This study aimed to examine differences in brain activation for various types of reward and feedback in adolescent Internet addicts (AIA) and normal adolescents (NA) using functional magnetic resonance imaging (fMRI).
This study aimed to examine differences in brain activation for various types of reward and feedback in adolescent Internet addicts (AIA) and normal adolescents (NA) using functional magnetic resonance imaging (fMRI).
AIA (n = 15) and NA (n = 15) underwent fMRI while performing easy tasks for which performance feedback (PF), social reward (SR) (such as compliments), or monetary reward (MR) was given. Using the no reward (NR) condition, three types of contrasts (PF–NR, SR–NR, and MR–NR) were analyzed.
In NA, we observed activation in the reward-related subcortical system, self-related brain region, and other brain areas for the three contrasts, but these brain areas showed almost no activation in AIA. Instead, AIA showed significant activation in the dorsolateral prefrontal cortex for the PF–NR contrast and the negative correlation was found between the level of activation in the left superior temporal gyrus (BA 22) and the duration of Internet game use per day in AIA.
These findings suggest that AIA show reduced levels of self-related brain activation and decreased reward sensitivity irrespective of the type of reward and feedback. AIA may be only sensitive to error monitoring regardless of positive feelings, such as sense of satisfaction or achievement.
Adolescents are vulnerable to addiction because their reward-seeking tendency is stronger than that of adults. Internet addicts also show craving, aggression, and impaired decision-making, with a sharpened sensibility for immediate reward.
The need for reward-related neurobiological research on Internet addiction is slowly increasing.[3, 4] Yuan et al. reported that the volume of reward-related prefrontal area decreased in adolescent Internet addicts (AIA). Furthermore, a study by Dong et al. showed that Internet addicts had higher activation in the orbitofrontal cortex (OFC) when they earned monetary reward, but a significant reduction was observed in the anterior cingulate gyrus (ACG) when they lost the monetary reward. ACG is associated with the anticipation of risk or pain. So these results suggest that the brain of AIA is also sensitive to immediate reward but insensitive to the psychological pain associated with loss of the reward.
However, monetary reward is not the only reward that adolescents seek on the Internet. They are sensitive to the praises or criticisms posted by other users on their Twitter pages or websites. Recently, the number of neurological studies on ‘social reward’, such as compliments or reproach, has increased.[7, 8] Social rewards are shown to trigger a significantly increased activation in the dorsal and ventral striatum, including the nucleus accumbens. Meanwhile, humans integrate external social feedback into their self-concept. For those receiving attention and praise, activation of the brain areas associated with primary reward and the self-related cortical midline structure were observed.
Additionally, there is ‘performance feedback’, which somewhat differs from social reward and involves an objective evaluation of a task (e.g., ‘That is correct’, ‘You are wrong’). Positive feedback reinforces goal-directed behavior and invokes positive feelings. Different brain regions, such as the striatum, OFC, ACG, and prefrontal cortex are involved in positive feedback processing. So, performance feedback is likely to be related to the brain's reward system. Choi et al. reported that social reward and performance feedback both activate the subcortical area in adolescents.
Our study aimed to examine the differences between AIA and normal adolescents in brain responses to various rewards and feedback. This kind of study would be a new approach for understanding the phenomenon of adolescent Internet addiction deeply. We anticipated that the AIA would show reduced brain activation in three kinds of rewards and feedback.
AIA and normal adolescent (NA) groups were recruited at general hospitals and middle schools in Chungbuk province, Korea. Inclusion criteria for both groups were right-handed male, middle school students, with a score of short-form Wechsler Intelligence Scale for Children-III ≥ 80, and absence of major psychiatric disorders, such as schizophrenia, affective disorder, conduct disorder, and other substance-related disorders by Kiddie-Schedule for Affective Disorders and Schizophrenia – Present and Lifetime Version – Korean Version.
