Making Sense of the Sense Unit Condition

1 Introduction

Traditionally, it has been assumed that listener preferences for intonational boundary placement fall under the domain of linguistic competence. Under this view, native speakers of a language possess specific linguistic knowledge that determines permissible intonational phrasings for a given utterance. Although a number of theories of this type have been proposed (see, e.g., Nespor and Vogel 1986, Hirst 1993), Selkirk's (1984) may be the most successful to date. Selkirk proposes that the distribution of intonational phrase boundaries can be accounted for by a semantic constraint called the Sense Unit Condition (SUC).

(1) The Sense Unit Condition of IntonationalPhrasing

The immediate constituents of an intonational phrase must

together form a sense unit. Two constituents C_i, C_j form a

sense unit if either (a) or (b) is true of the semantic interpretation

of the sentence:

1. a. C_i modifies C_j (a head)

2. b. C_i is an argument of C_j (a head)

The SUC makes the following predictions for the sentences in (2):

(2)

1. a. John gave the book // to Mary.

2. b. *John gave // the book to Mary.

3. c. John gave // the book // to Mary.

According to the SUC, (2a) is acceptable because both intonational phrases in the utterance form sense units. In (2b), the SUC is violated because the intonational phrase the book to Mary does not form a sense unit. The book and to Mary do not participate in a head-argument or head-modifier relationship. If an additional intonational boundary is added after book, as in (2c), the SUC predicts that the sentence should become acceptable because all three resulting intonational phrases form sense units.

An alternative account of the peculiarity of (2b) is provided by a theory grounded in processes involved with understanding and producing language. In particular, the Anti-Attachment Hypothesis (AAH; Watson and Gibson, in press) in (3) provides an explanation of the judgments in (2).

(3) Anti-Attachment Hypothesis

Listeners prefer not to attach an incoming word to a lexical

head that is immediately followed by an intonational phrase

boundary. As a result, the presence of a boundary at a local

attachment site increases processing difficulty, and the presence

of a boundary after a word that has no subsequent attachments

decreases processing difficulty.

The AAH accounts for the judgments in (2) as follows. Sentence (2b) is less acceptable than sentences (2a) and (2c) because it includes a misleading cue: the intonational boundary between gave and the book. The presence of this boundary suggests to the listener that the NP the book does not integrate with the preceding verb gave; but this is incorrect, leading to an increase in processing difficulty. Sentence (2a) does not contain the misleading cue, and so this sentence sounds better. Sentence (2c) contains the misleading cue, but it also contains an additional helpful cue: the intonational boundary between the book and to Mary. This cue improves the acceptability of the sentence, because no additional words attach to the immediately preceding site the book.

In Watson and Gibson, in press, we propose that the AAH follows from listeners' implicit understanding of the relationship between intonational phrasing information and syntactic structure during the production of a sentence, and that they use this knowledge to infer aspects of syntactic structure when comprehending a sentence presented auditorily. Support for the AAH comes from people's preferences in interpreting ambiguity and from complexity effects in unambiguous structures (see Watson and Gibson, in press, for a full discussion). For example, the intonational boundaries placed in the globally ambiguous sentences in (4) bias listeners toward one interpretation over the other.

(4)

1. a. The cop saw // the spy with the telescope.

2. b. The cop saw the spy // with the telescope.

In (4a), the boundary after the verb saw biases the listener toward an interpretation where the PP with the telescope modifies the noun spy (Schafer 1997, Carlson, Clifton, and Frazier 2001).According to the AAH, the intonational boundary acts as a cue not to attach incoming items to the verb saw, so the listener interprets the PP as a modifier of the direct...