Single-trial ERP modelling reveals how task constraints modulate early visual processing

Abstract

We used a parametric design to study how task constraints modulate the relationship between single-trial ERPs and image noise. We used two pictures of faces equated in amplitude spectra and colourized with red and green tones. We added varying amounts of noise to the phase spectra. Thirteen subjects saw these faces in identity and colour discrimination tasks. Behavioural accuracy followed a noise dependent sigmoid in the identity task; it was high and independent of noise level in the colour task. Using a similar design, Philiastides, Ratcliff & Sajda [J. Neurosci. 2006, 26(35), 8965–75], suggested that brain activity before 200 ms is not modulated by task constraints, followed by a time-window sensitive to top-down influences. We used a single-trial ANCOVA [Rousselet et al. Frontiers in Psychology 1:19] to assess task modulation of ERP noise sensitivity while regressing out the main ERP differences due to identity, colour and task. Group analyses showed a reduction in noise sensitivity in the colour task compared to the identity task around 140–300 ms post-stimulus onset. A bootstrap spatial-temporal statistical analysis in every subject revealed more complicated results: significant task modulation occurred in 8/13 subjects, one showing an increase and 7 showing a decrease in noise sensitivity in the colour task. Onsets and durations of effects also differed between group and single-trial analyses, so that at any time point only 4 subjects showed results consistent with group analyses. A shift function analysis revealed non-uniform task effects on ERP distributions: decreased noise sensitivity in the colour task was due mostly to a modulation of single-trial ERPs to faces, not noise. In conclusion, single-trial analyses suggest that [1] early face processing can be modulated by task demand, at least in some subjects; [2] substantial individual differences in the time-course of task modulations speak against group statistics to study these effects.