July 2013
Volume 13, Issue 9
Vision Sciences Society Annual Meeting Abstract  |   July 2013
Representations of action categories generalize across the phylum, genus, and species of the actor
Author Affiliations
  • Andrew C. Connolly
    Dartmouth College
  • James V. Haxby
    Dartmouth College\nUniversity of Trento
Journal of Vision July 2013, Vol.13, 185. doi:https://doi.org/10.1167/13.9.185
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Andrew C. Connolly, James V. Haxby; Representations of action categories generalize across the phylum, genus, and species of the actor. Journal of Vision 2013;13(9):185. doi: https://doi.org/10.1167/13.9.185.

      Download citation file:

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

  • Supplements

Neuroimaging studies of action representation frequently investigate activity associated with viewing human actions. However, many actions are general with respect to the kinds of animals that can perform them. It is just as easy to recognize "eating" whether done by beetle or bear. We investigate the generality of action representations and dissociate representational dimensions that reflect action category versus animal category. Undergoing fMRI, subjects viewed 24 video clips that depicted a range of animals---primates to ladybugs---performing different actions: eating, fighting, running, swimming. Using a voxel clustering procedure (Connolly et al. 2012) we produced "early" and "late" vision ROIs ("EV" and "LV"). SVM classification in EV and LV yielded mean accuracies of 0.75 and 0.60, chance = 0.04 (1/24). We analyzed the structure of representational spaces using multi-table MDS (Abdi et al. 2012). In LV, the first, third and fourth dimensions of the solution distinguished between different types of actions, with the 1st dimension distinguishing all action pairs except for fighting versus swimming. Dimension 3 distinguished fighting from eating and running, and 4 distinguished swimming from others. Dimension 2 separated mammals from non-mammals and contained no action information. EV carried information about action classes to a lesser degree: Dimension 1 distinguished between swimming versus eating and fighting, and dimension 3 between running and others. Again, dimension 2 separated the mammals from non-mammals. Next we used searchlight analysis to classify each clip into action categories. This analysis revealed maps of action representation networks including posterior middle temporal cortex---including pSTS and extending into LOC---and IPS. Searchlight analysis for mammal vs. non-mammals produced peaks in posterior lateral occipital cortex. In sum, these results provide evidence for abstract action representations that generalize across the species of the actors. Separate components of an overlapping distributed representational space reflected action and animal categories.

Meeting abstract presented at VSS 2013


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.