Tracking pattern learning with single-trial event-related potentials

Objective: The main aim was to track the dynamics of pattern-learning using single-trial event-related potentials (ERPs). A new `learning-oddball' paradigm was employed presenting eight random targets (the `no-pattern') followed by eight regular targets (the `pattern'). In total, six repetitions of the `no-pattern' followed by the `pattern' were presented. Methods: We traced the dynamics of learning by measuring responses to 16 (eight random-eight regular) targets. Since this alternation of the `no-pattern' followed by the `pattern' was repeated six times, we extracted single-trial responses to all 96 targets to determine if learning occurred more rapidly with each repetition of the `pattern.' Results: Following random targets, ERPs contained a marked P3-N2 component that decreased to regular targets, whereas a contingent negative variation (CNV) appeared. ERP changes could be best described by sigmoid `learning' curves. Single-trial analyses showed that learning occurred more rapidly over repetitions and suggested that the CNV developed prior to the decay of the N2-P3 component. Conclusions: We show a new paradigm-analysis methodology to track learning processes directly from brain signals. Significance: Single-trial ERPs analyses open a wide range of applications. Tracking the dynamic structure of cognitive functions may prove crucial in the understanding of learning and in the study of different pathologies. (c) 2006 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.