Effects of Partial Occlusion on Response Dynamics and Interregional Processing within Primate Superior Temporal Sulcus.

Abstract

Recognizing partially occluded objects is a critical visual function that primates perform with ease, yet the underlying neural mechanisms remain incompletely understood. Previous studies in macaque inferotemporal cortex have reported mixed results on whether occlusion delays and reduces responses to partially occluded objects. To address this, we recorded single-unit activity from body-responsive regions of the middle and anterior Superior Temporal Sulcus (STS) in male macaques while presenting body stimuli with varying levels of occlusion using a dot pattern. Occlusion reduced response strength and increased onset latency in both regions, and even low occlusion levels altered response dynamics by increasing the difference between the response trough and second peak. While body selectivity was preserved, body decoding accuracy declined and was delayed as occlusion increased. In contrast to some prior reports, we found no consistent enhancement of body decoding during the late response phase. By controlling for information loss and clutter introduced by the occluder, we found that reductions in response strength were partly due to the deletion of body features, whereas changes in response dynamics primarily reflected interactions between the occluder and the remaining body features. Occlusion delayed the first but not the second response peak, suggesting distinct mechanisms for these phases. Peak decoding at high occlusion levels emerged later in anterior than middle STS, indicating a feedforward component. However, representational similarity analysis combined with Granger causality suggested enhanced feedback from anterior to middle STS under high occlusion. Together, these results highlight the response dynamics supporting robust recognition under occlusion.Significance statement Recognizing objects under partial occlusion is fundamental to visual perception, yet the underlying neural mechanisms remain unclear. We aimed to clarify previous mixed results concerning the effects of occlusion on ventral stream responses and the contribution of feedback within the Superior Temporal Sulcus (STS) during occlusion. We recorded responses of body-responsive neurons to occluded bodies at two hierarchical levels in the macaque STS. Occlusion consistently reduced response strength and increased onset latency, with even low levels of occlusion altering response dynamics. We found no consistent evidence for enhanced body decoding under occlusion during later phases of the response. Temporal differences between middle and anterior STS representations under occlusion suggest an interplay between feedforward and feedback processes during occluded object recognition.