Figure 2.

Decoding of LFP. This figure represents the single-trial decoding accuracy as a function of task epoch for each LFP frequency band. Only correct trials from “Target” trials were included. Cutoff frequencies for each band are indicated on the y-axis. The results were aligned to cue onset or saccade onset, and merged along the x-axis between the 1000 and −400 time points. Insets, Decoding accuracy as a function of LFP band for three representative 400 ms time bins within the cue, attention, and saccade epochs. Thin blue bars on top represent statistically higher than chance decoding (p < 0.001). Non-zero decoding accuracies just before cue onset are due to a smoothing artifact caused by the integration window (400 ms). The time axis is truncated because of the time jitter between cue and response event onsets.