Reinforcement Learning (RL) and Active Inference (AInf) are related computational frameworks for modeling learning and choice under uncertainty. However, differences in how they account for human behavior on established decision tasks remain unclear. To address this, participants from two samples (Taiwan and US) were asked to complete a three-armed bandit task and several affective measures. In one sample, the task further manipulated gain/loss frame and reward magnitude. Choice data were used to fit several complexity-matched RL and AInf models in order to: 1) perform model comparison; 2) examine relationships between parameters of complexity-matched models and identify the unique choice patterns predicted by each model; and 3) compare external validity of the best-fit models by examining task manipulation effects on parameter estimates and associations with other psychological measures. Bayesian model comparison favored AInf models in both samples. However, RL and AInf models showed similar accuracy in predicting participant choices. Correlations in parameter estimates between models suggested both explained behavior in overlapping but partially distinct ways. Trial-wise analyses suggested RL better explained some patterns of exploitation behavior (stay choices), while AInf appeared to better explain specific patterns of exploratory behavior (switch choices). AInf parameters showed relatively greater sensitivity to task conditions in some cases, while both models showed similar effect size associations with other psychological measures. These results suggest that RL and AInf have similar levels of explanatory power with respect to behavior on this task. Yet, they provide partially distinct explanations and may offer complementary insights. Future research will be necessary to extend this work to other tasks and populations. Toward this end, we provide an open-source toolbox with this article including accessible code for each RL and AInf model variant for future adaptation.

Importance:Anorexia nervosa (AN) is a deadly psychiatric disorder with relapse rates approaching 50% after weight restoration. Disrupted gastrointestinal interoception may underlie persistent symptoms and relapse vulnerability. Objective:To examine behavioral, computational, neural, and physiological markers of gastrointestinal interoception in weight-restored individuals with AN and test their association with relapse. Design, Setting, and Participants:This crossover trial was a single-blind, within-participant, randomized (block-order) trial conducted at the Laureate Institute for Brain Research between August 2021 and February 2025. Participants were females with weight-restored restrictive AN and age- and sex-matched healthy comparators (HCs). All participants ingested a vibrating capsule that delivered counterbalanced blocks of normal- and enhanced-intensity gut stimulation. Behavioral detection performance, electroencephalography, peripheral physiology, and computational modeling were used to assess interoception. Main Outcomes and Measures:Experimental-session measures included interoceptive accuracy, prior beliefs, interoceptive precision, learning rates, gastric-evoked potentials (GEPs), and hunger. Main clinical outcomes at 6 months included relapse status and symptom severity. Results:The cohort included 62 female participants with weight-restored restrictive AN (mean [SD] age, 18.9 [4.5] years) and 57 age- and sex-matched healthy comparators (HCs; mean [SD] age, 20.7 [5.3] years). Six-month follow-up data were collected for 54 participants with AN. Compared with HCs, participants with AN showed lower perceptual accuracy (Cohen d = -0.98; 95% CI, -1.51 to -0.44; P = .001) and higher miss rates (Cohen d = 1.02; 95% CI, 0.55 to 1.48; P < .001). Computational modeling revealed in the AN group stronger prior expectations that capsule vibrations would not be present (Cohen d = -0.31; 95% CI, -0.67 to 0.05; P = .05), greater shifts in interoceptive precision between blocks (Cohen d = 0.38; 95% CI, 0.02 to 0.75; P = .01), and learning asymmetries (vibration: Cohen d = -0.40; 95% CI, -0.77 to -0.04; P = .007; no-vibration: Cohen d = 0.35; 95% CI, -0.02 to 0.71; P = .01). GEP amplitudes did not differ by group but were correlated with accuracy and learning in AN. Capsule stimulation induced greater hunger increases in AN (interaction: η2p = 0.04, P = .04; AN: Cohen d = 0.94; 95% CI, 0.57 to 1.30; HCs: Cohen d = 0.40; 95% CI, 0.03 to 0.77). At follow-up, relapse was predicted by initial priors (odds ratio [OR], 3.82; 95% CI, 1.02 to 15.91; P = .05), response bias (OR, 5.37; 95% CI, 1.15 to 32.04; P = .04), and stomach unpleasantness (OR, 5.73; 95% CI, 1.38 to 33.5; P = .03), while eating disorder symptom severity was predicted by miss rate (β = 1.05; R2 = 0.08; P = .05), difference in interoceptive precision (β = 5.84; R2 = 0.16; P = .004), and initial priors (β = -2.99; R2 = 0.09; P = .05). Conclusions and Relevance:In this study of weight-restored females with AN, gastrointestinal interoception was disrupted across multiple domains, including reduced accuracy detecting gut signals, maladaptive priors, rigid learning, and abnormal hunger rating. Several interoceptive markers predicted relapse and symptom severity at follow-up. These findings support the use of ingestible mechanosensory probes and computational modeling as scalable tools to monitor treatment response and guide relapse prevention in eating disorders. Trial Registration:ClinicalTrials.gov Identifier: NCT05111977.

