Fig. 3.

Task-relevant EEG components. (A) A linear classifier based on EEG data in 80 ms windows was used to discriminate between the two frequency conditions of interest. The smooth curve reflects group-averaged ROC values (Az) with SEM represented by shaded boundaries. Yellow asterisks highlight projections in which Az reached significance, and the dashed lines represents significance, based on randomisation tests (at p < 0.001). Coloured curve segments indicate the k-means clustering of scalp projections derived from the classifier topographies. Clustering revealed three distinct components, each systematically different temporally and topographically. The first cluster (black curve) spanned the epoch in which the stimulus was being presented; the second (purple) cluster comprises shorter-latency activity possibly originating from sensory-specific regions; and the third (orange) cluster reflects later-activated.