self-regulation measures an individual’s ability to organize their time spent on online activities. Items (N = 10) were adapted from Neo (2008). Respondents responded to a 4-point Likert scale (1 = strongly disagree, 4 = strongly agree) to items such as “I believe I can stop going online if I need to.” Cronbach’s α for the measure was .92.

Internalizing Symptoms

At Wave 1, internalizing symptoms (e.g., depression, loneliness, and anxiety) were measured. Depression was assessed using the Asian adolescents’ depression scale (Woo et al., 2004), which is composed of 22 items measured on a 5-point scale (1 = strongly disagree; 5 = strongly agree). The reliability of the scale was .96.

Loneliness was assessed with 16 items measuring peer-related feelings of loneliness by Marcoen and Brumagne (1985). It includes items such as “I feel sad because I have no friends” (1 = not true at all; 4 = really true). The reliability of the scale was .96.

Finally, anxiety was assessed with the Screen for Child Anxiety Related Emotional Disorders (Birmaher et al., 1997). There were 20 items on the questionnaires, and the respondents rated each statement on a 3-point Likert-type scale (0 = not true or hardly ever true; 2 = very true or often true). The reliability of the measure was .94.

Parental Factors

Parental factors were assessed at Wave 1. Parental factors included measures on (a) parental communication apprehension, (b) parental attachment, (c) parental involvement, and (d) warmth at home. Parental communication apprehension was measured using the Child–Parent Communication Scale (Lucchetti et al., 2002). The scale was intended to capture the children’s apprehension or anxiety about engaging in communication with their parents. There were 12 items in this scale, with Cronbach’s α = .78. Responses ranged from 1 = strongly disagree to 4 = strongly agree.

Parental attachment scale, adapted from the parental subscale of the Inventory of Parent and Child Attachment (Raja et al., 1992), consists of 12 items, measuring adolescents’ perceived quality of attachment to their parents. Responses were made on a 4-point Likert scale, ranging from 1 = strongly disagree to 4 = strongly agree. High scores indicated secure parental attachment, whereas low scores indicated poor parental attachment. Cronbach’s α was .80.

Parental involvement is a measure of parents’ monitoring, guidance, and education on the internet use and rule setting for different online activities. The items (N = 15) were adopted from the U.K. Children Go Online project (Livingstone & Bober, 2004) and EU Kids Online project (Livingstone & Haddon, 2008). Respondents answered whether their parents allowed them to engage in certain online activities (e.g., online games, social networking sites) or whether their parents have set rules about the amount of time spent on online activities. A total score was computed by tallying different parental involvement behaviors up. A higher score indicates that there was more parental involvement in children’s online activities.

At last, children’s perceived warmth at home was measured using items taken from Glezer (1984). Warmth at home was assessed using items such as “I feel accepted at home” and “Generally, there is nothing good about living at home (reversed).” Respondents rated six items on a 4-point Likert scale, 1 = strongly disagree, 4 = strongly agree. Higher scores indicate children’s positive perception of their home environment. Cronbach’s α for the measure was .84.

Problematic Internet Use

At Wave 3, Young’s internet addiction test (Young, 1998) was used to assess PIU. This questionnaire contains 20 items, and the respondents rated the statements (e.g., “How often do you find that you stay online longer than you intended?”) on a 5-point Likert scale (0 = not applicable; 1 = rarely; 5 = always). Cronbach’s α for the measure was .92.

A score of 50 or above is considered moderate addiction to internet (Young & Abreu, 2011), and this was the cutoff point used in the past studies using Asian samples to determine whether an individual is considered a problematic internet user (e.g., Ni et al., 2009; Tateno et al., 2018; Wu et al., 2016). Because we are interested in the clinical status of internet use disorder, the PIU variable was coded as a binary outcome, such that a score of 50 or above was PIU (score of 1), and a score below 50 was coded as non-PIU (score of 0).

Problematic Video Gaming

At Wave 3, PVG was measured using the Revised Pathological Video Gaming Scale (Choo et al., 2010; Gentile, 2009; Gentile et al., 2011), which contains measures of nine symptoms of internet gaming disorder as specified by DSM-5. Respondents answered whether they have experienced each symptom in the past year with a “no = 0,” “sometimes = 0.5,” or “yes = 1.” In compliance with the DSM-5 recommendations, a total score of 5 symptoms or more was considered PVG (score of 1). Cronbach’s α for the measure was .94.

A score of 4 or less symptoms was coded as non-PVG (score of 0). Although all respondents have used internet before, some respondents indicated that they have never played video games. These respondents did not have to answer the scale, and their scores were set to 0 (non-PVG).

