Decoding and Reconstructing the Focus of Spatial Attention from the Topography of Alpha-band Oscillations

Abstract

Many aspects of perception and cognition are supported by activity in neural populations that are tuned to different stimulus features (e.g., orientation, spatial location, color). Goal-directed behavior, such as sustained attention, requires a mechanism for the selective prioritization of contextually appropriate representations. A candidate mechanism of sustained spatial attention is neural activity in the alpha band (8-13 Hz), whose power in the human EEG covaries with the focus of covert attention. Here, we applied an inverted encoding model to assess whether spatially selective neural responses could be recovered from the topography of alpha-band oscillations during spatial attention. Participants were cued to covertly attend to one of six spatial locations arranged concentrically around fixation while EEG was recorded. A linear classifier applied to EEG data during sustained attention demonstrated successful classification of the attended location from the topography of alpha power, although not from other frequency bands. We next sought to reconstruct the focus of spatial attention over time by applying inverted encoding models to the topography of alpha power and phase. Alpha power, but not phase, allowed for robust reconstructions of the specific attended location beginning around 450 msec postcue, an onset earlier than previous reports. These results demonstrate that posterior alpha-band oscillations can be used to track activity in feature-selective neural populations with high temporal precision during the deployment of covert spatial attention.

Figure 1

Task design and behavioral data. (A) Covert spatial attention was directed with an omnidirectional symbolic cue (80% valid) that indicated in which of six spatial locations an upcoming target was likely to appear. Participants were instructed to attend to the cued location while maintaining fixation and then decide if the target was a “×” or a “+.” (B) RT decreased when targets appeared in cued, as contrasted with uncued locations (t(1, 7) = 5.62, p < .001), suggesting that participants were indeed attending the cued location. Each line represents one participant.

Figure 2

Classification and IEM. (A) The topography of alpha-band power (8–13 Hz) over posterior electrodes during sustained attention (1000–1900 msec) shows a clear spatial correspondence with the attended location (indicated by a red dot). (B) A linear SVM classifier trained to decode the attended location using the topography of power from 1000 to 1900 msec could successfully classify the attended location, but only from activity in the alpha band (t(1, 7) = 3.46, p = .010). As a control, classification did not differ from chance (dashed line) when precue alpha power was used (t(1, 7) = 0.29, p = .774). Error bars represent 95% confidence intervals. Each gray point represents a participant’s average classification accuracy. (C) The basis set used for the IEM contained six tuning functions, one for each spatial location, that represent the idealized response of the neural population tuned to each location.

Figure 3

Reconstructed tuning functions from alpha power. (A) Using the topography of 8–13 Hz power over time, we were able to reconstruct the spatial focus of attention. Tuning functions for each polar angle location were circularly shifted and averaged so that 0° corresponds to the response of the channel representing the attended location on that trial. The reconstructed tuning functions demonstrate a characteristic Gaussian shape, with a maximal response at the attended location (0°) and a fall off in response for channels tuned to locations further away. Time points with significant selectivity (p < .01), as measures by the slope of the tuning function at each time point, are indicated with black squares. (B) Comparison of channel tuning during the prestimulus baseline (−350 to −50 msec) and covert attention interval (1000–1900 msec). The amplitude of the attended channel response (0°) was significantly higher compared with both baseline (M_diff = 0.125 μV², p = .017) and to the channel tuned to the most distant location (i.e., 180°; M_diff = 0.279 μV², p = .001), indicating that the reconstructions depended on the allocation of spatial attention. Error bars indicate ± within-subject SEM. (C) As a metric of which regions were most informative for the reconstructions, we computed MI between alpha power and attended location for each electrode for 11 time windows. This analysis reveals a clear posterior distribution of electrodes that carried the most information about the attended location, beginning around 400 msec.

Figure 4

Reconstructed tuning functions from alpha phase. (A) In contrast to the alpha power results, reconstructions based on the topography of phase angles did not reveal sustained location tuning during spatial attention (black squares indicate p < .01). A transient response was observed immediately following cue onset, likely reflecting phase locking (e.g., an evoked response) from the cue. (B) The attended-channel response during attention was statistically indistinguishable from baseline and from the response in the channel tuned 180° away (ps > .54). Error bars indicate ± within-subject SEM. (C) MI analysis shows no clear topography of information between phase and attended location.