P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. 7 May 27, 2005 15:13 The birth of new dialects Paul Kerswill and Peter Trudgill 1 New Dialects as Heightened Dialect Convergence: Research Issues 1.1 Introduction New-dialect formation, as conceptualised by Trudgill and others (e.g. Britain and Trudgill 1999; Trudgill 2004), refers to the emergence of distinctive, new language varieties following the migration of people speaking mutually intelligible dialects to what, to all intents and purposes, is linguistically ‘virgin’ territory.1 As such, it is an extreme, and often very rapid, form of dialect convergence. Examples probably abound in world history, but only a few have been described in detail. There seem to be two main scenarios in which new-dialect formation takes place: the settlement of a relatively large territory, either previously uninhabited or in which a previous population is ousted or assimilated; and the formation of a new town in a geographically delimited area in which relatively intense interpersonal communication can take place. Examples of the former are the settlement of New Zealand largely by English speakers in the nineteenth century, and the transport of indentured labourers from the Hindispeaking areas of the Indian subcontinent to Fiji, Mauritius, Trinidad, and South Africa, also in the nineteenth century. Examples of the latter are the establishment of the Norwegian towns of Høyanger, Odda, and Tyssedal, and the English town of Milton Keynes. In this chapter, we start by outlining the processes and stages found in new-dialect formation. Then we review the sociolinguistic histories of a number of new dialects – or immigrant koines (Siegel 1985: 364; Kerswill 2002a) as they are also known. We follow this with a scrutiny of what is normally considered the crucial stage in the formation of a new dialect: the linguistic strategies of the first native-born speakers, that is, the children of the original speakers. 1 New dialects are also formed by a process of split and subsequent dialect divergence; we are not concerned with this here. A more precise term for Trudgill’s concept, but one which is not in common use, might be koine formation. 196 P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. May 27, 2005 15:13 The birth of new dialects 1.2 197 Processes and stages in new-dialect formation Because new-dialect formation is never instantaneous, we can expect to isolate chronological stages through which it must pass before reaching completion. At the same time, we can expect these stages to be characterised by a number of processes, both linguistic and social-psychological, which may be found in one or more of the chronological stages. To take an extreme case: we are asking how a disparate collection of migrants, thrown together at random in a new, unpeopled, place, start out as linguistic individualists with no social bonds to each other, and end up forming, within a couple of generations, a cohesive speech community (cf. Patrick 2002). This is both an oversimplification and a caricature, however, because each set of initial circumstances is different, as we have suggested. The type of linguistic input involved, especially in terms of the linguistic differences between speakers, may have an effect on the speed with which an embryonic new dialect ‘focuses’, or acquires norms and stability (though we will not be discussing the factor of linguistic differences in this article; see Kerswill 2002a). Non-linguistic factors will be crucial, and may override linguistic ones: a new settlement where people are isolated from each other will take much longer to form a ‘speech community’ than one where there is intensive contact. People may migrate as individuals or in groups. Clear social divisions in the new society will inhibit linguistic uniformity, but will instead promote the appearance of distinct sociolects with focused norms for their use. Thus, we are asking a complex question that has as much to do with forces of social integration and disintegration as with a purely linguistic account of language change. This chapter will attempt to isolate factors that affect the rapidity with which linguistic focusing takes root in new-dialect formation. We turn now to the processes themselves, in a discussion based on our earlier work (Trudgill 1986, 1998; Trudgill et al. 2000; Trudgill 2004; Kerswill and Williams 2000). 1.2.1 Mixing Mixing refers to the coexistence of features with origins in the different input dialects within the new community, usually because speakers have different dialect origins. Trudgill (1986: 129) notes features in modern Newfoundland English which are directly ascribable to its southwest English and Irish origins, respectively; while Siegel (1997: 115) lists grammatical and lexical forms in Fiji Hindi which have origins in different Hindi dialects. 