To apply the new approach to tracking the timecourse of attention described in the introduction, each timepoint in the unaveraged trial-by-trial data was fit to a linear model within each participant, predicting the values in the Oz channel additively from the values in the ipsilateral and contralateral channels, yielding regression weights for ipsilateral and contralateral channels that were then collapsed to a difference score. Finally, these difference scores from all participants was fit to a single GAM predicting this difference score as a smooth function of time, yielding the curve depicted in Figure 3, demonstrating a clear pattern of weights consistent with the expectation of facilitation (faster responses to cued targets) typically observed in behaviour at short cue-target intervals, and weights consistent with IOR (slower responses to cued targets) typically observed in bahaviour at longer cue-target intervals.