Syntactic Theory


The Specifier-Head Agreement Constraint (SHAC)

The Specifier-Head Agreement Constraint guarantees agreement between the specifier and the head. This is needed for:

  • person and number agreement between the subject and a finite verb in a sentence.
    Pat walks/*walk, I am/*are/*is from Germany.
  • number agreement between a determiner and a head noun in a noun phrase.
    this book/*books, those books/*book

The effect of the SHAC can be summarized as follows:

Informal characterization of the Specifier-Head Agreement Constraint:
A lexical element that requires a specifier agrees with its specifier.

The SHAC in Sag/Wasow/Bender 2003

In Sag/Wasow/Bender 2003 the SHAC is not a principle on the sort word or phrase, but a principle on a subtype of lexeme, on the type inflected-lexeme (p. 495).

The Specifier-Head Agreement Constraint
For each inflected-lexeme: The value of SYN HEAD AGR of the lexeme is identical to the SYN HEAD AGR of the element in the SPR list.

The effect of the SHAC belongs to the most complicated aspects of the book. Note that it is a constraint on lexemes, but refers to a valence value (SPR). Typically, valence values are not constrained on lexemes, because this is only done at the word level, where the Argument Realization Principle distributes the elements of the ARG-ST list over the valence values.

A simpler formulation of the SHAC

We can formulate the SHAC as a principle on words with specifiers. One possibility is the following:

The Specifier-Head Agreement Constraint:
if a word has a non-empty SPR value, then its SYN HEAD AGR value is identical to the SYN HEAD AGR value of the element on its SPR list.

This reformulation is not fully compatible with all analyses proposed in Sag/Wasow/Bender 2003. In particular, it might be problematic for the analysis of subject-auxiliary inversion in their Chapter 13.

Another question is whether we really want a general principle such as the SHAC. If we use the feature SUBJ for subjects and SPR for determiners of nouns, we need a second version of the SHAC to account for subject-verb agreement. But this is an open issue.

What is important is:

  • In HPSG we specify agreement at the level of the lexical head of a phrase, not as a principle on phrases.
  • There are several possibilities to specify a principle that ensures the relevant identities.

Back to the page on general principles!