Internal noise and transducer function in sensory detection experiments: Evaluation of psychometric curves and of ROC curves

Abstract

The theory of signal detection (TSD) is applied on sensory discrimination by human observers. The discrimination of stimuli without external noise can be described by two distinct models. The first model implicates additive neural noise at very low stimulus intensities and multiplicative noise at large intensities (i.e., noise level increasing with intensity). The second model implicates additive neural noise at all stimulus intensities and an attenuator which is adapted to the overall intensity level of an experiment (multirange meter model). The psychometric curves and the ROC curves bear information about the kind of probability distribution of the noise and about the transducer function, i.e., the relation between stimulus intensity and the relevant neural activity. First a theoretical investigation of the psychometric curves and the ROC curves according to both models is given. Thereafter experimental results of auditory and visual discrimination are presented. It appears that the noise which limits the auditory discrimination is of an additive nature and has a Gaussian distribution. This result combined with the fact that differential sensitivity decreases with the intensity of the stimuli can be explained by the second model and not by the first one. Accepting the transducer function to be a power relation the exponent n has been determined. The value of this exponent appears to be 0.8 and 0.2 for very low and moderate intensity levels respectively, with intensity in signal power. Visual discrimination yields also additive Gaussian noise at low intensity level and a power relation with exponent 0.6.