Abstract
Psychophysical contrast discrimination is believed to be mediated by internal responses, each characterized by contrast-dependent mean value and noise amplitude. The standard measure of contrast discrimination, TvC (threshold versus contrast) curve does not allow unambiguous characterization of these two components of the internal response, since many possible combinations could account for the same TvC curve. Here we propose a novel approach that is based on performing a larger number of pair wise contrast discriminations. The performance, measured as percent correct discrimination, is compared with the predicted one based on a model that assumes normally distributed responses. The two response components are then determined by matching the model predictions to the experimental results. This procedure requires a minimal number of stimulus pairs in order to derive a complete or over-complete system of equations for mean and noise response amplitudes. The method can also be used for other stimuli configurations, such as local stimuli surrounded by flankers, allowing to study the lateral interactions. We applied the method to a set of 6 isolated Gabor patches (9.2 cpd, sigma=0.11 deg) with different contrasts (randomly mixed) and 10 pair wise discriminations (temporal 2AFC) that resulted in the complete system of model equations. The preliminary experimental results (2 observers, ∼600 trials each) indicate that the shape of the TvC curve is determined by a nonlinear transducer function and a non-trivial contrast-dependence of the noise amplitudes. However we find that calculated noise amplitudes are more sensitive to experimental measurement errors. It is possible that we will be able to reduce this uncertainty by considering additional contrast discriminations that will result in an over-complete system of equations for response amplitudes.