On the left anterior negativity (LAN) in electrophysiological studies of morphosyntactic agreement

In a review of event-related potential (ERP) studies, Molinaro, Barber, and Carreiras (2011) lay out a theory of the temporal and neural dynamics of grammatical agreement compre- hension. As is clear from their review, ERPs are an excellent tool for studying agreement processing because they have a temporal resolution high enough to detect transient events in the brain, such as those that characterize language compre- hension, and their multidimensional nature allows one to make inferences about qualitatively dissociable cognitive processes engaged during real-time processing (Otten & Rugg, 2005; Rugg & Coles, 1995). In particular, their theory accounts for findings across domains of linguistic agreement (e.g., determinerenoun, subjecteverb) and agreement feature types (e.g., person, number, gender) and proposes three separable sub-stages in the processing of sentence-embedded agreement anomalies. While the latter two stages in their theory are indexed by a large positive-going wave in the anomalous relative to well-formed condition (the P600 effect), it is the ERP index of the first stage that is of concern here. According to their theory, this first stage occurs between approximately 300 and 500 msec after presentation of an anomalous word, and reflects a syntactic analysis where a morphosyntactic violation is detected based on a mismatch with predicted features. The ERP index of this stage is a negative-going wave in the anomalous compared to well- formed condition, most prominent over left anterior elec- trodes: the left anterior negativity (LAN).

Whereas the P600 effect is found nearly uniformly across languages, experimental tasks, and presentation modalities for agreement-violating words, the LAN shows considerable cross-study variability in its scalp topography, or even whether it is elicited at all (see Tanner & van Hell, 2014, for a review). Nonetheless, Molinaro and colleagues maintain that the LAN is one of the defining electrophysiological markers of morphosyntactic agreement processing, and discuss some issues that may relate to its presence or absence across studies. In addition to outlining several linguistic criteria that need to be met in order for a LAN to be elicited, they address two other issues that could potentially impact the presence or absence of a LAN: the reference electrode(s) chosen during data processing and the role that individual differences in ERP responses could play in giving the illusion of a LAN effect in the grand mean. These two issues will be addressed in this commentary.