Internet addiction was determined using the Korean Adolescent Internet Addiction Scale (K-AIAS). This is a standardized Internet addiction measure that is a modified Korean version of the most widely used Internet addiction questionnaire by Young. The scores over 50 indicate addiction and the scores of 20–49 indicate normal users. To clearly distinguish the NA from the AIA, subjects scoring 50 and above were categorized as AIA and those scoring below 40 were NA. The participants of NA were different from those of the study by Choi et al. The degree of addiction to Internet games was measured using the Internet Game Addiction Diagnostic Scale. Because adolescence is a time in which brain maturation occurs dynamically, age match was applied for the two groups. The final analyses included 15 participants for AIA and 15 participants for NA consisting of six pairs for age 13, five pairs for age 14, and four pairs for age 15.
This study was conducted in accordance with the final version of the Declaration of Helsinki and was approved by the Bioethics Committee at the Chungbuk National University Hospital. Written consent was obtained from each participant after the study objectives and methods were fully explained.
The characteristics of the participants are shown in Table 1. One participant in AIA showed inattentive type attention deficit hyperactivity disorder (ADHD) and another had dysthymia. However, both were included in the study because these problems frequently coexist in AIA.
|Variables||AIA group (n = 15)||NA group (n = 15)||t|
|Age (years)||13.87 ± 0.83a||13.87 ± 0.83||0.000|
|IQ||101.80 ± 13.89||103.40 ± 7.89||−0.388|
|K-AIAS||61.87 ± 7.89||29.07 ± 5.37||11.706**|
|IGADS||61.47 ± 22.87||29.87 ± 4.03||5.270**|
|Duration for internet use per week (h)||16.93 ± 20.27||2.47 ± 2.56||2.743*|
|Duration for internet game use per day (h)||3.80 ± 2.02||0.78 ± 0.77||5.402**|
This was a block-design study. The task was a right–left discrimination test in which the participant was required to answer whether the abstract symbol was on the right or left side of the screen. This kind of task had also been applied by Choi et al. Stimuli were delivered using E-prime software version 2.0 (Psychology Software Tools, Pittsburgh, PA, USA).
Each block belonged to performance feedback (PF), social reward (SR), monetary reward (MR), or no reward (NR). There were a total of eight blocks, with each block comprising 10 trials. Each block appeared randomly. Each trial was configured as follows: stimulus (200 ms) – blank screen (1000 ms) – reward-related word (1300 ms) – small cross (500 ms). Such configurations were similar to the study by Pan et al. Each block took 30 s with a 15-s rest between blocks. Total time of the entire task was 6 min.
In each condition, a reward-related word was presented as follows. In NR, the Korean word for ‘next’ was presented continuously after performing the task regardless of the result of each trial. If the trial was performed correctly, the Korean word for ‘correct’ or ‘right’ was presented in PF, the Korean word for ‘great’ or ‘good’ was presented in SR, and the Korean words for 300 won or 400 won were presented in MR. At that time, 300 won corresponded approximately to $US0.267 whereas 400 won corresponded to $US0.357. According to the overall results of the MR condition, participants were given the money corresponding to the correct answers. Finally, if the trial was performed incorrectly in all three conditions, the Korean word for ‘next’ was presented.
The magnetic resonance images (MRI) were taken using a 3.0 Tesla whole-body ISOL Technology FORTE scanner (ISOL Technology, Seoul, Korea) at Daejeon KAIST. The blood-oxygen-level dependent technique using the echo planar imaging sequence was applied. The MRI parameters were: slice thickness = 5 mm, TR = 3000 ms, TE = 35 ms, flip angle = 80°, field of view = 220 × 220 mm, and matrix = 64 × 64. Thirty slices were obtained through the axial section image. The MRI parameters in the T1 anatomical scan were: TR = 2800 ms, TE = 16 ms, flip angle = 80°, field of view = 192 × 220 mm, and matrix = 192 × 256.
Imaging data were analyzed using the SPM2 (Wellcome Department of Cognitive Neurology, London, UK) software, and a general linear model was applied. For motion correction and co-registration and for identification of the anatomical location of the functional image, brain imaging data were analyzed using normalization and smoothing processes that matched a standard brain template to the image data from the experiment. The size of the smoothing kernel was 8 mm.
