Key Concepts
Exploration 7.2: Hypercorrection
Labov makes much of this phenomenon of hypercorrection, particularly because it appears to relate to changes that are taking place in the language. However, a word of caution is necessary. Such displays as we ind in igure 7.4 are displays of group means. We have no information about the amount of variance about the means so we cannot be sure how comparable they are. We do know they are based on quite small numbers of informants in each case. In addition, we cannot be sure that any two means which difer do so signiicantly in the statistical sense. he cross-over shown in igure 7.4 could, theoretically at least, be the result of the way the data have been treated. However, the fact that it occurs for both word lists and pairs provides us with some assurance of the correctness of Labov’s claims.
176 Inherent Variety
informant showed that his own use of glottal stops shadowed their use by each informant. he more such stops an informant used, the more he himself used such stops. However, his use almost always trailed any other’s use. his is a good example of convergent accommodation (see discussion of this in chapter 4). It is also further evidence that the observer-observed relationship is not a neutral one!)
Trudgill (1995, 93–4) uses data such as those presented in a bar graph in igure 7.5 to demonstrate two very important points: First, when style is kept constant, the lower the social class the greater the incidence of the nonstandard variant; thus we see that for each group, the bars increase as the speech style goes from most to least formal. Second, when social class is kept constant, the less formal the style the greater the incidence of the nonstandard variant. hus we see that the irst bar is lower (zero for the irst two groups) for the middle-class speakers and gets higher as the social class goes from upper to mid to lower working class. Some increases are negligible and some are considerable. For example, middle middle-class speak- ers always avoid -in’ pronunciations in the two most formal styles but ‘relax’ con- siderably more in casual style. Upper working-class speakers make a very sharp diferentiation between the two reading styles and the two speaking styles. Lower working-class speakers make no real distinction between the two speaking styles and use -in’ pronunciations almost exclusively in both; however, just like middle working-class speakers, they are conscious that -ing pronunciations are used in reading styles and do manage to introduce them on many occasions.
Table 7.2 shows a further breakdown by gender of subjects’ performance on the (ng) variable with a score of 000 indicating exclusive use of [ŋ] and a score of 100 indicating exclusive use of [n]. Since the number of informants in each cell is quite Figure 7.5 Percentage of use of -in’ in four contextual styles of speech in Norwich
0 0 5
23
29
0
10 15
44
66
3
15
74
88
98
28
42
87
95 100
Mid Middle Class
Lower Middle Class
Upper Working Class
Mid Working Class
Lower Working Class Word List Reading Passage Formal Speech Casual Speech
small, there is no perfect linear change by gender across the social classes, but we can see that women have lower scores on the whole, indicating use of the more formal [ŋ] variant, and that the scores become higher as social class becomes lower, indicating more use of the informal [n] variant by speakers assigned to lower social- class categories.
Variation in Detroit
A Detroit study (Shuy et al. 1968) and Wolfram’s follow-up to that study (1969) are other irst wave studies. he Detroit study investigated the use of multiple negation as a linguistic variable in that city. It showed that there is a very close relationship between the use of multiple negation and social class. Whereas upper middle-class speakers used such negation on about 2 percent of possible occasions, the corre- sponding percentages for the other three social classes were as follows: lower middle class, 11 percent; upper working class, 38 percent; and lower working class, 70 percent. From such igures we can make a further observation: it is not that members of the upper middle class always avoid multiple negation and members of the lower working class always employ it; it may be our impression that such is the case, but the facts do not conirm that impression. No class uses one variant of the variable to the exclusion of the other, regardless of circumstances. Speech within any social class, therefore, is inherently variable, just as it is in society as a whole. However, the analyses of the diferent variables that were investigated in Detroit clearly show that, although individuals exhibit a certain amount of variation in their linguistic Table 7.2 he (ng) variable in Norwich
Styleb
Social classa No. Sex WLS RPS FS CS
MMC 6 M 000 000 004 031
F 000 000 000 000
LMC 8 M 000 020 027 017
F 000 000 003 067
UWC 16 M 000 018 081 095
F 011 013 068 077
MWC 22 M 024 043 091 097
F 020 046 081 088
LWC 8 M 066 100 100 100
F 017 054 097 100
a Social class: MMC (middle middle class), LMC (lower middle class), UWC (upper working class), MWC (middle working class), LWC (lower working class).
b Style: WLS (word list), RPS (reading passage), FS (formal), CS (casual).
