The syntax of (in)dependence (review)

This monograph explores the interaction between structural properties of sentences and the reference of nominal expressions occurring in them. Safir begins by stating that ‘one of the most important discoveries in modern linguistics’ (1) has been the interaction between syntax and interpretation in determining nominal reference. Indeed, this area is probably the most advanced topic in the syntax-semantics literature. Many nontrivial predictions of subtle principles have been obtained and a very rich theoretical understanding of the interaction between syntax and semantics can be found in the work of Noam Chomsky, Tanya Reinhart, Irene Heim, James Higginbotham, and many others including S himself. Some of this research is presented in a recent textbook by Büring (2005). At the same time, some tough problems loom in this area—hard problems that have withstood several attempts to find an elegant and convincing account over a number of years. These problems are the challenges this book takes on.

S’s central claim is that binding theory is not about binding per se, but about dependent identity. Dependent identity is indicated by the arrow notation of Higginbotham 1980 (S actually draws instead of ). The key difference between binding and S’s dependent identity concerns c-command. From common assumptions about the interpretative mechanism, such as those presented in the textbook by Heim and Kratzer (1998), it follows that binding is subject to a c-command requirement: the binder must c-command the bound element. S claims that dependent identity is subject to a weaker c-command prohibition, his extended independence principle (EINP). The EINP states that the dependent element or any nominal that contains it must not c-command the binder (52). The EINP predicts that the pronoun can be dependent on the name Bill in 1a and 1b, but not in 1c and 1d (56–57).

(1)

a.

b.

c.

d.

The EINP is designed to allow a dependency between a name and a pronoun exactly when ellipsis allows a sloppy interpretation. For example, the restriction that a dependency be blocked if a nominal containing the dependent c-commands the element it depends on is motivated by the absence of a sloppy interpretation in 2a. A small shortcoming here seems to be that examples like 2b are not considered. Here the EINP would allow a dependency, but a sloppy interpretation is nevertheless not possible. However, this is a minor flaw: The EINP could easily be revised to state that the dependent element or any PHRASE that contains it must not c-command the binder. With this revision, the EINP would correctly disallow a sloppy interpretation in 2b.

(2)

a.

b.

There is further support for this revision: On S’s account, the EINP must block the representation in 3a as part of the account of weak crossover. The analogous representation in 3b, however, is only ruled out by the revised EINP since the pronoun is contained within a prepositional phrase.

(3)

a.

b.

S’s departure from binding is a bold move. He motivates the departure from binding mainly on the basis of three empirical problems for binding analysis: (i) apparent binding from almost c-commanding positions, (ii) licensing of strict interpretations in ellipsis, and (iii) apparent binding from non-c-commanding positions. I discuss these three issues in more detail in the following paragraphs.

The first problem involves almost c-command. Almost c-command refers to the relationship between a possessive and the c-domain of the possessed nominal. As 4a shows, a quantifier seems to be able to bind a pronoun that it does not c-command in this case. At the same time, though, his and the boy in 4b do not show a condition C effect.

(4)

a.

b. His mother doesn’t think the boy is smart. (13)

S argues that the facts in 4 are problematic for the analysis of Grodzinsky and Reinhart (1993) since their Rule 1 would predict a condition C effect to arise in any configuration...