next up previous
Next: Summary Up: Jamaican Creole Previous: An Acoustic-Phonetic Grammar for

Stress Effects

While there appear to be factors other than stress involved in Jamaican prosody (which may be more like tonal phenomena), there are certainly many clear cases of prominent as opposed to non-prominent syllables. Therefore, despite the unresolved phonology of Jamaican prosody, I impressionistically coded this stretch of speech for phrasal stress, and in this section present the correlation of these classifications with formant-frequency variation.

The method for coding stress is discussed in Section [*]. After considerable time listening to the rhythm of speech of the two speakers studied, my intuitions about the location of stress seemed fairly reliable. While for reasons discussed below these developed intuitions are in some cases quite different from the naive intuitions of American listeners, they generally coincided with those of Peter Patrick in the joint coding we did.

A self-consistency test was conducted on 231 syllables, which I classified for three levels of stress on two occasions, 9 months apart. The results are presented in Table [*].

Table: Confusion matrix of phrasal stress classifications on two occasions.
  October, 1990
  Primary Secondary Unstressed
  Primary 55 6 6
July, 1991 Secondary 5 3 8
  Unstressed 1 3 144

The frequency of consistent classifications is 202/231 = 87%. If secondary stress is collapsed with either unstressed or stressed, the consistency level increases to 213/233 = 92.2%. This is almost identical to the levels of consistency found between another coder and myself in the same task on a Chicago speaker, discussed in the Methods chapter, page [*].

It wasn't difficult to apply this procedure, which was previously used with the other dialects studied, to Jamaican Creole.

In JC, prominence seems cued sometimes by an apparent pause before the stressed syllable. Pitch rise is not the only way to achieve pitch prominence: Pitch fall is also prominent. But my impression is that this is less frequent.

There are some problems with stress coding in this way. In some examples, there appears to be an unlikely stress clash: three or four consecutive syllables can have high pitch, each apparently stressed. Also, the location of the stress is sometimes unclear, because it may not make linguistic sense to stress closed class words, or because a single pitch prominence may seem to extend over more than one syllable. In fact, Wells (1973:22, 1982:573) says that stress may appear to shift to the right one syllable, from the expected (Received Pronunciation, presumably) stress pattern. This is consistent with the view that stress is a feature on an autosegmental tier, which can slide around from one syllable to the next or previous, or to stretch out across several syllables. This view is rather more reminiscent of tone than of syllable stress.

Let us consider the effects of impressionistically coded stress on vowel quality as shown in F1-F2 charts of nucleus measurements. Wells says that stress in JC is marked by lack of reduction in unstressed syllables. If we look at the charts in Figures [*],  [*], it turns out that on the contrary, most of the vowels have significant effects of stress.

Figure: JC Stress Effects (Juba): Stressed and unstressed vowels in non-clitic words only, and in both clitic and non-clitic words.

Figure: JC Stress Effects (Roasta): Stressed and unstressed vowels in non-clitic words only, and in both clitic and non-clitic words.

The upper charts in Figures [*] and [*] show the effects of stress (more precisely, of destressing) when clitics are excluded, while the lower charts show the effects of stress on vowels in when clitics are included. In some other dialects, including or excluding the clitic words makes the pattern more clear, so both figures are presented. Very little difference is observed between the upper and lower charts for either speaker. So I will discuss just the lower charts, which includes both clitic and non-clitic words, since it contains more data.

I will discuss the effects on long vowels first, and then the effects on short vowels. The most striking fact is that the high and low long vowels have no significant effect of stress for Roasta, and relatively small (though significant) effects for Juba.

