Personal tools
You are here: Home Events ANC Workshop: Nikos Gekas, Chair: Simon Lyons

ANC Workshop: Nikos Gekas, Chair: Simon Lyons

— filed under:

Priors formed over multiple time scales facilitate performance in visual perception

  • ANC Workshop Talk
When Mar 18, 2014
from 11:00 AM to 12:00 PM
Where IF 4.31/4.33
Add event to calendar vCal

Bayesian Inference provides a parsimonious model of perception, where sensory-stimuli are combined with prior beliefs in a near-optimal way. In this framework, a question of key importance is how the visual system updates its priors through interaction with the environment.

We used a visual search task to explore how expectations of different time scales (from the last few trials to hours to long-term statistics of natural scenes) interact to alter perception. In previous work, we demonstrated that the statistics of past visual motion stimuli can powerfully modulate the perception of new motion directions. Here, we presented human observers with low contrast white dots in twelve possible positions equally spaced on a circle, and we asked them to identify the presence and position of the dots, and manipulated subjects’ expectations by presenting stimuli at some positions more frequently than others. We compared the effect of expectations developed over a long sequence of trials, with priming, which we defined as positions being presented in the very few last trials (e.g. 3).

Our results can be modeled and understood within the Bayesian framework, in terms of a near-optimal integration of pre-existing long-term priors with rapidly learned statistical priors, which are skewed towards the very recent history of trials, and may help in understanding the time-scale of developing expectations at the neural level.