Andrew Radford - Linguistics An Introduction [Second Edition]

backing is characteristic of Burnouts. As noted above, Eckert’s work is important because she demonstrated the power of observing self-forming and self-defining groups of people, rather than simply assigning people to well-known global social categories and observing variation within them (exercise 3).

In summary, we have painted a picture of an intimate relationship between a

number of sociological variables – social class, educational achievement, gender, ethnicity, social network and community of practice – and a range of linguistic variables. It seems quite clear that our position in society can shape certain aspects of our linguistic behaviour. Linguistic variability is not divorced from social conditioning. We now turn to a different type of variation.

Sound variation

53

0.9

0.8

king

0.7

el bacw 0.6

o

0.5

0.4

0.3

0.2

eight – the nearer the score to 1.0, the

0.1

higher the frequency of v

actor w

0

F

Bur nout girls

Bur nout boys

Joc k girls

Joc k boys

Sex and social categor y

Figure 20 Degree of backing of /ʌ/ among students at a Detroit high school

(based on Eckert 2000: 118). Adapted and reprinted with permission of Wiley/

Blackwell Publishing.

Stylistic variation

We are all probably conscious that we speak differently to a teacher

than to our friends over a coffee. We tend on the whole to speak using a more standard dialect with the teacher, and use more non-standard or informal language when having a chat. Similarly, we may find that we speak in a more standard way when discussing some topics – say, politics or linguistics – than when discussing others – yesterday’s baseball game, or your neighbour’s latest antics. Linguistic variability that is dependent on the social context we find ourselves in or the topic of the conversation we are engaged in is usually termed stylistic variation. Allan Bell, a linguist from New Zealand, developed a model for the analysis of stylistic variation known as audience design. He claimed that in designing our style of speech at any particular time, we assess the sociolinguistic characteristics of our addressees and adapt the way we speak to conform to these characteristics.

Let’s look at an example. Nik Coupland investigated the extent to which an

assistant in a travel agency in Cardiff, Wales, shifted her speech to match that of the social class of her clients. One of the variables he studied was the flapping of (t) – i.e.

the use of [bʌɾə] instead of [bʌtə], and the results of this part of his study appear in

figure 21. These results show how the assistant altered her use of this variable quite radically when speaking to clients of different social classes (exercise 4).

The model of audience design helps to explain why people seem, to a non-

native ear, to ‘pick up’ the accent of places they stay in. A British or North American English speaker spending a couple of years in Australia would have a predominantly Australian English speaking audience and would accommodate to

that variety so often when conversing that, to non-Australians, they may ‘sound

54

sounds

90

80

70

60

50

40

30

% use of (t)-flapping

20

10

0

Upper Middle

Low er Middle

Upper W orking

Low er Working

Social class of client

client

travel agent

Figure 21 A travel agent’s style shifting to clients: (t)-flapping (from Coupland 1984: 63). Adapted and reprinted with permission of the author and Mouton de Gruyter.

like an Aussie’. What this indicates, then, is that variation in language is constrained not only by the social characteristics of the speaker, but also by those of the addressee in any conversation; variation is also interactionally determined.

Linguistically determined variation

We would be wrong to go on from the above to claim that it is only

social factors that determine the structure of variation within a speech community.

Linguistic factors, too, play a considerable role in determining the relative use of different variants of a variable. One variable which appears to behave in a similar way across the English-speaking world is so-called consonant cluster deletion or more specifically -t/-d deletion. This involves the variable deletion of word-final

[t] or [d] when it follows another consonant. So we find examples such as those in

(28), where the candidate for deletion appears in bold and the phonetic transcriptions give variant pronunciations depending on whether [t] or [d] delete:

(28)

Data set 1:

best friend

→ [bɛst frɛnd] – [bɛs frɛnd]

cold weather

→ [koʊld wɛðə] – [koʊl wɛðə]

Data set 2:

he stuffed the turkey → [hiː stʌft ðə tεːkiː] – [hiː stʌf ðə tεːkiː]

she seemed funny

→ [ʃiː siːmd fʌniː] – [ʃiː siːm fʌniː]

Data set 3:

most of the time

→ [moʊst əv ðə taɪm] – [moʊs ə ðə taɪm]

ground attack

→ [ɡraʊnd ətæk] – [ɡraʊn ətæk]

Data set 4:

he seemed odd

→ [hiː siːmd ɒd] – [hi siːm ɒd]

she passed a test

→ [ʃiː pɑːst ə tɛst] – [ʃiː pɑːs ə tɛst]

Sound variation

55

(Note that in these examples, we transcribe ‘r’ sounds as [r], a common practice unless more precision is needed.)

As you read this set of data, you will probably feel that the further you go down the sets, the less likely you would be to hear the second example in each phonetically transcribed pair, that is the example in which [t] or [d] is deleted. This is because in each set of data the word final [t] and [d] are in different linguistic contexts, and it is these contexts which are affecting whether or not deletion of [t] or [d] seems likely.

In data sets 1 and 2, [t] and [d] are followed by consonants, whereas in sets 3 and 4

they are followed by vowels. Research has shown that deletion is less likely before vowels than before consonants. In data sets 2 and 4, [t] and [d] are the realisation of the past tense ending -ed, whereas they don’t have this function in sets 1 and 3. We would expect, based on evidence from many English-speaking communities around the world, to find less deletion in the -ed examples, since phonetically the [t] and [d]

are the only indication of the tense of the verb. This means that linguistic factors (whether the candidate for deletion precedes a vowel or a consonant and whether it encodes past tense or not) predict most deletion in data set 1 and least in data set 4.

Table 5 provides evidence from a number of dialects of English to support this prediction. It is important to note that the ordering predicted on the basis of the linguistic factors is the same in each of the dialects investigated, despite the fact that there are quite considerable differences in the actual figures with the Puerto Rican speakers generally deleting final [t] and [d] much more frequently than speakers of Standard American English. What these differences show, of course, is that social factors as well as linguistic factors are playing a part in this variation.

The pattern that we see in table 5 illustrates what is known as an implicational scale. This notion is exemplified in a hypothetical case in table 6. Here ‘+’

signifies that a particular deletion always takes place and ‘−’ that it never takes place. Thus, in Dialect A, final [t] and [d] are always deleted, irrespective of linguistic context and in Dialect D they are always deleted when followed by a consonant so long as they do not encode tense – otherwise they are never deleted in Dialect D. Dialects B and C are intermediate between A and D. Now, we can look at table 6 and formulate the implicational statement in (29):