ABSTRACT
The description of English rhythm and intonation must obviously be concerned with ‘connected speech’ (this being in contrast, presumably, with single words spoken in isolation); but a particular conception of what constitutes connected speech – speaking in complete and grammatical sentences, and most typically in a dialogue – has dominated most descriptions. As a result, there is a large class of examples of English expressions whose rhythm and intonation are largely undescribed; examples are prices, telephone numbers, exchange rates, credit-card numbers, addresses, postcodes, flight numbers, times. Many of these have their own characteristic rhythmic and intonational treatments, which are important to the listener in processing the information conveyed. In this paper we describe a study of one such neglected ‘connected speech’ activity, that of spelling aloud. Seven subjects were recorded spelling aloud fifty-six words varying in length from four to fourteen letters. The recordings were analysed auditorily to identify word-group boundaries. Though there were differences between speakers in the way in which words were split into groups, there was also remarkable uniformity in the general pattern and in average length of groups. On the basis of these findings, we are able to propose an algorithm which generates an acceptable phrasing. The algorithm shows that a realistic word-group division (including idiomatic handling of the use of ‘double’ for double letters) can be produced by a mechanism which works from left to right looking at the word to be spelled through a window just three letters wide.
