Keele Research Repository

Performance of curvature discrimination was measured for a wide range of stimulus curvatures and sizes, using an adaptive version of the Method of Constant Stimuli. Performance is expressed both in terms of curvature difference thresholds, and efficiency, where efficiency is defined as the ratio ideal minimum response variance, based on the spatial statistics of the retina, to the observed response variance.
The results lead to three major conclusions.
Curved lines may be processed for curvature discrimination decisions with the same efficiency as straight lines, under appropriate conditions.
The results are not consistent with a high degree of common processing for straight and curved lines, but suggest the operation of two parallel processes, one for straight lines and one for curved lines. Each process has striot input limitations. Those for the process concerned with straight lines have already been determined by Andrews, Butcher and Buckley (1973).
The stimulus input limitations for the curved line process are determined in the present study, and suggest that this process is primarily concerned with local slope analysis, and is limited to a range of slopes of 40 degrees. Experiments using broken stimuli throw further light on the working of this process, as does a detailed study of the relationship between efficiency for curvature discrimination and stimulus orientation range (defined as the product of stimuLus curvature and length).
The orientation range limit is reduced to 30 degrees when oblique stimuli are used, or a single central gap is added to horizontal stimuli.
The implications of these results are discussed, and suggestions for further research made.