We may explain the stress effects on long vowels as an assimilation of gliding vowels to their phonetic off-glides. Thus /ie, uo/ have a strong shift from their stressed nuclei in the direction of their offglides. Note that /ie/ for both speakers glides downward along the edge of vowel space, not inward, towards [], consistent with the transcription as /ie/, as opposed to /i/, as might instead have been expected. The other long vowels show insignificant or small effects of stress, perhaps because they have no glides, or phonologically specified glides (that is, not phonetic glides). Long monophthongs /ii, aa, uu/ lack offglides and thus show no significant effect of stress for Roasta, and only small effects for /ii, aa/ for Juba. ``True'' diphthongs /ai, ou/ have two phonologically specified targets, and thus nucleus targets appear to be attained whether the vowel is stressed or not -- if the effect is limited to phonetic diphthongs. This rather subtle distinction between glides and diphthongs has previously been made on phonetic grounds by Lehiste and Peterson (1961). They listed English [I], [U], [I]6.17 as long, complex vowels with double targets, while the other long vowels were either monophthongs or were classified as ``glides'' -- i.e., as single-target vowels with a phonetic offglide that did not constitute a second phonetic target. This phonetic distinction corresponds to a phonological distinction, where the glide is specified underlyingly on /ai, ou/, but not on /ie, uo/ (underlyingly /e:, o:/, in the lexical sets FACE and GOAT). In the phonological analysis proposed above, the realization of /e:, o:/, as phonetic sequences of high nuclei followed by mid glides, is derived by one of the rules that specifies the surface phonetic facts.

The phonetically derived glides, /ie, uo/, are in a sense less strictly specified than glides which are specified in the underlying phonological form. A reasonable way to look at this data is to consider that the phonetic rules that create the derived glides interact with (phrasal) stress, in such a way that derived nucleus-glide sequences are less phonetically distinct when stress is reduced.

Another way of looking at the same data is to make the analogy to Gallo-Romance breaking, in which vowels diphthongize under stress.6.18 That is, rather than viewing the effect of stress reduction as nucleus-glide assimilation, perhaps instead we can view the breaking and raising of the nuclei of underlying /e:, o:/ to /ie, uo/ as a phonetic process that is triggered by the application of stress (see also LYS):

The nucleus of /e:, o:/ is raised and diphthongized when stressed.

If true, this explains why the unstressed nuclei are so much lower than the stressed nuclei: they have not been diphthongized and raised from their underlying positions. This view would also explain why it is that these two vowels, among all the long vowels, have the strongest effect of stress: the process of breaking and raising under stress is restricted to the long, underlyingly mid vowels /e:, o:/, and doesn't apply to /i:, a:, u:/.

Next let us consider the short vowels. Of these, /i, a, u/ shift a considerable distance towards the center of the vowel space, while /e, o/ shift a relatively small distance towards the same reduction target. Why aren't the short mid vowels also centralizing? Notice that /e, o/ are already mid vowels to begin with; because they are short and non-peripheral, a process of reduction which shifts vowels to a mid and central position need not move these vowels very far at all. So even if /e, o/ were to centralize, it would have only a slight effect, because they are already mid and relatively central at their stressed, target locations. This makes it reasonable to state that all the short vowels centralize under phrasal stress reduction. We may summarize the effects on the short vowels by the statement,

Centralization due to phrasal stress applies to short vowels.

If we explain the effects of stress on long vowels and diphthongs as suggested above (that is, not by a rule of centralization), then we may summarize the effect of the process of centralization on both short and long vowels by appending ``only'' to the above statement.

The effects of stress for L.A. Chicano English, and for Chicago White English, are characterized as shifts in the direction of a reduction target. That is, all the arrows, displaying the effect on formant values of stress reduction, point fairly precisely towards a single point, with a small number of exceptions that are explained for other reasons. In Jamaican, the picture is not so clear, until we look at the effects on the short vowels alone, as argued above. The short vowels, /i, e, a, o, u/, in Juba's and Roasta's clitics-excluded charts, point towards a single location in formant space (F1 48#48 420Hz, F2 48#48 1300Hz for both of them, though this point is relatively higher within the vowel space for Roasta than for Juba). Thus the phonetic process of stress reduction, shifting vowel nuclei in the direction of a single ``reduction target'', can be seen to apply in Jamaican Creole as well as in other English dialects, though in JC it appears to be restricted to applying to the short vowels.

next up previous
Next: Summary Up: Jamaican Creole Previous: An Acoustic-Phonetic Grammar for
Thomas Veatch 2005-01-25