1.2.2 Levelling Levelling is concerned with the selection of forms found in the mix. An early use of the term in the context of new-dialect formation is to be found in Blanc (1968), who describes its operation in the development of Modern Hebrew, which he refers to as a koine. Blanc talks of levelling, first, as P1: FXS/ABE P2: FXS 0521806879c07.xml 198 CU1837B-Auer et al. May 27, 2005 15:13 Macrosociolinguistic motivations of convergence and divergence applying to the children of the early migrants whose speech diverged from their parents’ non-native, substratum-influenced speech, especially in informal situations. Secondly, he refers to the ‘leveling of communal differentiation’ (Blanc 1968: 240), referring to the erasure of differences reflecting the speakers’ linguistic substrata. Dillard (1972; quoted in Siegel 1985: 364) gives the following definition, which is a useful starting-point: Dialect leveling is the process of eliminating prominent stereotypable features of differences between dialects. The process regularly takes place when speakers of different dialects come into contact, such as in migration. The key word here is ‘stereotypable’. Stereotyping of a feature may occur when it is either demographically a minority form in the new community, or when for some reason it has become ‘salient’, perhaps because it deviates linguistically or because it has become associated with a stigmatised social group (see Trudgill 1986: 11–12; Kerswill and Williams 2002a). An example of this is discussed by Kerswill (1994: 157), who argues that rural migrants in the Norwegian city of Bergen avoid a particular rural dialect vowel because it has attained the status of a stereotype. However, stereotyping is not a necessary condition for the demise of a feature. Trudgill (2004) shows how the purely demographically based proportions of features found in the mix are directly reflected in the features found in the new dialect. The mechanism can be explained in term of Giles and Powesland’s (1997/1975) speech accommodation theory: When people speak different varieties, as in a new settlement, the dialect differences are likely to be exploited – consciously or passively – as part of accommodation. This can explain the mechanism behind the survival of majority forms in a koine: There will be more ‘acts of accommodation’ involving the adoption of majority rather than minority variants simply because there are more conversational contexts in which this can take place. (Kerswill 2002a: 680) – though see Auer and Hinskens (this volume) for a critique of the idea that speech accommodation involving specific features prefigures dialect change involving the introduction of the same features. There are purely linguistic forces at work in addition to levelling, too. In particular, we find the process of simplification, a notion which refers to ‘either an increase in regularity or a decrease in markedness’ (Siegel 1985: 358). In practice, this means a decrease in irregularity in morphology and an increase in invariable word forms, as well as the loss of categories such as gender, the loss of case marking, simplified morphophonemics (paradigmatic levelling), and a decrease in the number of phonemes. The result of levelling is that, in a given location, there is a reduction in the number of exponents of linguistic units on the phonological level (nonpredictable – non-allophonic – variants of phonemes) and on the morpholexical level (distinct, non-predictable forms of lexical items and of exponents of P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. May 27, 2005 The birth of new dialects 15:13 199 morphological categories). Morphosyntactic and syntactic differences will also be levelled, though on the whole such differences are much fewer in number and crop up very infrequently in conversational speech (see Cornips and Corrigan, this volume; and Cheshire, Kerswill, and Williams, this volume). In general, the new dialect is also characterised by simplification in relation to all or most of its input dialects. Mixing, levelling, and simplification are the necessary precursors of newdialect formation. Together, they can be said to constitute koineisation. 1.2.3 Interdialect development Interdialect forms can be defined as those which were not actually present in any of the dialects contributing to the mixture but which arise out of interaction between them. Such forms are of three types. They may be (a) forms which are simpler or more regular than any of those present in the original dialect mixture. They may also be (b) intermediate forms (see Kerswill 1994: 161; Trudgill 1986), which are most usually those which are phonetically intermediate between two contributing forms in the mixture. Additionally, morphological or lexical units may combine variants from more than one contributing dialect, giving what are essentially novel coinages. Finally, they may be (c) forms which are the result of hyperadaptation. The best known type of hyperadaptation is ‘hypercorrection’, in which speakers attempt to use forms from higher status accents, but employ an incorrect analysis and extend changes to items where they are inappropriate. 1.2.4 Reallocation Even after koineisation (mixing, levelling, and simplification), some competing variants left over from the original mixture may survive. Where this happens, reallocation may occur, such that variants originally from different regional dialects may, in the emerging new dialect, become social-class variants, stylistic variants, or, in the case of phonology, allophonic variants (see Britain and Trudgill 1999). 