Capacity limitations of visual memory in two-interval comparison of Gabor arrays.

The capacity of short-term visual memory (VSTM) was assessed in a two-interval spatial frequency (SF) discrimination task. The cued Gabor target in a multi-element array either increased or decreased in SF across a 2 s inter-stimulus interval (ISI). Distracters as well as target were made to change across ISI so that memory of the individual SF of Gabor elements was required to solve the discrimination. The dynamics of the information loss from visual memory were analysed by manipulating the timing of spatial cues and masks. Cueing the target position before the first display gave thresholds comparable with those for a single Gabor patch. Cues placed after the first display gave higher thresholds indicating some loss of information. Within the ISI there was little increase in threshold or set size effect with cue delay. However there was a sharp rise in thresholds for cue positions after the second display. Gabor masks placed before a mid-ISI cue were more effective than noise masks or Gabor masks placed after the cue. With a cue placed late in the ISI, preceded by a Gabor mask, the masking effect decreased with increasing delay of the mask after the first display. This suggests a selective, dynamic but increasingly durable representation of the initial stimulus is built up in memory, and there is a graded form of "overwriting" of this representation by new stimuli.