Predictive processing of scene layout depends on naturalistic depth of field

Read the full article See related articles


Boundary extension (BE) is a classic memory illusion in which observers remember more of a scene than was presented. According to predictive processing accounts, BE reflects the integration of visual input and expectations of what is beyond a scene’s boundaries. According to normalization accounts, BE rather reflects one end of a normalization process towards a scene’s typically-experienced viewing distance, such that close-up views give BE but distant views give boundary contraction. Here across four experiments, we show that BE strongly depends on depth-of-field (DOF), as determined by the aperture settings on a camera. Photographs with naturalistic DOF led to larger BE than photographs with unnaturalistic DOF, even when showing distant views. We propose that BE reflects a predictive mechanism with adaptive value that is strongest for naturalistic views of scenes. The current findings indicate that DOF is an important variable to consider in the study of scene perception and memory.

Statement of Relevance

In daily life, we experience a rich and continuous visual world in spite of the capacity limits of the visual system. We may compensate for such limits with our memory, by filling-in the visual input with anticipatory representations of upcoming views. The boundary extension illusion (BE) provides a tool to investigate this phenomenon. For example, not all images equally lead to BE. In this set of studies, we show that memory extrapolation beyond scene boundaries is strongest for images resembling human visual experience, showing depth-of-field in the range of human vision. Based on these findings, we propose that predicting upcoming views is conditional to a scene being perceived as naturalistic. More generally, the strong reliance of a cognitive effect, such as BE, on naturalistic image properties indicates that it is imperative to use image sets that are ecologically-representative when studying the cognitive, computational, and neural mechanisms of scene processing.

Article activity feed