Individual differences in implicit learning abilities and impulsive behavior in the context of Internet addiction and Internet Gaming Disorder under the consideration of gender (2017)

Available online 7 February 2017

• Rayna Sariyska^{a, ,} ,
• Bernd Lachmann^a, ,
• Sebastian Markett^{b, c,} ,
• Martin Reuter^{b, c,} ,
• Christian Montag^{a, d,}

http://dx.doi.org/10.1016/j.abrep.2017.02.002

Highlights

• Higher Internet addiction scores were linked to deficient implicit learning.

• This association was found in two independent groups of male (excessive) gamers.

• Online Gaming addiction was linked to higher risk-taking in healthy participants.

• Implicit learning and risk-taking were assessed, using an experimental task.

Abstract

Introduction

In three consecutive studies, we aimed to investigate the relationship between problematic Internet use (PIU), Internet Gaming Disorder (IGD) and implicit learning abilities, and impulsivity/risk-taking among online video gamers and control participants.

Methods

In study 1, N = 87 male visitors, recruited at the “Gamescom” in Cologne (2013), filled in a short version of the Internet Addiction Test (s-IAT), the Online Gaming Addiction Scale (OGAS), and completed an experimental task to assess implicit learning abilities. In study 2, a group of WoW gamers and control participants completed the same set up, in order to replicate the results of study 1. Study 3 used a modified version of the experiment to measure impulsivity/risk-taking in a group of healthy participants.

Results

In study 1, results revealed a significant negative correlation between the s-IAT score and the measure of implicit learning among male Gamescom participants. In study 2, the s-IAT and WoW addiction scores were negatively correlated with implicit learning only in male WoW players, which mirrors the results from study 1. In study 3, the OGAS score was positively correlated with the experimental measure of impulsivity/risk-taking.

Conclusion

In the current research project, deficient implicit learning was linked to PIU only in male participants with (a tendency towards) IGD. These findings might help to disentangle some opposing results on this relationship, when considering the gender of participants. Furthermore, higher risk-taking tendencies were associated with IGD among healthy participants, thus, suggesting the potential of risk taking as a predictor of IGD in a non-gamer population.

Keywords

• Internet addiction;
• Internet Gaming Disorder;
• Implicit learning;
• Risk-taking

1. Introduction

The Internet has found its way into the daily life of many people worldwide, offering an easy way to gather information and to consume entertainment. With the growing number of Internet users, accounting for almost 50% of the world population at the moment (accessed on 07.09.16. http://www.internetlivestats.com/internet-users/), the number of reports on problematic Internet usage (PIU) is rising. In a representative study from Germany (N = 15,024 participants) Rumpf, Meyer, Kreuzer, John, and Merkeerk (2011) showed prevalences of 1.5% in Internet addiction, with younger users showing higher proportions (4% in the group of 14–16 year olds). First attempts to define and diagnose PIU¹ have been made by Kimberly Young in the year 1998 (see also first case report from Young, 1996). Since then numerous tests and screening instruments have been developed (e.g. Young, 1998b, Young, 1998a and Tao et al., 2010), in order to be able to calculate prevalences in different populations and provide patients with effective treatment. However, there is still no existing nosological classification of PIU. The research on online gaming addiction seems to be one step ahead, as recently Internet Gaming Disorder (IGD) was included in Section III of DSM-5, by this means encouraging further examinations before its consideration as a formal disorder (American Psychiatric Association). IGD is considered to be a specific form of PIU, which only overlaps in small parts with the generalized form of PIU described above (e.g. Davis, 2001 and Montag et al., 2015).

