December 2022
Volume 22, Issue 14
Open Access
Vision Sciences Society Annual Meeting Abstract  |   December 2022
The dorsal visual pathway represents object-centered spatial relations for object recognition
Author Affiliations
  • Vladislav Ayzenberg
    Carnegie Mellon University
  • Marlene Behrmann
    Max Planck Research School for Intelligent Systems
Journal of Vision December 2022, Vol.22, 3233. doi:https://doi.org/10.1167/jov.22.14.3233
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Vladislav Ayzenberg, Marlene Behrmann; The dorsal visual pathway represents object-centered spatial relations for object recognition. Journal of Vision 2022;22(14):3233. https://doi.org/10.1167/jov.22.14.3233.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Although there is mounting evidence that input from the dorsal visual pathway contributes to object recognition processes in the ventral pathway, the dorsal pathway is rarely included in conceptual or computational models of visual recognition. One reason for this exclusion is that the specific functional contributions of dorsal cortex to object recognition are poorly understood. Given its classic role in computing spatial relations, we hypothesized that the dorsal pathway may compute the object-centered relations among an object’s component parts (i.e., its topological structure) and propagate this information to the ventral pathway to support object categorization. Using multiple functional localizers, we found regions in the posterior and anterior intraparietal sulcus (IPS) with selectivity for part relations. These regions were found most strongly in the right hemisphere, and their selectivity was independent of allocentric relations and other dorsal object representations, such as 3D shape and tools. Importantly, these regions exhibited task-dependent functional connectivity with ventral cortex, such that connectivity increased in conditions where part relations differed. Next, we examined whether and how these regions might support object categorization. We found object category could be decoded in right pIPS at levels comparable to ventral object regions (e.g., LOC). Moreover, similarity analyses suggested that decoding in right pIPS was supported by a representation of object-centered relations, as approximated by a skeletal model, and not by low- or high-level image properties as approximated by Gabor or neural network models. Interestingly, we found that the multivariate response in right pIPS mediated representations of part relations in ventral cortex, a finding further supported by its multivariate connectivity to ventral cortex. Altogether these findings highlight how object-centered part relations, a property crucial for object recognition, are represented neurally. Moreover, these findings provide further evidence of a strong link between dorsal and ventral visual pathways in accomplishing object recognition.

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×