Internet use (IU) is a widespread and unique feature of modern human society. IU can involve various activities, such as obtaining information, gaming, online shopping, and communications via social networking services (SNS). Many aspects of daily life have been dramatically facilitated by IU.

A number of studies have suggested that excessive IU (EIU), particularly internet addiction and internet gaming disorder (American Psychiatric Association, 2013), have negative impacts on mental status (Chen et al., 2018). Among these two types of EIU, internet gaming disorders are particularly prevalent among younger populations, including adolescents, in East Asia (Liu J. et al., 2017).

Excessive IU is now regarded as a behavioral addiction that shares common pathophysiological mechanisms with compulsive shopping and pathological gambling. The reward-deficiency syndrome (RDS) hypothesis is the most influential predictive framework regarding the development of addiction, including behavioral addiction (Blum et al., 2008, 2014). The RDS hypothesis proposes that reduced dopaminergic neurotransmission, based on genetic background, underlies addiction (Blum et al., 2008).

Recent structural neuroimaging studies have demonstrated gray matter volume reduction in prefrontal brain regions including the orbitofrontal cortex (OFC) (Jin et al., 2016; Wang Z. et al., 2018), dorsolateral prefrontal cortex (DLPFC), anterior cingulate cortex (ACC), supplementary motor area (SMA) (Jin et al., 2016), and temporo-parietal regions (Wang Z. et al., 2018). A study has also reported decreased cortical thickness in the OFC and temporo-parietal regions (Wang Z. et al., 2018) in people with EIU. Additionally, diffusion tensor imaging has revealed lower structural connectivity in the ventral tegmental area (VTA)-nucleus accumbens (NAcc) tracts of people with EIU (Wang R. et al., 2018). These results indicate structural deficits in brain regions that are part of the reward system. These structural deficits have been demonstrated by different measures and imaging modalities, and are also consistent with positron emission tomography (PET) observations of decreased dopamine (DA) D₂ receptor availability in the striatum of people with EIU, which indicates an attenuated reward system (Kim et al., 2011; Tian et al., 2014). A functional magnetic resonance imaging (fMRI) study also demonstrated the involvement of reward-system deficiency in people who frequently play online games, reporting lower brain activity in the ventral striatum when anticipating small or large monetary rewards (Hahn et al., 2014).

Resting-state fMRI (rs-fMRI) has revealed a number of networks that are consistently found in healthy people and represent specific patterns of synchronous activity (Rosazza and Minati, 2011). In addition to task-based fMRI, evaluation of resting-state functional connectivity (RSFC) provides an opportunity to characterize distributed circuit abnormalities in EIU (Ko et al., 2015; Wang et al., 2017). Jin et al. (2016) conducted a rs-fMRI analysis using brain regions in which volume reductions were found (OFC, DLPFC, ACC, and SMA) as the seeding areas and found that people with EIU exhibited lower functional connectivity (FC) between the seeds and several cortical regions, including temporal cortices, occipital cortices, and the insula (Ins). Moreover, they also found that EIU was associated with significantly lower FC between the seeds and subcortical regions, including the dorsal striatum, pallidum, and thalamus. Finally, some of these differences were associated with the severity of EIU (Jin et al., 2016). Another rs-fMRI study also demonstrated lower FC between VTA-NAcc and the OFC within reward circuitry (Wang R. et al., 2018) and lower RSFC between the amygdala and the DLPFC (Ko et al., 2015). Taken together, these findings from structural and functional imaging studies indicate that reward-system function is attenuated in EIU.

