Fig. 3: Quantifying the RIFT response at the single-trial level.

a (left, top) The coherence between the signal-trial MEG data and RIFT signal was quantified using a sliding-window FFT approach, whereby the coherence was estimated by averaging over the 0.1 s in the 0.2–0.5 s interval (moved in steps of 0.025 s). (left, bottom) The topography of the coherence (combined planar gradiometers) suggests a response in the occipital sensors. (right) The RIFT response to the target and distractor was then concatenated into one vector, and submitted to a GLM with the factors target colour (T), unguided (U), distractor colour (D), and time-on-task (tot). b The contrast between the regressors associated with the target colour and unguided, and between distractor colour and unguided, was compared to 0 using a cluster-based permutation test (5000 permutations). The model fitted to the set size 16 conditions yielded no significant results but suggested reduced responses for distractors compared to unguided stimuli (p = 0.08). c The model fitted to the set size 32 conditions replicated the magnitude-squared coherence results reported above, with a significantly stronger response to the target colour compared to unguided and a significantly reduced response to the distractor colour.