Unaccusative Syntax and Verbal Alternations

Embick 2001

Embick demonstrates that in many languages, the morphological markings for unaccusative, passive, and reflexives are similar. He accounts for this similarity by saying that these constructions all share a common structural feature: they do not project an external argument; and that this structural feature is expressed morphologically.

He starts by presenting his syntactic analysis for each construction:

  • Passive:

    [vP [v {+AGENTIVE}] [equation1.pngP equation1.png DP]]

  • Unaccusative:

    [vP [v {-AGENTIVE}] [equation1.pngP equation1.png DP]]

  • Reflexive:

    [vP CL [v {+AGENTIVE CASE}] [equation1.pngP equation1.png DP]]

Where equation2.pngAGENTIVE indicates whether a verb is agentive; CASE indicates whether it includes accusative case; and CL is the reflexive clitic. The intuition behind the reflexive analysis is that CL is an anaphor ("himself"), that gets bound by the DP. In order for this to happen, DP raises to spec/vP. At the same time, CL needs to join with v. It's unclear to me whether the position that CL originates from is "spec/vP"; and if so, whether it's the same spec/vP as the position that the DP ends up in. After the movement, the reflexive looks like:

  • Reflexive:
    [vP DP1 [v t2 [v CL2+{+AGENTIVE CASE}]

    [equation1.pngP equation1.png t1]]]

After movement, none of the three constructions projects an external argument. But passive and unaccusative differ from reflexive, in that reflexive originally had an external argument.

After presenting the basic syntactic analysis, he shows how it accounts for data in Greek, Fula, and Tolkapaya. In each language, there are two morphological markings: (names are mine)

  • /-external/ is present for the three constructions that don't project external arguments. In all 3 cases, it attatches to the verb.
  • /+reflexive/ indicates reflexivity. It is typically glossed as "self," and might be an affix, a clitic, or some sort of affixed adverb. It is present on some (but not all) reflexives.

Also, in each language, there are two kinds of reflexives (names are mine):

  • inherent reflexives are just marked with /-external/. Only a limited set of verbs form inherent reflexives, and they tend to be verbs that might have some "inherent reflexivity" to them, such as "wash" or "dress" (c.f. "John washed," "John washed himself").
  • marked reflexives are marked with both /-external/ and /+reflexive/. They tend to be much more productive.

Embick is primarily concerned with accounting for the distribution of /-external/, which he explains using the absence of a projected external argument. For the actual implementation of the morphology, he relies on a distributed morphology framework, and proposes a morphological rule of the form:

  • v equation0.png v-X / no external argument

Where -X is an affix or a feature underlying the eventual morphological realization.

Finally, Embick considers two alternatives to his theory. First he considers the possibility of an explicity [-external] feature, which he dismisses on the grounds that it is not necessitated syntactically or semantically. Then he considers Reinhart's proposal that the reflexive is produced by a reduction operation on an unergative (and is not "unaccusative"). He dismisses her lexical reduction operation, on the grounds that it doesn't make sense to have two nearly identical rules in the lexicon and the syntax, and we know that we need the syntactic reduction operation. He then argues that most of Reinhart's arguments can be accounted for by the fact that the reflexive originally had an external argument. Thus, we should expect it to pattern with unergatives for some tests, since unergatives also originally had an external argument.

In his conclusion, Embick points out that it might make more sense to say that different syntactic constructions produce a single reflexive semantic operation (reduction), instead of saying that the reflexive semantic operation produces the syntactic constructs, because the relationship between syntactic and semantic constructs seems to be many-to-one.


  author =       {Dave Embick}
  title =        {Unaccusative Syntax and Verbal Alternations},
  note =         {to appear in Studies on Unaccusativity:
                  the Syntax-Lexicon Interface, CUP.},
  year =         2001