Key regions of the reward system constitute a functional network called the “motivation network” (Kinnison et al., 2012). This network includes brain regions activated by a reward-related paradigm (Padmala and Pessoa, 2011). It consists of 22 regions of interest (ROIs) that are mutually well-synchronized; the bilateral intraparietal sulcus (IPS), medial prefrontal cortex (MPFC), frontal eye field (FEF), middle frontal gyrus (MFG), anterior insula (aIns), midbrain (MB), putamen (Put), caudate (Caud), NAcc, left inferior parietal lobule (IPL), right rostral anterior cingulate cortex (rACC), SMA, and left precentral gyrus (PCG). In addition to midbrain and basal ganglia that constitute the central core of the reward system, these ROIs include several cortical regions that are well-synchronized with the core regions (Kinnison et al., 2012). Although the studies mentioned above have suggested abnormalities in multiple regions of the reward system in EIU, they have not explicitly investigated FC within the reward/motivation network.

At the same time, according to epidemiological surveys and their meta-analysis, EIU is associated with a variety of psychiatric disorders, typically depression, attention-deficit hyperactivity disorder, anxiety, obsessive-compulsive disorder, and autism spectrum disorder (ASD) (Ko et al., 2008; Carli et al., 2013; Ho et al., 2014; Nakayama et al., 2017; Kim et al., 2018; Kitazawa et al., 2018). Among these comorbidities, depression is the most commonly reported psychiatric feature. Kitazawa et al. (2018) demonstrated the relationship between EIU and depressive state in a general population of university students. Additionally, the short alleles of the serotonin transporter (5HTTLPR) polymorphism are known to be associated with depression (Wrase et al., 2006), and this common polymorphism was also demonstrated in EIU (Lee et al., 2008). Further, Kraut et al. (2002) introduced a “rich get richer” model in which the internet provides more benefits to those who are already well-adjusted, and in contrast, poorly adjusted adolescents with depression may suffer more deleterious effects from heavy IU, thereby creating a vicious circle.

Evidence also suggests a possible relationship between EIU and ASD (MacMullin et al., 2016; So et al., 2017), including subclinical autistic traits (Finkenauer et al., 2012; Romano et al., 2013, 2014; Liu S. et al., 2017). MacMullin et al. (2016) reported that individuals with ASD had greater compulsive IU than individuals without ASD. Finkenauer et al. (2012) reported that adults with high autistic traits could be accurately predicted to have compulsive IU. This finding is supported by several reports demonstrating positive correlations between IU and autistic traits in college students (Romano et al., 2013, 2014).

From a therapeutic point of view, treating psychiatric comorbidities of EIU is just as important as dealing with EIU itself. Nakayama et al. (2017) reported that psychosocial therapies (including cognitive behavioral therapy) for EIU and pharmacotherapies (including antidepressants) for comorbid psychiatric or developmental disorders have been effective at reducing the degree and symptoms of EIU. Likewise, Ko et al. (2008) suggested that psychiatric disorders comorbid with EIU should be evaluated and treated to prevent the worsening of EIU prognoses. They also proposed that recognizing the possibility of EIU is important when treating people with psychiatric disorders. In the current study, among the many possible comorbidities, we focused on depression (the most common comorbidity) and autistic traits that might be associated with restricted or repetitive/compulsive IU.

Despite the experimental evidence regarding EIU described above, the psychological correlates and neural underpinning of normal/adequate levels of IU have received relatively little research attention. Although the neural and psychological correlates of non-clinical IU are still unclear, we assumed two hypotheses: (1) the mechanisms of non-clinical IU are continuous with those of EIU, suggesting that IU, even at a mild level, constitutes a potential prodromal stage of EIU; and (2) the mechanisms of non-clinical IU are different from those of EIU, suggesting that, if not excessive, IU might have neutral effects or health benefits despite the potential for developing into a pathology. For example, a brain volumetric study demonstrated that online social network size was positively correlated with human brain size in healthy individuals (Kanai et al., 2012), indicating IU for social interaction might positively influence our brain. Furthermore, it is currently unclear how co-existent depressive tendencies or autistic traits affect the level of non-clinical IU, as well as the relationship between IU and brain function.

In the current study, we investigated the psychological correlates and neural underpinnings of normal levels of IU using resting-state fMRI (rs-fMRI), focusing on the FC within the reward/motivation network (Kinnison et al., 2012). We also investigated whether depression and autistic traits modulate the relationship between IU and FC. We hypothesized that the degree of IU would be associated with FC within the motivation network, and that depression and autistic traits would modulate the association between IU and FC.

