The influence of spatial pattern on visual short-term memory for contrast

Abstract

Several psychophysical studies of visual short-term memory (VSTM) have shown high-fidelity storage capacity for many properties of visual stimuli. On judgments of the spatial frequency of gratings, for example, discrimination performance does not decrease significantly, even for memory intervals of up to 30 s. For other properties, such as stimulus orientation and contrast, however, such "perfect storage" behavior is not found, although the reasons for this difference remain unresolved. Here, we report two experiments in which we investigated the nature of the representation of stimulus contrast in VSTM using spatially complex, two-dimensional random-noise stimuli. We addressed whether information about contrast per se is retained during the memory interval by using a test stimulus with the same spatial structure but either the same or the opposite local contrast polarity, with respect to the comparison (i.e., remembered) stimulus. We found that discrimination thresholds got steadily worse with increasing duration of the memory interval. Furthermore, performance was better when the test and comparison stimuli had the same local contrast polarity than when they were contrast-reversed. Finally, when a noise mask was introduced during the memory interval, its disruptive effect was maximal when the spatial configuration of its constituent elements was uncorrelated with those of the comparison and test stimuli. These results suggest that VSTM for contrast is closely tied to the spatial configuration of stimuli and is not transformed into a more abstract representation.

Fig. 1

VSTM task for the contrast of random binary noise patterns. The comparison stimulus (Stimulus 1) and the test stimulus (Stimulus 2) were separated by a memory interval of either 0.3, 1, 3, or 5 s. At the end of each trial, participants had to respond by pressing a button to indicate which stimulus had the higher contrast. (a) Noise patterns in both stimulus intervals were kept the same, or (b) the noise patterns in the two stimulus intervals were contrast-inverted. Subsequent trials started after a 1-s intertrial interval. (c) Control stimuli. Luminance levels of the individual elements of these stimuli were sampled from a uniform probability distribution with a broad range of grayscale levels (rather than just the two used in the binary stimuli) and were spatially uncorrelated between the two intervals. Here, Stimulus 1 has lower contrast than Stimulus 2, but participants could not use the luminance of individual elements to make a reliable judgment, since some elements became darker (top circles and arrow) and some brighter (lower circles and arrow). Participants had to extract and utilize the overall contrast of the patterns to perform the task

Fig. 2

(a) Discrimination thresholds (Weber fractions) for the contrasts of random noise patterns as a function of memory interval in the VSTM task. The light gray line indicates thresholds for trials in the identical-pattern condition, when the two stimuli had the same spatial pattern. The black line indicates thresholds for trials in the reversed-pattern condition, when the two stimuli were contrast-reversed. Error bars indicate SEMs across (n = 4) participants. (b) Discrimination thresholds (Weber fractions) in the control experiment using stimuli created with uniform random, rather than binary, noise. With such stimuli, it is impossible to perform the task reliably by basing judgments on the apparent brightness of individual elements within the display; observers must use the contrast across the image. Observers showed the same drop in performance with retention interval as when binary noise patterns were used. Error bars indicate SEMs across (n = 4) participants

Fig. 3

VSTM task for the contrast of random noise patterns with masking. On each trial, two stimuli of different contrasts were presented for 0.5 s each, separated by a 3-s memory interval. During the middle of the memory interval, a mask stimulus was presented. As before, participants had to indicate by pressing a button whether the comparison stimulus (Stimulus 1) or the test stimulus (Stimulus 2) had higher contrast. The bottom mask stimulus has the same spatial noise pattern as in the stimulus intervals, and the top mask has a different (uncorrelated) binary pixel noise pattern. The next trial started after an intertrial interval of 1 s

Fig. 4

Effect of masks on discrimination thresholds (Weber fractions) in the VSTM task. The gray line and symbols show the effect of masking on thresholds when the spatial pattern of the stimuli and mask were the same, whereas the black line and symbols show thresholds when the masks were different from the contrast stimuli (1 and 2). Error bars indicate SEMs across (n = 4) participants. The dashed line and shaded area indicate the threshold and ±1 SEM, respectively, across the same group of participants when no mask stimulus was presented