Sensation-seeking has long been defined as the pursuit of varied, novel, complex and intense experiences, even when those experiences carry potential costs. It is linked to a wide range of behaviours, from alcohol and substance use to antisocial conduct, but also to positive outcomes such as greater psychological well-being and constructive risk-taking. This duality has long puzzled researchers: the same trait can support healthy exploration in one setting and costly risk-taking in another. The new work suggests a possible reason — the consequences may depend on how stimulation value and uncertainty are jointly weighted in a given decision environment.

The study, led by Ern Wong, Tobias U. Hauser, Pietro Pietrini and Charley M. Wu, tackled a stubborn confound. In real life and in most laboratory tasks, options that promise stimulation are also uncertain: novel, exciting or poorly understood things tend to have unpredictable outcomes. Someone who repeatedly chooses a stimulating option might be attracted to the stimulation itself, or might simply be more willing to accept uncertainty, or both. Because these two quantities are usually correlated, standard designs cannot tell them apart.

The first experiment reanalysed previously published data from an operant sensation-seeking task in which participants chose between abstract stimuli associated with monetary rewards and probabilistic, non-painful electric stimulation to the hand. The researchers fit hierarchical Bayesian learning models that tracked, trial by trial, each participant’s beliefs about the probability of stimulation and the uncertainty around that probability. Because the original design made expected stimulation and uncertainty highly correlated, the team compared two single-process models: one assigning a bonus to expected stimulation and one assigning a bonus to uncertainty. Model comparison using leave-one-out cross-validation favoured the uncertainty model decisively (LOOIC 6712 versus 7673 for the expected-stimulation model). In that winning model, higher sensation-seeking scores were associated with less negative uncertainty weights — meaning that whereas most participants avoided uncertain stimulation, high sensation-seekers were more willing to sample it, even though the uncertainty carried no useful information about reward.

This reanalysis was exploratory and could not decisively separate the mechanisms. The second experiment was therefore preregistered on the Open Science Framework and designed to break the collinearity. A total of 619 participants were recruited through Prolific; after exclusions for failed attention checks and below-chance learning, the final sample comprised 527 adults aged 18 to 80. In a web-based adaptation of the task, participants completed four rounds, each split into a reward-learning phase with full feedback and a test phase in which they could self-administer a brief coloured spinning animation. Six objects per round were assigned reward levels and stimulation probabilities of 0.15, 0.50 or 0.85, with the two factors independent by design. Participants were explicitly told that stimulation did not influence rewards, so uncertainty about stimulation was non-instrumental — sampling it could not help them earn more money.

Behaviourally, sensation-seeking and impulsivity were only modestly correlated (r = 0.17), and they showed clearly different signatures. Higher sensation-seeking was associated with broader, more diffuse sampling, measured as higher Shannon entropy over choices (r = 0.12, p = 0.004), and with lower overall reward performance (r = −0.12, p = 0.005). This performance gap was stable rather than widening across the session, and it persisted even after statistically controlling for trial omissions, suggesting it was not simply fatigue or disengagement. Reaction-time analyses also ruled out a speed-accuracy trade-off. Impulsivity, by contrast, showed no reliable association with either performance or choice entropy.

Trial-level mixed-effects logistic regressions then revealed a striking opposition. Participants as a whole reliably tracked reward and avoided uncertainty: larger differences in expected reward increased the odds of choosing an option (β = 3.38), while larger differences in stimulation uncertainty decreased them (β = −1.70). But the traits moderated uncertainty in opposite directions. Higher sensation-seeking weakened avoidance of uncertain stimulation (β = 0.090), whereas higher impulsivity strengthened it (β = −0.099). A direct statistical contrast between the two interaction coefficients confirmed they differed significantly (Δβ = 0.189, p = 0.004). A three-way interaction further showed that the strongest uncertainty aversion occurred in people low in sensation-seeking and high in impulsivity, while for those high on both traits, uncertainty no longer systematically deterred choice at all.

