Page 234 - Encyclopedia of Philosophy of Language

Reference
In (30) it is shown that the appropriate elements can take VP (learn to lambada) or N' (gun) antecedents; (32) is an example with S serving as an antecedent; and (33) shows that, in some cases, nonconstituents (in this case, make...wash the floors) may serve as antecedents.
I know that [Pat was here],, but I don't know (32) when ,.
They may make Pat wash the floors, but they (33) wouldn't do it to Chris.
Hankamer and Sag (1976) distinguished two classes of anaphoric elements. Anaphoric elements of the first class, exemplified by ordinary third-person pronouns, can derive their interpretations from anything that is semantically appropriate and contextually salient. The others, exemplified by VP ellipsis, require a linguistic antecedent of a certain syntactic type. They called these 'deep' and 'surface' anaphora, respec- tively. The following examples provide a minimal con- trast between the two types:
Pat hat jemand gesehen, aber ich weiB (37b)
[Pat tries to jump over a fence and trips]
(a) Chris: I'll bet I could do it. (b) *Chris: I'll bet I could .
(34)
However, examples like (38) pose a problem for deletion analyses, because the would-be deletion site is inside its antecedent, so that trying to reconstruct the predeletion structure leads to an infinite regress. This problem is not insurmountable, provided that the notion of 'deletion under identity' is formalized with appropriate care, but there is considerable controversy regarding how best to handle such phenomena. (The problem presented by such 'ante- cedent contained deletions' is reminiscent of a cel- ebrated argument known as 'the Bach-Peters Paradox.' Examples like (i), it was claimed, show that pronouns cannot be transformationally derived from full copies of their antecedents, without positing infi- nite underlying structures, (i) [The pilot who shot at it,], hit [the MIG that chased himj,.)
Pat [reads everything Chris does (38)
Though the literature on these other types of ana- phora is rich, there is no treatment of them that enjoys the same sort of currency as the Binding Theory for pronouns with pronominal antecedents.
2. Semantics
The nature of 'identity' which anaphora is supposed to represent has also been the subject of much research. For example, the VP-ellipsis examples in (39) show that strict (syntactic) identity is not at the basis of such constructions.
A: Do you think they'll [like me],? (39) B: I'm sure they will ,-.
='like you'
^ 'like me'
As indicated, what is reconstructed is not the form like me, but rather some semantic unit 'like x,' where x is anchored to the speaker for the first sentence (and hence would be referred to as you by B).
2.1 Coreference
The usual idea about pronominal anaphora is that the pronoun refers to the same individual as the ante- cedent—thus, the two corefer. An example like (40a) then, would be interpreted as in (40b).
(a) John, read his, mail. (40) (b) John, read John.'s mail.
However, coreference is just one semantic relation between a pronoun and its antecedent.
Likewise, surface anaphora requires a greater degree of syntactic parallelism between anaphoric element and antecedent than is required for deep anaphora:
(a) Pat's phone was tapped by the FBI, though (35) they claim they didn't do it.
(b) ?*Pat's phone was tapped by the FBI, though they claim they didn't .
Hankamer and Sag argue that surface anaphora is the result of deletion under identity, but that deep anaphoric elements are base generated. There are a number of other arguments supporting a deletion analysis of surface anaphora. For example, the case marking in (36) depends on structure that can be miss- ing on the surface.
(a) Someone helped Pat, but I don't know (36) who/*whom (helped Pat).
(b) Pat helped someone, but I don't know who/whom (Pat helped).
This argument is even clearer in a language like German, where idiosyncratic case (such as the dative marking on the object of helferi) must be preserved.
212
Pat hat jemand geholfen, aber ich weiB Pat has someone helped, but I know
(37a)
(Pat geholfen hat). (Pat helped has)
nicht wem/*wen
not who-DAT/*who-ACC
Pat has someone seen,
(Pat gesehen hat). (Pat seen has)
but I know
nicht wen/*wem
not who-ACC/*who-DAT