Introduction: parameters and other kinds of variation

In the preceding chapters of this book, I concentrated on the putatively universal conditions on agreement – those conditions that shape agreement and concord in all natural human languages. But agreement is a relatively superficial phenomenon, where one expects to see variation as well as uniformity.

It is perfectly clear that there is significant variation at some levels. The most obvious type is the variation between languages with a great deal of agreement, where agreement is central to the expression of grammatical relationships, and languages where agreement is either absent or peripheral to the expression of grammatical relationships. The first class of languages are the head-marking languages of Nichols 1986, including the polysynthetic languages of Baker 1996; the second class of languages includes the isolating languages and pure dependent-marking languages. I have modeled this dimension of variation in part by saying that languages can vary as to whether or not functional heads like FA or FN are generated above APs and NPs, and in part by saying that functional heads may or may not be specified as being probes, which look for features to agree with. For example, Ps, Ts, and vs are functional heads with semantic content that agree with nearby DPs/NPs in some languages but not others. A language has lots of agreement if it includes suitable functional heads in its syntactic structures and those functional heads are designated as being probes. In contrast, a language has little or no agreement if either functional heads are not projected, or if the functional heads that are present do not count as probes.

A generalization to be explained

Even though agreement phenomena are some of the most familiar and well-studied aspects of grammar, there are certain basic questions that have rarely been asked, let alone answered. One such question concerns the fact that, in many languages, the three major lexical categories – noun, verb, and adjective – behave quite differently with respect to agreement. Words of all three categories can bear similar, even cognate, inflectional affixes. But verbs are consistently the most prolific agreers, often agreeing with their subjects in person, number, and gender features, and sometimes agreeing with their objects in these features as well. Adjectives clearly participate in agreement, but they do so more modestly: they rarely or never agree with more than a single nominal, and they can agree in number and gender but they typically do not show first or second person agreement forms. In light of these differences, traditional grammar often says that adjectives participate in concord, a distinct phenomenon from the agreement that involves verbs. Nouns are more conservative still. Although they are often inflected, sometimes with the same affixes that adjectives take, nouns do not need to agree with another NP in their environment the way that an adjective in a similar structural configuration does. My thesis in this book is that the true theory of agreement should be able to explain these robust cross-categorial differences, and that much can be learned about agreement by seeking a theory that meets this condition.

To illustrate the fundamental contrasts in agreement behavior that I have in mind, consider (1) from Swahili.

Most of the ingredients of the theory of agreement in the previous two chapters are either fairly standard or generalizations of notions that are fairly standard. The notable exception is the Structural Condition on Person Agreement (SCOPA), repeated in (1).

1. F can agree with XP in +1 or +2 only if a projection of F merges with a +1 or +2 element and F projects.

This highly specific condition is not part of standard theory. Section 3.3 gave a range of evidence showing that this condition correctly characterizes when verbs can and cannot agree in person with their arguments, in addition to ruling out first and second person agreement on adjectives quite generally. But as it stands, (1) is merely a descriptive generalization. The purpose of this chapter is to explore the SCOPA more thoroughly.

To accomplish this, I begin by investigating when person agreement is possible on categories other than nouns, verbs, and adjectives, confirming that the SCOPA is descriptively correct (section 4.1). I then go on to derive the SCOPA from more fundamental principles which go to the heart of what it is for any linguistic expression to be first or second person. The first step in this derivation is to distinguish the relationship of Agree that holds between a functional head and a nearby NP from the sort of agreement that holds between operators and the variables that they bind (section 4.2). I claim that first and second person agreement belongs properly to operator-variable agreement and not to Agree per se. It is not particularly surprising, then, that the two are subject to different formal conditions.