An Idiomatic Argument for Lexical Decomposition

1 Boots, Creeps, Flak, and Verb Parts

Idioms have long been regarded as conforming to some kind of locality requirement constraining the relations between their parts. This requirement has taken various forms in the literature (see Marantz 1996, Nunberg, Sag, and Wasow 1994, and references cited there), and it will not be important here to develop a specific version of it; in the interests of concreteness, we might assume (1), adopted from Koopman and Sportiche 1991.

(1) If X is the minimal constituent containing all the idiomatic
material, the head of X is part of the idiom.

Requirements like (1) have sometimes been used to argue that certain kinds of sentences involve more structure than is immediately apparent on the surface. For instance, given an assumption like (1), the data in (2) have been used to argue for the existence of NP-raising.

(2)

1. a. The cat is out of the bag.

2. b. The cat seems to be out of the bag.

If (1) must hold in (2b) at D-Structure (or at LF following reconstruction, depending on one's additional assumptions), then the well-formedness of (2b) is evidence that the idiom chunk the cat begins the derivation somewhere lower in the tree, closer to the rest of the idiom with which it is associated.

A similar argument was given by Koopman and Sportiche (1991) for the VP-Internal Subject Hypothesis, based on alternations like those in (3).

(3)

1. a. The shit hits the fan.

2. b. The shit has hit the fan.

3. c. The shit should have hit the fan.

The alternations in (3) show that inflectional elements (e.g., tense and agreement morphology) are not part of the idiom in question; the idiom contains only the material that all the sentences in (3) have in common, namely, the shit hit the fan. If (1) is correct, then, the idiom chunk the shit must have begun lower in the structure than its surface position, perhaps in the specifier of the constituent headed by hit. Crucial to this account is the belief that hits in (3a) is syntactically as well as morphologically complex; the verb hit is part of the idiom in question, but the affix -s is not, and is generated in a syntactically higher head.

In this squib I would like to develop an argument of a similar form, based on alternations like those in (4)-(6) ((6) is adapted from Larson 1988:341).

(4)

1. a. Mary gave Susan the boot.

2. b. Susan got the boot (from Mary).

(5)

1. a. Bill gave John flak (about his behavior).

2. b. John took flak from Bill (about his behavior).

(6)

1. a. The Count gives everyone the creeps.

2. b. You get the creeps (just looking at him).

I would like to suggest that the reasoning employed in (3) be used here as well. We decided to regard the sentences in (3) as involving an idiom consisting of the structure that all the sentences had in common. Applying this reasoning to (4)-(6), we might conclude that the idioms in these sentences involve a noun phrase together with a portion of verbal structure that the verbs give, get, and take all have in common; I will refer to this verbal structure, following Harley (1995, 1997, 1999), as HAVE. Thus, the idioms in (4)-(6) would be HAVE the boot, HAVE flak, and HAVE the creeps, respectively. HAVE can combine with other material to form verbs; we might regard give as CAUSE + HAVE, and take and get as BECOME + HAVE, for example.¹ On this theory (4a) and (4b) might be diagrammed as having (partial) underlying structures like those in (7), with the idiom boldfaced (following Harley (1995, 1997, 1999), I represent HAVE as a preposition).

(7)

1. a.

2. b.

In (3a) the properties of the idiom in question suggest that the phonological word hits should be divided into two parts, one of which is part of the idiom while the other is not. Similarly, on this theory, the idioms in (4)-(6) show that verbs like give, get, and take should be divided into several parts, one of which is...