• Tidak ada hasil yang ditemukan

Controversies

Dalam dokumen An Introduction to Sociolinguistics (Halaman 191-199)

In a previous section I noted that linguistic variables may show correlations not only with social variables but also with other linguistic features, i.e., they may be linguistically constrained too, as with the deletion of l in Montreal. In their discussion of linguistic variation, Wolfram and Fasold (1974, pp. 101–5) present data from an earlier study by Fasold (1972) to show that it is possible to state how two or more factors, or constraints, interact to affect the distribution of a variable. In this case they are concerned with deletion of final stops in clusters, e.g., the d in a word like cold, in speech among blacks in Washington, DC.

The data showed that the parenthesized stops were deleted as follows: san(d)

Table 7.9 Final cluster simplification among black speakers in Washington, DC

% deleted Example Environment

83.3 san(d) castle after sonorant, before non-vowel 68.8 fas(t) car after non-sonorant, before non-vowel 34.9 wil(d) elephant after sonorant, before vowel

25.2 lif(t) it after non-sonorant, before vowel Source: based on Wolfram and Fasold (1974, p. 102)

castle, 83.3 percent deletion; fas(t) car, 68.8 percent deletion; wil(d) elephant, 34.9 percent deletion; and lif(t) it, 25.2 percent deletion. If we look closely at the environments of these stops, we will find that sometimes the stop is preceded by a sonorant (a nasal or l) and sometimes by a non-sonorant (a stop or a fricative), and it is followed sometimes by a vowel and sometimes by a consonant (or non-vowel). We can see this distribution more clearly in table 7.9. Wolfram and Fasold point out that the constraint of appearing before a non-vowel has a greater effect than the constraint of appearing after a sonorant, i.e., appearance of the stop before a non-vowel leads to a greater amount of deletion than appearance after a sonorant. When both constraints are present we find the highest percentage of deletions: 83.3 percent in san(d) castle. When neither constraint is present we have the least: 25.2 percent in lif(t) it. In the intermedi- ate cases, appearing before a non-vowel is more important than appearing after a sonorant. Wolfram and Fasold, therefore, call appearing before a non-vowel a first-order constraint and appearing after a sonorant a second-order con- straint. That is, the former exercises a greater influence on a person’s linguistic behavior than does the latter.

Constraints may also mix phonological and grammatical features. Wolfram (1969, pp. 59–69) explains a situation in Detroit in which black speakers also delete final stops in clusters, but in this case make a distinction according to the grammatical function of the stop. In the final cluster in cold the d has no independent grammatical function – it is part of a single unit of meaning – but in burned it marks past tense and is grammatically the -ed ending, and therefore has its own meaning. The data are distributed as in table 7.10. In this variety of English the first-order constraint is once again appearance before a vowel or non-vowel (here consonant). Appearance before a vowel inhibits cluster simpli- fication in all cases and appearance before a consonant encourages it. The second-order constraint is appearance as the -ed ending. That is, such appear- ance has a lesser effect than whether or not the following sound is a vowel or consonant. Consequently, the greatest loss of [d] in these examples occurs when the following sound is a vowel and the [d] does not represent the -ed grammat- ical ending. The least loss occurs when the [d] is followed by a vowel and it is the -ed ending. This situation is the same for all social classes, but the actual amounts of deletion vary from class to class.

Table 7.10 Final cluster simplification among black speakers in Detroit Social class

Upper Lower Upper Lower

middle middle working working Example Environment 0.07 0.13 0.24 0.34 burn(ed) up -ed, before vowel 0.28 0.43 0.65 0.72 col(d) out not -ed, before vowel 0.49 0.62 0.73 0.76 burn(ed) coal -ed, before consonant 0.79 0.87 0.94 0.97 col(d) cuts not -ed, before consonant Source: based on Wolfram (1969, pp. 59–69)

Table 7.11 Final cluster simplification among black youth in New York City

Simplification

(%) Example Environment

24 pass(ed) eleven -ed, before vowel

59 pas(t) eleven not -ed, before vowel

74 pass(ed) five -ed, before non-vowel

91 pas(t) five not -ed, before non-vowel

Source: based on Labov (1972b, p. 222)