1.2.5 Focusing Focusing is the sociolinguistic process by means of which the new variety acquires norms and stability, and is implicit in the linguistic process of levelling. The term was elaborated by Le Page, who sees focusing as part of an individual’s more or less conscious behaviour when interacting with others. Thus, [w]e engage in activities I call projection and focussing: we project on to the social screen the concepts we have formed, by talking about them, so as to furnish our universe and try to get others to acknowledge the shape of our furniture; we in turn try to bring our concepts into focus with those of others, so that there is feedback from the social screen through language. (Le Page 1980: 15–16) Among the ‘concepts’ we form are, we argue, the language varieties which are perceived by the speaker to exist in the community, along with the social groups P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. 200 May 27, 2005 15:13 Macrosociolinguistic motivations of convergence and divergence associated with them and the norms for their use. Our emphasis on ‘perception’ is important, since these varieties and groups are likely to be simpler and more stereotyped than the detailed and complex picture a linguist or anthropologist might paint. This indeterminacy on the part of speech community members is in line with Le Page’s ‘hypothesis and four riders’ model of linguistic accommodation to groups with which we wish to be associated (Le Page 1980): one of the riders states that we are restricted by ‘the extent to which we have sufficient access to [the model groups] and sufficient analytical ability to work out the rules of their behaviour’. It is obvious that, in a stable speech community with clear norms, the ‘bringing into focus’ of these concepts, including language varieties, is more readily achieved than in a ‘diffuse’ community, where the social groups are not easily discerned, many language varieties are spoken, and the norms for language use are either not shared or are non-existent. In the early stages of new-dialect formation, the newly arrived migrants face a maximally diffuse situation. They engage in projection and focusing to make sense of the sociolinguistic melting pot. Over time, certainly over two or more generations, countless attempts lead to greater success in focusing. The outcome is what we call a ‘focused’ dialect, where there is a measure of uniformity as well as agreement on the social symbolism of the variation that exists. Under Le Page’s model, levelling is a direct consequence of speakers’ attempts to project and focus. And the model subsumes accommodation theory in that it allows not just for the interpersonal relations that are a prerequisite for accommodation but also for speakers’ orientations towards, and beliefs about, social groups and linguistic varieties. In a similar vein, Kerswill (2003) argues that levelling is not only the consequence of face-to-face accommodation but is also dependent on attitudinal and identity-based factors. We turn now to the stages of new-dialect formation. Trudgill has argued for the following three stages, roughly corresponding to the first three generations of speakers (Trudgill 1998; Trudgill et al. 2000): Stage I II III Speakers involved adult migrants (first generation) first native-born speakers (second generation) subsequent generations Linguistic characteristics rudimentary levelling extreme variability and further levelling focusing, levelling, and reallocation In Stage I, we find some levelling among the adult migrants, as they avoid what they perceive as ‘marked’ forms; motivations will be both the maintenance of intelligibility and the avoidance of social stigma. At the same time, people will accommodate to each other linguistically in their conversational interactions – though, as already alluded to, the evidence that this form of ‘short-term’ accommodation actually involves the adoption of the features that will later end up in the new dialect is far from clear. (See Coupland 1984; Auer and P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. May 27, 2005 15:13 The birth of new dialects 201 Hinskens, this volume; Hinskens 1996: ch. 11; and Kerswill 2002a: 680–682 for discussions of this point and the argument that accommodation takes place through ‘identity projection’ rather than pattern matching.) Stage II (the focus of this chapter) involves the children of the migrants. As we shall see, their speech shows great inter-individual variation. Surprisingly, there is also much greater intra-individual variation than we would normally find in an established community. The nature of the transition to Stage III, during which the new, focused, dialect appears, is at present rather unclear. What we can say is that it must involve further accommodation – both face-to-face and driven by social psychological factors. Because the focusing in Stage III can be achieved by the third-generation speakers, this accommodation must be rather extensive. On the other hand, the focusing may take several generations to occur, and, in some cases, it may never be completed. 