1.1. PIU and implicit learning/decision making

Deficits in decision making have been shown in numerous studies, investigating patients with substance and behavioral addictions (e.g. Bechara et al., 2001 and Schoenbaum et al., 2006). Because of similarities in the conceptualization of PIU and behavioral/substance addiction (Young, 1998a), the topic of decision-making is also of high relevance to better understand the nature of excessive Internet usage. When assessing decision making a differentiation between decision making under ambiguity and decision making under risk have been made (Brand et al., 2006 and Schiebener and Brand, 2015). While in decision making under ambiguity the rules for gains and losses and the probabilities of different outcomes are not explicitly explained (measured e.g. with the (first trials of the) IOWA Gambling Task or IGT), in decision making under risk explicit information about the potential consequences, and the probabilities for gains and losses is available or is calculable (measured e.g. with the Game of Dice Task or GDT) (Brand et al., 2006 and Schiebener and Brand, 2015). Based on this differentiation and on the dual-process models of decision making (e.g. Epstein, 2003), Schiebener and Brand (2015) proposed a theoretical model to explain decision making under risk. In this model the role of executive functions is highlighted as a key of relevance for decision making under risk, but not decision making under ambiguity. Emotional reward and punishment are supposed to accompany both forms of decision making. Thus, both reflective processes (controlled by cognition), along with impulsive processes (induced by the anticipation of emotional reward and punishment) may be involved in decision-making processes under objective risk conditions (Schiebener & Brand, 2015). Moreover, factors such as information about the decision situation, individual attributes and situational induced states and external influences have been proposed to have modulatory effects on decision making (Schiebener & Brand, 2015).

With respect to Internet addiction a new theoretical framework was proposed by Brand, Young, Laier, Wölfling, and Potenza (2016), called an Interaction of Person-Affect-Cognition-Execution (I-PACE), where an impairment of executive functions and inhibitory control also has been highlighted to be of relevance for the development of PIU. According to this model the development and maintenance of specific Internet-use disorders underlie interactions between predisposing factors (e.g. personality and psychopathology), moderators (e.g. dysfunctional coping style and Internet expectancies), and mediators (e.g. affective and cognitive responses to situational cues). These complex interactions, combined with experiencing gratification and positive reinforcement, as a consequence of the use of a certain feature of the Internet, and with reduced executive functions and inhibitory control, could result in a specific Internet use disorder.

So far, a few empirical studies have been conducted in the context of PIU, inhibitory control and decision making. Most of them are in accordance with the aforementioned theoretical framework by Brand et al. (2016). Sun et al. (2009) for example reported worse performance in a gambling task in excessive Internet users and slower choice of a successful strategy compared to control participants. In a more recent study, Pawlikowski and Brand (2011) reported reduced decision-making ability under risk in the GDT in a group of excessive World of Warcraft (WoW) players compared to control participants. Yao et al. (2015) used a modified version of the Go/NoGo task (where gaming-related stimuli were used next to neutral stimuli) and reported reductions in inhibitory control in participants with IGD, compared to control participants. Laier, Pawlikowski, and Brand (2014) found similar results with a modified version of the IGT, when using pornographic and neutral pictures on the advantageous and/or disadvantageous card decks. Here, in a sample of male pornography users, participants showed deficient decision making in trials where the pornographic pictures were associated with disadvantageous card decks. However, also mixed results concerning decision making in the context of PIU or IGD were reported. In a study by C. Ko et al. (2010) for example Internet addicted participants showed better decision making, measured with the IGT, compared to control participants. In the study by Yao et al. (2015) already cited above, no difference in decision making using the IGT could be found between healthy participants and those with IGD. To disentangle these conflicting results further studies, examining possible interfering variables, are necessary. One particular variable is described later in the current study.

1.2. PIU, risk taking and impulsivity

Due to the initial characterization of PIU as an Impulse control disorder, a number of studies were conducted to explore PIU in the context of impulsivity and risk-taking. Cao, Su, Liu, and Gao (2007) and Lee et al. (2012) showed that PIU was positively associated with trait impulsivity, measured with the Barratt Impulsiveness Scale (BIS-11). With respect to the theoretical framework by Brand et al. (2016), already introduced above, impulsivity is mentioned among the personality factors, showing most stable associations with PIU and is, thus, proposed to be one of the factors, influencing its development and maintenance. Broadly, impulsivity is characterized as “a predisposition toward rapid, unplanned reactions to internal or external stimuli, without regard to the negative consequences of these reactions to the impulsive individuals or to others” (Moeller, Barratt, Dougherty, Schmitz, & Swann, 2001; p. 1784). The related term of risk-taking is defined as “behaviors performed under uncertainty, with or without inherent negative consequences, and without robust contingency planning” (Kreek, Nielsen, Butelman, & LaForge, 2005; p. 1453). C. Ko et al. (2010) applied the Balloon Analog Risk Task (Lejuez et al., 2002) to measure risk-taking, but found no significant association with PIU. In the present study, we are once more looking into these associations, by applying both, self-report along with experimental measures of impulsivity/risk-taking.

