Decision Acuity and Mental Health

The Moutoussis et al. (2021) Neuron paper proposes that variation in decision-making ability is itself a measurable individual-difference dimension—called decision acuity—that is partially distinct from general intelligence and that carries its own associations with psychopathology and brain connectivity. The claim is operationalized through a substantial multi-task battery and resting-state fMRI in the Neuroscience in Psychiatry Network (NSPN) cohort, and it advances a particular computational-psychiatry program: that formal decision-modeling tasks, harvested at scale across a population, expose a latent dimension of decision quality that traditional cognitive measures do not capture.

The construct gap that decision acuity tries to fill

General cognitive ability (g) has dominated individual-differences psychology for a century, and the Wechsler-tradition IQ score remains the field’s most reliable cognitive predictor of life outcomes. But IQ tasks—Vocabulary, Matrix Reasoning, Block Design—do not directly measure how someone navigates probabilistic, delayed, or socially embedded choices. Decision-making research has accumulated dozens of task paradigms (gambling tasks, intertemporal-choice tasks, two-step reinforcement-learning tasks, social exchange tasks) that each tap distinct cognitive operations relevant to real-world decision quality, but until recently no work had asked whether scores across these paradigms factor onto a common dimension.

The computational-psychiatry program articulated by Huys, Maia, and Frank (2016) framed this gap as the central methodological project for translating decision neuroscience into clinical practice. Mental disorders, on the computational view, manifest as systematic biases in specific decision parameters—learning rates, perseveration, risk preferences, exploration versus exploitation—and a comprehensive computational characterization of an individual would map their position on each dimension. Whether those dimensions are independent or share a common factor is the empirical question that Moutoussis et al. (2021) take on directly.

A parallel ontology-discovery program by Eisenberg, Bissett, Enkavi and colleagues (2019) attempted the same kind of factor extraction across self-regulation measures using a different task battery and ultimately found a more dimensional structure with limited general-factor coverage. The two studies use different task selections and yield different factor solutions, which makes their comparison informative for the larger question of whether a unified decision-quality construct exists or whether decision-making is fundamentally heterogeneous across paradigms.

The NSPN study design

Moutoussis and colleagues recruited 830 participants ages 14-24 from the NSPN cohort, an age range deliberately spanning the developmental window in which decision-making capacities mature. Each participant completed a battery yielding 32 decision-making measures drawn from seven computational-psychiatry task paradigms:

• Go/NoGo task (8 measures: Pavlovian bias, response times, learning rates, motivational asymmetry, decision variability)
• Economic preferences task (4 measures: gambling preference, risk aversion, skewness sensitivity, expected-value sensitivity)
• Approach-avoidance conflict task (3 factor scores)
• Two-step reinforcement-learning task (5 measures: model-based weighting, learning rate, perseveration, reward sensitivity, eligibility trace)
• Information-gathering task (4 measures: sampling noise, information cost, fixed-cost variants)
• Investor-trustee social-exchange task (3 measures: initial trust, cooperativeness, responsiveness)
• Interpersonal-discounting task (5 measures: discounting coefficient, preference relevance, taste uncertainty, decision variability, irreducible noise)

A subset of 349 participants additionally completed resting-state fMRI on the same day as cognitive testing. The full cognitive and imaging protocol was repeated 18 months later in the same individuals, providing test-retest data for both the decision-acuity construct and its associated brain connectivity patterns.

The four-factor solution and the decision-acuity dimension

Factor analysis of the 32 decision measures yielded a four-factor solution. The first factor—decision acuity—drew loadings from across the task families and represented a generic decision-quality dimension that the authors interpret as reflecting speed of learning, sensitivity to cognitively distant outcomes, and low decision variability. The remaining three factors were task-specific in character: an interpersonal-discounting factor, an information-gathering factor, and an economic-risk-preference factor. These were narrower constructs that captured task-specific variance not absorbed by the general decision-acuity factor.

The substantive psychometric question is whether decision acuity is genuinely distinct from IQ. The 2021 paper provides two relevant statistics. IQ subscores combined with age accounted for r²adj = 0.31 of decision-acuity variance—substantial overlap but far from determinative. Discriminant validity, computed as D = 0.76, fell below the conventional .85 threshold, supporting the conclusion that decision acuity carries information beyond what IQ measurement captures. The two constructs share roughly a third of their variance and develop in parallel across the 14-24 age window, but they are dissociable at the population level.

Psychopathology associations

The mental-health correlates of decision acuity were specific rather than general. Adjusting for IQ, age, gender, and other covariates, decision acuity was strongly associated with adaptive sociality (β = +0.32, p < 0.001)—participants with higher decision acuity reported better social functioning. Decision acuity showed a marginal negative association with aberrant thinking (β = −0.10, p = 0.074), in the predicted direction but not statistically conclusive. Associations with general distress and worry were not reliable (p > 0.8 in both cases).

The pattern is informative. Decision acuity does not appear to be a generic mental-health-vulnerability marker—it is not associated with internalizing distress measures the way trait neuroticism would be. Instead, it is most strongly tied to social functioning and to the unusual-thought-content dimension that often appears in early-psychosis literature. This specificity argues that decision acuity captures decision-relevant cognitive variation rather than serving as a proxy for general psychopathology.

Distinct neural signatures

Resting-state functional connectivity analyses in the 349-participant fMRI subset showed that decision acuity and IQ have separable connectivity profiles. The connectivity patterns associated with decision acuity involved different network components than those associated with IQ, and the connectivity signature for decision acuity was reliably reproducible at the 18-month follow-up. The reproducibility is non-trivial: a brain-connectivity correlate that survives test-retest at 18 months in adolescents (a developmental window in which network organization is still maturing) suggests that the neural substrate is reasonably stable rather than a transient state correlate.

