A Computational Model of Mental Rotation in Aphantasia
Year: 2026
Type: preprint
Abstract: People with aphantasia seem unable to create visual imagery, yet they perform mental manipulation of objects in the mind. Different frameworks have different views on how this is possible. Mental rotation is a particularly useful test case for comparing these frameworks, as people with aphantasia show systematic differences and similarities on this task compared to people without aphantasia. Here, we propose a computational model of mental rotation based on the ‘broken rendering’ account of aphantasia (Balaban & Ullman, 2025), which separates physical/spatial representations from graphical, picture-like ones. Our model proposes that both people with and without aphantasia are similar in their identification and rotation of objects, but differ in the comparison stage: People with and without aphantasia can use spatial/physical similarity to compare rotated objects, but people with typical imagery can also render a rotated object back into a 2D, pixel-like representation, and carry out a fast-but-less-reliable comparison between such visual depictions. Both sources of comparison (physical and visual) provide evidence to a single drift-diffusion decision process. We used our model to generate simulated experimental data, which we compared to the detailed findings of Kay et al., 2024. We found that our model captures the qualitative and quantitative patterns of behavior in a mental rotation task for people with and without aphantasia, including the canonical increase in response time (and decrease in accuracy) with angle growth, the slower but more accurate responses in aphantasia, and the opposite within-group correlations. This lends support to the broader framework of ‘broken rendering’ over ‘different strategies’ and ‘lack of meta-access’ as approaches to aphantasia. The findings hold across a wide set of parameter settings, and the model makes novel and testable predictions for further study in mental rotation.
Source: PsyArXiv (OSF Preprints)
Authors: Shiu-Cheng Wang, Tomer David Ullman
Language: English
Subfield: Experimental and Cognitive Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7236066600
