Speaker: Richard Murray, York University, Brain, Behaviour and Cognitive Science area, Department of Psychology

Title: Spatial vision and the generalized linear model

Abstract: Template matching models are a default starting point for understanding many areas of visual perception.  For many years these models were studied using ad hoc methods developed by vision scientists, not statisticians, but in fact the simplest template matching models are instances of the generalized linear model, and it is now becoming more common for vision researchers to use standard GLM tools and methods.  I will describe three experiments that use template matching/GLM methods, and illustrate some of the modelling and inference problems that they raise. (1) I will describe experiments that show how template matching models can reveal the role of illusory contours in visual perception. (2) I will show how the generalized additive model, an extension of the GLM, uncovers new information about how we perceive at a glance the number of elements in a scene. (3) I will show how the template matching model can be used to resolve long-standing questions about how people make simple decisions about visual stimuli.