The distinct neural signature is the strongest argument that decision acuity is not just statistical novelty extracted from a heterogeneous task battery, but a construct with a reproducible biological correlate. The brain correlate does not, by itself, prove that decision acuity is causally upstream of decision-making behavior; it shows that the construct has the kind of trait-like neural footprint that other validated cognitive constructs have.

What decision acuity does and does not establish

The Moutoussis et al. (2021) paper makes three claims with varying empirical support.

The strongest claim, well-supported by the data, is that cross-paradigm decision-making variance has a substantial general factor. People who show high model-based reinforcement learning also tend to show low decision noise in economic-preferences tasks, low perseveration in two-step tasks, and adaptive information-gathering. The general factor is not an artifact of any single task family—it is loaded across paradigms with different cognitive demands.

A more modest claim, also well-supported, is that decision acuity is dissociable from IQ at the population level. The 31% shared-variance figure makes the dissociation real but not extreme; IQ remains a strong predictor of decision performance. The 0.76 discriminant validity is in the supportable range.

The most contested claim is the clinical-translational potential. The strong adaptive-sociality association and the marginal aberrant-thinking association point in clinically interesting directions, but the cross-sectional cohort design cannot establish whether decision acuity prospectively predicts mental-health outcomes, whether interventions on decision acuity would improve social functioning, or whether decision-acuity measurement adds incremental clinical value over standard psychiatric assessment. These remain open questions that the design enables but does not answer.

Comparison with the broader ontology-discovery program

The Eisenberg et al. (2019) self-regulation ontology project, using a different task battery and a different population (online crowdsourced sample of adults), reached a structurally different conclusion: self-regulation tasks produce a more multi-dimensional latent structure with limited general-factor variance. The disagreement between the two studies is informative rather than contradictory. Eisenberg’s task selection was weighted toward executive-control and self-regulation paradigms; Moutoussis’s was weighted toward decision-modeling paradigms with reinforcement-learning and economic-preference content. The two task collections may be tapping different aspects of cognitive function, and the latent structure each recovers is partly determined by which paradigms it includes.

The general lesson for computational-psychiatry measurement is that the latent factor structure of decision-relevant tasks is not a property of the brain alone but a property of the task selection. Moutoussis et al.’s decision acuity is a real construct in the NSPN data; whether it generalizes to a different task battery, a different population, or a different cohort age range is the next-decade research question.

Methodological implications for individual-differences cognitive science

The 2021 paper is a methodological proof-of-concept for a particular research strategy: assemble a large multi-paradigm decision-task battery, recruit a sample large enough to support stable factor extraction, estimate task-specific computational parameters using model-fitting machinery, and then ask whether the parameters factor onto interpretable dimensions. The strategy is expensive—32 measures from 7 tasks per participant in 830 participants is a substantial data-collection effort—but it produces individual-differences findings that single-task studies cannot.

The decision-acuity construct is therefore best read as the first usable output of a measurement program rather than as a finished psychometric instrument. Subsequent research will determine whether the construct replicates across cohorts, whether it has incremental predictive validity for mental-health outcomes over IQ and demographic covariates, and whether it can be measured economically with shorter task batteries. The conceptual frame—that decision-making ability is a measurable trait, partially distinct from IQ, with its own neural and clinical correlates—is a useful frame for the field whether or not the specific four-factor solution from this paper is the final word.

Frequently asked questions

What is decision acuity?

Decision acuity, as proposed by Moutoussis et al. (2021), is a latent individual-differences dimension extracted from a multi-task decision battery. It captures speed of learning, sensitivity to cognitively distant outcomes, and low decision variability across paradigms. The construct loads across reinforcement-learning, economic, social-exchange, and information-gathering tasks rather than being task-specific.

How is decision acuity different from IQ?

IQ subscores combined with age account for r²adj = 0.31 of decision-acuity variance, and discriminant validity is D = 0.76 (below the .85 threshold). The two constructs share roughly a third of their variance and develop in parallel across the 14–24 age window, but they are dissociable at the population level. Decision acuity also has a separable resting-state functional connectivity signature.

What was the NSPN sample?

The Neuroscience in Psychiatry Network cohort study recruited 830 participants ages 14–24, of whom 349 completed resting-state fMRI on the same day as cognitive testing. The full protocol was repeated 18 months later, providing test-retest data for both decision acuity and its associated brain connectivity patterns.

What does decision acuity predict in psychopathology?

Adjusted for IQ, age, and gender, decision acuity is strongly associated with adaptive sociality (β = +0.32, p < 0.001), marginally negatively associated with aberrant thinking (β = −0.10, p = 0.074), and not reliably associated with general distress or worry. The pattern argues that decision acuity captures decision-relevant cognitive variation rather than serving as a proxy for general psychopathology.

How does this relate to computational psychiatry?

The computational-psychiatry program articulated by Huys et al. (2016) frames mental disorders as systematic biases in decision parameters such as learning rates, perseveration, risk preferences, and exploration-exploitation balance. The Moutoussis et al. (2021) paper tests whether those parameters factor onto a common dimension and demonstrates that, within their task selection, they substantially do.

Does decision acuity generalize beyond this study?

Not yet established. Eisenberg et al. (2019) used a different task selection (weighted toward executive-control and self-regulation paradigms) and recovered a more multi-dimensional latent structure with limited general-factor variance. The latent factor structure depends on which paradigms a battery includes, so cross-cohort and cross-battery replication is the next-decade research question.