September 2018
Volume 18, Issue 10
Open Access
Vision Sciences Society Annual Meeting Abstract  |   September 2018
Testing the limits of identity recognition with mixed-identity faces
Author Affiliations
  • Isabelle Buelthoff
    Max Planck Institute for Biological Cybernetics, Human Perception, Cognition and Action
  • Mintao Zhao
    Max Planck Institute for Biological Cybernetics, Human Perception, Cognition and ActionUniversity of East Anglia
Journal of Vision September 2018, Vol.18, 157. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Isabelle Buelthoff, Mintao Zhao; Testing the limits of identity recognition with mixed-identity faces. Journal of Vision 2018;18(10):157. doi:

      Download citation file:

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

  • Supplements

Similarly to how we look for telltale signs of both parents' facial features in their children's faces, we are able to recognize two identities from one photo that mixes two persons' faces together. When more people's faces are used to create mixed faces, the identities of individual faces (i.e., "parent" faces) become less recognizable (i.e., identity information is degraded). In our study, we investigated the limit of identity recognition in such identity-degraded faces and whether familiarity with the "parent" faces enhances identity recognition from such mixed faces. We first tested whether people can extract the identities from a mix of three faces. Participants who were familiar with the "parent" faces performed better than those who were not. We then tested whether participants can extract the identities of mixed faces generated with more faces. We showed a mixed face of 2 to 10 "parent" faces together with a test face. Participants had to decide whether the test face was a parent of the mixed face. Both familiar and unfamiliar participants performed better than chance for mixed faces generated with up to eight faces. Finally, we tested at what level mixed faces lose their identity so that we cannot discern between two mixed faces generated with completely different "parent" faces. We presented two mixed faces in a trial and participants performed a same/different task. Both mixed faces had the same number of identities (2 to 32), but had no parents in common. Participants were better than chance even for the 32-face mixed faces. Together, these results indicate that our face processing system is extremely sensitive to facial identity information. Familiarity helps identity recognition, but this advantage becomes less evident when identity information degrades (i.e., with increased number of "parent" faces in a mixed face).

Meeting abstract presented at VSS 2018


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.