5.3 ANÁLISIS MODAL DE FALLOS Y EFECTOS (A M F E.)
5.3.9 DESCRIPCIÓN DEL PROCESO AMFE
To recap a previous discussion (4.1.2.2), in the Hopkins dialect of Garifuna, there is a set of words that may be subject of /r/-deletion and a set that is not. This type of word class split is often used as an argument for the lexical diffusion of sound change, that is, that sound change is phonetically abrupt, but lexically gradual (Wang and Cheng 1977, and summarized in a discussion in Labov 1994). In addition, as this is an internal change in progress with no discernible external cause of the change, it is impossible to attribute the split to dialect mixture unless we somehow consider the existence of more /r/-ful dialects (such as, for instance, those in Barranco and Honduras) to be exerting an external pressure against deletion of /r/. Clearly lexical selection is a factor determining /r/- deletion in Garifuna, and the question is whether this case offers evidence for lexical diffusion and against a Neogrammarian model of sound change (lexically abrupt and phonetically gradual).
The discrete nature of this type of sound change, of course, makes it particularly unsuitable for describing sound change as phonetically gradual, since it is either the case that the /r/ is deleted or not – there are no possible steps along the way (unless one considers deletion one step and vowel coalescence a second step, which I do not), and herein lies possibly the simplest answer to this question, from Labov (1994:543), in his disctinction between the types of phonological change that are likely to fall under the
realms of lexical diffusion or regular sound change. Labov does consider “vocalization of liquids” to be a type of change that is more likely to show regularity, however his generalization draws on studies of the vocalization of post-vocalic /r/ and /l/ in English, where the liquid is generally pre-consonantal, and not on the deletion of an intervocalic liquid.22 As the deletion of an intervocalic liquid is discrete and thus resembles more closely the deletion of an obstruent than a vowel shift or vocalization of a liquid, I argue that it should be included, along with “deletion of obstruents” in that category of sound changes that are more likely to exhibit lexical diffusion.
As this change is ongoing, we cannot predict the eventual outcome. It remains to be seen whether Garifuna will end up with /r/-deletion in all words, or with two word classes – one where /r/ has been deleted and another where it has not. What we can say that at this stage of the progression of the change is a) there is evidence of lexical diffusion and b) it may simply be due to the phonological nature of this particular type of change, that is, deletion.
Some researchers (most notably Bybee 2002) have proposed that if and/or when sound change proceeds through lexical diffusion, it will affect high frequency words before lower frequency words. This has led some variationists to propose using word frequency measurements in variationist analyses of sound change. Diáz-Campos and Ruiz-Sánchez (2008) for instance, promote the idea of using word frequency as an independent variable, and include it in their multivariate analysis of /r/-deletion in two Spanish dialects, one where the variable is stable, and one where it is undergoing a
22 /l/ vocalization in Philadelphia can be intervocalic (Ash 1982), and she does show increased probability of deletion in certain words (really, Philadelphia)
change in progress, to show that in both cases high frequency words exhibit more deletion than lower-frequency words (see also Diáz-Campos 2005).
In this analysis I have not tried to deduce the effect of frequency but rather tried to mitigate the effect of lexical selection by looking at only those words where deletion is possible for all speakers to see whether we still see evidence of a change in progress. Figure 4.3 shows the deletion rate for twelve frequently deleted words, by age. “Frequently deleted” is potentially a misnomer – the category is made up of all of those words that showed /r/-deletion in all age groups (including the oldest age group) and where more than one token of the word was found (i.e. I did not include any singletons, in which there was only one token of the word in my data). In other words, these are the words that frequently exhibit /r/-deletion in the speech community as a whole, and are also relatively frequent in my corpus. There were a total of 633 tokens of these 12 words, and I simply calculated the rate of deletion for each of these words for each speaker (Figure 4.3 shows the average rate of deletion for the twelve words combined for each generational group).
The conclusion to be made from the results displayed in Figure 4.3 is that even when we account for lexical selection by eliminating those words in which /r/ cannot be deleted we still see evidence of a change in progress. Although the differences between each age group are not significant here, at the very least Figure 4.3 shows us that there is a significant difference in average /r/-deletion rates between the oldest and the youngest speakers within the twelve words that show deletion. In other words, even if we only count those words where deletion is clearly acceptable for all speakers, young people are still deleting more than their parents and grandparents.
Figure 4.3: Average deletion rate by age group for frequently deleted words (interview data)