The hierarchical Bayesian model then decomposed choices into parameters for expected reward, expected-stimulation weighting (θ), uncertainty weighting (ω), motor stickiness and softmax temperature (τ), which indexes choice stochasticity. Higher sensation-seeking was linked to greater choice randomness (a positive partial correlation with log temperature, r = 0.137, p = 0.002), suggesting that broader sampling partly reflects a less deterministic choice policy rather than pure information-seeking. Yet stochasticity alone did not capture the full picture. In exploratory follow-up regressions, sensation-seeking showed a negative θ × ω interaction (β = −0.187, p = 0.003): higher sensation-seeking was associated with greater tolerance of uncertainty when expected-stimulation weighting was low, but with greater avoidance of uncertainty when expected-stimulation weighting was high. Impulsivity showed the opposite sign of interaction (β = 0.147, p = 0.024), and a joint path model confirmed that this θ × ω association differed significantly between the two traits (Δβ = −0.284, p < 0.001).

The authors also probed whether subjective feelings about the stimulation explained these patterns. Participants rated the visual stimulation as pleasant and moderately arousing, and sensation-seeking was positively associated with arousal ratings while impulsivity was associated with lower valence and lower arousal. But neither valence nor arousal predicted the computational parameters, and including them as covariates left the critical trait dissociation intact. In other words, the distinct profiles of sensation-seeking and impulsivity were not explained by how pleasant or arousing the stimulation felt.

The findings carry important caveats. The reanalysis was exploratory and its design correlated the two key quantities; the preregistered experiment used a simplified artificial stimulus, prioritising computational identifiability over ecological richness, so it remains unknown whether the same mechanisms govern clinically salient forms of sensation-seeking such as gambling, substance use or social risk-taking. Temperature-like parameters can also absorb multiple sources of unexplained variability, which is why the authors interpret the stochasticity effect as evidence for a broader, less deterministic choice policy rather than pure random exploration. Both experiments used adult samples, so no conclusions about developmental change can be drawn. Still, the framework generates testable predictions: environments such as gambling venues, where salient audiovisual cues raise the subjective value of stimulation, might bias high sensation-seekers toward persistent engagement with predictable but costly options — a speculative interpretation the authors flag for future experimental work. What the study establishes now is more fundamental: sensation-seeking and impulsivity do not merely differ in how much risk people take, but in how the value of stimulation and the weight of uncertainty are combined in the computations that drive every choice.

Subject of Research: Computational mechanisms separating sensation-seeking from impulsivity in reward-guided choice under stimulation uncertainty

Article Title: Stimulation uncertainty reveals distinct mechanisms of sensation seeking and impulsivity

Article References: Wong, E., Hauser, T. U., Chandrasekaran, A., Pietrini, P., & Wu, C. M. (2026). Stimulation uncertainty reveals distinct mechanisms of sensation seeking and impulsivity. Communications Psychology, 4(1), Article 141. https://doi.org/10.1038/s44271-026-00525-5

Image Credits: AI Generated

DOI: 10.1038/s44271-026-00525-5

Keywords: sensation-seeking, impulsivity, uncertainty, decision-making, computational modelling, reinforcement learning, exploration, risk-taking, Bayesian modelling, behavioral economics, personality traits, Stimulation

The study recruited 128 participants, evenly split between 64 men and 64 women, and assigned each of them to one of three experimental conditions. In the first, known as the pleasant affective induction, participants viewed a series of images chosen to evoke positive feelings. In the second, the neutral affective induction, the images were deliberately unremarkable. In the third, the unpleasant affective induction, the pictures were selected to stir negative emotions. The images came from the International Affective Picture System, a standardized and extensively validated library of photographs rated for emotional valence and arousal, which has been adapted for Spanish-speaking populations by several research groups over the past two decades. This standardization matters: it means the emotional pull of each image is not guesswork but a quantified property, allowing experimenters to induce affective states with reasonable confidence that participants across conditions experienced genuinely different moods.

To verify that the inductions actually worked, the researchers administered the Positive and Negative Affect Scale, or PANAS, before and after the experimental session. This widely used instrument asks participants to rate the extent to which they are currently experiencing a range of feelings, from enthusiastic and inspired to distressed and jittery, yielding separate scores for positive and negative affect. The results confirmed that the manipulations shifted participants’ emotional states in the expected directions, with pleasant images raising positive affect and unpleasant images raising negative affect. The authors are careful to note, however, that the before-and-after measurements reflect a compound effect: they capture not only the influence of the images themselves but also the experience of completing the decision-making task that sat between the two assessments. Disentangling those two contributions is a persistent challenge in this line of research, and the study’s design acknowledges it openly.

