Common independent components of the P3b, N400, and P600 ERP components to deviant linguistic events

Two of the most reliable and most studied neural correlates of language comprehension are the N400 and P600 ERP components, canonically elicited by semantically and grammatically anomalous words (respectively). Over two decades of research on these phenomena have yielded insight into the time course of sentence comprehension and language learning, and have established links between language comprehension and non-linguistic domains such as music perception. However, the ability of the N400 and P600 to inform cognitive neuroscience theory has been limited by the high likelihood that they are composed of multiple, functionally and anatomically distinct subcomponents whose contributions to these scalp potentials overlap in time and space. Methods for identifying and extracting these subcomponents are clearly needed to remove these limitations. One method for potentially extracting such spatio-temporally overlapping ERP subcomponents is independent component analysis (ICA). In this talk I will present the results of my dissertation relevant to its primary focus, the extraction of one putative subcomponent of the N400 and P600, the P3b ERP component, using ICA. The P3b is popularly believed to index the update of a probabilistic model of the sensory environment or the engagement of attentional resources, and is canonically elicited by rare targets in simple target detection tasks. Identifying a P3b subcomponent was done by applying ICA separately to N400/P600 and P3b datasets to find independent components (ICs) that appeared to replicate across the two decompositions. The results of this procedure suggest considerable though far from total overlap between the sources of the N400, P600, and P3b. In particular, there appears to be a common subcomponent with a predominantly midline central distribution that accounts for around 50% of the variance of the N400, P600, and P3b effects. This midline central (MCe) subcomponent is characterized by a late positivity that is enhanced to rare/anomalous stimuli, habituates over the course of the experiment, and tends to peak later for stimuli that presumably take longer to evaluate. In light of these characteristics, I have tentatively interpreted the MCe subcomponent as an index of the allocation of attentional resources. As there are several reasons why these results might be misleading, it is critical to validate them and my interpretation of them with additional research. If these results prove valid, they would improve our ability to analyze the brain's response to language via ICA, add much needed subtlety to our functional understanding of this response, and potentially enhance our understanding of the neural mechanics of this response by making it possible to relate language comprehension to animal models of the P3b