Spontaneous EEG oscillations reveal periodic sampling of visual attention

Abstract

An important effect of sustained attention is the facilitation of perception. Although the term "sustained" suggests that this beneficial effect endures continuously as long as something is attended, we present electrophysiological evidence that perception at attended locations is actually modulated periodically. Subjects detected brief light flashes that were presented peripherally at locations that were either attended or unattended. We analyzed the correlation between detection performance for attended and unattended stimuli and the phase of ongoing EEG oscillations, which relate to subsecond fluctuations of neuronal excitability. Although on average, detection performance was improved by attention--indicated by reduced detection thresholds at attended locations--we found that detection performance for attended stimuli actually fluctuated over time along with the phase of spontaneous oscillations in the (≈7 Hz) frequency band just before stimulus onset. This fluctuation was absent for unattended stimuli. This pattern of results suggests that "sustained" attention in fact exerts its facilitative effect on perception in a periodic fashion.

Fig. 1.

(Left) illustration of the experimental paradigm. Each run of trials consisted of a symbolic cue followed by four targets at the cued location (attended), four targets at the uncued location (unattended), and one catch trial without a target in randomized order. Note that size and luminance of the targets are magnified in this illustration. Right: (Right) 50% detection thresholds as estimated by the staircase procedures as a function of hemifield (left/right) and attention (attended/unattended). Thresholds were lower for attended stimuli and for stimuli presented in the left hemifield. Significant effects are marked with asterisks; error bars represent SEM.

Fig. 2.

(A) ERPs (grand average over 13 subjects). Time courses show data averaged across electrodes in a posterior–parietal region of interest (indicated by white markers in the topographies). Shaded time windows indicate the time ranges used for statistical analysis of the ERP. Topographies show ERP distributions in these time ranges averaged across attended and unattended hits. (B) GFP (colored lines) and area under the ROC curve (black/gray line). The time course of this area measure for each subject provides information about when the GFP contains the most information about perception (hits vs. misses).

Fig. 3.

Circular–linear correlation between EEG phase and poststimulus global field power. (A) Correlation computed across all conditions and channels. The color map represents P values, and the white outline delimits significant effects corresponding to a FDR of 5%. Correlations are most evident around 7 Hz at 400–100 ms prestimulus, and at frequencies <5 Hz after 100 ms poststimulus (corresponding to the early portion of the ERP). The topography shows the distribution of the correlation in the time–frequency range from 4 to 10 Hz and −400 to −100 ms. White markers indicate channels of the region of interest, used for subsequent analyses. (B) Correlations separately for unattended and attended stimuli computed for the fronto-central region of interest. Stronger and more sustained correlation between prestimulus phase and GFP is found for attended stimuli.

Fig. 4.

GFP and perceptual performance as a function of prestimulus phase at 7.1 Hz and −224 ms. (A) Relationship between prestimulus phase and GFP for attended and unattended stimuli. Single trials were pooled in 11 bins, centered on the preferred phase bin. Statistical tests compare the 10 nonzero bins, leaving aside the central bin. The curves indicate that GFP amplitudes for attended, but not for unattended stimuli, decrease monotonically the more phases deviate from the preferred phase angle. Error bars represent SEM. (B) Relationship between prestimulus phase and perceptual performance (hit rate) for attended and unattended stimuli. Conventions as in A. Detection performance for attended stimuli (but not for unattended stimuli) decreases monotonically the more phases deviate from the preferred phase angle.