Introduction 9correlated to consonant duration, in both perception and production rather
than an enhancement model, in which an increase in one factor is accompa-
nied by an increase in the other. This sophisticated multivariate treatment of
what are often taken to be single variables (e.g., voicing) leads to the conclu-
sion that hearers are attuned to very low-level features of the signal. Purnell
goes on to show that the different groups of hearers use these elements differ-
ently. In recalculation of previous work, Purnell also shows that identi¿ cation
of African American, Chicano, and European American voices in a matched-
guise experiment is dependent on low-level aspects of the features presented
in the signal (the word “hello”). In this case ¿ lter (e.g., formant values), source
(e.g., pitch) a nd sou rce -¿ lter (e.g., intensity) characteristics of the signal were
considered. He concludes that the relative position of /͑/ and /o/ in the vowel
space, particularly vowel height, is the major contributing phonetic factor in
ethnic identi¿ cation.
Chapter 14 introduces an interesting twist on the problem of gender iden-
ti¿ cation. Although fundamental frequency (f0) is usually the best clue, it
can be unreliable in cases of overlap and idiosyncratic voices and is, there-
fore, particularly unreliable in the case of young children. Foulkes, Docherty,
Khattab, and Yaeger-Dror investigate the possibility that gender related pho-
netic details other than F0 in a speech community are salient to listeners as
gender identity clues. They conducted an experiment in Tyneside, the large
conurbation in the northeast of England, with the city of Newcastle upon Tyne
as its hub, to investigate this possibility. They ¿ rst report on production stud-
ies that establish gender related voiceless stop variation. In Tyneside, laryn-
gealized (i.e., partially voiced) versions of voiceless stops in medial position
are much more common among men, and in word-¿ nal (pre-pause) position,
preaspirated voiceless stops are preferred by women. Tyneside, non-local UK
respondents, and speakers of American English judged lexical items spoken
by young children (three to four years of age) as “boy” or “girl” on the basis
of samples that incorporated both these variables and were also coded for
F0, amplitude, rate of articulation, and vocal quality. F0 returned an interest-
ing result—higher F0 was associated with the “boy” rather than the “girl”
response, but amplitude was clearly associated with boys (and these two
factors may have interacted). Nevertheless, the features themselves (laryn-
gealization and preaspiration) were signi¿ cant for locals; Tynesiders more
frequently identi¿ ed laryngealized medial tokens as “boy” and did not reverse
the pattern of identi¿ cation for preaspirated ¿ nal tokens (where the nonlocal
UK and American respondents identi¿ ed them as “boy”). As in many of the
other studies reported here, untrained listeners, particularly local ones, seem
remarkably sensitive to low level phonetic variation.