First, whole brain analyses were conducted to analyze the part of the brain activated in each contrast. The three main contrast pairs in this experiment were PF–NR, SR–NR, and MR–NR. After conducting an individual level analysis, within-group analyses were conducted for each group, followed by between-group analyses to examine the areas that were relatively more activated in each contrast for AIA and NA. Clusters that passed the voxel-level uncorrected threshold of P < 0.0005 and exceeded 10 voxels in size were considered activated regions. However, in AIA, within-subject analyses did not show any activation area and between-group analyses showed activation only in one area (the medial globus pallidus) in MR–NR condition for this threshold. Therefore, the threshold in AIA was adjusted and the regions that exceeded the threshold of P < 0.001 and the range exceeding 10 voxels were also reported.
The regions that were significantly activated in each contrast were selected as regions of interest (ROI). The effective values (5-mm radius) were extracted from the normalizing image using the MarsBaR (http://marsbar.sourceforge.net) toolbox. The correlation between the level of activation in the ROI area in each contrast and the variables related to Internet usage was analyzed using correlation analyses.
The reaction time and task accuracy for each trial in the two groups were compared using Student's t-tests. For analyses of the correlation between the level of activation in the ROI area and the variables related to Internet usage, Spearman's correlation analyses with Bonferroni's correction were applied. spss 12.0K for Windows was used for statistical analyses (spss, Chicago, IL, USA). Because the number of ROI was different in each contrast, the significance levels according to Bonferroni's correction were as follows: (i) in PF–NR, P < 0.00050; (ii) in SR–NR, P < 0.00062; and (iii) in MR–NR, P < 0.00096.
Both the accuracy of performing each task were as follows: (i) in NR: NA 100%, AIA 100%; (ii) in PF: NA 98.67%, AIA 99.33% (t=–0.894, P = 0.379); (iii) in SR: NA 100%, AIA 100%; and (iv) in MR: NA 99.00%, AIA 99.33% (t=–0.339, P = 0.737). The reaction times for each task were as follows: (i) in NR: NA 173.51 ± 50.14 ms, AIA 164.43 ± 56.22 ms (t = 0.467, P = 0.644); (ii) in PF: NA 179.30 ± 35.58 ms, AIA 160.40 ± 56.22 ms (t = 1.182, P = 0.248); (iii) in SR: NA 187.12 ± 54.29 ms, AIA 157.06 ± 50.31 ms (t = 1.573, P = 0.127); and (iv) in MR: NA 165.49 ± 48.19 ms, AIA 156.48 ± 61.25 ms (t = 0.448, P = 0.658). In all cases, statistically significant differences were not observed.
In NA, significant activated areas were as follows: (i) in PF–NR, right culmen (T = 6.11), left fusiform gyrus (BA 37, T = 5.29), left claustrum (T = 4.98), left hippocampus (T = 4.88), right putamen (T = 4.74), and right superior temporal gyrus (STG) (BA 38, T = 4.73); (ii) in SR–NR, the left fusiform gyrus (BA 37, T = 5.68), right cuneus (BA 23, T = 5.11; BA 18, T = 4.83), left cuneus (BA 18, T = 5.01), left tuber of vermis (T = 5.05), right declive (T = 4.77), left posterior cingulate (BA 30, T = 4.84), left culmen (T = 4.61), and left lingual gyrus (BA 18, T = 4.52); and (iii) in MR–NR, the left pons (T = 4.95) and left midbrain (T = 4.80) (Fig. S1).
However, AIA did not show any areas of significant brain activation in PF–NR and SR–NR, whereas significant activation in the left culmen (T = 4.99) was observed in MR–NR.