Individuals with anxiety-related disorders (AD) often sacrifice positive outcomes in order to avoid or neutralize threats. Mixed findings to date suggest that individual differences in threat-related behavior patterns may vary based on specific parameters of the behavior being measured. Repetitive threat-neutralization behaviors, such as repeated checking or reassurance-seeking, are a prominent feature of many AD but remain sparsely studied. In this preliminary study, we employed a novel fear conditioning paradigm to assess repetitive threat-neutralization behavior in adults (n=35) with and without AD. The paradigm included a threat cue (CS+) paired with shock, safety cues (CS-) never paired with shock, and safe stimuli varying in similarity to the CS+, followed by an extinction phase with no shocks. Participants were instructed that they could repeatedly tap a button to reduce risk of shock, while also reducing accumulation of reward points. Those with greater self-reported anxiety sensitivity showed greater threat expectancy and more threat-neutralization behavior to the CS+ while under true threat, and during extinction in the absence of threat. Those with versus without AD showed more threat expectancy and threat-neutralization during extinction only. We discuss clinical implications and future directions for assessment of threat-neutralization behaviors.

IntroductionExisting experimental threat-related paradigms focus primarily on active or passive avoidance behavior, but do not model the common behavioral pattern of repetitive, effortful actions aimed at neutralizing perceived threats. Here, we describe and provide initial validation for the Tap-To-Safety Task, a novel human paradigm designed to experimentally elicit repetitive threat-neutralization behavior during functional magnetic resonance imaging (fMRI).MethodsAdult participants completed the Tap-To-Safety Task; one sample completed the task online and an additional sample completed the task in person, with a subsample completing fMRI. Task stimuli included a threat cue (CS+) paired with an aversive unconditioned stimulus (US), safety cues (CS-) never paired with the US, and safe generalization stimuli (GSs) varying in similarity to the CS+. During an extinction phase, the CS+ was no longer paired with the US. Trials included passive viewing trials, without a neutralization option; and choice trials, in which participants could tap a button repeatedly to gain protection from the US (i.e., repetitive threat-neutralization) while reducing accumulation of reward points. Linear mixed-effects models were used to assess behavioral and neural responses. For fMRI analyses in a subset of participants, a priori regions of interest (ROIs) were used with Bonferroni correction.ResultsBehavioral results (n=49) demonstrated increased threat expectancy, anxiety, and repetitive threat-neutralization behavior were higher to the threat cue than to safety cues (ps<.001, ηp2>.42), and generalized across safe stimuli resembling the threat-cue (ps<.001, ηp2>.42). During extinction, risk and anxiety ratings gradually decreased (ps<.015, ηp2>.01), whereas neutralization behavior persisted (p=.10). Greater trial-wise neutralization predicted lower post-neutralization threat-expectancy and anxiety ratings (ps<.005, ηp2>.11). Behavioral results were largely replicated in an online sample (n=89). Analyses of fMRI data (n=31) indicated that neural activity pre-neutralization in anterior insula, dorsal anterior cingulate cortex, and dorsal striatum scaled with threat-relevance of stimuli (ps<.001, ηp2 >.30) and with magnitude of neutralization (ps<.003, ηp2 >.06). ConclusionThese findings support the use of the Tap-To-Safety Task for quantifying the behavioral and neural mechanisms of repetitive threat-neutralization. Results point to a key role of the salience network and dorsal striatum. Future research in clinical populations is warranted.

