August 2014
Volume 14, Issue 10
Free
Vision Sciences Society Annual Meeting Abstract  |   August 2014
Representational geometry measures predict categorisation speed for particular visual objects
Author Affiliations
  • Ian Charest
    Medical Research Council Cognition and Brain Sciences Unit, Cambridge, UK
  • Thomas A. Carlson
    Department of Cognitive Sciences, Macquarie University, Sydney, Australia
  • Nikolaus Kriegeskorte
    Medical Research Council Cognition and Brain Sciences Unit, Cambridge, UK
Journal of Vision August 2014, Vol.14, 909. doi:https://doi.org/10.1167/14.10.909
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Ian Charest, Thomas A. Carlson, Nikolaus Kriegeskorte; Representational geometry measures predict categorisation speed for particular visual objects. Journal of Vision 2014;14(10):909. https://doi.org/10.1167/14.10.909.

      Download citation file:


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

      ×
  • Supplements
Abstract

The choice reaction time reflects the rate of accumulation of sensory evidence. For categorisation, Carlson et al. (2013) showed that an animate object can be more rapidly recognised as such when it is further away from the animate-inanimate boundary in human-inferior-temporal (hIT) representational space (as reflected in fMRI data from Kriegeskorte et al. 2008). Here, we extend these results by considering multiple category dichotomies, multiple measures of representational geometry, and multiple ventral-stream brain regions. Subjects categorised objects according to four dichotomies: animate vs. inanimate; face vs. body; human vs. animal face; natural vs. artificial inanimate object. We examined a range of representational-geometry measures, including each object's representational centrality (i.e. its average distance to other members of its own category), its distinctness from the other category (i.e. its average distance to members of the other category), and the proportion of category peers within its local neighbourhood. We assessed whether these measures predict subjects' reaction times for particular objects using Spearman's rank correlation (ρ, computed within each subject and then averaged across subjects) combined with non-parametric inference and control of the false-discovery rate. For most tasks and ventral-stream brain regions, an object's within-category representational centrality was the best predictor of reaction time among the tested representational-geometry measures. Three example results are: (1) The hIT centrality predicted reaction time for animate vs. inanimate categorisation (ρ = 0.2; p<0.0001). (2) Centrality in the fusiform face area predicted reaction time for human vs. animal face categorisation (ρ = 0.2; p <0.01). (3) Centrality in the parahippocampal place area predicted reaction time for natural vs. artificial inanimate categorisation (ρ = .29; p <0.001). Our results suggest that the representational geometry as reflected in human fMRI can explain aspects of the computational processes leading to a decision about a particular visual object.

Meeting abstract presented at VSS 2014

×
×

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.

×