September 2019
Volume 19, Issue 10
Open Access
Vision Sciences Society Annual Meeting Abstract  |   May 2019
The effect of distractor statistics in visual search
Author Affiliations & Notes
  • Joshua M Calder-Travis
    New York University
    University of Oxford
  • Wei Ji Ma
    New York University
Journal of Vision May 2019, Vol.19, 318d. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Joshua M Calder-Travis, Wei Ji Ma; The effect of distractor statistics in visual search. Journal of Vision 2019;19(10):318d.

      Download citation file:

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

  • Supplements

Behaviour and mechanisms in visual search tasks – where observers must detect or identify a target amongst distractors – have been extensively studied. Animals frequently perform visual search (e.g. detecting camouflaged prey), however naturalistic stimuli are far more complex than the stimuli usually studied. Here we aim to characterise the behaviour of humans under the more realistic condition of heterogeneous distractor items. While heterogeneous distractors have been studied before, how distractor statistics affect behaviour has not been explored with stimuli that can be varied parametrically. We presented participants with arrays of Gabor patches. The target was a Gabor patch of a specific orientation, whilst the distractor orientations were independently drawn from one of two distributions, depending on the current ‘environment’. In one of the environments, distractors were likely to be similar to the target (von Mises distributed around target), whilst in the other, distractors took a wider range of values (uniformly distributed). For each trial we computed the distractor sample mean, sample variance, and most similar distractor to the target. All three statistics were predictive of behaviour. The effect of sample variance depended on the environment. This dependence suggests that observers accounted for the statistics of the environment, however a Bayesian model provided a poor fit to the observed data. In contrast, a simpler heuristic model, which only uses the item most likely to be the target to make decisions, outperformed the Bayesian model (AIC difference 85; 95% CI [47, 140]). The interaction between distractor variance and environment may be a consequence of the fact that increasing variance can push distractor orientations both towards and away from the target orientation. In addition to providing novel parametric descriptions of human behaviour in heterogeneous visual search, these results also suggest ways in which human visual search algorithms may deviate from optimality.


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.