Against the Unaccusative Analysis of Reflexives

Reinhart, Siloni 1999

Reinhart and Siloni argue that reflexives should be analyzed as unergatives (i.e., the subject is not an underlying object). Under their account, a "reduction" operation is applied to the reflexive verb, which merges the two equation0.png roles, and eliminates the verb's accusative/dative case feature. Thus, the verb is left with a single equation0.png role (the external role), which is assigned to the subject. (The subject's case is checked by infl). The reduction operation can be applied either in lexicon, or in the syntax (at LF). Reinhart and Siloni claim that this difference in the module where reduction applies accounts for differences between how reflexives pattern in different languages. E.g., reduction applies in syntax in French (which has productive reflexives, and can apply reflexives to ECM verbs); but applies in the lexicon in Hebrew or English (which have non-productive reflexives, and can't apply reflexives to ECM verbs).

Reinhart and Siloni begin the article by presenting arguments against a simple unergative account for reflexives: that reflexives are basically just transitives; and that the reflexive clitic originated as the object. E.g., "Jean se lave" would have the DS structure "[Jean [lave se]]." But this analysis seems untennable, given the numerous ways in which "Jean se lave" and "Jean le lave" seem to operate different symantically.

But Reinhart and Siloni point out that this is not the only unergative account for reflexives, and go on to give their reduction-based account. They impose a restriction on reduction, to help account for the distribution of reflexives: it must apply to a pair of free co-equation0.png-roles, one of which is external.

They then go on to discuss the unaccusative account, and present some problems with it. In particular, they claim that it's coverage is no better than their account; and show several ways in which reflexives pattern with unergatives (and not with unaccusatives). For example, you can make agent nominals from unergatives ("he is a good runner") and reflexives ("he is a good dresser") but not unaccusatives ("he is a good mover"). (But c.f. "The plant appears to be a good grower and producer.", found via google.)

Finally, they argue that reduction applies in the lexicon in some languages, and in the syntax in others. In particular, they use this to account for the productivity of reflexives, and for whether it can occur in ECM constructions:

  • Jean se considere intellegent

Where the two merged equation0.png roles come from two different verbs (external equation0.png from consider, internal from intellegent).


  author =       {Tanya Reinhart, Tal Siloni},
  title =        {Against the Unaccusative Analysis of Reflexives},
  note =         {to appear in Studies on Unaccusativity:
                  the Syntax-Lexicon Interface, CUP.},
  year =         1999