| |
Abstract:
There are seemingly contradictory observations: on the one
hand, the acoustic data (Peterson and Barney 1952) indicate that
vowel space defined by the four corner vowels of American English
can be linearly and homogeneously scaled between adults and
children. The validity of the linear scaling suggests that an
adult vocal tract is a uniformly scaled up version of a child
vocal tract. Proposals for vowel normalization (Fant 1975;
Nordström 1977) and studies of vocal tract growth (Goldstein
1980; Fitch and Giedd 1999), on the other hand, have confirmed a
non-uniform scaling. In fact, it is a common observation that the
oral cavity of newborns and children is much longer than the
pharyngeal cavity. But for adults, especially for males, the oral
cavity becomes shorter than the pharyngeal cavity. The
non-homogeneous growth would predict non-uniform acoustic scaling
of vowels. Moreover, in recent papers, we have shown that the
maximal vowel space of a newborn is potentially as! large as that
of adults or even greater, after a uniform length normalization.
If newborn had the same sensorimotor control capabilities as
adults, their vocal tracts enabled them to produce the same range
of [i a u] vowel contrast.
In this paper, we propose and test the following hypothesis to
explain this apparent contradiction: in order to produce vowel
contrasts comparable to adult men (if we arbitrarily assume the
vocal tract morphology of a man as a reference), speakers with a
shorter vocal tract (children and women) employ somewhat
different articulatory strategies to cope with their
morphological differences. Using the VLAM model (Boë 1997)
and the concept of maximal vowel space, we determined the
prototypical articulatori-acoustic realizations that a newborn, a
child, a woman and a man would have if, for each growth stage,
the speaker were to display the same control capacities as an
adult. Vowels [i], [u], and [a] have been studied. Results show
that articulatory adjustments are made on constriction size as
well as constriction location.
|