1.3. The role of gender for PIU/IGD

Another important issue in the context of Internet addiction is the preference of specific features of the Internet (e.g. online shopping, online gaming), depending on gender. A representative study from Germany showed that 77.1% of Internet addicted females at the age of 14–24 years use social networking sites compared to 64,8% males at the same age (Rumpf et al., 2011). In the same study 7.2% of Internet addicted females at the age between 14 and 24 years reported using the Internet to play online video games, compared to 33.6% of males at the same age (Rumpf et al., 2011). Thus, it seems that with respect to IGD, male participants show higher preference for online-gaming, compared to female participants and were reported to be more at risk to develop IGD. Moreover, Ko, Yen, Chen, Chen, and Yen (2005) observed that older age, lower self-esteem and lower daily life-satisfaction were associated with more severe IGD among males, but not females. Despite these results, there are still just a few studies, which systematically consider the gender of participants as a moderator/mediator variable in the context of PIU. However, it is possible that these differences account for some opposing results in the field and, thus, in the following studies they will be taken into consideration.

The aim of our research project was to investigate the link between PIU, as well as IGD and implicit learning in a group of male participants with proneness to IGD (study 1). In study 2 we aimed at replicating these results, by comparing healthy participants and excessive WoW players under the consideration of gender. The purpose of study 3 was to explore the relationship between PIU, IGD and Impulsivity/risk-taking (self-report and experimental data) in healthy participants.

Based on the aforementioned literature, we formulated the following hypotheses:

Hypothesis 1.

We expect negative associations between PIU/IGD and implicit learning abilities (Study 1).

Hypothesis 2.

We expect negative associations between PIU/IGD and implicit learning abilities (Study 2). We expect this negative association to be strongest in the group of male WoW players.

Hypothesis 3.

We expect positive associations between PIU/IGD and the self-report and experimental measures of impulsivity/risk-taking in healthy participants (Study 3).

2. Study 1

2.1. Methods

2.1.1. Participants

N = 107 participants (99 males, 8 females, age M = 19.52, SD = 3.57) were recruited at the “Gamescom 2013” in Germany, the world’s largest gaming event. However, because the very low number of female participants in the present sample (n = 8) and the above reported gender differences in the context of IGD (e.g. Rumpf et al., 2011), we excluded the female participants from the further analyses of the study. After also excluding participants with missing data, the sample resulted in n = 79 male participants (age M = 19.81, SD = 3.62). Regarding their education, 8.9% reported having university or polytechnic degree, another 40.5% reported having A-level or vocational baccalaureate diploma and 26.6% reported having secondary school leaving certificate or secondary modern school qualification, while 24.1% reported having no school diploma.

2.1.2. Measures

Participants answered questions about their age, gender and education, filled in a short version of the Internet addiction test (s-IAT, Pawlikowski, Altstötter-Gleich, & Brand, 2013; Cronbach’s Alpha in the present sample was 0.70), containing 12 Likert-scaled items (1 = never to 5 = very often) and the Online Game Addiction Scale (OGAS, a modified version of the Gaming Addiction Scale by Lemmens, Valkenburg, & Peter, 2009, where the word “online” was added to every item; Cronbach’s Alpha in the present sample was 0.66), consisting of 7 items ranging between 1 = never and 5 = very often. Additionally, participants rated their computer gaming experience (e.g. “For how many years have you been playing computer games?” or “How many hours on average per week do you play online computer games?”). A self-report measure of risk-taking was administered, including one item on overall risk-taking tendencies (“How would you describe yourself from 0 (not at all willing to take risks) to 10 (absolutely willing to take risks?”); German Socio-Econimic Panel (SOEP; Siedler, Schupp, Spiess, & Wagner, 2008). We used a slightly adjusted experimental task (“Devil’s chest”), incorporated from a study by Eisenegger et al. (2010), in order to measure implicit learning. On each of a total of 36 trials, we presented ten pictures of closed wooden boxes on the computer screen. The boxes were aligned in one row and participants had the opportunity to subsequently open a self-selected number of boxes, working from left to right. Participants were instructed that nine of the boxes contained a virtual monetary reward (5 cents) and one contained a “devil”. If participants opened only reward boxes on a given trial, they proceeded to the next trial by gaining the sum of the rewards. If they opened a box, containing the devil among with the other boxes, they lost everything on the current trial. The upcoming position of the devil was randomized among the 36 trials, but appeared on each position from 2 to 10² exactly four times. Although this was not mentioned to the participants, participants with higher cognitive skills might have worked out an implicit understanding for this rule and might have learned to perform better in the course of the experiment. The total of monetary rewards by the end of the experiment is further referred to as “GAIN” and will be used as a measure of implicit learning. The experimental set up is depicted in Fig. 1.