Depression and anxiety are common, highly co-morbid conditions associated with maladaptive learning and decision-making processes. While the computational mechanisms underlying these deficits have received growing attention, the transdiagnostic vs. diagnosis-specific nature of these mechanisms remains insufficiently characterized. In this discovery-focused study, we aim to better characterize these mechanisms and generate novel hypotheses. To do so, we employed a commonly used, domain-general decision-making task, combined with computational models of learning, to assess differences in patterns of choice and reaction times in individuals with affective disorders (iADs; i.e., depression with or without co-morbid anxiety; N = 168 and 74, respectively). To establish diagnostic specificity, we further incorporated data from individuals with substance use disorders (iSUDs; N = 147) and healthy comparisons (HCs; N = 54). Computational modeling afforded separate measures of learning and forgetting rates, among other parameters. Bayesian analyses indicated that forgetting rates (reflecting recency bias) were elevated in both iADs and iSUDs compared to HCs (posterior probabilities [pp] = 0.99 and 1, respectively). In contrast, iADs showed faster learning rates for negative outcomes than iSUDs (pp = 0.98), but they did not differ from HCs. Reaction times in iSUDs also showed less sensitivity to uncertainty than both iADs and HCs using model-based metrics. Finally, exploratory dimensional analyses suggested possible links between learning rates for negative outcomes and early adversity. These findings demonstrate two model-based metrics that differentiate iADs from iSUDs (learning from negative outcomes and sensitivity to uncertainty) as well as a third metric (forgetting rate) that appears transdiagnostic, differentiating both disorders from HCs. This pattern of results points to distinct cognitive mechanisms that could inform disease models for each disorder and paves the way for future work investigating their potential clinical utility.

Effective emotion regulation plays a critical role in mental health. Identifying factors that contribute to this ability could have clinical relevance. Two candidate factors are emotion recognition and emotional awareness. Specifically, if one does not recognize emotions in the self or others-and does not have a fine-grained understanding (trait awareness) of those emotions-it may be difficult to identify effective regulation strategies. Here, we tested whether individual differences in awareness moderate the association between recognition and regulation. Two independent samples of participants (Sample 1: N = 375, 72.8% female, 63.2% White [collected between 2018 and 2020]; Sample 2: N = 196, 67.3% female, 63.8% White [collected between 2020 and 2021]) completed a battery of commonly used self-report and performance-based (PB) measures of each construct. Self-report measures showed the expected pattern of significant positive associations; however, correlations between self-report and PB measures were largely absent. In both samples, two self-reported aspects of emotional awareness-greater emotion vocabulary and a better understanding of how emotions relate to bodily sensations-were significant moderators. In particular, they each strengthened the link between better self-focused emotion recognition (interoception, internal attention) and greater tendencies to use mindfulness-based regulation strategies. Novel positive associations were also observed between PB measures of each construct (e.g., tests of emotion recognition in faces/voices, emotion concept knowledge, and choice of regulation responses within hypothetical scenarios). However, hypothesized moderation patterns in PB measures could not be replicated. These results highlight new pathways and specific measurement approaches that might be utilized in future studies aiming to improve mental health and well-being. (PsycInfo Database Record (c) 2026 APA, all rights reserved).

Interoceptive interventions offer a promising avenue for improving mental health conditions, which commonly feature bodily or interoceptive symptoms. Perceptual accuracy for interoceptive signals, such as heartbeats, varies across individuals and presents a potential target for treatment. Adult participants (N = 28, 20F) completed eight sessions of a cardiac interoception training protocol, and their anxiety reduction was compared to that in a passive control group (N = 26, 22F). Bayesian computational models were compared to identify mechanisms of perception and learning that best explained participants' responses during the heartbeat discrimination task. Parameter estimates from the best-fitting model were used as computational phenotypes to explain anxiety reduction due to interoceptive training. Interoceptive training improved perceptual accuracy in two tasks of heartbeat perception and reduced self-reported trait anxiety. Computational modelling indicated that accuracy improvement in the heartbeat discrimination task was explained by increases in the internal reliability estimate for interoceptive signals - their precision weighting - while a lower-level parameter representing stable sensory noise moderated this precision weighting improvement by influencing the speed of learning. Reductions in both state and trait anxiety scores in the training group were uniquely explained by computational parameter estimates, and not by conventional accuracy measures. These findings indicate that cardiac interoceptive accuracy is modifiable and can be targeted to alleviate anxiety symptoms, and that interoceptive interventions may be best guided by a computational phenotyping approach.

