Internet addiction, first described by psychologist Kimberly Young in 1998 as a clinical disorder involving loss of control over online activity and interference with daily functioning, has since become a major public health concern. Recent studies report prevalence rates of 20.6 percent in the United States and 18 percent in Canada, while Spanish data place problematic Internet use among people aged 15 to 24 at between 9.5 and 11.3 percent, peaking in 2020. Social media, the dominant online activity for young adults, is engineered around behavioural reinforcement loops, notifications, likes, and algorithmic personalisation that exploit social validation and comparison, and users in Spain reportedly spend an average of nearly six hours per day on these platforms. Against this backdrop, researchers have long hypothesised that heavy digital exposure might leave detectable traces in the body’s stress and arousal systems, producing measurable differences in how heavy and light users respond physiologically to cognitive challenges.

To test that hypothesis rigorously, a team led by Vicente Villalba-Palacín of the University of Barcelona, together with Lydia Sánchez and Caterina Calderón, recruited 46 undergraduate students of Audiovisual Communication, whose mean age was just over 21 years. After one participant was excluded because of persistent technical artefacts in the physiological signal, the final analytical sample comprised 45 participants, 28 women and 17 men. Each participant completed the Internet Addiction Test, a 20-item instrument scored from 20 to 100 with high internal consistency, and was classified into either a mild or a moderate-to-severe category. Notably, 62.2 percent of the sample fell into the moderate-to-severe range, a figure consistent with current prevalence trends in university settings.

The experimental design was deliberately structured to probe the autonomic nervous system under three distinct cognitive conditions of increasing demand. In the first task, participants passively viewed 12 static news articles selected from widely read national newspapers. Static text was chosen over video to minimise uncontrolled emotional variability, and the news items were screened for content neutrality and standardised format so that the researchers could isolate baseline arousal during sustained screen fixation. In the second task, participants completed a sociodemographic questionnaire and the Internet Addiction Test itself, engaging in introspection and self-referential processing, a domain considered central to models of addiction-related dysregulation. In the third and most demanding task, participants answered three binary-choice questions requiring them to discriminate between authentic and synthetically generated news items, a judgement that, according to revised cognitive taxonomy frameworks, constitutes genuine analytical reasoning. Relaxation periods of three to five minutes were interposed between tasks to allow physiological levels to stabilise and to establish baselines.

Throughout the session, the researchers recorded electrodermal activity using a device called the Sociograph, which measures both the tonic and phasic components of the signal via the constant current method, expressed in kiloohms and sampled at 20 hertz, one measurement every 0.05 seconds. The tonic component, known as the electrodermal level, reflects general cortical arousal associated with sustained information processing, with lower resistance indicating higher activation and optimal performance typically occurring at moderate activation levels. The phasic component, the electrodermal response, captures rapid fluctuations in autonomic reactivity to specific stimuli, with larger amplitudes indicating stronger reactions. The raw dataset comprised roughly 1,143,400 data points, which were preprocessed in Python, aggregated at the participant level to avoid pseudoreplication, and converted into change scores relative to the relaxation baselines.

The statistical analysis rested on linear mixed-effects models, a framework well suited to repeated-measures data because it models within-subject dependency across tasks while remaining robust under unbalanced subgroup sizes. Fixed effects included task, Internet addiction level, and gender, along with all their interaction terms, and a random intercept for participants accounted for inter-individual variability. Parameters were estimated by restricted maximum likelihood, significance was set at an alpha of 0.05, and 95 percent confidence intervals were computed for all fixed-effect estimates.

The results were striking in their asymmetry. For tonic activity, no significant effects emerged at all, for task, addiction level, gender, or any of their interactions. Descriptive trends, such as a reduction in tonic values during introspection and a rise during the analytical task, did not survive statistical testing. Neither addiction level nor gender contributed meaningfully to variability in the tonic measure, and the random effects revealed considerable inter-individual variability, suggesting that tonic arousal is a highly heterogeneous, person-specific signal that resists group-level prediction under these conditions.

For phasic activity, however, one effect stood out clearly. The analytical task elicited a significant increase in phasic responding compared to passive viewing, with a coefficient of 0.19, a standard error of 0.08, and a p-value of 0.024, yielding a 95 percent confidence interval from 0.02 to 0.35. The introspective task, by contrast, did not differ significantly from the passive condition. In plain terms, when participants were asked to exercise analytical judgement, distinguishing real news from synthetic content, their autonomic systems responded with measurably sharper reactivity than when they were merely looking at screens or reflecting on themselves. Neither addiction severity nor gender, nor any interaction among the factors, significantly modulated this pattern, indicating that the surge in phasic reactivity was tied to what the task demanded of the mind, not to who was performing it.

The authors are careful to note that this finding must be interpreted in light of the fixed task sequence. Because the analytical task was always administered last, its effects cannot be fully disentangled from order-related influences such as fatigue, habituation, anticipatory arousal, or general session progression. The elevated phasic response may therefore reflect a combination of genuine cognitive demand and the temporal position of the task within the experiment. Even so, the magnitude and consistency of the effect, contrasted with the complete absence of addiction-related or gender-related effects across all three conditions, strengthens the paper’s central conclusion that electrodermal activity is primarily modulated by task characteristics rather than by individual differences in problematic Internet use.

The study also examined gender as a potential moderator, motivated by prior literature showing that motivational drivers for Internet use diverge by gender, with women typically prioritising emotional connection and social validation while men gravitate toward gaming and entertainment, and with men showing heightened impulsivity in digital behaviour. Descriptive differences in tonic and phasic activity between gender groups did appear during the introspective condition, but these patterns were not supported by inferential analyses once inter-subject variability and repeated-measures dependencies were accounted for. The gender imbalance in the sample, 28 women versus 17 men, further constrains the generalisability of this aspect of the findings.

The methodological implications may prove to be the study’s most durable contribution. Electrodermal activity has been validated across diverse contexts, from assessing attention and emotion during Instagram viewing to predicting cognitive load and mental effort in academic and workplace settings, and it offers an objective alternative to self-reports, which are vulnerable to subjectivity and social-desirability bias. Yet the present results suggest that researchers investigating the psychophysiological correlates of problematic Internet use, or indeed any behavioural phenomenon, must design tasks with sufficient cognitive demand to elicit measurable autonomic engagement. Under passive or minimally demanding conditions, autonomic responses appear to remain close to baseline regardless of who is being tested, rendering the measure effectively blind to the individual differences it is often deployed to detect.

The authors acknowledge several limitations beyond task ordering: the aggregation of data at the participant level reduced statistical power, particularly for higher-order interactions; the controlled, news-based stimuli, while standardised, limited ecological validity; variables such as anxiety, cognitive ability, and executive functioning were not measured; and emotional responses to digital stimuli are known to be modulated by social identification and individual regulation strategies that were not assessed here. Future research, they suggest, should incorporate more demanding and ecologically valid tasks, potentially including immersive contexts, to explore how autonomic reactivity scales with real-world digital engagement.

For a field increasingly concerned with the physiological consequences of constant connectivity, information overload, fear of missing out, and attentional fragmentation, the message is both sobering and clarifying. The body’s electrodermal signature, at least in this experimental context, does not readily distinguish heavy from light Internet users. What it does, reliably and significantly, is respond to the mind being put to work, a reminder that in psychophysiology, as in so much of science, what you find depends on what you ask the brain to do.