1.3 Related contact-induced changes: pidginisation, creolisation, and regional dialect levelling The processes observed in new-dialect formation are similar to those found in other contact-related changes. New-dialect formation has much in common with both pidginisation and creolisation, in that new language varieties emerge from relatively intense face-to-face contacts between people speaking different varieties. New-dialect formation has more in common, however, with creolisation than with pidginisation, since both lead directly to a new native-speaker generation. As with creolisation, new-dialect formation involves a disturbance in the ‘normal’ cross-generational transmission of language. Normal transmission is defined by Thomason and Kaufman (1988: 9–10) as taking place when ‘a language is passed on from parent generation to child generation and/or via peer group from immediately older to immediately younger’. However, the analogy should not be taken too far, because, in new-dialect formation, there is no need for individuals to abandon their original language varieties, and the process can take many generations. Creolisation, by many definitions, entails language shift by individuals and, for linguistic communication to take place at all, must be rapid – this is essential to Bickerton’s (1977) view of creolisation as ‘first-language learning with restricted input’. More recent work on creoles suggests, however, that creolisation may take place over more than one generation (Arends 1995). (See also Kerswill 2002a: 695–696.) As we have seen, new-dialect formation involves levelling. In much of Europe, a related phenomenon of ‘regional dialect levelling’, or ‘dialect supralocalisation’ has been reported for local speech generally (Auer and Hinskens 1996; Hinskens 1996, 1998a; Sandøy 1998b; Thelander 1980, 1982; Foulkes and Docherty 1999; Williams and Kerswill 1999; Milroy 2002a; Britain 2002a; Torgersen and Kerswill 2004; Hinskens, Auer, and Kerswill, this volume). By P1: FXS/ABE P2: FXS 0521806879c07.xml 202 CU1837B-Auer et al. May 27, 2005 15:13 Macrosociolinguistic motivations of convergence and divergence this is meant that, within a region where mutually intelligible dialects are spoken, there will be a decrease in linguistic differentiation associated with location; the result is the disappearance, or attrition, of local dialects. Differentiation comes to be on a regional, rather than a local, basis. These outcomes, along with simplification, are shared with the levelling found in new-dialect formation, as can be demonstrated in regions where both are taking place at the same time (Kerswill 2002a: 684); in such cases, new dialects are in advance of the levelled regional dialects, anticipating their shape. However, in the discussion of regional dialect levelling, the mechanism that underlies it is rarely examined: we are still at the descriptive stage. It seems to us that the mechanism cannot be the same as that for levelling in new-dialect formation, since the latter relies entirely on the reduction of variants brought about by face-to-face interaction. Across a whole dialect area, this is patently impossible. Instead, we see a continuation of the ‘old’ process of diffusion (Chambers and Trudgill 1998), with features spreading from large urban centres at a measurable rate, doubtless now accelerated by the increase in mobility and (perhaps) a priming effect caused by the broadcast media (Stuart-Smith 2001). For example, the spread of the use of /f / for /θ/ in British English varieties can be shown to be geographically gradual, despite being a recent change in much of Britain (Kerswill 2001, 2003). In what follows, we review a number of cases of new-dialect formation, seeking out factors, linguistic and social, which favour focusing, and those which hinder it. 2 From Melting Pot to Speech Community: Previous Research Studies of new-dialect formation have been conducted within a number of frameworks. As we shall see shortly, the advantage of this methodological eclecticism is that it allows us to tease out a rather wide range of mainly social factors influencing the progress of new-dialect formation. One tradition is linguistic and descriptive in orientation, and deals in traditional units of analysis such as the phoneme. It shows a concern for careful linguistic description; origins of features are sought in input dialects and in such processes as levelling. Social explanations are post hoc and concentrate on demography. A parallel trend has been much more interpretive, dealing with the social symbolism of the varieties from the speakers’ point of view. There is a strong emphasis on ethnography, with members’ concepts (cf. Le Page) being sought through ethnographic interviews and observation. We deal with each of these two types in turn. 2.1 Linguistic-descriptive studies and the role of demography Studies of the first kind (the linguistic-descriptive) have given us insights into the processes and stages of new-dialect formation, and are indeed the source of our P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. May 27, 2005 15:13 The birth of new dialects 203 previous discussion of these. Omdal’s brief (1977) study of Høyanger, discussed in Trudgill (1986), demonstrates that Stage II speakers there are still relatively unfocused, focusing being achieved by the third generation (Stage III). Both Siegel’s (1987) research on Fiji Hindi and Mesthrie’s (1992) investigation of South African Bhojpuri show the importance of demography for the outcomes in the new dialects, by showing how statistics on the origins of the first-generation migrants are reflected in the proportions of morphological and lexical features from different Hindi dialects found in the