The heart of the experiment was the Iowa Gambling Task, a legendary instrument in the psychology of decision-making. Developed in the 1990s by Antoine Bechara, Antonio Damasio, and colleagues to study patients with damage to the prefrontal cortex, the task presents participants with four decks of cards. Each card draw yields a monetary gain or loss, but the decks differ in their underlying payoff structures. Two decks are advantageous over the long run, offering modest rewards with occasional small penalties, while the other two are disastrous, dangling large rewards that are ultimately swamped by punishing losses. Healthy participants gradually learn to favor the good decks, often before they can articulate why, a phenomenon that helped inspire Damasio’s famous somatic marker hypothesis: the idea that emotional signals from the body guide our choices, and that decision-making without emotion is not purer reasoning but impaired reasoning.

Because the task’s rules are never explained, performance on the Iowa Gambling Task is typically decomposed into two phases. Early draws are considered decisions under conditions of ambiguity, when participants cannot yet know the consequences of their choices. Later draws, once the payoff structure has implicitly been learned, are treated as decisions under risk. This distinction is more than a technicality. Research by Brand, Recknor, Grabenhorst, and Bechara has shown that the two phases correlate with different cognitive abilities, with decisions under risk engaging executive functions and explicit knowledge to a greater degree. The Spanish team analyzed performance across both phases, and their statistical power analyses indicated the design was sensitive enough to detect even moderate interaction effects, with power reaching 0.92 for the decisions-under-risk model.

The headline finding is strikingly asymmetric. Men who had been shown pleasant images scored significantly higher on the task than both women in the same pleasant condition and men in the neutral and unpleasant conditions. In other words, positive affect acted as a performance enhancer, but only for male participants. Women’s scores, by contrast, remained stable across all three affective inductions. Whatever mood the images stirred up, their decision-making on the gambling task did not budge. The researchers interpret this pattern as evidence that men are more susceptible to the facilitative influence of positive affect, which appears to improve their decision-making performance, while women exhibit greater behavioral stability, suggesting their decision-making processes are less vulnerable to affective fluctuations.

Why might this be? The authors situate their findings within a rich theoretical landscape. Barbara Fredrickson’s broaden-and-build theory holds that positive emotions expand the scope of attention and cognition, encouraging exploration and flexible thinking, which could plausibly help a participant learn the hidden structure of a gambling task faster. If that mechanism operates more strongly in men, perhaps due to differences in how emotions interact with reward circuitry, the advantage becomes explicable. Hormonal evidence offers another thread: studies have linked testosterone to risk-taking on the Iowa Gambling Task specifically, and neuroimaging work has connected testosterone to activity in the orbitofrontal cortex, a region central to the somatic marker framework. Meanwhile, the stability observed in women resonates with a substantial literature on sex differences in emotion regulation. Research by McRae, Ochsner, and Gross has shown that men and women recruit overlapping but not identical neural machinery when reappraising negative images, and work on emotion regulation flexibility suggests women may deploy a broader repertoire of strategies with greater context sensitivity, potentially buffering their cognition from mood swings.

The new study also converges with, and helps to refine, a growing body of work on sex differences in the Iowa Gambling Task itself. A 2024 meta-analysis by Zanini, Picano, and Spitoni in Neuropsychology Review confirmed that men and women perform differently on the task as a rule, but the sources of that difference have remained contested. Earlier findings complicated the picture further: a 2019 study by Flores-Torres and colleagues found that induced humor improved women’s performance on the task while impairing men’s, the opposite direction of the present results for positive affect. Such contradictions underscore that affect is not a single lever. Different positive states, induced by different means, may act through different pathways, and the current study’s use of standardized affective images rather than humor provides one specific and reproducible route to the effect. What both studies share is the conclusion that sex cannot be treated as a statistical nuisance variable; it moderates how emotion and cognition interact.