Fig. 1. 

Experimental set up of the Devil’s chest – opening the chest with the devil led to losing all collected coins of a given trial.

Figure options

2.1.3. Procedure

All questionnaires in English were translated into German by our own work group. The participants first filled in the questionnaires and then completed the Devil’s chest experiment. Please note, that participants in study 1 did not receive any monetary reward after completing the experiment and that they were informed about this fact prior to completing the experiment.

2.1.4. Statistical analyses

For the following analyses the normality of the data was examined by applying the rule of thumb, suggested by Miles and Shevlin (2001; p. 74), considering the skewness of the investigated variables. Correlation analyses were computed with Pearson’s or Spearman correlations, depending on the distribution of the data, and bootstrap bias-corrected and accelerated confidence intervals (BCa 95% confidence intervals) were computed for every correlation coefficient to further test their significance. Repeated measures ANOVA was used to test for implicit learning effects, when comparing the gain in the first 18 trials with the gain in the last 18 trials of the experiment.

2.1.5. Ethics

The research project (studies 1, 2 and 3) was approved by the Local Ethic Committee of the University of Bonn, Bonn, Germany. All subjects provided informed consent before completing the study.

2.2. Results

Means and standard deviations of the variables under investigation are presented in Table 1.

Table 1.

Mean, standard deviation (SD) and possible/actual range for the variables gaming experience (years), online gaming hours per week, s-IAT, OGAS, GAIN and risk taking (self-report).

N = 79, risk taking (self-report) n = 64.

a

Please note that the maximal possible range for the variable GAIN was estimated under the assumption that the devil would appear on every of the 36 trials on position 10 and the participant would stop the current trial at position 9. Thus, the devil would not interrupt the process of opening boxes and the participants would win the highest possible amount of money per trial (= 45 MU) in every consecutive trial. However, realistically there is a very low possibility that this event would occur.

Table options

2.2.1. Correlation analyses

Only the variable GAIN was not normally distributed. The age of the participants was positively correlated with GAIN (ρ = 0.27, p < 0.05). Moreover, GAIN showed a negative correlation with the s-IAT score (ρ = − 0.26, p < 0.05). In addition, we computed partial correlations for GAIN and the s-IAT score to control for age. The correlation remained significant (r = − 0.28, p < 0.05). The negative correlation between GAIN and the OGAS score did marginally not reach significance (ρ = − 0.20, p = 0.073) and remained non-significant after controlling for age (r = − 0.12, p = 0.292). All significant correlations remained significant after the inspection of the BCa 95% confidence intervals. Please see Table 2 for an overview of the results. (See Fig. 2 and Fig. 3.)

Table 2.

Correlations between GAIN in the “Devil’s chest” experiment and the s-IAT, OGAS score and risk taking (self-report).

N = 79, risk taking (self-report) n = 64; Spearman correlations are depicted in Italic.

⁎⁎

p < 0.01.

⁎

p < 0.05.

Table options

Fig. 2. 

Means and the standard error for the GAIN in the first 18 trials vs. the GAIN in the last 18 trials of the “Devil’s chest” experiment. MU = monetary units.

Figure options

Fig. 3. 

Means and the standard error for the GAIN during the first 18 vs. the last 18 trials of the “Devil’s chest” experiment, for control participants (left graph) and WoW-players (right graph). MU = monetary units.

Figure options

2.2.2. Manipulation check of the “Devil’s chest” experiment as a measure of implicit learning

The results of the repeated measures ANOVA showed a significant mean difference between the GAIN in the first 18 trials of the experiment, compared with the last 18 trials (F(1,78) = 17.303, p < 0.01), showing that participants won more money in the second part of the experiment (M₁ = 192.34 and M₂ = 221.27 respectively).

