Anaphora

Introduction

In this paper I will try to examine how different academic fields approach anaphoric phenomena. The fields involved here are primarily philosophy and linguistics, with its sub- fields of semantics, syntax, and psycholinguistics. Especially in linguistics, Anaphora is tradi- tionally thought of in opposition to deixis. Philosophy is not so much directly concerned with anaphora, but rather more generally with the processes of reference; however this larger dis- cussion adds to an understanding of anaphoric processes too. Formal semantics, again, has developed a new dichotomy of bound anaphora versus referential ones – instead of the classi- cal one between anaphora and deixis. Given these different approaches, I will try to find those within the two broad fields which seem to be best compatible. I will especially present one linguistic approach that might be compatible with at least one strand of the philosophical lit- erature, and also with psycholinguistic insights into the matter. My hope is that Sag & Hank- amer’s paper of 1984 fulfills these criteria and thus allows for an interdisciplinary view of anaphora.

Sag & Hankamer’s approach will however only be addressed later. First, I will give an oversight of what anaphora is supposed to mean linguistically and then go on to present what philosophy makes of the phenomenon.

Anaphora as a Linguistic Phenomenon

Anaphora is part of the phenomenon of indexicality, which also includes deixis. This family of phenomena shares the fact that they cannot be easily accounted for by formal se- mantic theories. This is because they involve reference to a point outside of the linguistic realm of individual sentences, which is usually the basis for a semantic, truth-conditional analysis. Sentences that include indexicals cannot be easily evaluated as to their truth values, because these suddenly depend on extra-linguistic (or at least extra-sentential) features. Be- cause of this problem, indexicality has since long been regarded as the core domain of prag- matics. At the same time it was believed to be a relatively contained problem, which “in- volves a small set of terms and a fairly trivial mapping from the spatio-temporal context of an utterance to a (possibly unique) interpretation for each.” (Green 34). This belief is proved wrong as soon as one studies indexicality in more detail. Today it is clear that the problems evoked in this field are neither trivial nor contained. Rather, “the interpretation of indexicals

1

and related anaphora is shown to require assumptions of inferences about the speaker’s beliefs and/or intended referents, and thus requires a broader interpretation of pragmatics.” (Green 17).

Anaphora is the linguistic process whereby the speaker uses a relatively vacuous word to refer to an entity that has already been established in previous discourse. This can best be illustrated with an example:

(1) John likes pizza. He rode down to Wholefoods to get some for dinner.

The pronoun he refers to a previously established entity ‘John’. In contrast to John, which in- troduces the entity to the discourse, he only reinstates that entity. The mystery of anaphora is not how the entity enters the discourse, i.e. becomes a part of a specific context of utterance (token) from being a mere type (i.e. general word without context-specifics). This is rather a question for deixis. “Deictic reference involves a relation between an object in the world and a linguistic form with no semantically determined reference (some form X is used to refer to A), while anaphoric reference involved relations between such a form and some other linguis- tic expression (the form X is used to refer to something indicated or evoked by some other ex- pression).” (Green 23). Anaphora gives us the problem of what anaphoric expressions like he refer to, the word John or the concept ‘John’.

Anaphors can be more or less vacuous. Anaphoric pronouns (he, her), adverbs of time (then, now) and space (there, here), and demonstratives (this, that) are particularly vacuous. Other anaphoric expressions are less so, like epithets (the idiot, where idiot does not refer to a person that is categorically idiotic) and definite descriptions (the gray sweater, to refer to a previously introduced gray sweater). The most vacuous anaphora is ellipsis, where the ana- phoric phenomenon is linked with a void linguistic expression. 1

Syntactic and Semantic Approaches to Anaphora

Generative grammar can partly account for some anaphoric phenomena, namely pro- nouns and reflexives. However, these anaphors must exist within the same sentence as their antecedent, because the theory is based on a binding relationship between the two. Reference is thus understood as binding in this syntactic framework. The antecedent must be co-indexed with the pronoun and it must c-command it. 2 Because this binding relationship is prone to in- terference by other projections in the syntactic structure, antecedent and anaphor must be

1 These examples are taken from Green 25.

2 A node α c-commands a node β iff neither node dominates the other, and the first branching node dominating α

dominates β.

2

fairly close to each other. In that way, syntax can only account for a small number of ana- phoric phenomena.

Formal semantic theories like that of Heim & Kratzer take up this syntactic interpreta- tion of anaphoric expressions and try to build it into their theory. One asset of their theory is that they can account for anaphors that have antecedents in a different sentence, mostly the one preceding that of the anaphor. 3 They can also show the ambiguity of such sentences as (2) John hates his father.