In PF–NR, compared to AIA, NA showed a significantly higher activation in right thalamus (T = 5.05), left thalamus (T = 4.00), right insula (BA 13, T = 4.16), right culmen (T = 4.54), right declive (T = 4.09), right putamen (T = 4.28), left putamen (T = 4.28), right fusiform gyrus (BA 19, T = 4,00), left fusiform gyrus (BA 37, T = 4.28), left inferior temporal gyrus (ITG) (BA 37, T = 3.71), left cerebellar tonsil (T = 4.24), and right cuneus (BA 17, T = 4.02) (Fig. 1). Compared to NA, AIA showed significantly higher activation in the right middle temporal gyrus (MTG) (BA 39, T = 3.88) and left middle frontal gyrus (MFG) (BA 10, T = 3.73) (Fig. 2a,b).
In SR–NR, compared to AIA, NA showed significantly higher activation in the right cuneus (BA 17, T = 5.28; BA 23, T = 5.08; BA 18, T = 4.13), left precuneus (BA 31, T = 4.75), left fusiform gyrus (BA 37, T = 4.73), left STG (BA 22, T = 4.06), and left lingual gyrus (BA 19, T = 3.97) (Fig. 1). Compared to NA, AIA showed significantly higher activation in the right MTG (BA 21, T = 4.11) (Fig. 2c).
In MR–NR, compared to AIA, NA showed significantly higher activation in the right medial frontal gyrus (BA 6, T = 5.34), left medial frontal gyrus (BA 6, T = 4.72), right insula (BA 13, T = 4.92), right declive (T = 4.84), left paracentral lobule (BA 5, T = 4.53), right precentral gyrus (BA 6, T = 4.32), left precentral gyrus (BA 43, T = 4.29), and left putamen (T = 4.05) (Fig. 1). In contrast, compared to NA, AIA showed a higher level of activation in the left medial globus pallidus (T = 4.22) (Fig. 2d).
In AIA, a negative correlation was found in SR–NR between the level of activation in the left STG (BA 22, X = −59, Y = −28, Z = 6) and the duration for Internet game use per day (r = −0.789, P = 0.00047).
The most important finding was that the activation of AIA brain was significantly lower than that of NA in all three types of contrasts.
These results are similar with the previous studies about addiction.[19, 20] Additionally, our study chose an easy task and low level of reward. According to a previous study, ADHD showed significantly lower activation in the nucleus accumbens and thalamus for low monetary reward compared to the control group, but the difference between groups disappeared for high monetary reward.
The only brain imaging study in the reward system of Internet addiction conducted by Dong et al. found a significantly higher OFC activation in the adult Internet addicts group for monetary rewards. This study used the ‘win–loss condition’ and our study used a design similar to the ‘win–no-win condition’. Our results suggest that the brains of the AIA group show a significantly reduced sensitivity compared to normal adolescents, especially when there is no possibility of losing the reward earned or when the direct feedback or punishment for incorrectly performed task does not exist.
In the NA group, activation was found in a wide area, including the thalamus, striatum, insula, STG, posterior ITG, hippocampus, and cerebellum. Activation in the subcortical area suggests that feedback also activates the brain area associated with reward.[22, 23] The insula is involved in decision-making along with STG by integrating previous responses and outcomes. Meanwhile, the activated STG area in our study is BA 38 area. This area was activated in the task in which the participant's own voice was provided as a feedback. The hippocampus is also associated with self-referential memory.
In the AIA group, significant activation in the dorsolateral prefrontal (DLPFC) (BA 10) and posterior temporal area (BA 39) was observed in between-group analyses. DLPFC was not activated even when the threshold was adjusted to P < 0.001, k > 10 in NA. This area serves an important function in motor planning, regulation, and action and was also implicated in conflict resolution.[28, 29] Mayer et al. reported hyperactivation of the cognitive control network, including DLPFC during Stroop task in cocaine use disorder patients, and their results suggest that compensatory activation within the cognitive control network in cocaine use disorders achieves similar levels of behavioral performance. Therefore, our results suggest that the AIA group only focuses on error monitoring when performing the PF task. AIA would not regard PF as a kind of reward.
In the NA group, the SR-related area was found to be the precuneus, cuneus, the area surrounding the temporo-parieto-occipital (TPO) junction, STG (BA 22), fusiform gyrus (BA 37), and posterior cingulate (BA 30).
