ABSTRACT
Sentences like (1)-(3) present a dilemma in international analysis. On the one hand, the vocative in (1), the reporting clause in (2), and the tag in (3) tend to be marked off by a pause from the stretch of speech they follow, while on the other hand they do not appear to be accented and thus do not have a nuclear tone. On the basis of the first characteristic, an analysis of each of (1), (2) and (3) into two tone units would be called for, but on the basis of the second, an analysis as a single tone unit would be more appropriate, since the intonation of the whole utterance represents a single nuclear tone.
Were you THERE, Jonathan?
'We're not GOing', he said
Missed the BUS, has he
If we adopt the first analysis, it becomes difficult to explain why in the second tone unit of (1) to (3) fewer intonation contours are possible than in the first, or indeed than in other tone units in general. If we adopt the second analysis, the problem is how to account for the prosodic break at the position indicated by the comma. In this paper, I argue that the solution should be based on the recognition that the association domain of a tone cannot be identified with any one constituent in the prosodic hierarchy. Divorcing association domains from prosodic constituency does not imply that association domains do not respect the boundaries of prosodic constituents. Rather, the claim is that, instead of there being one particular prosodic constituent that can be identified with the association domain of a tone, any one of a number of constituents from the Foot onwards can define the rightmost boundary of the association domain. Below, I briefly discuss earlier analyses of (1) to (3). Then, in section 3, a new solution is presented. There, it is shown that the tonal association domain is not coextensive with any one prosodic constituent, and an account of prosodic phrasing is given in which the prosodic hierarchy and tonal association domains play separate, though interdependent, roles. Also, I will argue that the intonational and prosodic status of reporting clauses (cf. (2)) differs from that of vocatives and constant polarity tags, a difference which in the account given here finds a natural explanation.