December 2022
Volume 22, Issue 14
Open Access
Vision Sciences Society Annual Meeting Abstract  |   December 2022
Contextual Influences of Perceptual Inferences
Author Affiliations
  • Emily A-Izzeddin
    Queensland Brain Institute, The University of Queensland
  • Jason Mattingley
    Queensland Brain Institute, The University of Queensland
    The School of Psychology, The University of Queensland
  • William Harrison
    Queensland Brain Institute, The University of Queensland
    The School of Psychology, The University of Queensland
Journal of Vision December 2022, Vol.22, 3811. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Emily A-Izzeddin, Jason Mattingley, William Harrison; Contextual Influences of Perceptual Inferences. Journal of Vision 2022;22(14):3811.

      Download citation file:

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

  • Supplements

Humans have well-documented priors for many features present in nature. These priors develop over the lifespan and are thought to be learned, at least in part, from statistical regularities in our environment. In a series of experiments, we investigated the influence of such priors on peoples’ interpretations of naturalistic stimuli. Specifically, we determined how priors for low- and high-level image features influence participants’ ability to infer the “upright” orientation of naturalistic image patches. To disentangle the contributions of low- and high-level features to perceptual inferences, targets were cropped to eliminate most of the high-level content that could give explicit clues about “true” target orientations. When individual patches were presented with no additional contextual information, participants’ inferences of “upright” were well-approximated by an ideal-observer model incorporating heuristics based on known perceptual priors for low-level features (namely orientation and lighting). This suggests participants use priors for features as seen in naturalistic environments to interpret novel scenes when no contextual information is given. In separate experiments, we investigated the influence of contextual information by pre-cueing target patches with relevant naturalistic context. When context was provided, observers’ performance improved, suggesting people incorporate and use contextual information to inform their judgements. Interestingly, observed improvements in performance appeared to be consistent with participants using the low-level features present in the provided context. Taken together, our findings suggest that inferences about novel naturalistic stimuli can largely be driven by how low-level features are judged relative to our existing priors for natural scenes.


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.