The posterior cingulate is associated with social reward. In addition, this area has been reported to be a self-referential area along with the TPO junction. In particular, it is a part of the cortical midline structure, which is the neural correlate of the self. Thus, activation of the self-related area in SR contrast was remarkably intense in the NA group.
In the AIA group, very weak activation was observed in the right posterior MTG (BA 21) only in the between-group analyses. AIA have been reported to show depressed mood or low self-esteem. We think that the decreased brain activity in response to compliments and encouragement in AIA might be due to low self-esteem and decreased confidence.
In the NA group, wide activation was not observed, but signs of activation were observed in the subcortical area, including the pons, midbrain, and putamen, as well as the insula, dorsal medial frontal area (BA 6), precentral gyrus, and adjacent areas.
Activation of the subcortical area for MR has been reported frequently in previous studies. In addition, intense activation of the BA 6, including the dorsomedial and precentral areas, was observed in MR only. This area might be related to the ‘embodiment’ of external situations. Another area related to embodiment is the insula, which integrates the interoceptive state into conscious feeling. The dorsomedial frontal area is also associated with self-referential thought.
Only a small area of the cerebellum and medial globus pallidus were activated in the AIA group. Unlike the NA group, only certain areas of the brain associated with immediate rewards were activated and a reduction in overall brain activity was observed in the AIA group.
In the SR–NR contrast, the AIA group showed a negative correlation between the activation in the left STG (BA 22) and the duration of Internet games per day. As previously mentioned, this area corresponds to the area associated with self-referential memory and inner speech. In addition, this area is associated with temporal action planning and judgment tasks. So, the positive effect of compliments and encouragement on enhancing planning and judgment skills might be severely impaired in AIA subjects, if they overly immerse themselves in Internet games.
We divided the participants into the AIA and the NA groups in accordance with the scores for K-AIAS. If a structured interview about Internet addiction to assess withdrawal symptoms and daily-life impairments was additionally adopted, the distinction between AIA and NA might be clearer. Although the structured interview form about Internet addiction was developed in Korea, its validity has not yet been clearly proven. Also, there have been no studies adopting the structured interview about Internet addiction disorder among the 18 brain imaging studies about Internet and gaming addiction. All those studies used the scales about Internet addiction. We tried to distinguish clearly between the AIA and the NA group by setting a criterion that subjects with scores of 50 and above were categorized as AIA and those scoring below 40 were NA according to K-AIAS.
First, the results about NA on within-group analyses in our study were quite different from the results in the study of Choi et al. However, there are some different points between the two studies. The average IQ (108.93 ± 14.01) of participants in the study by Choi et al. was higher than that of the NA in our study. The number of 15-year-old participants in Choi et al. was smaller than that of the NA in our study. The size of smoothing kernel and the threshold conditions for analyses were also different. Second, although reward seeking or recognizing would be modulated by personality or temperament, our study did not consider this point. Third, although we used SPM2 for the analyses, it is an old type of software. Also, the adopted thresholds were uncorrected ones and the adopted cluster size was relatively small. Fourth, in spite of many reports about the relation between rewards and orbitofrontal region, our results did not show OFC activation. Therefore, replication researches overcoming these limitations of our study would be needed in the future.
Regardless of various rewards or feedback for easy tasks, activation in the reward-related subcortical area, the self-related area, and other areas were significantly reduced in AIA compared to NA, and AIA might be slightly sensitive to error monitoring only as evidenced by significant activation in the DLPFC area in the PF contrast. That is, they might have little experience of positive feeling, such as sense of satisfaction or achievement in their life.
There are several limitations of this study. But this was the first study to examine the brain responses in AIA on various rewards and feedback. We expect this kind of study to be widespread from now on, which will help people's comprehensive understanding about AIA and developing new treatment methods.
The authors report no biomedical financial interests or potential conflicts of interest. The authors gratefully acknowledge the assistance of all persons and volunteers whose participation was essential in the successful completion of the study. This study was supported by the research grant of the Chungbuk National University in 2011.