2.3. Discussion

To sum up, as proposed in our hypotheses, in study 1 Internet addiction was associated with deficient implicit learning abilities. This result delivers further evidence for the role of poor decision making in the context of PIU (e.g. Brand et al., 2016). The association with IGD was in the same direction, however, did not reach significance. This might be explained by the relatively small sample size and/or the relatively low internal consistency (0.66) of the OGAS scale in this study. In order to further investigate these relationships and compare the results between male and female participants and between gamers and non-gamers, study 2 was conducted.

2.4. Study 2

The aim of the second study was to replicate the results of study 1, by using a sample of World of Warcraft (WoW) players and control participants, who were naive to WoW. Given that the association between the s-IAT and GAIN as a measure of implicit learning could be observed in male participants with proneness to IGD, we were interested to see replication of study 1’s results particularly in male WoW gamers.

2.5. Methods

2.5.1. Participants

WoW players and control participants, took part in the study. The WoW players were recruited, using following criteria: WoW gaming experience for a minimum of two years. An exclusion criteria was playing other games than WoW for > 7 h per week, however, participants with no experience in other games were preferably recruited. Control persons needed to be WoW naïve, hence had no experience of playing this game before. Exclusion criteria for both groups of participants were visual impairment, difficulties in reading and writing, dyschromatopsia, concussion, long-term medication, neurological and psychiatric diseases, hearing disability and high substance use. After a thorough inspection of the sample we excluded one participant due to an eating disorder and daily cannabis consumption, one participants due to neurological and psychiatric disorders and one participant due to extreme values, and participants with missing data, which resulted in n = 77 control participants (39 males) and n = 44 WoW players (28 males). 6.5% (n = 5) of control participants reported casual usage of online role-playing games (< 3 h computer gaming per week) and 23.4% (n = 18) reported casual usage of Ego-shooter games (< 1 h gaming per week). The mean age of the total sample was M = 23.70 (SD = 3.93). Regarding their education 10.7% reported having a university degree, another 85.9% reported having A-level or vocational baccalaureate diploma and 2.5% reported having secondary school leaving certificate or secondary modern school qualification. One person did not answer the items regarding education.

2.5.2. Measures

Here again the s-IAT (Pawlikowski et al., 2013; Cronbach’s Alpha in the present sample was 0.76), OGAS (a modification of the GAS by Lemmens et al., 2009; Cronbach’s Alpha in the present sample was 0.88) and the computer gaming experience were assessed. Additionally, the World of Warcraft Specific Problematic Usage-Engagement Questionnaire (WoW-SPUQ), consisting of 27 items, rated on a scale from 1 = “completely disagree” to 7 = “completely agree” (Peters & Malesky, 2008; Cronbach’s Alpha in the present sample was 0.89) was filled in by the WoW group only. Moreover, the Barratt Impulsivity Scale (BIS-11; Patton & Stanford, 1995; Cronbach’s Alpha in the present sample was 0.85) was administrated as a measure of impulsivity (30 items are scored on a scale, ranging from 1 = “rarely/never” to 4 = “almost always/always”). With this scale, three second order factors can be assessed: Attentional impulsivity is defined as an inability to focus attention or concentrate; Motor impulsiveness involves acting without thinking, while Non-planning impulsiveness involves a lack of “futuring” or forethought (Stanford et al., 2009). Internal consistencies for the subscales in the present study were 0.73, 0.69 and 0.69 respectively.

2.5.3. Procedure

The participants took part in a large longitudinal study to investigate biological factors next to psychological variables and their role for IGD. For the present study, only the data from the first measurement point was used to test and replicate the findings from study 1 (completing the Devil’s chest experiment for a second time (T2) is clearly not comparable to being naïve with it as in study 1). The questionnaires and the experiment were completed in the same order as in study 1. Compared to study 1, however, in study 2 participants were paid the amount of money that they won in the “Devil’s chest” experiment and they were informed about this fact prior to completing the experiment.

2.5.4. Statistical analyses

The data assessment was conducted analogously to study 1.

