Abstract
The study of eye movement patterns is a valuable tool for psychological research in several domains such as reading and scene analysis. Here we outline a new methodology for the analyses of fixation data that can be used to examine the representation and recognition of three-dimensional object shape. In principal, fixation patterns can be used to infer properties of local image features that support shape recognition. However, a number of methodological problems must be overcome. For example, it is unclear how predicted fixation patterns from different theoretical models can be statistically compared to observed data. In addition, where analyses are based on the definition of a priori areas of interest (AOIs) the spatial precision, and theoretical validity of the AOIs, can limit the validity of the analyses. To address these issues we outline a new approach, known as the fixation region overlap analysis, which uses observed fixation patterns to generate AOIs that can be subject to analyses for shape information content. These analyses statistically contrast the degree of spatial overlap between the observed AOIs, and those predicted by a random distribution and any number of theoretical models of shape analysis, including mathematical and those derived from lesion data. This methodology provides a quantitative and statistically valid technique for the analysis of fixation patterns in studies of shape recognition, and has applications in other research domains.