This is the first study to investigate the neural underpinnings of IU in healthy participants, focusing on the association between IU and the motivation network. We also investigated whether depressive states and autistic traits mediated the relationships between IU and FC values within the motivation network.

Importantly, the current results revealed a positive correlation between IU and FC values within the reward/motivation network. In past EIU studies, evidence suggested that the degree of IU is associated with various neurobiological measures. For example, reports indicate that the severity of IU is associated with lower D₂ receptor binding potential (Kim et al., 2011; Tian et al., 2014), lower neural activation during anticipation of small and large monetary reward in ventral striatum (Hahn et al., 2014), and reduced RSFC between VTA-NAcc and OFC (Wang R. et al., 2018). These findings are all in line with the concept of attenuated reward-system function. Our current finding of a positive correlation between IU and FC within the motivation network indicates that subclinical levels of IU might act to maintain reward/motivation-network integrity.

Among the reward/motivation network ROIs that we studied, MFG, MPFC, SMA, and aIns were the regions among which FC values between two regions were positively correlated with the degree of IU. Considering that the frontal cortex receives DA projection A10 from the VTA (Fujiwara et al., 2012) and synchronizes with aIns within the reward/motivation network (Kinnison et al., 2012), greater IU in healthy individuals appears to be associated with enhanced reward/motivation-network function among these regions. Although the brain parameters investigated in previous neuroimaging studies of EIU were varied, most (but not all) of them suggest weakened or disturbed function within or connected to regions of the reward/motivation system. Jin et al. (2016) performed multi-modal imaging and reported cortical gray matter volume reduction in the SMA of individuals with EIU. Using SMA as a seed region, they demonstrated lower FC between the SMA and the Ins. Wang R. et al. (2018) focused on investigating FC in the reward system and found reduced RSFC between mesolimbic and cortical regions, supporting the RDS hypothesis for EIU.

The assumption that IU will gradually develop into EIU over a continuum is reasonable. However, our finding of a positive correlation (not negative) between IU and FC might be related to the characteristics of our participants, namely, subclinical levels of IU if any, and without psychiatric comorbidities. Additionally, in most previous studies, participants were pathologically heavy users of online gaming. Therefore, the discrepancy between our current results and those from previous reports might be related to differences between light and heavy Internet users and/or differences in the reasons they used the internet, as most previous studies focused on internet gaming.

Among the GPIUS2 subscales, we found that Mood Regulation (MR) and Deficient Self-Regulation (DSR) were associated with FC within the motivation network. Using the internet for mood regulation has been suggested previously (Caplan, 2002, 2007; LaRose et al., 2003). Caplan (2002) found that MR was a significant cognitive predictor of negative outcomes associated with IU. In a later study, Caplan (2007) argued that socially anxious individuals may prefer online interaction (that is, POSI) because it represents a way to mitigate their anxiety about self-presentation in interpersonal situations. LaRose and colleagues’ work on the socio-cognitive model of unregulated IU also emphasizes the role of MR in the development of DSR (LaRose et al., 2003). DSR refers to a failure to adequately monitor one’s use, judge one’s usage behaviors, and adjust one’s pattern of use (Bandura, 1991). Thus, DSR conceptually represents a higher-order construct that reflects the interplay between compulsive behavior (corresponding to CU) and symptoms of obsessive thought (corresponding to CP) in GPIUS2 (Caplan, 2010). Intuitively, higher scores on the MR and DSR subscales would be expected to lead to negative outcomes in daily behavior and mental health. However, as long as their scores do not reach pathological levels, as we found here, higher scores on these subscales might be expected to have positive effects that encourage individuals to engage in various activities through the internet, which might increase their general levels of motivation. In contrast, the other two subscales (POSI and “Negative Outcomes”) were not associated with FC within the reward/motivation network. One possible reason is that a floor effect emerged due to the smaller mean scores and smaller variances of these two subscales.

