Shared variance across FrACT visual tasks: Robust positive associations but inconclusive latent structure

Abstract

A long-standing question in vision science is whether individual differences in visual performance reflect a general visual ability or a set of largely task-specific processes. Previous studies have often reported weak associations across heterogeneous perceptual tasks, challenging the idea of a common visual factor. Here, we asked whether shared variance becomes more evident when visual tasks are conceptually and procedurally similar. One hundred fifty-eight adults completed eight low-level acuity and hyperacuity paradigms implemented in FrACT₁₀: single letter, Landolt C, tumbling E, row of optotypes, flanking rings, frame ring, Vernier two-line, and Vernier three-line tasks. After outlier exclusion, analyses were conducted on 132 participants. Within-session test–retest reliability was moderate for most tasks. Pairwise correlations revealed a broad positive manifold: all correlations were positive, 27 of 28 survived false-discovery-rate correction, and Bayesian analyses provided strong evidence for nonzero associations in most task pairs. However, association strength depended strongly on task similarity, with robust correlations among optotype-based tasks and weaker links between optotype and Vernier paradigms. Exploratory factor analyses indicated substantial shared covariance but failed to support a stable latent structure: A one-factor model fit poorly, whereas two-factor and three-factor solutions produced improper estimates. These findings support an intermediate account of visual individual differences. Visual performance is neither fully independent across tasks nor reducible to a strong general visual factor; instead, commonality appears partial, task-dependent, and shaped by the composition of the task battery.

The data, data dictionary, analysis code, and rendered analysis reports that support the findings of this study are available on OSF (https://doi.org/10.17605/OSF.IO/AKSN3). The study materials are described in the Method section and are based on FrACT₁₀. This study was not preregistered.