Abstract
To determine what visual objects' representations in memory “look like”, the remembered spatial frequencies of Gabor patches were measured with the psychophysical method of adjustment.
Experiment One: Subjects viewed a single Gabor whose spatial frequency varied over trials, and reproduced its spatial frequency (i) while the Gabor remained visible on the display, or (ii) 1400 or 2400 msec after it had disappeared. In all three conditions, subjects' clustered tightly around the Gabor's actual frequency, although trial-to-trial variability of the reproductions increased by ∼2× when responses drew on memory rather than perception.
Experiment Two: To determine how a memory changes when multiple Gabors are being remembered, subjects saw and tried to remember two sequentially-presented Gabors on each trial. A post-stimulus cue identified which Gabor, first or second, had to be reproduced from memory. From trial-to-trial, the reproduced frequencies were about as variable as when just one Gabor had to be remembered (Experiment One). However, reproductions of the first Gabor, which had been held longer in memory, were more variable than that of the second Gabor. Moreover, reproductions of the target Gabor tended to be skewed toward the spatial frequency of the non-target Gabor, and this skew increased with the difference between the Gabor's frequencies. The reproduced frequencies' skew toward the non-target might reflect the extent to which the non-target frequency had successfully infiltrated memory.
Experiment Three: To modulate the non-target Gabor's influence on memory for the target Gabor, a cue presented prior to the two study items rendered the trial's non-target Gabor task-irrelevant. As predicted, this manipulation dramatically reduced, but did not eliminate, the reproductions' skew toward the frequency of the non-target Gabor. We hypothesize that distortions in remembered spatial frequency arises relatively late in processing.
Supported by NIH grant MH-068404.