Biological systems must regulate competing needs under limited perceptual bandwidth, where sharpening one estimate costs the capacity to sharpen the others. Any fixed-budget system therefore has to decide where to allocate its perceptual precision. We study this in a foraging agent that must keep several bodily needs satisfied to survive, modelled with active inference. At each step it reads its own body-state beliefs, identifies the most-needed channel, and reallocates a fixed budget of interoceptive precision toward it, so that the same precision-shaped likelihood feeds both belief update and planning. In AffectWorld, a four-channel foraging gridworld, this selective allocation more than doubles learning-phase survival at matched budget against a uniform-precision agent (0.414 vs 0.199 across 11 layouts, n=32 seeds each, paired cluster-bootstrap p ≤ 10^-4). Two further results sharpen the mechanism. The benefit runs through planning as well as perception, since denying the shaped likelihood to the planner alone removes about half of it. It is also need-aligned, since aiming precision at the least-needed channel does worse than spreading it evenly. The attended channel additionally learns its own dynamics about twice as fast, and stays ahead even at matched observation count, a behavioural trace of the same precision routing, visible in learning speed, not survival.

Threat-related behaviors are highly relevant to anxiety-related and obsessive-compulsive disorders. However, many such behaviors are not well-modeled by existing experimental paradigms. Here, we describe and provide initial validation for a novel paradigm, termed the Tap-To-Safety Task, designed to elicit repetitive threat-neutralization (RTN) behavior during functional magnetic resonance imaging (fMRI). Stimuli included a threat-cue (CS+) paired with shock, safety-cues (CS-) never paired with shock, and generalization stimuli (GSs) forming a continuum of similarity between CS+ and CS-, also never paired with shock. During an extinction phase, the CS+ was no longer paired with shock. On choice trials, participants could tap a button repeatedly to gain protection from shock (i.e., RTN) while relinquishing reward points. Adult participants (n=49) demonstrated increased threat-expectancy, anxiety, and RTN behavior to the threat-cue (ps&lt;.001, ηp2&gt;.42) which generalized across safe stimuli resembling the threat-cue (ps&lt;.001, ηp2&gt;.42). During extinction, risk and anxiety ratings gradually decreased (ps&lt;.015, ηp2&gt;.01), whereas RTN behavior persisted. Greater trial-wise RTN behavior was associated with lower post-RTN threat-expectancy and anxiety ratings (ps&lt;.005, ηp2&gt;.11). These results support the use of the Tap-To-Safety Task in eliciting repetitive threat-neutralization behavior. Future research is warranted to examine the clinical relevance of the task.

Current theories suggest individuals with methamphetamine use disorder (iMUDs) have difficulty considering long-term outcomes in decision-making, which could contribute to risk of relapse. Aversive interoceptive states (e.g., stress, withdrawal) are also known to increase this risk. The present study analyzed computational mechanisms of planning in iMUDs, and examined the potential impact of an aversive interoceptive state induction. A group of 40 iMUDs and 49 healthy participants completed two runs of a multi-step planning task, with and without an anxiogenic breathing resistance manipulation. Computational modeling revealed that iMUDs had selective difficulty identifying the best overall plan when this required enduring negative short-term outcomes - a mechanism referred to as aversive pruning. Increases in reported craving before and after the induction also predicted greater aversive pruning in iMUDs. These results highlight aversive pruning deficits as a novel mechanism that could promote poor choice in recovering iMUDs and create vulnerability to relapse.

Studying psychiatric illness has often been limited by difficulties in connecting symptoms and behavior to neurobiology. Computational psychiatry approaches promise to bridge this gap by providing formal accounts of the latent information processing changes that underlie the development and maintenance of psychiatric phenomena. Models based on these theories generate individual-level parameter estimates which can then be tested for relationships to neurobiology. In this review, we explore computational modelling approaches to one key aspect of health and illness: affect. We discuss strengths and limitations of key approaches to modelling affect, with a focus on reinforcement learning, active inference, the hierarchical gaussian filter, and drift-diffusion models. We find that, in this literature, affect is an important source of modulation in decision making, and has a bidirectional influence on how individuals infer both internal and external states. Highlighting the potential role of affect in information processing changes underlying symptom development, we extend an existing model of psychosis, where affective changes are influenced by increasing cortical noise and consequent increases in either perceived environmental instability or expected noise in sensory input, becoming part of a self-reinforcing process generating negatively valenced, over-weighted priors underlying positive symptom development. We then provide testable predictions from this model at computational, neurobiological, and phenomenological levels of description.

