Abstract

The prevailing assumption is that an Research underlying this study was conducted with support of the Russian Foundation for Basic Research grant #17-06-00460.agreement feature originates in one linguistic element, that is a controller, and is copied onto another one, a target. This form-to-form approach encounters massive difficulties when confronted with data, such as missing controllers or feature mismatches. A cognition-to-form mapping approach is proposed instead, suggesting that agreement features, such as person, number, and gender, are associated with referents in the cognitive representation. They serve to specify referents on either notional or conventional grounds, and are thus referential features. Referential features are mapped onto various sites in linguistic structure, including inflections. Parallel agreement between various sites is observed, as a side effect of mappings from the same cognitive source. Languages differ in which and how many sites for marking referential features they require. Analysis of Russian evidence suggests that the cognition-to-form mapping approach has a much greater explanatory force than the traditional form-to-form approach. There are only peripheral classes of instances in which form-to-form agreement may be needed as a supplementary factor. In general, the roots of agreement lie in cognitively motivated discourse processes associated with reference.