Further study of this phenomenon by Labov (1972b, p. 222) showed that different sub-groups in society may order two constraints differently. Among black speakers in New York City, Labov found that, whereas more adolescent groups order the above constraints in the way reported in table 7.10, upper working-class adults reverse the order. With most adolescents the situation is as in table 7.11; however, with upper working-class adults we have the situation shown in table 7.12. In this example the first-order constraint is the status of the [t]: whether or not it represents the -ed ending; the second-order constraint is the next phonological segment: whether or not it is a vowel. Whereas adolescents are inhibited in their simplification of final clusters, first by whether the following segment is a vowel and only then by the nature of the [t], adults are inhibited in their simplification first by the status of the [t], i.e., they are reluctant to omit it if it represents -ed, and only then by the presence of a following vowel.

Using information similar to the kind just presented, Wolfram and Fasold (1974, pp. 133– 4) go on to show how it is possible to take a phenomenon like cluster simplification and predict certain kinds of linguistic behavior. They dis- tinguish between speakers of Standard English (SE), white nonstandard English (WNS), and what they call Vernacular Black English (VBE). They consider four

Table 7.12 Final cluster simplification among black upper working-class adults in New York City

Simplification

(%) Example Environment

9 pass(ed) eleven -ed, before vowel

19 pass(ed) five -ed, before non-vowel

40 pas(t) eleven not -ed, before vowel

90 pas(t) five not -ed, before non-vowel

Source: based on Labov (1972b, p. 222)

Table 7.13 Final cluster simplification in several varieties of English Variety of English Cluster simplification

SE sometimes always always always

tes’ program test idea testing risking Most VBE and some usually or sometimes always always

WNS speakers always tes’ idea testing risking

tes’ program

Some WNS always always always always

Some VBE speakers tes’ program tes’ idea testing risking

Some WNS always always usually usually

Some VBE speakers tes’ program tes’ idea or always or always testing risking

Some VBE speakers always always always always

(especially Deep tes’ program tes’ idea tessing rissing South children)

Source: Wolfram and Fasold (1974, p. 134)

environments in which cluster simplification can occur: (1) before a word begin- ning with a consonant (test program); (2) before a word beginning with a vowel (test idea); (3) before a suffix such as -ing (testing); and (4) involving a final consonant other than t (e.g., k) before a suffix such as -ing (risking). They report their findings (p. 134) as in table 7.13, with ‘always,’ ‘sometimes,’ and ‘usually’

in that table referring to the pronunciation which they predict will occur.

According to such a display, tes’ idea (for test idea) is not a feature of SE; there is a considerable overlap between features found in WNS and VBE; but it is only in the latter that you find tessing (for testing) and rissing (for risking). Tes’

program (for test program), however, is found in all varieties of English but with

a different incidence of usage: only ‘sometimes’ in SE, but ‘usually’ or ‘always’

in the other varieties.

From the foregoing discussion we can see that it may be possible to predict certain kinds of linguistic behavior if we know the various constraints that operate in connection with a particular variable and the relationships between that variable and factors such as social class, level of formality, age, gender, and race. Labov has suggested that we should attempt to state what we know by writing variable rules. A variable rule is a modified version of the kind of rule found in grammars modeled on Chomsky’s ideas. It recasts such a rule in the form of a statement that introduces probabilities: do this or that at a certain frequency or frequencies according to the presence or absence of factors a, b, c, . . .n. Wolfram and Fasold (1974, p. 110) actually hypothesize that speakers

‘can identify variable rules, which linguistic factors favor rule operation, the hierarchical order in which they are ranked, the extent to which higher-order constraints are stronger than lower-order ones, and the probabilities toward rule operation contributed by each.’

