Kinds of Reduction

Theoretical background for this study of vowel reduction is discussed in Chapter 4. There are many possible effects of stress reduction on vowel nuclei:

1) No effect
2) Raising
3) Lowering
4) Fronting
5) Backing
6) Shift towards a mid-central reduction target
7) Shift towards a high-central reduction target
8) Shift towards any given reduction target

To determine which of these are closer to the truth, consider the charts of stressed and unstressed means in Figures , , . In these figures, the tails of the arrows are the means of the stressed tokens; the heads are the means of the unstressed tokens.

Figure: Arrows point from mean of stressed tokens to mean of unstressed tokens. ` means that only stressed tokens of that category occur.

56#56

Figure: Arrows point from mean of stressed tokens to mean of unstressed tokens. ` means that only stressed tokens of that category occur.

57#57

Figure: Arrows point from mean of stressed tokens to mean of unstressed tokens. ` means that only stressed tokens of that category occur.

58#58

The two charts in each figure are only slightly different; their purpose is to test the hypothesis that the inclusion of vowels in clitic words might have a distorting effect on the vowel-reduction pattern. The patterns seem more clear for Jim and Judy: vowels shift fairly consistently in the direction of a single central, upper-mid location, which for Jim is in the area of the realization of unstressed //. The picture is most precise for Judy, but the pattern fairly consistently applies to Jim's vowels also. Rita's pattern is more clear with clitic words excluded; all the significant effects are in the direction of high-central //, excepting the effect of reduction on /or/. Rita's pattern, when clitics are included, becomes more murky, since /æ, er/ undergo backing rather than shift in the direction of a common reduction target. The greater precision of the pattern found for Judy and Jim may be due to the fact that the speech analysed for these two speakers was casual-style speech. That is, the data for these two speakers was taken from narratives of personal experience, in which self-monitoring is minimized. The data analysed for Rita, on the other hand, constituted the entire 25-minute interview, and much of it was careful-style speech. This may be one reason that the patterns of stress reduction for Jim and Judy are quite clear, while Rita's pattern has more significant shifts that don't fit the pattern.

Clearly there are strong effects in this data, so (1) above is ruled out. Most low vowels rise, but most mid and high vowels do not, so (2) can also be ruled out. (3) is a quite unlikely model for vowel reduction, since very few vowels undergo lowering. (4) applies to the phonologically back vowels for Judy and Jim, but its opposite (5) applies to the phonologically non-low front vowels for Jim and to all front vowels for Judy. (6) is incorrect, since the high vowels do not lower to phonetically mid height. (7) seems to be an accurate description of the overall pattern for both Judy and Jim, and also for Rita's non-clitic words. Most of the arrows for both Jim and Judy display shifts quite precisely in the direction of a single location. The arrows here suggest that there is a ``reduction target'' toward which vowels shift in this dialect. The phonetic quality of this target is high and central, a quality that may be written as []. Is it a coincidence that the apparent reduction target is where it is, namely, in high-central position? One might generalize from this pattern and propose that the universal reduction target is high and central. To be precise, vowel reduction might be characterized as a universal phonetic process of gradient shifts in the phonetic quality of vowel nuclei in the direction of a high-central reduction target. In short, the process might be properly characterized universally as (7), the general tendency of raising and centralizing towards a high-central target. However in the Jamaican data examined in the last chapter, it was shown that a similar reduction target for the short vowels explains the Jamaican vowel shifts that are due to stress reduction. This reduction target was mid and central, not high and central, as found in the data for these speakers. Thus if the process of vowel reduction as shift towards a reduction target is universal, the exact phonetic target appears not to be. So far we have evidence from two dialects that patterns (6) and (7) above represent the possibilities for vowel reduction. The last possibility, (8), would seem to be too general and unrestricted a statement, since not just any point in vowel space will do. In the next two chapters we will see if more possibilities need to be added, and begin to further explore the space of possible reduction targets.