Source: based on Trudgill (1974, 94). Copyright © 1974 Cambridge University Press. Reprinted with the permission of Cambridge University Press
178 Inherent Variety
behavior, there is nevertheless a pattern to that behavior. For example, as the situ- ation becomes more formal, an individual’s linguistic usage comes closer to standard usage, and the higher the social class of the speaker, the more standard too is the speaker’s behavior.
Wolfram’s study was an attempt to show how the distribution of linguistic vari- ables correlated with such factors as social class, gender, age, and racial category in Detroit. Wolfram wanted to identify varieties of speech which might be associated with speciic social groups in the city, for example, upper middle-class Whites or lower working-class Blacks. His work is based on data collected from forty-eight Black research participants drawn from 702 people who initially took part in the Detroit study. here are two male and two female research participants in each age group (10–12, 14–17, 30–55) in upper middle-, lower middle-, upper working- and lower working-class groups, along with twelve White research participants (two male and two female in each age group, all upper middle class).
Having identiied his groups, Wolfram then attempted to show characteristic diferences in linguistic behavior. He investigated four phonological variables: word inal consonant cluster simpliication; medial and inal th, as in nothing and path;
syllable inal d; and the occurrence of r ater vowels. He also investigated four gram- matical variables: the zero copula, as in He tired; invariant be, as in He be tired; the -s plural, possessive, and third-person singular verbal suixes, as in girls, boy’s, and goes; and multiple negation. Figure 7.6, for example, shows group means for the absence of the third-person singular tense-marking (z). A close inspection of the
Figure 7.6 Percentage of [z] absence in third-person singular present tense agreement in Detroit Black speech
% of [z] absence
100
80
60
40
20
0 Upper middle
Lower middle
Upper working Social class
Lower working 1.4
9.7
56.9
71.4
igure shows that, whereas it is quite possible that the diferences between the two groups at each of the ends, that is, between the upper middle and the lower middle classes and between the upper working and the lower working classes, may not be signiicant, there being only twelve subjects in each group, the diference between the top two groups as a whole and the bottom two groups as a whole, that is, between the middle class and the working class, almost certainly is, and probably at a very high level of signiicance. here does therefore appear to be a great diference in usage of the (z) between middle-class and working-class people in Detroit.
We can contrast this graph with another from the same study, this one concerned with (r) absence (Wolfram 1969, 110). Figure 7.7 gives us the information we need.
Here we ind a progressive step-like set of diferences. However, without statistical testing we cannot be sure that there is a signiicant diference between adjacent means, particularly when the groups are small (twelve subjects) and the diference in means is of the order of 61.3 and 71.7 percent. hat there is a signiicant difer- ence between the two groups at each end does seem very likely, but we cannot be sure of the signiicance of the diference between any adjacent pairs. he data do, however, fall into a very clear pattern and it is such patterns that sociolinguists seek to explain.
Wolfram and Fasold (1974, 80–1) argue that in the case of (r) absence in igure 7.7 we have an example of what they call gradient stratiication, that is, a regular step-like progression in means which matches social groupings. In the previous case of (z) (igure 7.6), we have sharp stratiication, that is, a clear break between a particular pair of social groupings. he irst kind of stratiication is said to be typical
Figure 7.7 Percentage of (r) absence in words like farm and car in Detroit Black speech
% of [r] absence
100
80
60
40
20
0 Upper middle
Lower middle
Upper working Social class
Lower working 20.8
38.8
61.3
71.7