Non-commitment in mental imagery is distinct from perceptual inattention, and supports hierarchical scene construction

Resumen de los autores

When calling an image before the mind's eye, people often do not include basic properties of the mental scene. For example, when imagining a person knocking a ball off a table, people may neglect to consider the color of the ball. This "non-commitment" in imagery has been part of ongoing debates on the nature of mental imagery in philosophy and cognitive science, but the central question of its origin remains open. It is possible that this lack of detail in the imagination reflects actual non-commitment, as one constructs a hierarchical mental scene. But it is also possible that people do form highly detailed mental images, but fail to notice, encode, or recall properties of the image, for the same reasons that they do not notice, encode, or recall details of real scenes in direct perception. We tested these contrasting hypotheses across three experiments (N=3,895 total). First, we compared non-commitment patterns in mental imagery with the details that participants failed to notice in briefly presented real images. We found no correlation between properties that were non-committed to in the imagination, and those missed under rapid perception, indicating that non-commitment does not simply arise from perceptual noise. Next, we examined the temporal unfolding of mental imagery, and found consistent evidence of a hierarchical pattern: core entities and spatial relations, then physical properties, then surface details like color and texture. Attempts to interrupt or constrain scene construction did not change the hierarchical ordering. We suggest the results support a hierarchical partial-scene construction account of mental imagery.

Áreas y palabras clave

Cómo citar (APA 7)

Li, Y., Hammond, V., & Ullman, T. (2026). Non-commitment in mental imagery is distinct from perceptual inattention, and supports hierarchical scene construction. https://doi.org/10.31234/osf.io/yc9jw_v2