Statistical Analyses

Descriptive statistics and zero-order correlations of the interest variables were calculated. Binary logistic regression was used to examine whether Wave 1 measures were significant predictors of subsequent PIU and PVG. Given that many Wave 1 measures were correlated with one another, to prevent overcontrolling for the variables, each Wave 1 measure was fitted as the explanatory variable for PIU or PVG on a one-on-one basis, respectively. Because multiple hypothesis tests were conducted, Benjamini–Hochberg corrected significance level (p = .036) was used to adjust for the false discovery rate. All analyses were conducted using SPSS Version 22.0. The data set used for this study is available upon reasonable request from the second author. The data set is not publicly available due to privacy and ethical restrictions.

Results

Descriptive Statistics

Means and standard deviations of Wave 1 measures and zero-order correlations among all variables are presented in Tables 1 and 2. There were 1,080 respondents who completed Wave 3 measure of PIU and 976 who completed Wave 3 measure of PVG. About 9.53% of the respondents met the clinical level of PVG. About 4.42% of the respondents reported both PIU and PVG.

Longitudinal Correlates of Problematic Internet Use

Sex (coded as 0 = male and 1 = female) was a significant longitudinal correlate of PIU measured at Wave 3, such that boys (vs. girls) were more likely to report PIU. Age and average exam grade (e.g., school performance) at Wave 1 were significantly associated with Wave 3 measure of PIU. Age of first exposure to internet was also associated with PIU. Of the basic psychological needs satisfaction, autonomy was the only significant correlate of PIU; competence and relatedness were not associated with PIU. Sensation seeking and impulsivity were associated with a higher odds ratio of PIU. All internalizing symptoms (e.g., loneliness, anxiety, depression) were significantly related to PIU. Finally, most parental measures (e.g., parental communication, parental attachment, warmth at home) were associated with lower odds ratios for PIU. Parental involvement, however, was not related to a lower odds ratio for PIU. The summary of the binary logistic regression results is shown in Table 3.

Longitudinal Predictors of Problematic Video Gaming

Sex was a significant longitudinal correlate of PVG: Boys were more likely than girls to be problematic video gamers. Age at Wave 1 and the age of first exposure to the internet were not related to subsequent PVG, but average grade at Wave 1 was associated with Wave 3 PVG (e.g., lower grade predicted a higher odds ratio of PVG). Autonomy was the only predictor of PVG among the basic psychological needs satisfaction measures; competence and relatedness were not associated with PVG. Sensation seeking and online self-regulation were not related to PVG. However, impulsivity and online social comfort were associated with higher odds ratios for PVG. All internalizing symptoms (e.g., loneliness, anxiety, depression) at Wave 1 were significant correlates of Wave 3 measure of PVG. Parental communication apprehension was not related to PVG, but all other parental factors (e.g., parental attachment, parental involvement, and perceived warmth at home) were associated with subsequent PVG. Table 4 summarizes the results of the analyses.

Discussion

The present study examined different correlates of PIU and PVG with a representative sample of Singaporean children recruited from a 2-year longitudinal study. In this sample, the overall prevalence rate for a moderate level of PIU was 22.68%. This prevalence rate was higher than the pooled prevalence rates reported in previous meta-analyses (Cheng & Li, 2014; Chia et al., 2020; Pan et al., 2020; Tang et al., 2018). For PVG, about 9.53% of the respondents who completed the measure met the criteria as problematic gamers. This prevalence rate was higher than what has been identified in some meta-analytic reviews (e.g., Pan et al., 2020; Stevens et al., 2021) but was lower than the rate reported in other studies based on Singaporean young adults (Tang et al., 2017, 2018).

Sex was a significant longitudinal predictor for both PIU and PVG. As evidenced in many past studies, boys were more likely to display patterns of PVG than girls (Gentile, 2009; Paulus et al., 2018; Rehbein et al., 2010). This finding is consistent with the general trend that more boys play video games than girls do (Chou & Tsai, 2007; Ko et al., 2005). Contrary to some previous findings where PIU was equally likely to affect both males and females (Holdoš, 2017; Wartberg et al., 2021), the present study found that boys were also more likely than girls to report PIU. This is similar to findings based on Korean children and adolescents, which also found that the prevalence rate of PIU was higher for boys than for girls (Ha & Hwang, 2014). Given that many studies examining the sex-correlated difference in PIU focused on young adults, it is worthy to note that the present study suggests that there may be sex-correlated differences for different age groups.

Age of first exposure to internet was a significant predictor of PIU, but not PVG. Early exposure to internet could be associated with PIU because those who start using the internet early on are more likely to incorporate technology into their day-to-day functioning, which may increase the risk of online activities replacing other hobbies and interests. This is consistent with the finding that newer (vs. older) generations tend to show more PIU patterns (Lozano-Blasco et al., 2022). However, the significance of early exposure does not necessarily mean that younger children are at a higher risk of PIU or PVG. We found that age was a significant predictor of PIU but not of PVG. Specifically, children who were older (vs. younger) at Wave 1 were more likely to report PIU at Wave 3, possibly because internet skills and access to different online activities tend to increase with age.

