On Argument Structure and the Lexical Expression of Syntactic Relations

Hale, Keyser 1993

Hale and Keyser discuss how predicate argument structure, equation0.png-roles, and verbal alternations can be accounted for with a syntactic theory of the lexicon. They assume that most verbs have an underlying VP-shell structure (similar to Larson 1988 and Harley 2000). They also claim that x-bar theory and binary branching follow from Unambiguous Projection, but I don't find their argument convincing.

Argument Structure.

They wish to explain why the set of equation0.png-roles is limited, and why the UTAH might hold (identical equation0.png-role relationships correspond to identical DS structural relationships). They claim that we can explain both facts by assuming that "equation0.png-roles" are really just names for DS structural configurations. They assume a limited set of phrasal categories (V, N, A, and P) to account for the limited number of equation0.png-roles. They claim that the semantic content of equation0.png-roles (e.g., agentivity) follows from the constructions they represent. They propose the following (partial) list of definitions for equation0.png-roles:

  • Agent: Specifier of a VP with a VP complement.
  • Theme: Specifier of a VP with a PP or AP complement.

External Subjects.

In order to explain why denominals can't form transitives, Hale and Keyser assume that denominal VPs have no spec/VP. They justify this on the grounds that NP's don't semantically license an argument, and so Full Interpretation won't permit one. They then go further to say that the subject is generated "externally" at the level of IP. This seemed misguided to mee; I think it would make more sense to say that the denominals are part of a VP-shell configuration, and the subject comes from the outer VP (which would mean, according to their theory, that it is in "agent" position, and would be seen as "causing" the action).


Hale and Keyser use their framework to explain several alternation patterns (middle, transitives, inchoatives, etc.) for certain classes of verbs. In part, they base their explanations on the idea that verbs can require an outer spec/VP, if their meaning relates to the agent. For example, they say that "smear" indicates the manner of action of the agent, and thus, the verb can't be used without an agent. As a result, we can't get the inchoative ("Mud smeared on the wall").


author =       {Kenneth Hale and Samuel Jay Keyser},
title =        {On Argument Structure and the Lexical Expression
               of Syntactic Relations},
chapter =      2,
publisher =    {MIT Press},
year =         1993,
pages =        {53-109}