The practical implications stretch well beyond the laboratory. Decision-making performance on tasks like the Iowa Gambling Task has been linked to real-world outcomes ranging from relapse in stimulant dependence to financial choices and clinical conditions such as major depressive disorder. If mood shifts translate into better or worse decisions differently across sexes, then contexts that manipulate emotion, from high-pressure workplaces to trading floors to therapeutic settings, may affect men’s and women’s choices in systematically different ways. The findings also carry methodological weight for researchers: experiments that induce moods before measuring cognition should analyze sex as a moderating variable, not merely control for it, or they risk averaging away effects that are real but opposite or absent in half their sample.

The authors are candid about their study’s limits. The sample was recruited by convenience rather than by an a priori power analysis, though subsequent sensitivity analyses showed the design could reliably detect effect sizes of a moderate magnitude. The PANAS measurements cannot fully separate the effect of the mood induction from the experience of the task itself, and the study examines a single moment in time rather than tracking how durable these mood effects might be. The dataset has been made publicly available in Mendeley Data, inviting other researchers to probe the patterns independently. Still, the core message stands with unusual clarity: emotion does not act on decision-making uniformly across people. A lift in spirits may open the doors of opportunity for some minds while leaving others untouched, and understanding why may require looking not just at the brain’s emotional circuitry, but at how that circuitry is wired differently across the population. In the ongoing effort to map the tangled relationship between feeling and thinking, sex has now earned its place as a critical variable, one that future models of affective decision-making ignore at their peril.

Subject of Research: Sex differences in how induced affective states influence decision-making performance on the Iowa Gambling Task

Article Title: Sex differences in the influence of affective state on decision-making

Article References: Martínez-Maldonado, A., Makeienko, M., Ramírez, E. S., & Martín-Azañedo, C. (2026). Sex differences in the influence of affective state on decision-making. Current Psychology, 45(20), Article 1603. https://doi.org/10.1007/s12144-026-10152-2

Image Credits: AI Generated

DOI: 10.1007/s12144-026-10152-2

Keywords: decision-making, Iowa Gambling Task, sex differences, affective state, emotion regulation, positive affect, PANAS, somatic marker hypothesis, prefrontal cortex, risk-taking, Current Psychology, mood induction

The research team recruited 60 healthy adults and asked them to complete a computerized decision-making task consisting of 50 trials. On each trial, participants faced three options. One was a safe choice, offering a modest but reliable payoff. The second was a probabilistic high-reward choice, which could deliver a large payoff but only with uncertain probability that participants had to learn from experience. The third was a costly information-seeking option: paying a small price to glimpse how the risky option was likely to behave. This design captures the classic explore–exploit dilemma in a laboratory setting, because learning about the risky option requires sampling it, yet sampling carries the danger of losses. Crucially, participants never received full descriptions of the probabilities; they had to infer them from the outcomes of their own choices, which is exactly how most real-world decisions work.

To manipulate cognitive load, the researchers used a trivia-based procedure. In the load condition, participants answered trivia questions of varying difficulty, while a control group completed the task without any trivia demands. Within the load subgroup, the trial-level analysis distinguished lower-load and higher-load trivia questions, allowing the team to examine whether harder mental work produced different decision patterns than easier mental work. The second manipulation was physiological: half of the participants received anodal transcranial direct current stimulation, or tDCS, over the left dorsolateral prefrontal cortex, a region long implicated in executive control, working memory, and the regulation of exploratory behavior. The other half received sham stimulation, which mimics the sensations of tDCS without delivering meaningful current. This produced a 2 × 2 between-subjects design crossing stimulation with cognitive-load condition.

tDCS is a non-invasive technique that passes a weak electrical current through the scalp, subtly shifting the excitability of neurons in the targeted cortical region. Anodal stimulation is generally thought to increase cortical excitability, making the underlying neurons more responsive. The left dorsolateral prefrontal cortex is a frequent target in such studies because it supports cognitive control processes that govern when we persist with familiar strategies and when we switch to alternatives. By combining this neuromodulatory approach with a behavioral load manipulation, the researchers could ask whether the effects of brain stimulation depend on how hard the brain is already working—a question that additive models of cognition would answer with a simple yes or no, but that the data answered in a more nuanced way.