We also asked whether depression or autistic traits mediate the association between IU and FC. Unlike other studies that have suggested significant relationships between EIU and depression (Cheng et al., 2018; Kim et al., 2018; Kitazawa et al., 2018), we did not find any correlations between BDI-II and GPIUS2 scores in the current study. Additionally, our results indicated that depression levels do not mediate (positively or negatively) the relationship between IU and FC. This discrepancy with past results could be due to the low levels of depression in the current study sample.

In contrast, AQ total score and the “Attention switching” subscale did mediate the association between IU and FC values. GPIUS2 scores were positively correlated with AQ scores (Table 3, “Effect of independent variable to mediator”), regardless of whether total or subscale scores were analyzed. AQ scores were negatively correlated with FC (Table 3, “Direct effect of mediator,” Figure 2). Importantly, correlation coefficients between IU and FC values increased after controlling for autistic traits, indicating that autistic traits at subclinical levels might have attenuated the positive effects of IU on the integrity of the motivation network. Indeed, autistic traits in non-clinical populations have been suggested to be associated with EIU (Finkenauer et al., 2012; Romano et al., 2013, 2014; Liu S. et al., 2017; So et al., 2017). Thus, non-clinical levels of IU appear to share common factors with EIU, in the sense that autistic traits negatively impact the ability to maintain and foster motivation in both situations. Because the internet is an important and growing source of information and of connectedness with others, and because it might help individuals with autistic traits by offering a safe environment in which to interact with others (Burke et al., 2010), approaches to encourage safe use of the internet should be considered. Previous research regarding IU has reported that autistic traits in the general population were associated with compulsive IU (Finkenauer et al., 2012). Other studies have reported that obsessive traits in particular (i.e., restricted and/or repetitive tendencies) were associated with EIU in clinical and non-clinical populations with greater autistic traits (MacMullin et al., 2016; Shane-Simpson et al., 2016). Lower social connectedness has been reported to be associated with EIU in individuals with non-clinical (but higher) autistic traits (Liu S. et al., 2017). Taken together, these findings suggest that exacerbation of “restricted and/or repetitive tendencies” might lead to compulsive use of the internet in situations in which individuals are socially isolated. If so, IU while maintaining various offline activities might prevent EIU. In addition, given the effect of gender on EIU characteristics (Finkenauer et al., 2012) and the relationship between EIU and decreased emotional regulation and trait-anxiety in individuals with higher autistic traits (Romano et al., 2014; Liu S. et al., 2017), individual factors should be considered when developing interventions for EIU.

The current study has several limitations that should be considered. First, we did not investigate the relationship between IU and FC in participants with a clinical disorder. Determining appropriate levels of IU in terms of mental health promotion will require further investigation of clinical cases and comparison with healthy volunteers. Second, the current study had a cross-sectional design. A longitudinal follow-up study is necessary to clarify the causal relationship between IU and FC. Finally, we did not examine the details of participant internet use, such as its purpose (e.g., internet browsing, shopping, gaming, or SNS), or the devices each participant used on a daily basis (e.g., PC vs. mobile device). Future studies are needed to determine whether the purpose or type of device used for IU have an effect on brain functional connectivity within the reward/motivation system.

We examined the neural underpinnings of IU in healthy participants and found a positive effect of IU on brain function and mental health. However, the results also revealed that the presence of autistic traits changed this relationship. A longitudinal follow-up study investigating both clinical (internet-addicted individuals) and subclinical populations is needed to clarify appropriate levels of IU.

HF conceived, designed, and conducted the experiments, acquired and analyzed the data, and drafted the manuscript. TM, SY, and HF contributed to the conception of the study, interpretation of data, and revisions for critically important intellectual content. KK, TU, and NO contributed to the design and data acquisition, interpretation of data, and drafting the manuscript. All authors approved the final manuscript for submission and agree to be accountable for all aspects of the work, including the assurance that questions related to the accuracy or integrity of any part are appropriately investigated and resolved.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.