This sentence can either mean that John i hates his own father i or that John i hates someone else’s father 2 . The approach can also show that the “referential construal” of anaphors is better suited to account for sentences like (3) The dog that greeted his master was fed.

than the “bound variable” approach. This is because the latter approach, like syntax, demands a c-commanding relation between antecedent and pronoun, while the referential approach leaves it to the context of utterance to make this connection. However, this only shows that neither syntactic nor semantic approaches 4 can account for the crucial point within anaphora: how do hearers know immediately what the speaker wanted the anaphoric expression to refer to? The mystery is still left to be solved.

Anaphora as a Philosophical Phenomenon

As far as I can see, philosophy is not primarily concerned with anaphora. However, the philosophical literature has quite extensively debated the question of reference, which has considerable influence on how one perceives the phenomenon of anaphora. Two strands of literature stand opposed: the first believes that linguistic expressions refer directly to objects in the world; the other rejects this claim because it can show that there is not always direct reference to the object in the world (the “referent”), and thus stipulates the existence of the mediating plane of “sense” 5 , “connotation” 6 or “intension” 7 . The impact of this quarrel for

3 Cf. Heim & Kratzer, ch. 9.

4 Heim & Kratzer also realize that some anaphora, so-called E-type anaphora, cannot be accounted for by either the bound variable approach or the referential construal because they are neither bound variables nor referential. Cf. Heim & Kratzer, ch. 11.

5 This term was introduced by Frege (1892) 6 This term is Mill’s (1843) as to Green 38.

7 This term goes back to Carnap (1947) as to Green 38.

3

anaphora is that the anaphoric expression could either refer to the world directly or as medi- ated by a layer of “intension”.

John Stuart Mill postulated in a paper in 1843 that proper names “are names of things, not of our ideas” (Mill 24). He argued that people were expressing beliefs about things and not about their beliefs concerning their ideas of them (24/25). Words are just names of things, “that which we intend to be understood by it when we use it” (25). In that way, one name or word can only have one meaning. 8 Mill also makes another claim about proper names, which will reoccur later in my discussion of Lambrecht’s psycholinguistic approach. “Proper names are not connotative: they denote the individuals who are called by them; but they do not indi- cate or imply any attributes as belonging to those individuals” (33). The words in the lexicon (for proper names) “are simply marks used to enable those individuals to be made subjects of discourse” (idem). They only pick out a referent, without attributing any characteristics to it. In his paper “On Sense and Reference” 9 of 1892, Frege challenged Mill’s approach as to his first point. If two distinct names (i.e. linguistic terms) a and b referred to the same refer- ent, a=a and a=b must be equal propositions under Mill’s theory. However, Frege can show that differences can arise “in the mode of presentation” (Frege 143) of the referent. In his by now classic example, “the morning star” and “the evening star” (146), Frege shows that, while in principle both denote Venus, they do not depend on this knowledge to have meaning. Someone not having this knowledge can legitimately assume at t 0 that a=1 while b=0, al- though Venus cannot exist and not exist at the same time. a and b can thus not mean the same thing – there must be an intermediate plane between the linguistic expression and the referent, which allows for the interpretation of the meaning. This is what Frege calls “sense” (146). In how far is this philosophical discussion fruitful for an understanding of anaphora? There are two problems with anaphora that might be understood within Frege’s framework. Anaphors apparently do not denote anything if not applied to a context. For example, the pro- noun “he” does not refer to a specific object in the world. Frege’s approach makes clear that this is an anomaly in language, because basically all other NPs do have some referent. “he” is however not completely void; it tells us the following:

8 One can wonder whether Mill’s approach does not already get in trouble with figurative uses of words, e.g. “He is an idiot.” Here idiot can denote either a mentally handicapped or an imbecile person. Mill could get around that problem by stipulating that the word idiot had really two meanings: idiot 1 = a mentally handicapped person; idiot 2 = an imbecile person. Such a solution however seems counterintuitive in the face of people’s creativity at understanding figurative speech. Also, there is a decision to make between the two readings. Mill’s theory just seems to miss a place where such decisions of hearers could come in when the linguistic input is ambiguous. One cannot discard this problem by saying that Mill’s theory only wanted to account for proper names; the same problem holds when I utter “Sam is nice.” denoting Sam 1 while my hearer believes I referred to Sam 2 .

9 Originally „Über Sinn und Bedeutung“.

4