Abstract
The literature on processing of person and number agreement contains some apparently contradictory results. On the one hand, some ERP studies do not find a qualitative difference between person and number when an agreeing verb does not match the features of its subject, the controller of the agreement relation (Silva-Pereyra and Carreiras, 2007; Zawiszewski et al., 2016). On the other hand, an ERP study reported in Mancini et al. (2011b) did find a qualitative difference between agreement violations in person and agreement violations in number, a result further corroborated by an fMRI study reported in Mancini et al. (2017). At the same time, there is also a trend on which the literature appears to agree: on the whole the response to agreement violations in person is stronger than the response to number agreement violations. In this paper we argue that the constellation of reported results can be accounted for by adopting a theory of person and number features that has the following two core properties: (i) pronouns are specified for both person and number, but regular NPs are specified for number only and do not carry any person specification; (ii) all of first, second and third person are characterized by one or more person features, whereas, in contrast, one of the numbers (singular) corresponds to the absence of number features.