Quantal theory of speech timing

This is a summary view of temporal patters we have observed in the analysis of Swedish text reading. Special attention is devoted to trends of quanta1 structures in the timing of vowels and consonants, syllables, interstress intervals and pauses. It is well known that pause durations increase with increasing syntactic order of boundaries. A recent study supports our previous findings of multiple peaks in the histograms of pause durations and our theory of neural coordination of pause durations and prepausefinal lengthening. These add to an integer of a basic timeconstant of about 0.5 sec which reflects a local average of interstress duration and preserves a quasi-rhythmical continuity of interstress intervals spanning a pause. Regularities in the timing of syllables and phonetic segments with due regard to relative distinctiveness and reading speed will be discussed and, on a higher level, tempovariations within a sentence. Segments and syllables The main source of data to be discussed here derives from studies by Fant & Kruckenberg, (1989) and Fant, Kruckenberg & Nord (1991A,B). The text was a passage from a Swedish novel of about 10 minutes duration read by our reference subject, a Swedish language expert. A databank search system organized within a linguisticframe was developed for the processing. Our analysis has been concerned with individual vowels and consonants, syllables, interstress intervals and pauses. In addition we have data from 15 other subjects reading a limited part of the text. Durations were measured by hand from broad band spectrograms. The concept of quantally structured durational data is not new. Girding (1981), in a study of contrastive prosody, proposed a model for read Swedish in which the duration of an unstressed CV syllable is the unit. Two such units are allotted syllables with either a long vowel or a long consonant, i.e. stressed syllables. Phrase final lengthening is allotted one extra unit. Our modelling supports these simple rules but is more extensive and shows a clear tendency of octave relations between major categories. Interstress intervals, measured from the onset of the vowel in a stressed syllable to the onset of a vowel in the next stressed syllable, excluding those spanning a pause or a syntactic boundary, averaged 540 ms. The average duration of primary stressed syllables as well as those of secondary stress in compound words was 270 ms. Unstressed syllables averaged 132 ms. Mean phoneme duration was 70 ms. Unstressed vowels averaged 59 ms and unstressed consonants 51 ms. There is thus approximately a 1:2:4:8 relation in the timing of phonemes, unstressed syllables, stressed syllables and interstress intervals. The data above refer to contexts excluding prepause locations. Within this regular frame there exists a continuity of variations of segment durations and positional variants but one still finds regularity traits. Thus consonants after short stressed vowels are about twice the length of unstressed consonants which holds for voiced as well as unvoiced consonants, in Fant & Kruckenberg (1989, page 81), 44 and 87 ms for voiced and 67 and 135 for unvoiced consonants. A basic distinction in Swedish phonology is that of "vowel quantity". A stressed syllable contains either a long or a short vowel. Their relation is not 2 to 1 but of the order of 1.6 to 1. Lexically stressed vowels in function words generally lose their stress in connected speech. As a mean trend over all contexts and tempos and several data corpora we have found a relation between long and short stressed vowels as follows The durational distinction is lost when Vshort approaches 50 ms. A fully stressed VC: is about 10 % shorter than a V:C and of the order of 2 10 ms. The average number of phonemes per syllable is close to 2.9 for stressed and 2.2 for unstressed syllables, but text specific variations occur. In our standard passage we noted 3.0 phonemes per stressed syllable. With a nonconventional definition of syllables to be constrained by root morphemic criteria, e.g. "leg-at" versus the conventional "le-gat", the average number of phonemes per stressed syllable will increase to 3.3. One argument in favour of the Fonetik 96, Swedish Phonetics Conference, Nasslingen, 29-31 May, 1996 morphemic definition is that the duration of the N consonant following the stressed vowel is pro20 longed. In our statistics, retaining the conventional definition of syllables, we have N accordingly noted a special category for initial 15 15 consonants of unstressed syllables that are preceded by a stressed vowel in an open syllable. The mean duration of such syllables is 192 ms 10 10 and the average number of phonemes is 2.55, i.e. substantially greater than for the main category of unstressed syllables. 5 5 Pauses and rhythmical o 500 loC0 ms 50