Purchase this article with an account.
Peter Kok, Lindsay Rait, Nicholas Turk-Browne; Distinct neural sources of expectations about features and objects. Journal of Vision 2018;18(10):315. doi: https://doi.org/10.1167/18.10.315.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Sensory processing is strongly influenced by our prior knowledge. In line with this, expectations about both simple features (e.g., orientation) and complex objects (e.g., abstract shapes) modulate processing in visual cortex. However, it is unclear whether or not these two types of expectations operate via the same underlying mechanism. For example, expectations about features may result from computations within visual cortex and induce global modulation, whereas expectations about objects may require feedback from higher-order, domain-general memory systems like the hippocampus. To address this question, we compare the influence of predictions about the orientation of grating stimuli to the influence of predictions about abstract shapes (i.e., Fourier descriptors). Using high-resolution fMRI in conjunction with inverted encoding models, we found that the neural representation of both gratings and shapes was delayed when they were incorrectly predicted. However, the effect for shapes was present throughout the visual cortical hierarchy (V1, V2, and LO), whereas for gratings it was limited to V1. Moreover, when oriented gratings were expected but omitted, the pattern of activity in V1 (but not higher-order visual cortex) reflected the expected orientation, suggesting that such expectations can evoke a template of the expected feature in sensory cortex. In contrast, this did not occur for expected (but omitted) shapes, suggesting that top-down shape expectations and bottom-up shape stimuli are represented via different neural codes. Finally, we found that the hippocampus represented expected shapes, but not expected gratings. In fact, the hippocampus signaled unexpected orientations, more consistent with coding of prediction errors. Altogether, our findings suggest that expectations about low-level features and higher-level objects may involve distinct neural mechanisms.
Meeting abstract presented at VSS 2018
This PDF is available to Subscribers Only