Brief reply to Zwart

Zwart’s (2015) reply to my 2014 Language article (Bruening 2014) seeks to refute the claim in that article that precede-and-command offers much better empirical coverage than c-command for the conditions that regulate coreference (binding principles A, B, and C). However, nothing in what Zwart says addresses the question of whether there is a structural condition on coreference, much less what that structural condition is if there is one. The empirical part of Zwart’s paper simply repeats what we already knew: that principle C can be violated in certain discourse situations. This was already allowed in the formulation of principle C in Bruening 2014:372, ex. 109, which was based on Schlenker 2005.

1. (1). Principle C (Minimize restrictors): A definite description of the form the A may not refer to a discourse referent in active set C if A could be dropped without affecting either (i) the denotation of the description or (ii) its various pragmatic effects.

Clause (ii) permits a full R-expression if there are pragmatic effects that would be affected if the R-expression were instead realized as a pronoun. Schlenker (2005) discusses some of these pragmatic effects.

All of Zwart’s discussion only illustrates what some of these pragmatic effects might be. Since clause (ii) of principle C allows principled exceptions to principle C, all of the data in Zwart’s reply are consistent with the analysis in Bruening 2014 and do not affect it in any way, as far as I can see.¹ Now, Zwart does seem to imply that perhaps there is no structural condition at all on coreference and that all apparent effects of structure are actually these pragmatic effects. To show that, however, he would have to address every single type of example in Bruening 2014 and the voluminous literature on principle C, and this he has not done. I would welcome a thoroughly researched paper arguing that structure is irrelevant to coreference and all apparent structural effects are actually due to discourse. Zwart’s reply is not such a paper.

Zwart further claims that there are two different bans on coreference. Examples like the following are claimed not to admit discourse manipulation.

1. (2).

   1. a. *He₁ loves John₁.         (Zwart’s example 6a)

   2. b. *He₁ said that John₁ is an idiot.         (Zwart’s example 6c)

This is not true, however. Even Zwart acknowledges that coreference is allowed in 2a in certain circumstances (his n. 8). Coreference is also allowed in 2b in certain discourse situations, for instance where clarity of reference is required (No, no, no! John said that John is an idiot!). But these are the same circumstances that make violations of condition C acceptable everywhere.

Regarding reflexivity, Bruening 2014:376–82 showed that precede-and-command is indeed superior to other formulations out there, at least for English. Zwart goes on to suggest that c-command is the right condition for reflexivity. But c-command is both too strict (We talked to Bobby about himself/*him, where Bobby does not c-command the anaphor) and too permissive (it would permit long-distance reflexivity). Again, what Zwart says does not actually address the issues raised in Bruening 2014.

As for ‘conceptual necessity’, ideas about it simply have to take a back seat to empirical coverage. Bruening 2014 is just one work in a long series demonstrating the irrelevance of c-command to syntactic dependencies. C-command is not relevant to variable binding or weak crossover (e.g. Barker 2012, Bruening 2014:§7.3), negative polarity item licensing (e.g. Hoeksema 2000), superiority (e.g. Shan & Barker 2006), the each … the other construction, and possibly even extraction (Bruening 2014:383–84). No one disputes the claim that c-command is a structural relation that falls out from binary Merge; it just seems not to be one that is used by syntax.