The drive to seek information through exploratory behavior is widespread in both humans and other animals. This can be adaptive in reducing uncertainty about the best course of action within novel or changing environments. However, exploratory behaviors can also become maladaptive if subjective uncertainty levels remain too high or too low, as may happen in states of elevated anxiety. In this article, we review recent studies investigating the influence of anxiety on information-seeking behavior. We focus primarily on studies using cognitive computational models and associated behavioral tasks designed to test specific exploratory strategies, which could each be affected by anxiety in distinct ways. Results of current studies remain mixed and highlight the importance of distinguishing potential effects of task, state vs. trait anxiety, somatic vs. cognitive anxiety, and clinical vs. sub-clinical anxiety. There are also a range of different information-seeking strategies that are necessary to consider. At present, many findings could be taken to support a picture in which cognitive anxiety, and/or trait anxiety more broadly, may increase information-seeking, while somatic and/or state anxiety could have opposing effects. However, a number of previous results also appear inconsistent or task-dependent. Future studies are needed to resolve these apparent inconsistencies and more directly disentangle effects of different dimensions of anxiety on the adaptive and maladaptive use of information-seeking.

Psychiatric disorders are highly heterogeneous and often co-morbid, posing specific challenges for effective treatment. Recently, computational modeling has emerged as a promising approach for characterizing sources of this heterogeneity, which could potentially aid in clinical differentiation. In this study, we tested whether computational mechanisms of decision-making under approach-avoidance conflict (AAC) – where behavior is expected to have both positive and negative outcomes – may have utility in this regard. We first carried out a set of pre-registered modeling analyses in a sample of 480 individuals who completed an established AAC task. These analyses aimed to replicate cross-sectional and longitudinal results from a prior dataset (N = 478) – suggesting that mechanisms of decision uncertainty (DU) and emotion conflict (EC) differentiate individuals with depression, anxiety, substance use disorders, and healthy comparisons. We then combined the prior and current datasets and employed a stacked machine learning approach to assess whether these computational measures could successfully perform out-of-sample classification between diagnostic groups. This revealed above-chance differentiation between affective and substance use disorders (balanced accuracy > 0.688), both in the presence and absence of co-morbidities. These results demonstrate the predictive utility of computational measures in characterizing distinct mechanisms of psychopathology and may point to novel treatment targets.

Repetitive threat-neutralization is a common behavior pattern across anxiety disorders and obsessive-compulsive disorder (OCD) that contributes to functional impairment. However, laboratory studies of this behavioral phenomenon are scant. In the present study, we examined effects of anxiety-related disorders and anxiety sensitivity on repetitive threat-neutralization behavior. Adults recruited from the community (n=40) with and without anxiety-related disorders completed clinician-administered and self-reported anxiety assessments and the Tap-to-Safety Task. Task stimuli included a threat cue (CS+) paired with shock, safety cues (CS-) never paired with shock, and safe stimuli varying in similarity to the CS+. The task included choice trials, in which participants could repeatedly tap a button to reduce risk of shock, while also reducing accumulation of reward points. In an extinction phase, shocks were no longer administered. We conducted linear mixed-effects models to examine differences related to anxiety sensitivity and anxiety disorder diagnosis in repetitive threat-neutralization behavior across stimulus types. Participants with greater anxiety sensitivity showed more adaptive neutralization behavior to the true threat-cue as well as more unnecessary neutralization to the former threat-cue. Those with anxiety-related disorders showed more neutralization behavior to the former threat-cue. Anxiety sensitivity and anxiety-related disorder status were both associated with less-steep declines in neutralization behavior across trials of extinction, consistent with the hypothesis that neutralization behaviors would persist in safe situations for anxious individuals. These findings suggest that the TTS task captures anxiety-related individual differences in extinction of repetitive threat-neutralization behavior, and provides an adaptable tool for probing threat-related behaviors.