2.6. Results

The OGAS score and the online gaming hours per week were non-normally distributed in the groups of male and female control participants. Furthermore, the s-IAT score and age were non-normally distributed in the group of female control participants. The correlation between GAIN and the s-IAT score in the group of male WoW players was tested one-sided, based on the findings in study 1.

Descriptive statistics for control participants and WoW players are presented in Table 3. Here male and female control participants had significantly lower gaming experience, online gaming hours per week, and OGAS score, compared to male and female WoW players (see Table 3). Moreover, female WoW players showed significantly higher scores on the s-IAT, compared to female control participants. All other variables did not differ significantly between control participants and the WoW players.

Table 3.

Means, standard deviations (SD), possible/actual range, t-/U value and significance for differences in means between the control and WoW group (p) for the variables gaming experience (years), online gaming hours per week, GAIN, s-IAT, OGAS, WoW-SPUQ and BIS-11 for the WoW and control participants.

Table options

2.6.1. Correlation analyses

For the groups of male or female control participants, age of participants was not significantly correlated to GAIN, s-IAT or the OGAS score. All other correlations are presented in Table 4. Here, GAIN was not significantly linked neither to the s-IAT nor to the OGAS score for male and female participants. Furthermore, the s-IAT score was positively linked to the BIS-11 subscale attentional impulsiveness in male control participants. All significant correlations remained significant after the inspection of the BCa 95% confidence intervals.

Table 4.

Spearman and Pearson correlations for the variables GAIN, s-IAT, OGAS and BIS-11 for the group of control participants.

Spearman correlations are depicted in Italic.

n(males) = 39, n(males, BIS-11) = 38, n(females) = 38, n(females, BIS-11) = 36.

⁎⁎

p < 0.01.

⁎

p < 0.05.

Table options

For the group of male and female WoW players, age was not significantly correlated with GAIN, s-IAT, OGAS or the WoW-SPUQ score. All other correlations are presented in Table 5. Here, GAIN was negatively associated with the s-IAT, as well as the WoW-SPUQ score only in the group of male WoW players. However, these correlations only showed a trend towards significance (r = − 0.30, p = 0.063, one-sided test and r = − 0.313, p = 0.104, two-tailed test). All significant correlations remained significant after the inspection of the BCa 95% confidence intervals.

Table 5.

Spearman and Pearson correlations for the variables GAIN, s-IAT, OGAS, the WoW-SPUQ score and BIS-11 for the group of WoW players.

Spearman correlations are depicted in Italic. For male participants, the correlation between the gain in the experiment and the s-IAT score was tested one-sided.

n(males) = 28, n(males, BIS-11) = 27, n(females) = 16, n(females, BIS-11) = 15.

⁎⁎

p < 0.01.

⁎

p < 0.05.

Table options

2.6.2. Manipulation check of the “Devil’s chest” experiment as a measure of implicit learning

The results of the repeated measures ANOVA did not show a significant mean difference between the GAIN during the first 18 and the last 18 trials of the “Devil’s chest” experiment in the group of male (F(1, 38) = 1.949, p = 0.171; M1 = 232.56 and M2 = 218.21) and female (F(1, 37) = 0.594, p = 0.446; M1 = 221.18 and M2 = 209.87) control participants. For the group of male WoW players, the difference between trials 1–18 and 19–36 reached significance (F(1,27) = 5.377, p = 0.028, M1 = 235.54 and M2 = 205.54; hence with a lower outcome in M2 compared to M1), whereas for female WoW players it remained non-significant (F(1,15) = 0.295, p = 0.595, M1 = 225.31 and M2 = 213.75).

Note to users:
Accepted manuscripts are Articles in Press that have been peer reviewed and accepted for publication by the Editorial Board of this publication. They have not yet been copy edited and/or formatted in the publication house style, and may not yet have the full ScienceDirect functionality, e.g., supplementary files may still need to be added, links to references may not resolve yet etc. The text could still change before final publication.

Although accepted manuscripts do not have all bibliographic details available yet, they can already be cited using the year of online publication and the DOI, as follows: author(s), article title, Publication (year), DOI. Please consult the journal’s reference style for the exact appearance of these elements, abbreviation of journal names and use of punctuation.

When the final article is assigned to volumes/issues of the Publication, the Article in Press version will be removed and the final version will appear in the associated published volumes/issues of the Publication. The date the article was first made available online will be carried over.