LECTURETTE #19 INTRO TO BINDING THEORY I. Reference, Coreference, and Binding In this Lecturette we will begin looking at REST's formal account of binding relations. This is an important area of the framework, involv- ing what often prove to be (or at least are often regarded as) the most critical constraints on `Move-Alpha', most notably the `Empty Category Principle' (ECP). `Binding' has a lot to do with `coreference', but they must not be con- fused. `Reference' and `coreference' are semantic relations, `binding' a syntactic one. As we shall see, it is possible for two constituents to be coreferent, i.e. to be in a semantic relation of coreference, without there being any syntactic binding relationship between them. However, if a binding relationship obtains between two constituents, then by defini- tion they are coreferential. We probably all have some sense of what `reference' means in linguis- tics. The assumption is that certain linguistic expressions *refer* to extra-linguistic entities. Archetypically, such expressions are assumed to be NPs, or at least some kind of projection of [+N -V], or what in a more theory-neutral context we might call `nominals'. Let us further acknowledge that verbs and their projections may not refer to extra-linguistic *entities*, but may refer to extra-linguistic *events* or *conditions* or something like that; at any rate, the typical verb has real semantic content and hence refers to *something*. And the same goes for adjectives; in fact there is a wide range of types of linguis- tic expressions that are at least potentially referential. Referentiality is a relation between a linguistic expression (E) and something (X) that is, properly speaking, not linguistic at all. X may be something, an entity or an event or a condition, in the `real world'; or it may `exist' only in some fantastic or hypothetical universe. (X may also be an `entity' in the `discourse context' -- an `objectified' linguistic expression. But that's a matter for a discussion of the grammar/pragmatics interface.) The important thing to bear in mind is that, when we are considering referentiality relations, X is generally *not* a linguistic entity. `Coreferentiality', on the other hand, is a relation that can exist be- tween two linguistic expressions A and B; all it entails is that, if A refers to X, then B also refers to X. Two linguistic expressions that have a common referent are ipso facto coreferential with each other. Again, coreferentiality is not a syntactic but a semantic relationship. In the extended passage in (1), the expressions `Bob Dole', `the Senate Majority Leader', `the senior Senator from Kansas', and `the front-runner for the Republican presidential nomination' all refer to the same entity and are therefore coreferential. But each is in a sentence by itself, and there is no evident syntactic relationship between them. (1) Bob Dole is a busy man these days. The Senate Majority Leader has been working hard hammering out ideological and procedural differences between the Senate and the House on the one hand and between Congress and the White House on the other, in order to maximize the passibility of certain pieces of legislation. The senior Senator from Kansas has also been touring the whole coun- try, making many appearances most notably in New Hampshire, the site of the first presidential primaries. The front-runner for the Republican presidential nomination has no intention of let- ting his position slip for lack of any diligence on his part. The archetypal `binding' relationship is that between a pronoun and its antecedent, as in the sentences in (2). If a given clause contains both a pronoun (or pronoun-like element) X and an antecedent A for X, then *a priori* there is a binding relationship between A and X. As we shall see, this is an oversimplification; it can occur that both A and X are in the same clause without any relationship between them that meets the criteria for a formal binding relationship. Binding relations can also exist between constituents that do not meet the pre-theoretical criteria just outlined. (2) a. Gefion(i) is looking for her(i) plow. b. Clio(i) bought herself(i) a book. >From the syntactic point of view, at least within REST, the relations in- volved in coreferentiality fall into two areas. To the extent syntactic theory has anything to say about coreferentiality in multi-clausal situa- tions like that in (1) it falls under the heading of `Control' Theory, about which i'm afraid i will have little to say. More critical is the kind of coreferentiality represented in (2), in which the two coreferen- tial constituents are within the same clause. This is the province of Binding Theory. Formally, the binding relation itself is defined as fol- lows: A constituent A is understood to be `bound' (in a binding relation) by another constituent B iff (if and only if) B c-commands A and A and B are coreferential. The last clause must always be borne in mind; obvi- ously, in any given tree diagram there are going to be lots of c-command relations, lots of constituents c-commanding or c-commanded by other con- stituents. But a binding relation can only exist between two constitu- ents that are in a c-command relation and in addition are semantically coreferential. II. Binding Theory Features and Principles In REST, Binding Theory is based on a taxonomy of constituents which are able to refer to extra-linguistic entities or concepts. Their *categori- al* classification is irrelevant here; they may be nouns, verbs, whate- ver. This taxonomy is in turn based on two binary features: pronominal and anaphoric. These two features are logically independent of each other; no entailments link them. Therefore, together they produce a four-way taxonomy: (i) There are constituents that are +ANAPHORIC but -PRONOMINAL; these are referred to as `anaphors'; examples are so-called reflexive and reciprocal `pronouns' such as `yourself' and `each other'. (ii) There are constituents that are +PRONOMINAL but -ANAPHORIC; these are referred to simply as `pronominals'; garden-variety personal pronouns fall into this category. (iii) There are constituents that are both -PRONOMINAL and -ANAPHORIC. The most obvious examples are ordinary noun phrases. These