Abstract
It is known that spatial and temporal properties have a complex interplay in processing of the visual stimulus. Here, we investigated the effect of the temporal duration of the signal by Classification Image method.
Classification Images were measured using a yes/no detection task. Target was a 1-D Gaussian spatial profile (s.d. 0.25 deg) in a symmetric Gaussian temporal envelope having s.d. of 10, 20 or 40 ms. Stimulus was masked by spatiotemporal white noise lasting for 600 ms, shown at 100 Hz on a monitor with a fast phosphor.
Classification images for four stimulus-response categories (hit, miss, correct rejection, false alarm) were analysed separately. In all conditions, linear analysis of noise masks in target-present trials reveal a compact template, slightly elongated in time compared to the target stimulus. In the 40 ms condition, linear analysis of the noise masks in target-absent trials showed a template roughly corresponding to the target. However, Classification Images for shorter presentation times were featureless (flat).
Next, Classification Images based on Fourier energy were investigated. Noise masks were windowed with a Gaussian profile roughly corresponding to the greatest extent of the linear effects. Mean amplitudes of the Fourier components were computed for each stimulus-response category. Analysis of false alarm trials, whose ordinary CI's were flat, revealed a Fourier energy template having a tuning close to the target in all durations. This indicates that observers detect signals in phase-insensitive manner.
Our results show that subjects are able to use both the spatial and temporal information of the stimulus remarkably well. However, when detecting short duration targets, they cannot utilize target phase information. Furthermore, spatial tuning of the detecting mechanisms was found to be very stable across the target duration, as no differences in the spatial profiles of the CI's were found.