In addition, early exposure to the internet can foster greater online social comfort, which predicted both PIU and PVG in this study. Individuals who report higher levels of online social comfort prefer online interactions over in-person interactions (Davis et al., 2002). When online interactions substitute for offline interactions, quality of life may decline (Burrows et al., 2000; Cheng & Li, 2014). Despite the advancements in technological development, empirical studies demonstrate that having offline social support is better than online social support for emotional well-being (e.g., Longest & Kang, 2022). Such evidence may explain why PIU and PVG are both associated with negative consequences such as emotional distress (McNicol & Thorsteinsson, 2017; Wartberg et al., 2019).

Of the basic psychological needs satisfaction, autonomy (but not competence and relatedness) statistically predicted both PIU and PVG. Contrary to past studies, which found a negative association between real-world needs satisfaction and internet gaming disorder (Allen & Anderson, 2018; Li et al., 2016; Scerri et al., 2019), the present study unexpectedly found a positive association between autonomy and PIU and PVG, such that more satisfaction of autonomy needs in real life was associated with a higher odds ratio of problematic outcomes. For school-aged children, their autonomy needs may be satisfied by the perceived freedom in their decision-making processes, especially in how they spend their free time after school. When parents set rigid rules on their children’s activities, children may feel their autonomy needs are not fulfilled. In contrast, when children are allowed to engage in activities of their choice (e.g., using the internet, playing video games), they could perceive autonomy needs to be satisfied, which may explain why autonomy was positively correlated with later PIU and PVG in this sample. Future studies should consider whether such findings replicate in different populations with different backgrounds (e.g., race, ethnicity, age, cultures).

Sensation seeking was a significant predictor for PIU but not PVG. Adolescents who are identified as problematic internet users often score high on sensation seeking (Lin & Tsai, 2002). Unlike problematic video gamers, problematic internet users may be more likely to be high on sensation seeking because the internet allows for immediate changing of stimulation to something novel, unexpected, or similar to other things one likes. In contrast, video games may not fulfill novelty-seeking needs as well. Games are often similar in format with similar game mechanics, and many require “grinding” (doing the same task over and over to advance to the next game stage), which may explain why sensation seeking was not a significant correlate of PVG.

Higher impulsivity and lower level of online self-regulation were associated with an increased likelihood of PIU. Given that PIU arises when the person fails to control or regulate the amount of time spent on the internet (Tao et al., 2010), this result is not surprising. Higher impulsivity was also related to PVG, but online self-regulation was not. The fact that the ability to regulate time spent online was not associated with a decreased likelihood of PVG highlights a possible conceptual distinction between PIU and PVG, such that self-regulation of online activities is not necessarily the same as self-regulation of video game play.

As demonstrated in past findings, all internalizing symptoms (e.g., loneliness, anxiety, depression) were significant longitudinal correlates of both PIU and PVG (e.g., Hou et al., 2019; Király et al., 2014; Lemmens et al., 2011; Moreno et al., 2022). From the perspective of the self-medication hypothesis (Khantzian, 1997), it is possible that the children who are experiencing internalizing symptoms were using the internet or video games to regulate their negative mood. Symptoms of both internet use disorder and internet gaming disorder include the use of media (e.g., internet, video games) as a coping mechanism to alleviate negative affect, and they are both correlated with dysfunctional emotional skills (Kuss et al., 2017). One pair of studies found some cross-sectional evidence for this, where young adults with more psychological problems were more likely to use video games as a coping strategy, which in turn statistically predicted more gaming disorder symptoms (Plante et al., 2019). Further studies are needed to test this hypothesis.

Finally, parental attachment and warmth at home predicted both PIU and PVG. This is consistent with the findings that poor parent–child relationship is associated with internet use disorder and internet gaming disorder (Bonnaire & Phan, 2017; Jeong et al., 2020; Shek et al., 2018; Teng et al., 2020). On the other hand, parental communication apprehension was significantly related to PIU, which is in alignment with the past studies that identified parent–child communication as a protective factor for PIU (e.g., Cai et al., 2021; Shek et al., 2018). However, parental communication apprehension was not significantly associated with PVG. Interestingly, parental involvement, or parents’ rule-setting and guidance on media use, was associated with a lower likelihood of Wave 3 PVG but not PIU. Some studies have found that parental control over adolescents’ time spent on internet may increase rather than decrease the risk of internet use disorder (van den Eijnden et al., 2010), possibly due to psychological reactance and the needs for autonomy during adolescence, which may explain the lack of association observed in this study. Nonetheless, the current findings should not be interpreted as evidence against the use of parental control over media use to reduce problematic media behaviors because studies also report that a lack of parental control over internet use is associated with a higher level of PIU (Martins et al., 2020). There are also several studies showing that parental limits can reduce the risk of gaming disorder symptoms (e.g., Gentile et al., 2014, 2017). Additional studies on the role of parental guidance on media use, especially among adolescents, would be useful to determine when and what types of parents’ interventions are useful in curving the risk of behavioral addictions to technology.