The primary findings split cleanly along outcome lines. Cognitive load slowed participants down and made them less willing to take risks. When working memory was occupied by trivia, people responded more slowly and gravitated toward the safe option, consistent with the idea that mental strain pushes decision-makers toward conservative, low-effort strategies. Anodal tDCS over the left DLPFC produced the opposite pattern on the choice side: stimulated participants made more risky choices, sought more information, and alternated more frequently between the different choice types. Notably, stimulation did not significantly change reaction times, suggesting that the current influenced what people chose rather than how quickly they decided. This dissociation between choice content and response speed is important, because it implies that prefrontal neuromodulation alters the underlying decision policy rather than simply speeding up or slowing down general processing.

The most intriguing results emerged from the interaction analyses. Stimulation and cognitive load interacted significantly for two of the outcome measures: information seeking and alternation between choice types. In other words, the effect of anodal tDCS on these behaviors depended on whether participants were under trivia load, and the pattern of load effects differed between the stimulated and sham groups. For reaction time and risky choice, by contrast, the interactions were not significant, meaning the effects of load and stimulation on those measures were essentially additive. This outcome-specific pattern suggests that different components of experience-based decision making recruit the prefrontal cortex to different degrees and respond to neuromodulation under different cognitive circumstances. Information seeking—a behavior that reflects curiosity and the willingness to pay for knowledge—appears especially sensitive to the joint state of cortical excitability and mental workload.

The secondary, trial-level analysis added further texture. When the researchers compared trials immediately following higher-load trivia questions with those following lower-load questions, they found that participants took longer to decide after the harder questions. However, this trial-level load difference did not extend to the choice outcomes themselves: harder trivia did not reliably change risky choice, information seeking, or alternation on a trial-by-trial basis. This suggests that the burden of a single difficult question manifests mainly as slower deliberation, while the broader load condition shapes strategy at a more global level. The distinction between moment-to-moment cognitive strain and sustained cognitive demand may therefore be critical for understanding how mental workload influences decisions in everyday settings, from a surgeon working through a long operation to a trader processing a stream of market data.

Why should these findings matter beyond the laboratory? The explore–exploit tradeoff is one of the most fundamental problems in behavioral science, economics, and machine learning, and this study shows that its components can be pulled apart experimentally. Risk-taking, information seeking, and behavioral flexibility each responded to load and stimulation in their own way, which cautions against treating exploration as a single psychological quantity. The results also carry practical implications for the growing field of cognitive enhancement. If anodal DLPFC stimulation boosts risk-taking and information seeking in one context but interacts with mental workload in another, then claims about stimulation as a universal performance booster need refinement. The same current that encourages healthy exploration in a rested mind might behave differently when attention and memory are already taxed.

The study also speaks to clinical questions. Altered exploration and risk sensitivity are hallmarks of several psychiatric and neurological conditions, including addiction, attention-deficit disorders, and damage to the frontal lobes. Understanding how prefrontal neuromodulation shifts these behaviors in healthy adults provides a baseline for future work asking whether tDCS could help restore adaptive decision-making in clinical populations, or whether cognitive load might be used diagnostically to probe the integrity of prefrontal control systems. Because the current study involved 60 healthy adults and a single session, the authors’ conclusions are appropriately modest: they demonstrate outcome-specific joint effects of load and stimulation, including both additive and context-dependent patterns, rather than establishing long-term or therapeutic effects.

What emerges from Singh, Dutt, and Rao’s work is a picture of the deciding brain as a system whose exploratory urges, appetite for risk, and hunger for information are governed by partly independent levers. Fill the mind with trivia, and people retreat to safety and slow down. Excite the left dorsolateral prefrontal cortex with a gentle current, and they reach for the risky option, pay for information, and shuffle more freely among strategies. But the two levers do not simply add together; for some behaviors, the state of one changes how the other operates. As non-invasive brain stimulation moves closer to everyday use, studies like this one serve as a reminder that the brain’s decision circuits cannot be understood—or tuned—one factor at a time.