are called `referring expressions' or `R-expressions'. (iv) There are constituents that are both +PRONOMINAL and +ANAPHORIC. This is an interesting category which we shall discuss at length later. Note that, while in traditional grammatical terminology expressions like `yourself' are considered merely a special type of pronoun, in this theo- ry a hard and fast distinction is made between them and ordinary pronouns; from the REST point of view they are not pronouns at all. It is possible for a constituent to be both a pronoun and an anaphor but, as we shall see, such a constituent is *forbidden* to have any phonological manifes- tation! Apart from this peculiar fact, it is in theory possible for any of the other three categories mentioned above to be either overt or phono- logically null. We are approaching the point at whikch we can discuss the core of REST's Binding Theory, the Binding Principles themselves. I need first to clar- ify a couple of terms. The Binding Principles refer to constituents being `bound' or `free' within the definition of a `governing cateory'. In the early 80's, the definition of a `governing category' was rather complex, requiring the identification of (i) the constituent itself, (ii) its `proper governor' (in the sense of Government Theory, some other constituent capable of assigning it, e.g., a theta-role or Case), and (iii) a `SUBJECT' (i.e., some third constituent that could function as a `subject' if the governor were of an appropriate sort -- in gene- ral, this boils down to a Specifier of some type). The smallest category dominating all three of these was the `governing category' for the constituent in (i). Since the developments in Govern- ment Theory in the mid-80's, we can simply say that a `governing cate- gory' is one surrounded by a barrier. At least, that's the ideal. There are some occasional complications or difficulties with trying to get the details of Binding Theory right on this definition, which we can discuss further, but in general this simplification works pretty well. Remember that `binding' is a syntactic relation between a constituent and a possible c-commanding antecedent. The Binding Principles address how close to each other these two constituents can be. They are always referred to as Principles A, B, and C as follows: A. An anaphor is bound within its governing category. B. A pronominal is free within its governing category. C. A referring-expression is free. By `anaphor' and `pronominal' here are meant constituents that are +anaphor and +pronominal, respectively. Principle A means that an anaphor *must* be bound within its governing category, i.e., it must be c-commanded by a coreferential antecedent that is not separated from it by a barrier. (Please note that neither anaphor nor antecedent need be overt phonologically.) Principle B means that a pronominal may have an antecedent, but that said antecedent must be separated from the pronominal by a barrier. Whereas an anaphor can never be too close to its antecedent, there is a definite limit to how close a pronominal may be; within its governing category, it must be free (i.e., not bound by anything within said category). The empirical point motivating this distinction is the dichotomy represen- ted in the sentences in (3): (3) a. Martha painted her face blue. b. Martha painted herself blue. c. Martha painted her blue. At least in certain colloquial registers, these two sentences may be syno- nymous; it is quite possible to say `Martha painted herself blue', when all she has actually done is coloured her face. Nevertheless, there is a clear Binding-Theory distinction in the two words referring to `Martha' in the respective VPs. In (3a), `her' is contained within the NP `her face', which constitutes an adequate barrier for its Specifier (though not for itself); since the antecedent of `her' is outside this `governing category', it takes the form of a pronominal. In (3b), however, `herself' is dominated directly by the governing verb, whose subject is `Martha'; the governing category is the entire clause, which includes the antece- dent. Therefore, the two constituents involved in the binding relation are too close for the lower one to be anything but an anaphor. A prono- minal in this position, as in (3c), is only acceptable under the inter- pretation that it refers to someone else, i.e, in which `Martha' and `her' are not coreferential and therefore not in a binding relation at all. Under such an interpretation and only under such an interpretation is `her' in (3c) free within its governing category. Unfortunately, such a relatively straightforward analysis fails to ex- plain the distinction between the sentences in (4), since `a picture of herself' should be as much of a barrier-surrounded governing category as `her face' is. This is symptomatic of one of the problems in Binding Theory, to which we will have to return, i expect, in Lecturette #21. (4) a. Martha showed us a picture of herself. b. Martha showed us a picture of her father. If interpreted along the analogy of Principle B, Principle C declares that a constituent that is neither a pronominal nor an anaphor ought not to have an antecedent at all, at least in the Binding-Theoretic sense of another constituent that both c-commands it and is coreferential with it. Obviously it is impossible to aply this principle strictly, witness equational clauses like (5). Instead, Principle C must be understood to mean that an R-expression is free to have or not have an antecedent with- in its governing category. Ultimately, what is meant by the principles of Binding Theory is that the grammar won't bother searching for antece- dents for R-expressions but will make a push to finding them for ana- phors. And a failure to find a suitable antecedent within the governing category will result in a serious problem. (5) Bob Dole is the Senate Majority Leader. In Lecturette #20, we will get into what is probably the most interesting part of Binding Theory: the typology of empty categories. Best, Steven --------------------- Dr. Steven Schaufele 712 West Washington Urbana, IL 61801 217-344-8240 fcosws@prairienet.org **** O syntagmata linguarum liberemini humanarum! *** *** Nihil vestris privari nisi obicibus potestis! ***