1. PARTICIPANTS

Participants should ideally be familiar with the basic notions of syntax and morphology, as can be gained from any general introduction to linguistics.

Agreement is a remarkable phenomenon: it is widespread in the languages of the world, it occurs in various different constructions and yet it often serves no obvious function, since the information made available is typically available from elsewhere. Agreement shows dramatic variation (across languages and within languages), caused largely by the conflicting effects of morphology, syntax and semantics.

We shall starting from the definition and scope of the phenomenon, looking in turn at the range of controllers, targets, domains, features and conditions, and giving special attention to instances of variation, since these provide a window into the working of agreement. Examples will be taken from a wide range of languages: Indo-European provides well-attested data, while the languages of Daghestan are a gold mine, in terms of pushing the boundaries of the typology.

We shall begin with core instances of agreement, that any linguist would accept as examples, and then move on to the problem cases, showing that the boundaries of agreement are not clear-cut. This ‘canonical’ approach has applications more generally in typology.

We call the element which determines the agreement (say the subject noun phrase) the controller. The element whose form is determined by agreement is the target. The syntactic environment in which agreement occurs is the domain of agreement. And when we indicate in what respect there is agreement, we are referring to agreement features. Thus number is an agreement feature, it has the values: singular, dual, plural and so on.

We call the element which determines the agreement (say the subject noun phrase) the controller. The element whose form is determined by agreement is the target. The syntactic environment in which agreement occurs is the domain of agreement. Giving examples of possible controllers is relatively straightforward, but possible targets are more numerous and varied than most linguists realize. We then ask why these lists of the possible controllers and the possible targets are not sufficient, and why we also need the notion of ‘domain’. Exotic domains will be included at this point, to show just how different agreement can be from the standard examples.

First we survey the means of morphological expression of agreement, including the question of whether different features can be realized together, in all possible combinations of values. Then we examine critically the simple but not unreasonable view of agreement which would have the syntax establish the domains of agreement, the agreement features and their values, and leave the morphology with the apparently simple task of ‘spelling out’ those feature values. Things are somewhat more complicated, since the agreement forms available depend on the agreement target and on its type. The question then arises as to the outcome when the demands of the syntax in terms of agreement cannot be met by the morphology.

The three indisputable agreement features, namely person, number and gender will be dealt with first. Examples will be given of systems with larger sets of feature values than those found in the most familiar languages, that is, with more than three persons, two numbers and two genders. Then the less obvious and more disputable features will be considered. Finally we consider the feature values which occur as defaults (for instance, for agreement with absent controllers, as with ‘weather expressions’ in various languages) and here the importance of looking at larger systems will become clear.

Agreement may be determined in part by factors which do not themselves register as agreement markers. For instance, in Hungarian agreement in number with conjoined noun phrases depends on the animacy of the noun phrases. But there is no agreement marker which registers animacy directly. The other main factor which works in this way is precedence. Sometimes these factors are treated as covert features (thus NPs are marked as +/- animate), but this somewhat obscures the difference between them and genuine agreement features. These factors can have an absolute effect, making one agreement outcome obligatory or impossible, or else they may just favour a particular outcome.

Resolution rules determine the form of agreement to be used when the controller consists of conjoined noun phrases. They are required for all possible combinations within person, number and gender. The rules for person and number are relatively straightforward, while those for gender show remarkable diversity. For instance, in Slovene, the conjoining of feminine and neuter nouns gives rise to masculine agreement, as does the conjoining of neuter nouns only (two noun phrases headed by neuter singular nouns take masculine dual agreements). There will be an account of the motivation for these systems of resolution rules; those for person and number are anchored in the semantics of those features, while those for gender require careful examination of the gender assignment systems in the different languages.

While in the simplest cases the form of agreement in a given context is completely determined, there are many instances in which a choice is possible, and these choices can tell us a good deal about how agreement systems work. Thus if the controller is Mary, we find only Mary has decided (not *have decided); yet with some controllers there is a choice: the committee has decided or have decided. The factors which have an influence will be isolated and analysed in turn. First there is the type of controller: there are unique lexical items with special agreement properties, groups of such items (committee-type nouns in English), constructions which are lexically restricted (those based on a small set of quantifiers) right up to free constructions, such as conjoined noun phrases. Then the target has a substantial effect; we discuss the Agreement Hierarchy and the Predicate Hierarchy, which determine implicational relations between different target types. Being able to isolate these factors in different languages under difference circumstances, and then to generalize across the different environments gives a substantial insight into how agreement systems work.

Another important view of agreement systems is provided by the ways in which they develop. There is some interesting work on the rise of verb agreement systems, which is suggestive but which has not been followed up to give a fully convincing set of studies. We shall evaluate this account and then discuss the rise of other types of agreement, the loss of agreement, and then less dramatic changes in agreement systems, involving interesting shifts within existing systems.

While our main perspective is typological, it is worth considering further questions: what does agreement do?, how do children acquire it? and what can psycholinguistic research tell us about it?