Subject of Research: Effects of cognitive load and prefrontal tDCS on exploration, risk-taking, and information seeking in experience-based decision making

Article Title: Cognitive load and prefrontal neuromodulation shape exploration, risk-taking, and information seeking in experience-based decision making

Article References: Singh, A., Dutt, V., & Rao, A. K. (2026). Cognitive load and prefrontal neuromodulation shape exploration, risk-taking, and information seeking in experience-based decision making. PLOS One, 21(10), e0360059. https://doi.org/10.1371/journal.pone.0360059

Image Credits: AI Generated

DOI: 10.1371/journal.pone.0360059

Keywords: cognitive load, tDCS, dorsolateral prefrontal cortex, decision making, exploration, risk-taking, information seeking, neuromodulation, working memory, PLOS One, Cognitive, load

The design of the study is what gives it its power. Rather than comparing a single patient group with controls, the team recruited 111 adolescents spanning a gradient of risk: 25 with a diagnosis of bipolar I disorder who were currently euthymic, meaning their mood was stable at the time of testing; 32 classified as high-risk, typically because of a family history of the illness; 24 deemed ultra-high risk, showing attenuated or subclinical manic symptoms identified through instruments such as the Bipolar Prodrome Symptom Scale-Prospective; and 30 healthy controls. This four-tier structure allowed the investigators to ask whether neurocognitive alterations deepen as one moves along the continuum from familial risk toward full-threshold illness — a question that cross-sectional comparisons of patients and controls alone cannot answer.

Each participant completed a battery of behavioral tasks and questionnaires targeting four cognitive-affective domains. Risk-taking was probed with the Modified Balloon Analog Risk Task, in which participants inflate virtual balloons for points while weighing the danger of an explosion, and with the Iowa Gambling Task, a classic decision-making paradigm that requires learning, often without conscious awareness, which of several card decks yields long-term gains. Impulsivity was measured behaviorally with a Go/No-Go Task, which tests the ability to suppress a prepotent motor response, and subjectively with the Barratt Impulsivity Scale-11, a widely used self-report instrument that decomposes impulsivity into attentional, motor, and non-planning facets. Reward sensitivity was assessed with the Behavioral Inhibition/Behavioral Activation Scale, which quantifies the strength of approach and avoidance motivational systems. Finally, emotional conflict resolution was tested with an Emotional Stroop Task, in which participants must name the color of words while ignoring their emotional content, forcing the brain to resolve interference between affective salience and task demands.

The statistical approach was deliberately conservative. Group comparisons used the Kruskal-Wallis test, appropriate for non-normally distributed data, with Benjamini-Hochberg false discovery rate correction to control for multiple comparisons, and the most robust findings were further checked against the stricter Bonferroni threshold. Multinomial logistic regression, adjusted for age and sex, then tested which measures best classified adolescents into the four groups, and receiver operating characteristic analysis quantified diagnostic discrimination. This layered pipeline matters because the field of psychiatric biomarkers is littered with findings that evaporate under multiple-comparison scrutiny; here, the authors explicitly flagged which results survived which level of correction.

The standout result was accuracy on the Emotional Stroop Task. Conflict accuracy — the ability to correctly respond despite emotional interference — was the most consistent discriminator in the entire study. It separated the high-risk, ultra-high-risk, and bipolar groups from healthy controls, and, crucially, it also distinguished the bipolar group from both risk groups. In other words, deficits in resolving emotional conflict appear not merely as a consequence of the illness but as a feature that intensifies across the risk continuum, making it one of the most promising candidate markers identified. The regression analysis underscored this: emotional conflict accuracy discriminated the combined ultra-high-risk and bipolar groups from the combined healthy and high-risk groups with an area under the curve of 0.923, a level of discrimination that, if replicated, would be remarkable for a single behavioral measure in psychiatry. The authors are careful to note, however, that this cut-off is exploratory and derived from a modest sample.

Decision-making under risk also told a coherent story. Lower net scores on the Iowa Gambling Task differentiated all three clinical groups — bipolar, high-risk, and ultra-high-risk — from healthy controls, suggesting that impaired affective decision-making is a broad vulnerability feature present even before illness onset. On the Modified Balloon Analog Risk Task, a higher average number of pumps — a direct behavioral index of risk-taking — differentiated the ultra-high-risk and bipolar groups from controls in the adjusted models. Notably, this elevated risk-taking in the ultra-high-risk group correlated with concurrent subclinical manic symptoms, hinting that the balloon-pumping behavior may capture something phenomenologically close to the emerging hypomanic temperament. Group differences on the Emotional Stroop, Iowa Gambling Task, Barratt Impulsivity Scale, and Go/No-Go task survived both false discovery rate and Bonferroni correction, whereas the balloon task differences survived false discovery rate correction only.