koines. As we have seen, Trudgill (2004) argues in a similar vein (see also below, section 3.1). Demographic factors of a different sort are adduced in Britain’s (1997a,b) study of the English Fens: that of the opportunities for children to form social relationships among themselves. Britain deals with the focusing of two vocalic variables following migrations as long ago as the seventeenth century; interestingly, he finds that one of the features (which is linguistically complex) is only now beginning to resolve the conflicting norms brought about by the migrations. He proposes a partial explanation as follows. In the days before widespread schooling, children living in the sparsely populated Fens had little opportunity to mingle with other children, and so the potential to develop new norms in the mixed dialect situation they found themselves in was severely limited (Britain 1997a: 165). This argument chimes well with Eckert’s (2000) contention that, in the developed world, the school is now the chief locale for the socialisation of children, and is therefore the primary setting for language change. We shall have much more to say on the role of children later in this chapter. Finally, we can observe a third key type of demographic information. This is to do with population stability versus transience. Like Britain’s study, Sudbury’s (2000, 2001) study of Falkland Islands English is quantitative and makes use of the linguistic variable. She finds that the English spoken there is ‘distinguishable from other English varieties’ (2001: 64). For example, it has developed distinct allophones of the vowels of /a i/ as in price and /aυ / as in mouth 2 before voiced and voiceless consonants, in the manner of Canadian Raising (Chambers 1979). Yet the allophones are not as distinct as those in Canadian English, and this is perhaps symptomatic of the lack of focusing she finds in Falklands English, despite a settlement history going back to the mid-nineteenth century. Instead, she finds considerable variation within groups defined by sex, age, and residence groups, as well as more-than-expected variation within individual speakers (Sudbury 2001: 64). In explaining this situation, Sudbury’s main argument concerns, first, the isolation of communities in the Falklands, leading not only to low rates of inter-community contact but also to a lack of opportunities for children to mix. Second, she points to the transience of the population caused 2 These keywords are used mnemonically following Wells (1982). P1: FXS/ABE P2: FXS 0521806879c07.xml 204 CU1837B-Auer et al. May 27, 2005 15:13 Macrosociolinguistic motivations of convergence and divergence by high rates of in- and out-migration, due to the fact that many settlers were contract workers (Sudbury 2000: 374–375). In addition to these demographic factors, there is a continued desire to maintain links with Britain – a factor which contrasts sharply with other southern-hemisphere English-speaking countries. Falklands Islands English is, she argues, still at the transition between Stages II and III (2000: 372). Transience and attitudinal factors will be explored further below. 2.2 Ethnographic studies: dialect as a local identity marker In this section, we deal with two recent studies of a somewhat different type of new dialect: these are what may be termed dialect islands, by analogy with ‘language island’ (see Rosenberg, this volume). Dialect islands are the result of the migration of people from a single area to another part of the same language area, while language islands are the consequence of migration to an entirely different language area. The distinction is important because, in dialect islands, we can expect levelling to occur in relation to the surrounding dialects, made possible by mutual intelligibility. At the same time, of course, there will be internal levelling to the extent that there are dialectal differences between the input speakers – as there almost invariably are. Mæhlum’s (1997) study of a dialect island in northern Norway is mainly ethnographic in approach. Over some three decades from 1791, farming families from the east of Norway were encouraged to move to two adjacent regions in the north of the country, Bardu and Målselv, which had not previously been farmed, though Saami reindeer herders were present. Local speech in these regions is to a great extent coloured by the eastern dialects of the settlers, with some speakers even today showing virtually no northern features. The study is especially valuable because of the existence of a previous description published some seventy years after the main settlement (Reitan 1928), which allows us to make some inferences about Stage II, as well to make comparisons with presentday speech. Furthermore, a good deal is known about where the original settlers came from. As Mæhlum points out, Reitan (1928) provides evidence that at least one majority form found in the dialect mix in Bardu had already won out in the speech of Stage II speakers (those of Reitan’s informants who were the children of settlers). This is the form of the first person singular pronoun, which took the form /ε i/, this being the variant found in the southeastern district of Tynset, from where Bardu was mainly settled. In the 1920s, there was no evidence of the more general eastern /je / or northern /e /, which must also have been present to some extent in the mix. However, there was also levelling away from strongly localised and therefore stereotypable Tynset forms, such as /v œt œ/ for ‘to know’, which was replaced by northern and more general Norwegian /vit ə /. P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. May 27, 2005 The birth of new dialects 15:13 205 Despite all this, focusing had clearly not taken place even as late as the 1920s: The struggle between southern [i.e. southeastern] and northern dialects has, however, not ceased, and the language of the two colonies has therefore not reached any degree of uniformity. There is vacillation both with respect to the form and the inflectional morphology of words, and it is not easy to determine what should be regarded as ‘genuine’ in the dialect. (Reitan 1928, cited in Mæhlum 1997: 19; our translation) Moreover, individual speakers would show instability in their use of particular forms (Mæhlum 1997: 19). Demography clearly played a part in the slow koineisation in these two districts: most farms were isolated, and until recently there were no compact villages or towns in which norms could be developed. Instead, according to Mæhlum, norms would have been much more local, perhaps based on an individual valley. The main thrust of Mæhlum’s study, however, lies in its emphasis on finding reasons for the maintenance of a continuum of eastern-derived dialects in the north of Norway for two hundred years. Historical records reveal that endogamy was practised right until the Second World War; in the early days, some men apparently travelled the 1,400 kilometres to their original districts to find wives. Mæhlum talks of the long-standing ‘positive self-image’ of the døl, or ‘valley person’, and the high status that was accorded to døler (as pioneers and farmers) by the authorities. In these communities, members of the out-group (the coastal residents of north Norwegian descent) were known disparagingly as skolp (a local word referring to coastal people (Mæhlum 1997: 13)). The valley people’s positive self-image continues to this day, as witnessed by Mæhlum’s account of schoolchildren’s self-reports, as well as (more recently still) by the existence of a local information website which døler are invited to make use of. Dyer’s (2000, 2002) study of the new town of Corby in the English East Midlands takes a similarly interpretive approach. Like Mæhlum, she sees the development and maintenance of a new, distinctive dialect as a reflection of the maintenance of locally relevant social categories, particularly those concerned with making distinctions between in- and out-groups. The new town of Corby began in the 1930s with the setting up of an iron and steel works by a Scottish company. Workers were directly transferred from the depressed industrial region around Glasgow, with the result that the population rose from 1,500 to 36,000 in thirty years. In 1971, Scottish-born people accounted for some 30 per cent of the population of the town (Dyer 2002: 101). Dyer shows that, while most pronunciation features today have an English origin, some Scottish features have been maintained by the most recent generation of speakers. One such is the monophthongal pronunciation, typically [o], of /ə υ / as in goat, a clearly Scottish variant that differs strongly from local, southern diphthongs such as [ə υ ]. Many young males use this monophthong, while young females tend to use a fronted variant of the southern diphthong, [ə  ], which P1: FXS/ABE P2: FXS 0521806879c07.xml 206 CU1837B-Auer et al. May 27, 2005 15:13 Macrosociolinguistic motivations of convergence and divergence is spreading and is popular among young females in the south of England (Kerswill and Williams 2000: 106; Cheshire, Kerswill, and Williams 1999; Kerswill and Williams 2005). This gender-differentiated pattern holds, regardless of whether a speaker’s family is of Scottish origin. Thus, although there has been considerable levelling within the town, with the effect of linking it with its region both linguistically and sociolinguistically, the town’s dialect has in certain crucial respects not become part of the regional dialect levelling. Dyer’s interpretation of these results, informed by detailed and extensive interviewing, is as follows. She finds that the oldest speakers, most of whom are Scots, perceive both themselves and the younger inhabitants of Corby as speaking a Scottish variety. Outsiders also perceive the accent as Scottish. However, the young people themselves do not see themselves as Scots but as local Corby people, claiming, with some justification, that they all speak the same. Thus, she argues, ‘Scots features have been reassigned or reallocated in the new Corby dialect to index local, rather than ethnic identity’ (Dyer 2002: 113). Dyer’s findings probably have a consequence for the way we should approach local speech and identity, even in areas which are arguably under strong pressure from the diffusion of London (or, elsewhere, other metropolitan) features, leading to the regionalisation of local speech (that is, levelling leading to distinctiveness at the regional, not local, level). If we find a strong linguistically marked local identity even in a small town in a relatively densely populated region, then