In practice we face considerable difficulties in trying to write even a single variable rule. For example, what kind of rule could cover all instances of the variable (h)? Some people nearly always say happen and after, while instances of ’appen are rare indeed. Others show considerable use of ’appen, and still others never say anything but ’appen. None of these say hafter for after. Not a few say ’appen and after, but occasionally some of these say happen and also hafter. That is, there are h-pronouncers, h-droppers, and h-inserters. Yet one variable rule is supposed to cover all speakers; either that, or there are two variable rules which interact. But this latter proposal would suggest that people communicate not through the same set of rules but through intersecting sets. It would therefore raise still other issues.

One very serious criticism is that, while the concept of ‘probability’ is often useful in life in explaining the chances of certain things happening, it offers no guide to conduct in specific instances. Categorical rules, i.e., rules which say ‘if X then Y,’ do offer a guide to, and therefore an explanation of, conduct, but variable rules do not and cannot. They do no more than summarize general trends, tendencies, or probabilities found within groups. Variable rules are stat- istical generalizations based on surveys of language use and they indicate trends or norms in populations.

Additional concerns have been voiced about the concept of ‘variability’ as some kind of rule-governed behavior that can also be ascribed to individuals.

We must ask what kinds of mental processes would be necessary to handle that kind of statistical information and how children could acquire it during their language learning (see Kay and McDaniel, 1979). More recently, Fasold (1991, p. 9) has admitted that, ‘Variable rules were proposed as a way of understanding how variation works within a theory of human language.’ He adds (p. 18) that in reality the variable rule ‘was never any more than a display device.’ In recent years there has been throughout linguistics a general decline in rule-writing;

sociolinguistics has seen much less of it too.

Bailey (1973) and Bickerton (1971) have been particularly critical of such attempts to use variable rules. They acknowledge variability in language but

insist that it can be explained if we look closely at the environments in which variation occurs and are prepared to relate the environments to one another using some kind of scale. We must note, of course, that they are concerned with individual speech behavior, what they call the isolect, whereas Labov and others have been concerned with group behavior, the sociolect, insisting that such behavior is important in studies of how people actually use language not only to communicate verbally but for a variety of other purposes too.

Bailey and Bickerton have proposed that each individual controls an isolect of the language, an individual array of linguistic usages which others may or may not share. Each isolect is a lect. The lects of a language differ from one another along a continuum, which forms a polylectal or panlectal grid such that there is an implicational relationship among the various lects: that is, if lect A has feature X, then it will also have features Y and Z, but if lect B has feature Y but not feature X, it will still have feature Z. Lect C may have only feature Z and it could not acquire feature X until it first acquired feature Y. As Petyt (1980, p. 190) points out, if a Yorkshireman pronounces grass with the first vowel of father [ap], he will pronounce cut with [Î]. According to the theory, this ‘dynamic’

view of language structure is valid both synchronically, i.e., as a description of the structure of a language at any specific moment in its history, and diachronically, i.e., over an extended period of time. It is a new variation of the old wave theory of linguistic change, but one that incorporates synchronic matters.

The theory proposed by Bailey and Bickerton tries to reduce the amount of variation in language that linguists must consider by requiring an investigator to look at individual lects and consider all linguistic behavior as categorical, i.e., fully determined by this or that factor or set of factors. They claim that, when the linguistic behavior of an individual is graphed for a particular linguistic variable, that behavior is far more likely to show an ‘all-or-none’ characteristic, i.e., to be categorical in nature, than to show some kind of statistical distribu- tion around a mean, i.e., to be variable in nature, if the particular circumstances that occasion the behavior are known. Individual lects may then be arranged on various continua that can be related implicationally to one another. At any moment it should also be possible to say exactly what the status of any linguistic variable is, i.e., how speakers are using it over a period of time, in regional or social space, and along various dimensions of the latter, i.e., according to social class, age, gender, ethnicity, and so on. Such panlectal grammars therefore would account for variation; moreover, Bailey and Bickerton suggest that much of the variation that sociolinguists actually talk about is a creation of the methods they employ, i.e., is a methodological artifact.

Bailey and Bickerton also appear to be making a claim which Labov does not make. Their dynamic model suggests that all variation in language results from changes in progress: variation is the mark of linguistic change. On the other hand, Labov does not regard every bit of linguistic variation as being associated with changes in progress. While some variation is associated with changes in progress, he regards variation as an inherent property of language, i.e., you cannot have language without variation but only part of that variation ‘goes somewhere,’ i.e., results in change.

