Abstract
In another paper (Hill, Witten and Jassem, 1978) we
reported a preliminary study of rhythm in Spoken British English which, among
other things, showed that there was a tendency for segments to be of
shorter duration, on average, in rhythmic units containing more
segments than in rhythmic units containing fewer segments. Such an
effect would tend to make rhythmic units more nearly equal in duration
than would be expected from a straight segment count and is referred to
as "a tendency towards isochrony". Much recent research in acoustic
phonetics, especially in the US< has tended to discount the importance
of such an effect. In view of our finding that it forms the third most
important factor in accounting for segment duration variation, we wished
to go further and to try to produce some reasonable measure of the
isochrony effect, partly as a useful descriptor for the rhythmic
character of utterances and partly to provide a more objective basis for comparing
different theories of rhythm that incorporated notions of such a
tendency.
In this paper we go some way towards a measure of isochrony which
expresses the percentage of the mean duration of rhythmic units that may be
regarded as fixed, regardless of rhythmic unit size (segment count). A figure of
0% would suggest no tendency towards isochrony whilst a figure of 100%
would suggest strictly isochronous speech. We applied this measure to
the body of speech data resulting from the analysis of exercises recorded
for M.A.K. Halliday's book written for students of spoken British English
(Halliday, 1970), on the basis of two different frameworks for rhythmic
description, one commonly associated with the names of Jones,
Abercrombie, Halliday and Ladefoged and based on what Halliday calls "feet", the due
to Jassem, based on "rhythm units". The differences between the two
frameworks were marginal as far as goodness of fit between data and
theory was concerned, the latter theory leading to somewhat less scatter of the
data and actually excluding just those portions of speech exhibiting
the least tendency towards isochrony. The measure did, however, show
very clearly another effect that was intuitively obvious in the data,
namely that marked rhythmic units showed a far greater tendency towards
isochrony than unmarked units. The work is continuing. (Work supported by the
National Research Council of Canada).