this should hold for other places, too, even if the dialect at first hearing is not so distinctive as that of Corby. However, we should be cautious in the expectation that linguistically marked local identities are powerful everywhere. A related approach to identity factors is to ask whether speakers of new, levelled dialects, such as that in Milton Keynes, can recognise other native speakers. Kerswill and Williams (2002b) find that dialect recognition rates in levelled dialects, including those affected by regional dialect levelling (e.g. Reading), is in fact low, and that this is especially true across generations in cases where the levelling is rapid. Conversely, the study showed high recognition levels for non-levelled urban dialects, such as that of Hull. This suggests that the strength of linguistically signalled identities varies greatly, and that these identities are not necessarily even shared across generations. The studies we have reviewed (as well as a number of others discussed in Kerswill and Williams 2000 and Kerswill 2002; cf. also levelling in language islands – Rosenberg, this volume) are remarkably consistent in their findings, suggesting that, at least for new-dialect formation, we really can speak of a finite set of processes and stages. We have also seen how demographic and local identity factors (accessed using both quantitative and qualitative methodologies) appear to account fairly comprehensively for differences in the application P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. May 27, 2005 15:13 The birth of new dialects 207 of the processes and differences in the rapidity with which the new dialect passes through the stages. We now look in much more detail at the relationships between the stages in terms of the transitions between them. We do this by reporting two of our own studies. 3 The Individual Speaker in the Second Stage of New-Dialect Formation: The New Zealand and Milton Keynes Studies So far, our picture of new-dialect formation is a rather detailed one, with findings converging in a satisfying manner. However, the studies reviewed are largely silent on Stage II, the highly variable speech of the children of the migrants, which we would argue is central to the whole process. The linguistic choices made collectively by people at this stage are necessarily crucial to the outcome of koineisation, since they provide the input to the next (or a subsequent) generation’s focused variety in which the number of available features has been whittled down and final selections made. The studies have, however, given us some insight into the kinds of social factors which favour or inhibit focusing during Stages II and III; these are largely to do with the degree of social interaction which is possible between children (at the micro-level) and across the new community as a whole (on a wider level). We will be examining in some detail the second stage, particularly the transitions into and out of it. That is to say, we are interested in how Stage II speakers differ from those at Stage I, both as a group and in terms of individual parent–child relationships. Secondly, we are interested in the same comparisons between Stage II and Stage III speakers. Previous studies have looked at the outcomes of new-dialect formation, and have identified the nature of the linguistic changes between the dialect input and the stable new dialect. Necessarily, the second stage has largely been by-passed because access to such speakers was not possible, simply because they are long dead. But two recent projects have been able to examine the second stage in more detail: 1. The Origins of New Zealand English (ONZE) project3 is unique in that it draws on a substantial historical recorded corpus of second-stage speakers of a known new dialect. The data come from recordings made by the National Broadcasting Corporation of New Zealand in 1946–1948. The recordings were oral history pioneer reminiscences, mostly from people who were the children of the first European settlers in New Zealand. The style of the recordings is relatively informal, often with family members present. Because of 3 Directed by Elizabeth Gordon at the University of Canterbury, Christchurch, and funded by the New Zealand Foundation for Research, Science and Technology, and by the University of Canterbury. Additional funding has also been made available by the British Council and the Marsden Fund. P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. 208 May 27, 2005 15:13 Macrosociolinguistic motivations of convergence and divergence the subject-matter, fairly extensive incidental information about family history appears on the tapes. About 325 speakers born between 1850 and 1900 were recorded (for more details see Gordon et al. 2004). The data presented here are derived from analyses by Peter Trudgill of 84 New Zealand-born speakers representing many regions of New Zealand, and born between 1850 and 1889. This generation of people represents the first native-born speakers of English in New Zealand – though they were elderly by the time they were recorded. From this data, inferences can be made about the way in which this generation’s highly mixed speech provided input to the focused variety, as well as ways in which their speech reflects the known demographic origins of the settlers. 