BACKGROUND:Substance use disorders (SUDs) pose significant societal challenges, and underlying mechanisms remain poorly understood. Work within the growing field of computational psychiatry has begun to offer novel insights into these underlying mechanisms, including impairments in learning from negative outcomes and less deterministic decision-making, among others. However, the longitudinal stability and predictive utility of these computational measures remain underexplored, limiting their clinical applicability. METHODS:A confirmatory longitudinal study was conducted with 144 participants (75 with SUDs and 69 healthy comparisons [HCs]) from the Tulsa 1000 project. Participants completed a three-armed bandit task at baseline and 1-year follow-up. Computational modeling assessed parameters including learning rates and action precision, among others. Bayesian and frequentist approaches tested group differences, stability, and associations with symptom severity (Drug Abuse Screening Test [DAST] scores). Machine learning analyses also evaluated out-of-sample predictive accuracy when combining this sample with an earlier exploratory dataset (83 SUDs, 48 HCs). RESULTS:Computational measures showed moderate stability over 1 year (ICC range: 0.4-0.58). Learning rates for losses were consistently lower in individuals with SUDs than HCs (posterior probability > 0.99), replicating prior findings. Baseline computational parameters did not significantly predict follow-up DAST scores. Out-of-sample classification achieved modest accuracy (59 %, AUC = 0.62). CONCLUSION:Findings confirm moderate longitudinal stability and group differences in computational parameters, supporting their mechanistic relevance but raising questions about their predictive value. This highlights the need for experimental designs and enhanced reliability in computational psychiatry. Future work should integrate neurophysiological measures and dimensional approaches to improve clinical relevance.

Interoception, the process of detecting, perceiving, and interpreting signals from within the body, is essential for physiological regulation and adaptive behavior. A growing body of research underscores important potential links between interoceptive dysfunction and psychiatric disorders. Parallel advancements in the field of computational psychiatry have led to the development of biologically plausible models of information processing in the brain. This review surveys the current state of traditional and computational research approaches to study interoceptive processes in psychiatry. We also provide a foundational description of predominant computational approaches and theoretical models of interoception. Finally, we discuss the potential molecular foundations of interoceptive computation and consider future directions for incorporating computational models to enhance clinical insights and inform personalized treatments. We conclude that combining interoception and computational modeling approaches holds considerable promise in moving the field forward, both in addressing unresolved mechanistic questions and identifying novel potential therapeutic targets.

Active Inference is a recently developed framework for modeling decision processes under uncertainty. Over the last several years, empirical and theoretical work has begun to evaluate the strengths and weaknesses of this approach and how it might be extended and improved. One recent extension is the "sophisticated inference" (SI) algorithm, which improves performance on multi-step planning problems through a recursive decision tree search. However, little work to date has been done to compare SI to other established planning algorithms in reinforcement learning (RL). In addition, SI was developed with a focus on inference as opposed to learning. The present paper therefore has two aims. First, we compare performance of SI to Bayesian RL schemes designed to solve similar problems. Second, we present and compare an extension of SI - sophisticated learning (SL) - that more fully incorporates active learning during planning. SL maintains beliefs about how model parameters would change under the future observations expected under each policy. This allows a form of counterfactual retrospective inference in which the agent considers what could be learned from current or past observations given different future observations. To accomplish these aims, we make use of a novel, biologically inspired environment that requires an optimal balance between goal-seeking and active learning, and which was designed to highlight the problem structure for which SL offers a unique solution. This setup requires an agent to continually search an open environment for available (but changing) resources in the presence of competing affordances for information gain. Our simulations demonstrate that SL outperforms all other algorithms in this context - most notably, Bayes-adaptive RL and upper confidence bound (UCB) algorithms, which aim to solve multi-step planning problems using similar principles (i.e., directed exploration and counterfactual reasoning about belief updates given different possible actions/observations). These results provide added support for the utility of Active Inference in solving this class of biologically-relevant problems and offer added tools for testing hypotheses about human cognition.

Emotional awareness (EA) is a valuable cognitive skill relevant to understanding human behavior and social relationships. The Levels of Emotional Awareness Scale (LEAS) was developed in 1990 and has been utilized extensively in research to measure individuals' EA. However, there is no published normative information on this instrument in a U.S. sample. We report normative LEAS data (n = 381, Mage = 42.71, 51% females, 85.30 % White) for demographic variable levels (age, sex, education, socio-economic status [SES], marital status, ethnicity, religion). We also examine interactions among these variables in predicting LEAS scores. Results showed main effects of age, sex, education level, SES, ethnicity, and marital status on LEAS scores; higher scores were observed for female, younger, more educated, higher SES, white, and single participants. There were significant interactions between sex/education level, sex/marital status, and age/marital status. In conclusion, analysis of archived normative data revealed that age, sex, education, SES, and marital status uniquely impact LEAS scores; interactions between these factors also provided further insights into sources of variability. Future research in U.S. samples can utilize these normative data to better understand findings within and between specific populations of interest.