Agreement (review)

This is a book on the typology of agreement. In addition to defining agreement and describing what the author considers agreement phenomena, it provides a host of data from a wide range of languages. Greville Corbett is already well known for his books on grammatical categories such as Gender (1991) and Number (2000), and the present book fits nicely into this series of studies of basic grammatical phenomena from a typological perspective. In a way, the book can be seen as a summary and a conclusion of C's earlier work on agreement, published in journals and edited volumes since the late 1970s. The approach is typological and data-oriented rather than formal and theoretical. There is little if any discussion of how the data—of which there is an abundance—may support or refute hypotheses about universal grammar or a formal theory of agreement. All C does is review different morphological theories, while being cautious to push his own view.

The book has nine chapters. Ch. 1 ('Introduction') is, as the rest of the book, very well written and composed. After having reviewed several definitions in the literature, C presents the three major components of agreement: the controller, the target, and the domain. He then introduces the notion of 'canonical agreement', instances of which 'fix a point from which occurring phenomena can be calibrated' (9). This turns out to be a very useful concept as a point of departure for understanding all sorts of varieties and deviations from what we would expect from an agreement pattern.

Ch. 2, 'Controllers, targets and domains', returns to those three main elements of agreement. With canonical agreement as the point of departure, C explores the way in which the three elements may vary, thus accounting for a wide range of agreement phenomena. The notion of domain is particularly important in C's typology. He lists four types of domains, the NP, the clause, the sentence, and beyond the sentence. The wider the domain, the less canonical is the agreement relation.

Ch. 3, 'The morphology of agreement', deals with how information in the controller is represented on the target. The chapter starts with a brief introduction to current morphological theories, presenting a two-dimensional typology with a distinction between incremental and realizational theories on the one hand and lexical and inferential on the other. C's own preferred network morphology falls within the inferential-realizational type. Like the book in general, the chapter is rich in illustrating examples of all kinds of conceivable phenomena that fall within the topic of agreement the way it is defined.

Ch. 4, 'Features', is a real exercise in terminology and categorization, but useful and clarifying. There are three 'indisputable agreement features': gender, number, and person (125). These C calls phi-features, using a term that has become current in minimalist syntax. Gender and number are interpretable features on nouns, person, on pronouns; gender and person are inherent features (on nouns and pronouns respectively), but C is less clear about number. He does not discuss the possibility of a finite list of features. But it would be interesting to know what features exist universally, if any.

Ch. 5, 'Mismatches', treats all those cases where the features of the controller and the target do not match the way they canonically should. An important distinction is made between 'normal case default' and 'exceptional case default', the latter applying when 'something goes wrong' (148), as in the case of a controller (e.g. a clause) with no gender. This chapter is almost a catalog of funny grammars. It is amazing how C has been able to dig out the most esoteric and eccentric cases of agreement from all corners of the world.

Besides features and other elements of agreement treated so far, there are other conditions that determine agreement. In Ch. 6, 'Conditions', C argues convincingly that conditions such as animacy should be distinguished from features proper. Animacy (and other less obvious factors) may influence whether a certain possible target will agree, without...