Top-down effect on pupillary response: Evidence from shape from shading

Ayelet Sapir, Ronen Hershman, Avishai Henik

Research output: Contribution to journalArticlepeer-review


Shaded 2D images often create an illusion of depth, due to the shading information and assumptions regarding the location of the light source. Specifically, 2D images that are lighter on top usually appear convex while images that are darker on top, usually appear concave, reflecting the assumption that light is coming from above. The process of recovering the 3D shape of a shaded image is called Shape from Shading. Here we examined whether the pupil responds to the illusion of depth in a shape from shading task. In three experiments we show that pupil size is affected by the percept of depth, so that it dilates more when participants perceive the stimulus as concave, compared to when they perceive it as convex. This only happens if participants make a judgment regarding the shape of the stimulus or when they view it passively but are aware of the different shapes. No differences in pupil size were found with passive viewing if participants were not aware of the illusion, suggesting that some aspects of shape from shading require attention. All stimuli were equiluminant, and the percept of depth was created by manipulating the orientation of the shading, so that changes in pupil size could not be accounted by changes in the amount of light in the image. We posit, and confirmed it in a behavioral control experiment, that the perception of depth is translated to a subjective perception of darkness, due to the “darker is deeper” heuristic and conclude that the pupillary physiological response reflects the subjective perception of light.

Original languageEnglish
Article number104664
StatePublished - 1 Jul 2021


  • Depth perception
  • Illusion
  • Pupil light reflex
  • Pupillometry
  • Shape from shading


Dive into the research topics of 'Top-down effect on pupillary response: Evidence from shape from shading'. Together they form a unique fingerprint.

Cite this