Reduced Vowels

• We have already considered the schwa (Worldbet &) as the classic neutral vowel; that is, its resonant frequencies are the closest of any American English vowel to those predicted by considering the vocal tract as a uniform tube (500 Hz, 1500 Hz, 2500 Hz).
• The schwa is the most commonly encountered example of a reduced vowel. This vowel occurs in an unstressed word or syllable; examples are the words the, a, the first syllable of about, and the last syllable of sofa. In all such cases, the quality of the vowel is much more central than when the phoneme is stressed. Thus a good definition of a reduced vowel is the following: ``an unstressed, central variant of a fuller vowel" (Olive, pp. 321-327).
• There are two other reduced vowels which are commonly distinguished from the schwa: the high reduced vowel (Worldbet I_x as in the second syllable of the word roses) and the retroflex reduced vowel (Worldbet &r) such as in the last syllable of butter. I_x has a formant structure similar to I, while &r has the inevitable hallmark of a retroflex sound: F3 at or below 2000 Hz.
• One characteristic of a reduced vowel is that its formants cannot readily be pinned down to the same degree of accuracy that the non-reduced vowels can. One reason for this is the extremely short duration of the reduced vowels. Sometimes a schwa will last for only three pitch periods. With such a short period of production, the articulators cannot be moved into the ``ideal" configuration and the reduced vowel often undershoots its target formants.
• Reduced vowels also lack the sustained period of stability provided by other vowels, and for the same reason cited above: they are too short to reach or stay at the target frequencies.
• The vocal tract is relatively open in reduced vowels, just as it is with full vowels. However, because of their short duration and the fact that they occur in unstressed syllables where the energy level is lower in general than in stressed syllables, reduced vowels often lack the energy of their longer-lasting cousins, and therefore they may only exhibit F1, or F1 and F2, at the normal decibel levels which we use to display our spectrograms.
• In spite of all these differences, however, it is relatively easy to recognize reduced vowels when they are segmented off by consonants. It is when a reduced vowel is part of a longer sonorant stream that the problems begin.

}