Dual Process Coding of Recalled Locations in Human Oscillatory Brain Activity

Abstract

A mental representation of the location of an object can be constructed using sensory information selected from the environment and information stored internally. Human electrophysiological evidence indicates that behaviorally relevant locations, regardless of the source of sensory information, are represented in alpha-band oscillations suggesting a shared process. Here, we present evidence from human subjects of either sex for two distinct alpha-band-based processes that separately support the representation of location, exploiting sensory evidence sampled either externally or internally.SIGNIFICANCE STATEMENT Our sensory environment and our internal trains of thought are coded in patterns of brain activity and are used to guide coherent behavior. Oscillations in the alpha-frequency band are a predominant feature of human brain activity. This oscillation plays a central role in both selective attention and working memory, suggesting that these important cognitive functions are mediated by a unitary mechanism. We show that the alpha oscillation reflects two distinct processes, one that is supported by continuous sampling of the external sensory environment, and one that is based on sampling from internal representations coded in visual short-term memory. This represents a significant change in our understanding of the nature of alpha oscillations and their relationship to attention and memory.

Keywords: EEG; alpha; attention; working memory.

Figure 1.

Delayed spatial estimation task. a, Trial procedure for each experimental condition. b, Precision of spatial estimation and guess rate. Error bars indicate SEM.

Figure 2.

Results of the alpha-based inverted encoding model estimation. a, Estimated neural population response, or “channel” (offset of channel's preferred location from stimulus location: ±0°, 45°, 90°, 135°, or 180°), response starting 500 ms before onset of stimulus until end of retention period. b, Linear regression weights of estimated channel response folded around 0° offset (“slope”), indicating the degree of spatial selectivity. c, BF of t test comparing slope at each time point to slope of response estimated using permuted location labels. d, BF of ANOVA for effects of eye closure and mask on slope. e, ITC indicating stimulus- and mask-evoked potentials and scatterplots of correlation between ITC and slope during the period of the mask-evoked potentials for masked conditions.

Figure 3.

IEM generalization. a, Temporal generalization within experimental condition. Slopes for a spatially selective response are plotted. Those with less than moderate evidence (BF < 3) plotted as uniform dark blue. Encoding endurance indicates the sequential extent of temporal generalization backward and forward in time, where BF ≥ 3. Processes 1 and 2 are indicated in the EO and EC plots. b, Intercondition generalization. Points with moderate or greater evidence (BF ≥ 3) for a difference between intracondition generalizations shown in a and intercondition generalizations are outlined in white.

Figure 4.

Linear discriminant classifier generalization. a, Temporal generalization within experimental condition. Classifier accuracy is plotted. Accuracy with less than moderate evidence (BF < 3) plotted as uniform dark blue. Encoding endurance indicates the sequential extent of temporal generalization backward and forward in time, where BF ≥ 3. Processes 1 and 2 are indicated in the EO and EC plots. b, Intercondition generalization. Points with moderate or greater evidence (BF ≥ 3) for a difference between intracondition generalizations shown in a and intercondition generalizations are outlined in white.

Figure 5.

Investigation into the role of alpha enhancement with eye closure on temporal generalization within and between conditions. a, Alpha power by condition demonstrating the alpha enhancement with eye closure and its recreation in the eyes open conditions for the simulation. b, Scalp topography of alpha power by condition demonstrating that alpha is largely occipital in all conditions, and the simulation of alpha power topography in the eyes open conditions. c, Scalp topography of electrode weight rank, where higher values indicate larger weights for those electrodes resulting from the IEM. d, Results of the temporal generalization with the simulated recreation of alpha enhancement in the eyes open conditions. The pattern of temporal generalization is unchanged by this simulated alpha enhancement (Fig. 3).

Figure 6.

Control study into the effects of eye closure (blink). a, Trial procedure for each experimental condition. b, Estimated neural population response, or “channel” (offset of channel's preferred location from stimulus location: ±0°, 45°, 90°, 135°, or 180°), response (alpha-band power) starting 500 ms before onset of stimulus until end of retention period. c, Linear regression weights of estimated channel response folded around 0° offset (“slope”), indicating the degree of spatial selectivity. d, BF of t test comparing slope at each time point to slope of response estimated using permuted location labels. e, Results of temporal generalization.

Figure 7.

a, Topography of normed alpha power during process 1 and process 2 time windows as a function of spatial location and condition. Alpha power is normalized to the range of alpha at P-O electrode sites (i.e., proportion of max value of alpha power among P-O electrodes for that condition) to be able to observe local topography for those critical sites. b, Alpha lateralization index (alpha power at ipsilateral sites − contralateral sites/ipsilateral + contralateral) by condition and time window overlapping with process 1 (T1 = 198–250 ms) or process 2 (T2 = randomly selected 52 ms window from 750 to 2000 ms).

Figure 8.

Investigation into the nature of the weight change. a, In the eyes open condition, average weight change was ~0 throughout the retention interval (iii–v), following the initial average increase in weights from 250 ms after stimulus onset forward in time (i and ii). This same increase in weights was also present in the eyes closed condition (i and ii). However, there was an additional average change unique to the eyes closed condition: a decrease from around the time of eye closure (~300 to 750 ms after stimulus onset) forward into the rest of the retention interval (iv). Otherwise average weight change in the eyes closed condition was ~0 (iii and v). b, Weights were more dynamic throughout the trial in the eyes closed condition, including before eye closure, than in the eyes open condition. c, Changes in the eyes closed condition after eye closure, although greater than that during the same time period in the eyes open condition, are largely in parietal and occipital electrodes as they are in the eyes open condition.