The relationship between working memory storage and elevated activity as measured with functional magnetic resonance imaging

Abstract

Does the sustained, elevated neural activity observed during working memory tasks reflect the short-term retention of information? Functional magnetic resonance imaging (fMRI) data of delayed recognition of visual motion in human participants were analyzed with two methods: a general linear model (GLM) and multivoxel pattern analysis. Although the GLM identified sustained, elevated delay-period activity in superior and lateral frontal cortex and in intraparietal sulcus, pattern classifiers were unable to recover trial-specific stimulus information from these delay-active regions. The converse-no sustained, elevated delay-period activity but successful classification of trial-specific stimulus information-was true of posterior visual regions, including area MT+ (which contains both middle temporal area and medial superior temporal area) and calcarine and pericalcarine cortex. In contrast to stimulus information, pattern classifiers were able to extract trial-specific task instruction-related information from frontal and parietal areas showing elevated delay-period activity. Thus, the elevated delay-period activity that is measured with fMRI may reflect processes other than the storage, per se, of trial-specific stimulus information. It may be that the short-term storage of stimulus information is represented in patterns of (statistically) "subthreshold" activity distributed across regions of low-level sensory cortex that univariate methods cannot detect.

Figure 1.

Behavioral task. Participants maintained the direction and speed of a sample motion stimulus over a long delay period. Midway through this delay period, they were cued as to the dimension on which they would be making an upcoming comparison, either direction or speed. At the end of the delay period, they were presented with a probe motion stimulus and had to indicate with a button press whether it matched or did not match the sample stimulus on the cued dimension.

Figure 2.

Univariate GLM results. Sample-evoked (A) and delay-related (B) activity, as estimated from a group-level GLM, thresholded at p < 0.05, FDR corrected, and displayed on a representative subject's inflated surface. Note that images are for illustrative purposes only, because all analyses were performed on single-subject data. Superimposed is an outline of MT+ as defined by the localizer. Note for this region that it is robustly activated by the sample but that its activity does not differ from the baseline during the delay period. A qualitatively similar pattern is observed in calcarine and pericalcarine cortex. LH, Left hemisphere; RH, right hemisphere.

Figure 3.

Whole-brain direction-decoding results. Decoding time courses after training on whole-brain data from direction-cued trials. A, Decoding of direction information from direction-cued trials. B, Decoding of direction information from speed-cued trials. Each waveform represents the mean direction-decoding accuracy across subjects (n = 7) for a classifier trained with data limited to a single time point in the trial and then tested on all time points in the holdout trials (e.g., the green line illustrates the decoding time course from a classifier trained on only data from time point 4, indicated by the small green triangle along the x-axis.) Horizontal bars along the top indicate points at which the decoding accuracy for the corresponding classifier was significantly above chance (p < 0.05, permutation test). Schematic icons of trial events are shown at the appropriate times along the x-axis. Data are unshifted in time.

Figure 4.

ROI BOLD and decoding time courses. A–D, Average ROI BOLD activity. Data from direction-cued trials use solid lines, and speed-cued trials use dashed lines; bands cover average SE across subjects. ROI stimulus-direction decoding results (E–H) and ROI trial-dimension decoding results (I–L). Graphical conventions same as Figure 3. All averaged across individual data from seven subjects.