Phonetic Vowel Reduction

Phonetic vowel reduction refers to phonetic effects on vowels of reductions in other phonetic dimensions -- that is, to the changes in phonetic vowel quality associated with decreased stress, sonority, duration, loudness, or articulatory effort. Vowel quality is a crucial phonetic variable because the identity of a vowel phoneme is perceived via the phonetic quality of its phonetic realization. The effects of reduction on acoustic vowel quality are largely unknown (though, once again, see Delattre 1968). Particular proposals exist for certain limited circumstances, but no general picture of reduction effects has been given.

One proposed effect of reduction is the assimilation of vowel quality to context, often called the undershoot effect. In this view, if adjacent segments are rounded, a reduced vowel would also become rounded(lowering formant frequencies, shifting back and up in (F1 vs. F2) vowel space); if fronted velar segments are adjacent, the vowel is front-velarized (shifting toward a high, front [I] quality). In a study of Stockholm Swedish short vowels, using nonsense syllables in fixed phrases repeated by a single speaker, Lindblom (1963a, 1963b) showed that vowels underwent contextual assimilation of vowel quality as duration reduced. He argued at the time that his findings failed to support the widely held theory that vowels centralize in quality when reduced in stress, since the effects could be entirely accounted for as a function of duration, which is itself partly function of stress.

Lindblom basically found that as a speaker produces a syllable faster and faster, the articulators simply cannot move fast enough to reach target positions for the consonant and the vowel. The result is that as vowel duration decreases, formant nuclei shift away from the ``target'' frequencies, towards the patterns characteristic of the phonetic context. This undershoot phenomenon is an example of ``hard'' coarticulation, which is governed by physical limitations of the vocal tract, such as the mass of the tongue, the distance between the vowel and consonant articulations, the amount of energy that can be expended by tongue muscles, etc.

Vowel undershoot of this kind is a phenomenon of general phonetics: when a syllable is so short that the tongue cannot physically move to a vowel target before a return gesture begins, then undershoot occurs. This universal physical constraint holds no matter what the language is. On the other hand, the whole range of linguistic fast-speech phenomena may be thought of as planned ways to avoid reaching the physical limits of the speech articulators. It may be expected that such phenomena extend to vowels, so that planned simplifications of articulatory activities may be incorporated in faster or more relaxed speech. If this is the case, then vowel undershoot may be relatively insignificant among the vowel shifts that occur across various contexts. Thus, it is possible that there are linguistic forms of phonetic vowel reduction which do not follow simply from the brute physical sluggishness of the tongue. In such cases, the tongue may arrive at different inflection-points (phonetic nuclei) on its trajectory when stress or other environmental factors are varied, according to linguistic intent rather than physical constraint. It is quite likely that there is room for language-specificity in the vowel shifts associated with shortening and stress reduction,etc. Such phenomena belong to linguistic phonetics. They militate against extending Lindblom's result to conversational speech and to other languages. But this is an empirical matter, which we will investigate in later chapters.

Phonetic vowel reduction was originally described as a process of centralization: ``The average effect of consonants on vowels'' is ``a centralizing effect..:all vowels are on the average shifted toward [] in contact with consonants.''(Joos 1948, quoted in Lindblom 1963b). Later it was described in laboratory studies as both centralization and contextual assimilation (eg., Stevens & House, 1963). But Lindblom's study found that centralization occurs ``only insofar as the immediate context contains schwa [central] elements'' (p1780) -- that is, only through assimilation, or coarticulation. In short, reduction in duration had the effect of increased coarticulation. For this reason, future research should also consider the coarticulatory effects of adjacent segments, examining the extent to which increased vowel reduction is simply increased coarticulation.

Another, possible effect of reduction is phonetic shift along the path of ongoing sound change. For example, Labov, Yeager & Steiner (1972) found that upward-shifting /æ/ in polysyllabic words (i.e., in relatively short, unstressed syllables), was raised even higher than the most advanced fully-stressed /æ/ realizations. An early, small pilot study for this work (Mazzie & Veatch 1986) found a similar result: in the speech of a male adolescent Southampton (England) speaker, unstressed vowels were more distant from conservative forms than stressed vowels. One may speculate in this case that unstressed vowels have a quality advanced along the course of sound change. That is, a vowel's distinctive quality isn't reduced if it lacks stress, rather it tends to occur farther along the trajectory of a vocalic sound change than its stressed co-allophones. The small amount of data examined in Mazzie & Veatch (1986) could be no more than suggestive.

Vowel quality could conceivably shift in any direction in vowel space under the influence of reduction, or none at all. Lindblom (1963b) points out that in Jassem's (1959) study of Polish stress, ``no relationship between stress and vowel quality could be demonstrated.'' LYS and the Mazzie & Veatch pilot study for the current research suggest that other kinds of reduction besides simple consonantal assimilation or centralization might exist. Indeed, given the great variety in phonological reduction, one might expect variety in phonetic effects as well. The actual differences between dialects in the effects of stress discovered in this work are discussed in each of the dialect chapters, and summarized briefly in the concluding chapter, section .