The kinds of variation we have seen raise important issues about the very nature of language itself. Labov and others have argued that the kinds of gram- mars preferred by Chomsky and his associates must be modified to recognize vari- ation and, particularly, that the famous (or notorious) competence–performance distinction (see pp. 2–3) made in such grammars must be reformulated, weak- ened, or abandoned. Chomsky himself has never appeared to find much value in Labov’s work: for him it is a study of linguistic performance and has very little to offer to a better understanding of language.

There is no denying, however, that recent studies employing the linguistic vari- able have added an important new dimension to our understanding of language.

It is now well documented that variation is a linguistic fact and that it is not haphazard. Previously, dialect geographers had amply documented the differ- ential but systematic distribution of linguistic forms, and the new techniques of investigation have revealed similar patterns that can be related to a variety of social and even linguistic factors. It is also apparent that people are aware, some- times consciously and sometimes not, that certain variants have more (or less) prestige than others. They are also able to modify their speech to reflect chang- ing circumstances, and do so quite systematically.

The distribution of the variants of variables also seems to be clearly related to changes that languages undergo. No longer is it possible to separate synchronic and diachronic matters into two mutually exclusive domains; descriptive and historical matters are interrelated. Moreover, some findings, such as Labov’s dis- covery of the cross-over phenomenon, appear to indicate not only the direction of change but possibly also some of its motivation. We will turn our attention to some of these matters in the chapter that follows.

Discussion

1. Try to devise a small-scale study focusing on the pronunciations of final clusters before words beginning with vowels and consonants and controlling for the grammatical function of the final stop in the cluster, e.g., past five, passed five; past eleven, passed eleven. Try to collect data from subjects having different social backgrounds and attempt to vary the formality of usage. Do your results correspond in any way to those reported in this chapter?

2. In comparison with sociologists, linguists who have studied linguistic variation have used very simple ways of determining the social-class mem- bership of individuals. They have argued that more sophisticated approaches are hardly necessary since the results they have achieved have been ‘very satisfactory’ (see Chambers and Trudgill, 1998, p. 49, and Chambers, 2003, pp. 47–54). How justified is such a claim about their methodology in gen- eral? In thinking through your answer, consider how linguists are inclined to treat specialists in other disciplines who treat linguistic data with a similar elementary approach.

3. The rule that tells you to pronounce the t at the beginning of top with a little puff of air, as [th], is a categorical rule, as is the rule that tells you to

pluralize man as men; such a rule always operates. The rule that allows you to pronounce the first vowel in either to resemble that in beet or in bite is an optional rule. However, the rule that allows you to say singing on one occasion and singin’ on another is a variable rule. Try to clarify the concept of ‘rule’ in each case. Look for other kinds of behavior in which the same kinds of distinction can be made. Try to assess the compatibility of these different notions of ‘rule’ within a single theory of behavior.

Further Reading

The journals Language in Society, Journal of Sociolinguistics, and Language Variation and Change often report studies conducted in this ‘quantitative’

tradition.

8 Change

Recent work in sociolinguistics has raised once again a long-standing question:

can linguistic change be observed while it is actually occurring? In modern linguistics the answer to that question has usually been a resounding negative.

Following the example of two of the founders of the modern discipline, Saussure (1959) and Bloomfield (1933), most linguists have maintained that change itself cannot be observed; all that we can possibly hope to observe are the consequences of change. The important consequences are those that make some kind of dif- ference to the structure of a language. At any particular time, it certainly may be possible for linguists to observe variation in language, but that variation is of little importance. Such variation must be ascribed either to dialect mixture, that is, to a situation in which two or more systems have a degree of overlap, or to free variation, that is, to unprincipled or random variation. Linguists therefore attached little or no theoretical importance to variation. Only in recent decades have some of them seen in it a possible key to understanding how lan- guages change.

Dalam dokumen An Introduction to Sociolinguistics (Halaman 191-199)