A neural measure of precision in visual working memory

Abstract

Recent studies suggest that the temporary storage of visual detail in working memory is mediated by sensory recruitment or sustained patterns of stimulus-specific activation within feature-selective regions of visual cortex. According to a strong version of this hypothesis, the relative "quality" of these patterns should determine the clarity of an individual's memory. Here, we provide a direct test of this claim. We used fMRI and a forward encoding model to characterize population-level orientation-selective responses in visual cortex while human participants held an oriented grating in memory. This analysis, which enables a precise quantitative description of multivoxel, population-level activity measured during working memory storage, revealed graded response profiles whose amplitudes were greatest for the remembered orientation and fell monotonically as the angular distance from this orientation increased. Moreover, interparticipant differences in the dispersion-but not the amplitude-of these response profiles were strongly correlated with performance on a concurrent memory recall task. These findings provide important new evidence linking the precision of sustained population-level responses in visual cortex and memory acuity.

Figure 1

Behavioral task. On each trial, participants were required to remember the orientation of a sample grating over a 12-sec interval and then adjust a randomly oriented probe to match the remembered sample.

Figure 2

Characterizing stimulus-specific patterns observed in visual cortex during WM storage. (A) A forward encoding model (see Experimental Procedures) was used to generate an orientation-selective population response profile (CRF) based on patterns of activation observed in V1 and V2v during a period from 8 to 12 sec following the start of each trial. CRFs peaked in the channel preferring the remembered orientation and decreased monotonically as the angular distance from this orientation increased. Error bars are ±1 SEM. (B) Each line depicts the mean CRF observed at a different point during the course of a trial. Error bars have been omitted for exposition.

Figure 3

Estimates of CRF amplitude and dispersion are largely unchanged during the memory delay. Mean amplitude (red; right ordinate) and dispersion (black; left ordinate) are plotted as a function of time (abscissa) relative to the start of each trial. Once CRFs began to emerge (approximately 4 sec following the start of each trial; see Figure 2B), estimates of dispersion and amplitude remained largely unchanged.

Figure 4

Individual differences in CRF dispersion are strongly correlated with memory performance. Each panel plots mean CRF dispersion (ordinate) from visual areas V1 and V2v as a function of memory performance (mean recall error, abscissa) during different temporal windows (e.g., 8–12 sec; top) or time points (4, 6, 8, 10, 12, or 14 sec; bottom) following the start of each trial. *p < .05, **p < .01. Qualitatively similar results were obtained when V1 and V2v were considered separately (see text).

Figure 5

Individual differences in CRF amplitude are uncorrelated with memory performance. Plots mean CRF amplitude (ordinate) from visual areas V1 and V2v (ordinate) as a function of memory performance (mean recall error, abscissa) during different temporal windows (e.g., 8–12 sec; top) or time points (4, 6, 8, 10, 12, or 14 sec; bottom) following the start of each trial. Qualitatively similar results were obtained when V1 and V2v were considered separately.

Figure 6

CRFs are contingent on observers’ intent to remember the sample orientation. In a second experiment, the sample orientation was followed by a postcue (a change in the color of the fixation point) that instructed observers to remember the sample orientation (“store” trials) or passively wait for the next trial to begin (“drop” trials). A forward encoding model was used to construct CRFs based on patterns of activation observed in V1 and V2v during a period from 8 to 12 sec following the start of each trial. During “store” trials, activation peaked in the orientation channel (0°) and fell off in a graded fashion as distance from this orientation increased. During “drop” trials, this pattern was eliminated. Error bars are ±1 SEM. Data have been pooled and averaged visual areas V1 and V2v; qualitatively similar results were observed when each area was considered separately.