Abstract
Object size is systematically related to shape properties, like curvature—big objects tend to be boxier to withstand gravity, while small objects tend to be curvier for comfortable handling. Here we examined whether internal object representations are sensitive to this typical covariance. To do so, we created a 2×2 stimulus set of big and small, boxy and curvy objects. While big boxy and small curvy objects are more prevalent than the opposite combinations, we ensured that across items, the conditions did not differ by familiarity, and were matched in real-world size and curvy-boxy ratings, aspect ratio, and pixel area. In the behavioral experiment, participants (N=12) were presented with an isolated object and judged its real-world size (small vs large) on some blocks of trials, or its shape (curvy vs boxy) on other blocks of trials. Overall, people were faster and more accurate at judging the typical size-curvature combinations relative to the atypical combinations (mixed-effect model interactions; size judgement: t = −6.12, p< .001; curvature judgement, t = −2.59, p = 0.01). In the neuroimaging experiment, participants (N=16) viewed a subset of these images in a blocked design. The typical covariance contrast—big boxy vs. small curvy—yielded more reliable topographies than the atypical covariance contrast—big curvy vs small boxy (typical split-half map correlations: M=0.52, SD=0.13; atypical: M=0.29, SD=0.27; t(15)=3.87,p=0.002). These two topographies were moderately positively correlated (typical vs. atypical map correlation: M=0.39, SD=0.13), indicating that the size organization is not fully explained by curvature and that other covarying mid-level statistics are required to fully account for the object-size topography. Broadly, these results provide both behavioral and neural evidence that the visual system is sensitive to the natural covariance between object size and curvature, supporting more efficient behavioral judgments and more robust neural responses.