From behaviour to latent processes: mapping the computational architecture of the mental rotation gender gap across sport expertise

Introduction The gender gap in mental rotation (MR) is a robust individual difference in spatial cognition, yet its association with sport expertise and the cognitive mechanisms that may underlie such an association remain poorly understood. The present study examined how long-term sport practice relates to the latent cognitive architecture of spatial decision-making and to the attenuation of this traditional performance disparity. Methods Behavioral outcomes and latent processes were evaluated across 100 young adults (50 athletes, 50 non-athletes; balanced by gender; mean age = 22.91 ± 2.86 years, range: 18–27) using a multi-stage analytical framework: linear mixed models (LMM) for reaction times, generalized linear mixed models (GLMM) for accuracy, and hierarchical Bayesian drift diffusion modeling (HDDM) for a mechanistic decomposition of the processing stages underlying mental rotation. Results Behaviorally, males were faster than females among non-athletes, whereas among athletes the gender difference in response time was not significant and female athletes were more accurate than male athletes. At the latent level, athletic expertise was associated with an advantage across all latent parameters: higher evidence-accumulation efficiency (drift rate, v ), a behavioral signature consistent with the neural efficiency hypothesis; lower response caution (boundary separation, a ), consistent with a speed-emphasizing decision strategy; and shorter non-decisional components (non-decision time, t ), consistent with more streamlined perceptual–motor processing. Importantly, the gender difference in drift rate shifted in favour of females from non-athletes to athletes, so that female athletes had higher drift rates than male athletes; the corresponding change in non-decision time was credible in direction but not practically conclusive. This pattern was stable across the alternative model specifications tested, but it is still model-dependent and exploratory. Discussion Similar response speed can therefore be associated with different latent parameter configurations. Because the design is cross-sectional, these differences are interpreted as associations rather than training-induced effects; whether they reflect experience-dependent plasticity, pre-existing differences, or both remains to be tested.