Problems With Pronoun Acquisition

Experimental studies have often shown that children have more problems with pronouns than was originally expected. One recent explanation of the discrepancy between empirical results and linguistic theory is that the Binding Theory is in need of reformulation to include only bound variable interpretations of pronominals as well as reflexives, with intended coreference relegated to the domain of pragmatics. The Dutch studies presented here explore children’s sensitivity to bound variable interpretations of reflexives and pronominals by investigating their understanding of VP-deletion sentences. The results show errors with pronominals that do not support the reformulated version of the Binding Theory as a better description of children’s acquisition problems. (DRAFT: non-final version!!) PROBLEMS WITH PRONOUN ACQUISITION 1. The traditional Binding Theory Knowledge of anaphora in relation to reflexives and pronominals is traditionally summed up in two of the principles of the Binding Theory (Chomsky, 1981): Principle A a bound anaphor, like the reflexive himself, must be bound in its governing category, and Principle B a free anaphor, like the pronominal him, must not be bound in its governing category. Based on the principles and parameters view of language acquisition, it has been proposed that the principles of the Binding Theory should have an innate basis, or be part of what is often called Universal Grammar. Children’s knowledge of the Binding Theory principles has been quite extensively investigated during the last decade, but the empirical evidence does not give as straightforward a picture as was originally expected. Children’s performance on anaphora tasks is more of a mixed success than had been predicted. Correct understanding of anaphoric reference with pronominals and reflexives is not instantaneously visible in very young children; but it does improve over time. Developmental patterns for sentences with reflexives and pronominals also do not necessarily run parallel. One of the more consistent results is that children relatively quickly and easily come to understand bound anaphors, such as reflexives. In the case of pronominals, development is usually (but not always) slow, often stagnating and problematic for many years, even up into middle childhood. In some experimental studies, children seem to be interpreting pronouns as if they were reflexives, and this is specifically the one error that they should not be making. To make the picture even more confusing, there are also situations in which children perform equally well on sentences with pronominals and reflexives, or even better on sentences with pronominals (Kaufman, 1992). If both Binding Theory principles are presumed to be part of innate linguistic knowledge, how is it possible that children often, but not always, seem to be able to start using one principle earlier and better than the other? Several proposals have been made as to what the problem is with the Binding Theory and the acquisition of anaphoric elements in general, or with Principle B and the acquisition of pronouns in particular (Grimshaw & Rosen, 1990, Koster, 1988). The main focus of this paper is the proposal that the traditional Binding Theory may be incorrectly stated and, specifically, that Principle B is in need of reformulation, separating grammatical binding from intended coreference interpretations. This separation would entail making a strict distinction between knowledge of a grammatical sort and knowledge of a pragmatic sort; the former presumed to be part of Universal Grammar and available from an early age onward, the latter presumed to be based on knowledge of the world which may be acquired only gradually, via experience. In the traditional version of the Binding Theory, reflexives are always given a bound variable interpretation, while pronominals are open to either a bound variable interpretation which is grammatically determined or to an intended coreference interpretation which is partially pragmatically determined. The fact that pragmatic rules are brought into play for pronominals and not for reflexives could result in an imbalance between children’s success with reflexives and pronouns. 2. The reformulated Binding Theory A recent attempt to reformulate the Binding Theory centers specifically on problems with pronouns: what must be explained grammatically and what can be explained pragmatically (Montalbetti & Wexler, 1985, Reinhart, 1983, 1986). The suggestion is that the core issue of a syntactic Binding Theory should be bound variable interpretation, and not intended coreference, of pronominals as well as reflexives. The reformulated Principle B, therefore, would also be limited to grammatical binding. Constraints on intended coreference of pronouns, in this view, should not be considered part of a theory of grammar, but limited to the domain of pragmatics. In Reinhart’s version of the reformulated Binding Theory, the anaphoric element must be syntactically bound, c-commanded by its antecedent, either within its governing category (for reflexives) or outside its governing category (for pronominals) in order to receive an interpretation. Reinhart’s (1983) definition of the reformulated Binding Theory is as follows: Binding Theory: Coindex a pronoun P with a c-commanding NP alpha (alpha not immediately dominated by COMP o r S’). Conditions: A if P is an R-pronoun alpha must be in its minimal governing category. B if P is a non-R-pronoun, alpha must be outside its minimal governing category (p. 71) For coreference, Reinhart (1983) describes a Speaker’s Strategy and a Hearer’s Strategy, which has been formulated as Rule I (Grodzinsky & Reinhart, to appear): Rule I (inference): A free NP, A, can be intended as coreferential with NP, B, in the same sentence, iff either a. it is impossible to replace A with a (distinct) anaphoric expression that can be bound by B or b. the coreference interpretation needs to be distinguished from the bound Recently, Chien and Wexler (1990) stated the reformulation of Principle B only in terms of the local domain: pronouns cannot have a bound variable reading with a c-commanding antecedent in the same local domain. The key to reformulating the Binding Theory, according to them, lies in the role of indices: anaphors must have the same index as the local c-commanding antecedent, while pronouns may not have the same index as the local c-commanding antecedent. The semantics and pragmatics must deal with how these indices are to be interpreted. In the case of Principle A, anaphors and their antecedents have the same index, and are intended to refer to the same entity; coindexation implies intended coreference here. Principle B rules out the possibility of a pronoun and a local c-commanding NP having the same index. The pronominal cannot be locally bound. The critical question for pronouns now is: when a pronoun has a different index from the local c-commanding NP, what does this noncoindexing imply? It can be demonstrated that it does not necessarily imply disjoint reference. Chien and Wexler give the following example to demonstrate that disjoint reference is not always the case: (1) Thati must be Johnj (2) a. *At least hei looks like himi b. At least hei looks like himj Coindexing in (2a) is ruled out because him is bound in its governing category. In (2b), where the pronoun is not bound and no binding principle is violated, noncoindexing does not necessarily demand disjointness: he and him can both refer to the same entity, the NP John. While two coindexed NPs must corefer, (2a), two noncoindexed NPs are simply free in reference, (2b). Chien and Wexler propose a Principle P (P=pragmatics), like Reinhart’s Rule I, to handle the pragmatic coreference cases like (2b) above, where he and him are not coindexed but nevertheless still can be coreferentially interpreted. The reformulated Principle B applies only to bound variable readings of pronouns and not to (2), which allowed a pragmatic, intended coreference interpretation of two noncoindexed NPs. In order for a pronoun to be interpreted as a bound variable, it must be coindexed with its antecedent and, as Principle B states, this must not occur in the local domain. Compare Reinhart’s example (1986, p 124) below with (2) above. (3) Each of the boysi brought hisi bear For all x (x is a boy), x brought x’s bear The pronoun his is coindexed with the quantified phrase Each of the boys, and is a variable bound by this antecedent. It does not have a fixed value here, since this value depends upon the choice of quantified antecedent. Quantified NPs have no reference. Now, look at an example of a quantified NP and an object pronoun c-commanded by that NP. Compare Chien and Wexler’s example (4) with (2) and (3) above: (4) Every womani looks like her*i/j