Next, recall that predicate NPs have no index on the referential tier. This predicts that predicate NPs should be
possible antecedentsof identity-of-sense anaphora, since a descriptionis present, but notof referentialanaphora, since
there is no referential index. The prediction is correct:Joan became a doctor, and Sue became one/it/her too.
Both referential and identity-of-sense anaphora can be applied to Event-variables as well.
The second sentence of (33) is syntactically parallel to the second sentence of (28). But the antecedent of the pronoun
is different: this time it is the entire event of Sam's falling out of a tree, as noted in section 10.10. So we have a clear
analogy to standard identity-of-reference anaphora. Next consider (34).
The second sentence in (34) asserts that a distinct event occurred, with the same description as thefirst sentence.
Hence it parallels identity-of-sense anaphora.^211
12.4 Referential dependence and referential frames
So far the referential tier consists of a list of indices or indexical features that
398 SEMANTIC AND CONCEPTUAL FOUNDATIONS
(^211) The notationin (34) is notquite preciseenough.In particular,it is possibleto usedo itanaphora withnon-coreferentialsubjects:...and then Fred did it too. Working thisdetail
intotheanalysisrequiresa better accountoftopic-focus structure.Somehintsappear insection12.5.Againandtoointheseexamplesseemtohaveaneffectsimilar toanother:
they mark events of similar or related description as referentially distinct. The temporal use ofthenlikewise marks two events as temporally distinct, hence referentially
distinct.AnimportantproblemwithVPanaphorsis howtheir understoodconstituentsacquirereferentialstatus. For instancein(34),thesecond eventoffalling outofa tree
makes a claim about there being a tree, not necessarily the same one as in thefirst event. A robust tradition (e.g. Fiengo and May 1994) treats this in terms of“syntactic
reconstruction”of the anaphor. Herewe are committedto there beingno second eventexpressed at any levelof syntax; Culicoverand Jackendoff(1995) show this leads to
insuperable technical problems. However, at the moment I have no explicit alternative.