To put this study into a broader context, some researchers have suggested that it is currently parsimonious to consider PIU and PVG to be two different morphologies of a similar underlying technology addiction (e.g., Rokkum et al., 2018). The argument is that people can be addicted to gambling despite doing almost nothing that looks the same—one puts money in a slot and pulls a lever, whereas other researches track conditions, horses, and jockeys. Nonetheless, both would be diagnosed with gambling disorder using similar definitions, we would expect similar symptoms and outcomes, and treatment would share many characteristics. That is, despite morphological differences, they are not distinct slot machine and horse racing addictions. Similarly, the vast majority of the research on PIU and PVG tends to show similar patterns of dysfunctional symptoms, similar outcomes, and similar responses to treatment. So how would we know if they are different morphologies of the same underlying disorder or distinct? Different prevalence rates would not be good evidence. The amount of overlap in populations of those with symptoms of each similar would not be good evidence. It is likely that there are different prevalence rates and populations who have problems with roulette, slot machines, poker, and horse racing, yet they all have gambling disorders. We would be convinced that gaming and internet disorders are clearly different if they:•demonstrated clearly different risk factors for those who become addicted to the internet or to games, • demonstrated clearly different protective factors between PIU and PVG, • had distinct etiologies or time courses, • had distinct patterns of comorbidity with other mental health problems, • had different outcomes of each, and/or • they needed different styles of treatment. At present, there is not enough research to answer most of these questions, although the research on outcomes seems to show very similar outcomes for both (e.g., depression, anxiety, poor school performance), which makes sense given that both are defined largely by dysfunction.

The present study examined whether longitudinal correlates were similar or different for each of the two disorders. There were common risk factors that were significant longitudinal correlates of both PIU and PVG, such as sex, academic performance (proxy measure of average grade), impulse control, online social comfort, internalizing symptoms, parental attachment, and perceived warmth at home. Given the shared risk factors of PIU and PVG, especially internalizing symptoms and impulse control, it could be argued that similar dysfunctional psychological factors contribute to the later development of PIU and PVG. However, the findings from the present study also point to the fact that PIU and PVG do not share all risk factors. For example, the results suggest that children, regardless of their age and first contact with the internet, could be at risk of PVG, whereas older children who were exposed to the internet earlier in life were at higher risk of PIU. That said, we did not measure the age of beginning gaming in this study and other studies that have found that earlier gaming predicted a higher risk of gaming disorder. Sensation seeking predicted PIU but not PVG, which points to the possibility that different personality traits could be related to PIU and PVG, respectively. Finally, parental factors may play different roles in PIU and PVG. For example, parental communication apprehension was related to PIU but not PVG. Parental involvement was a protective factor against PVG, but not PIU, perhaps because it is easier to monitor gaming than internet use. These results indicate that different parent–child dynamics may be involved in the developmental trajectories for PIU and PVG. The results suggest that the differences in risk and protective factors of PIU and PVG do not outweigh the similarities. That is, the data provided here do not provide strong evidence that PIU and PVG are different in their pattern of risk and protective factors.

Limitations and Future Directions

There are some limitations of the study that should be noted. First, we did not have baseline measures of PIU and PVG, which could have been used as control variables. Although beyond the scope of the purpose of this article, baseline measures can be useful to examine the developmental trajectory of PIU and PVG. It is unclear whether Wave 1 variables predict changes in PIU and PVG (or vice versa) based on the data used in the study. Future studies could consider controlling for the baseline measure of PIU and PVG.

Second, the study is based on children from Singapore. Having a sample drawn from a non-Western population could expand our current understanding of technology addiction across different cultures. However, it should also be noted that it is unclear how far these results generalize to different populations. Specifically, given that most respondents in the study were ethnic Chinese, we cannot draw inferences about risk factors for PIU and PVG for non-Chinese Asian populations. However, it is also important to keep in mind that we are dealing with psychological and behavioral patterns for which there is little theoretical reason to believe that Singaporean children would be particularly different from other children.

Finally, this study only included a selected list of widely examined risk factors for PIU and PVG. There are more risk factors that are relevant to both PIU and PVG that were not examined in the study, such as attention problems, peer relationships, and different motivations for internet use/gameplay (Bender et al., 2020). Future studies can include these risk factors as additional measures to replicate and further expand the research on predictors for PIU and PVG.

Copyright © 2023 The Author(s)

Received September 20, 2022
Revision received March 14, 2023
Accepted March 19, 2023