Perhaps the most clinically useful aspect of the findings is the differential pattern across impulsivity subdomains. Non-planning impulsivity, reward sensitivity, and emotional conflict resolution deficits distinguished all three clinical groups from healthy controls, positioning them as potential trait markers of vulnerability that are present from the earliest identifiable risk stage. By contrast, motor and attentional impulsivity, fun-seeking, and response inhibition deficits were more specific to the ultra-high-risk and bipolar groups, appearing only as adolescents move closer to or into full-threshold illness. This dissociation suggests a staged model of neurocognitive deterioration: a baseline vulnerability layer shared across the risk spectrum, overlaid by a later-emerging layer of inhibitory and attentional dyscontrol that may signal imminent transition. If longitudinal studies confirm this architecture, clinicians could in principle use the profile of deficits — not just their presence — to estimate where along the risk pathway a given adolescent sits.

The theoretical implications reach into long-standing models of bipolar disorder. Heightened reward sensitivity and approach motivation have been central to psychological theories of mania, and the present findings align with the idea that a hyperactive behavioral activation system, combined with impaired top-down regulation of emotional interference, creates a cognitive environment in which mood dysregulation can escalate. The anterior cingulate cortex, a region heavily implicated in conflict monitoring and error processing, is a plausible neural substrate for the Emotional Stroop deficits observed, and its dysfunction could plausibly link the affective and cognitive disturbances that characterize the bipolar prodrome. The authors themselves frame these alterations as candidate markers rather than established ones, a distinction that reflects both the strength and the limits of the evidence.

Those limits deserve emphasis. The study is cross-sectional, meaning it captures a snapshot rather than a trajectory; it cannot prove that the observed deficits predict who will convert from risk to illness, only that they covary with risk stage at a single point in time. The sample, while thoughtfully stratified, is modest — 111 adolescents across four groups — and the ultra-high-risk subgroup in particular is small. The authors also note that the striking area under the curve of 0.923 for emotional conflict accuracy should be treated as a hypothesis-generating estimate, since classification metrics derived from the same dataset used for discovery are prone to optimism. Longitudinal follow-up of high-risk and ultra-high-risk cohorts, ideally with neuroimaging and repeated cognitive assessment, will be needed to determine whether these candidate markers genuinely forecast onset and whether early intervention targeting emotional regulation could alter the course of the illness.

Even with those caveats, the study represents a meaningful step toward a long-sought goal in child and adolescent psychiatry: identifying measurable, task-based signals that stratify bipolar risk before the first full manic episode. Bipolar disorder in young people is frequently misdiagnosed, often as unipolar depression or attention deficit hyperactivity disorder, and treatment delays carry substantial costs in academic, social, and neurobiological terms. A validated cognitive profile — impaired emotional conflict resolution, compromised affective decision-making, and stage-specific impulsivity changes — could eventually complement clinical interviews and family history to sharpen risk assessment. For now, the message from Ankara and Izmir is one of cautious optimism: the cognitive fingerprints of bipolar vulnerability appear to be detectable in adolescence, arranged along a gradient that mirrors the illness itself, and waiting to be tested against the passage of time.

Subject of Research: Neurocognitive vulnerability markers across bipolar disorder risk stages in adolescents

Article Title: Neurocognitive candidate markers across bipolar disorder risk stages in adolescents

Article References: Akpınar, S., Kafali, H. Y., Kabasakal, İ., Kaya, H., Akman, A. Ö., Bora, E., & Dinç, G. Ş. (2026). Neurocognitive candidate markers across bipolar disorder risk stages in adolescents. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08732-4

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08732-4

Keywords: bipolar disorder, adolescents, ultra-high risk, emotional conflict, reward sensitivity, impulsivity, risk-taking, Iowa Gambling Task, Emotional Stroop Task, neuropsychology, prodrome, predictive markers