So far we have been regarding the vowel sounds mainly as ways of
representing the contrasts that occur among words or sentences. But
they can also be thought of in a completely different way. That is, we
may regard them as different vowel qualities representing the
articulation. Generally speaking, it is called vowel diagrams, and the
simple vowel diagrams in Figure 2.2 shows two of the
dimensions of vowel quality. Different vowel sounds can be produced
by changing their formant structure. Hence, we would expect a
correspondence between vowel sounds and formant frequencies. This
expectation is born out by the evidence: (1) consistent formant
frequencies can be recovered from speech and mapped onto vowel sounds,
and (2) artificial speech with selected formant frequencies is
perceived as having the intended vowel quality[22]. If we
set up a coordinate system using 
and 
as a basis, vowels
lie in specific regions (see Figure 4.2). The exact
partitioning of the 
space varies with age, gender,
language, and from one speaker to another, but the overall pattern
does not vary. If we reverse the directions of the and
axes, we see that the vowel loci correspond roughly with the positions
assigned to these vowels in the vowel diagram, i.e., is the
open/close of jaw and is the front/back of tongue.
図 4.1: Block diagram of automatic pronunciation instruction as part of the CAPL system
