August 2023
Volume 23, Issue 9
Open Access
Vision Sciences Society Annual Meeting Abstract  |   August 2023
Contextual coherence increases perceived numerosity independent of semantic content
Author Affiliations
  • Chuyan Qu
    University of Pennsylvania
  • Michael F. Bonner
    Johns Hopkins University
  • Elizabeth M. Brannon
    University of Pennsylvania
Journal of Vision August 2023, Vol.23, 5620. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Chuyan Qu, Michael F. Bonner, Elizabeth M. Brannon; Contextual coherence increases perceived numerosity independent of semantic content. Journal of Vision 2023;23(9):5620.

      Download citation file:

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

  • Supplements

The ability to rapidly represent visual number is supported by the Approximate Number System (ANS) and is thought to provide a foundation for mathematical cognition. Characterizing visual features that systematically bias our numerosity perception promises to uncover the processes that form ANS representations. Recent work demonstrated that reducing coherence of low-level visual attributes such as color and orientation systematically reduces perceived numerosity (DeWind, Bonner, & Brannon, 2020; Qu, DeWind, & Brannon, 2022). Here we ask whether the coherence effect is exclusive to low-level visual features or instead whether it can be driven by higher-level features. To this end, we tested adults in an ordinal numerical comparison task with contextual coherence mathematically manipulated using a statistical model of visual object co-occurrence (“object2vec”) (Bonner & Epstein, 2021). We found that arrays with high contextual coherence were perceived as numerically larger than arrays with low contextual coherence (Experiment 1). In Experiments 2a & 2b, we tested whether box scrambling and diffeomorphic scrambling the objects attenuated the coherence effect. In both cases the strength of the coherence effect was similar in the intact and scrambled conditions, suggesting that this effect was not driven by high-level semantic information and instead relied on lower-level processing. In Experiment 3, we conducted the first test of whether mid-level visual features can drive the coherence effect. We compared intact greyscale arrays with images modified by a texform generation algorithm and controlled for luminance and contrast (Deza, Chen, Long, & Konkle, 2019). We found that the mid-level features preserved in texform stimuli were sufficient to elicit the coherence illusion. Taken together, contextual coherence, which arises from the natural co-occurrence statistics of visual scenes, interacts with ANS representations at both low-level and mid-level visual processing, bypassing higher-level processing of semantic content.


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.