Abstract
Attention is rapidly deployed in response to external or internal cues. However, experimental measurements of the speed of this deployment find large variability across trials. Beyond neuronal and measurement noise, some of this variability could depend on the specific state, or phase, of an ongoing “attention cycle”, with specific moments leading to faster deployment than others. To test this, we present a ‘clock’ with a single hand rotating at 1 Hz. After a variable period, the rim of the clock briefly turns red. Observers (n = 13) indicate the position of the hand at this cue onset. The difference between the reported and veridical positions, called latency, specifies the speed of attentional deployment for this trial. We find large within-observer variance in attentional latencies across trials (SD = 54 ± 5 ms). As predicted by the cyclical hypothesis of attention, this variance can be partly explained by the strong linear-to-circular correlation between trial-by-trial latencies and the corresponding phase of EEG oscillations. Significant (p < 10−12) correlations are found at 7 Hz in occipital cortex, 50 ms before cue onset, and at 16 Hz in frontal cortex, 60 ms after cue onset. Each area accounts for roughly 5–6 ms of trial-by-trial variability. Compensating for the variance contributed by one area leaves the influence of the other intact and vice-versa, indicating that the two contributions are independent. Based on these findings, we propose that attentional selection involves two distinct stages. First, under sustained attention, the state of occipital activity in the theta (5–10 Hz) range determines the speed of attentional capture by the cue. Following this, attention samples the position of the moving clock hand for later report; the timing of this sampling depends on the state of frontal areas in the high-alpha, low-beta (12–20 Hz) range. Together, the two effects determine the timing of attentional allocation.
This research was suppported by a EURYI grant and an ANR grant JCJC06-154 to RV.