2. The Milton Keynes project4 examines the speech of children and their parents in a koineising new town. It is unique in recording Stage II speakers while they are still children, and in being able to make a direct comparison with the speech of their parents, who are among the original migrants. Milton Keynes, situated some 70 kilometres northwest of London at the boundary between the south Midlands and Home Counties modern dialect areas (Trudgill 1990: 63), was designated a new town in 1967. Between that date and the 1991 Census, the population increased from 44,000 to 176,000, rising further to 207,000 by the 2001 Census. The rationale for the project is the view that children’s and adolescents’ linguistic choices are the prime sources at least of phonological change (following Eckert 1988; and Aitchison 1981: 180). A socially homogeneous group of 48 children (8 boys and 8 girls from each of three age groups: 4, 8, and 12 years), who had been born in Milton Keynes or who had moved there within the first two years of life, was recorded in 1991. Almost all the children were born to parents who had moved to Milton Keynes as young adults, mainly from elsewhere in the southeast of England, but also northern England and Scotland. This distribution turns out to match rather closely the regional origins of the population as a whole (Kerswill and Williams 2000: 79). One caregiver (in all but two cases the mother) was recorded for each child. The children attended a nursery, a first school, and a middle school, respectively, in two adjacent neighbourhoods which were among the first developments in the new town, with a high proportion of the housing available for rent and with a population showing a relatively high rate of signs of social deprivation (as measured by the proportion of children receiving free school meals). The selection 4 ‘A new dialect in a new city: children’s and adults’ speech in Milton Keynes’, directed by Paul Kerswill at the University of Reading, 1990–1994, and funded by the Economic and Social Research Council (ref. R000232376). Ann Williams was the research fellow working on the project. Further data are from the project ‘The role of adolescents in dialect levelling’, jointly directed by Jenny Cheshire, Paul Kerswill, and Ann Williams, 1995–1999 (ESRC ref. R000236180). P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. May 27, 2005 15:13 The birth of new dialects 209 procedures, which involved approaching schools and parents and requesting volunteers, meant that a quasi-random sample was obtained within the broad parameters of age and sex. The children were recorded in same-sex pairs by Ann Williams, with a number of elicitation tasks being conducted in addition to an interview. The variables quantified were phonological. 3.1 Variability in Stage II 3.1.1 New Zealand The most striking fact about the New Zealand archive is the variability that exists, both between and within individuals. (By contrast, the variability in the Milton Keynes data is much less – for reasons we will return to.) Differences between the speakers, many of whom are siblings or class-mates, are much greater than the expectation would be for an established settlement (village or small town) of similar size. This effect is most likely accentuated by the absence of a stable peer-group variety. In such a situation, adults, especially parents and other caregivers, will have a greater than usual influence on children’s speech. We can take the example of Arrowtown, which was settled from about 1860 during a gold rush by people from various parts of the British Isles, other parts of New Zealand, and Australia. Among nine Arrowtown people born between 1863 and 1886, we find the following vowel variants: face: price: goat mouth: near: [e] [ai] [o] [ u−] [iɾ ] [ei ] [ɑ i] [oυ ] [ε u−] [iəɹ ] [ei] [ai ] [oυ ] [ευ ] [ij ə ] [ε i] [ɑε ] [ɔυ ] [ευ ] [iə ] [æi] [ɑε ] [θ u−] [æυ ] [iəɹ ] [ɐυ ] [æə ] [iə ] The fact that most of these can be traced to a British Isles provenance suggests that their users have ‘inherited’ a British dialect from their parents. This is not the case, however, since none of these speakers uses a variety which is unequivocally localisable to a single area of the British Isles. Instead, we find that speakers appear to have made an idiosyncratic selection of features from the available choice in the dialect mixture. This leads to great inter-individual variability of a kind that is unexpected from people with broadly similar backgrounds. For example, Mr Malcolm Ritchie has the following features: 1. /θ / and /ð / are realised as dental stops, [t] and [d], as in Irish English 2. Syllable-final /l/ may be clear (i.e. non-velarised), as in Irish English 3. He has h-dropping in words like home, an English feature absent in Ireland 4. He has a distinction between // and /w /, thus distinguishing which and witch. This feature is never combined with h-dropping in the British Isles. 5. He distinguishes between [ei] reflexes of ME /a / as in gate and [ε i] reflexes of ME ai as in chain. P1: FXS/ABE P2: FXS 0521806879c07.xml 210 CU1837B-Auer et al. May 27, 2005 15:13 Macrosociolinguistic motivations of convergence and divergence Mr Ritchie’s sister-in-law, Mrs H. Ritchie, attended the same school at the same time as he did, yet has some quite different features in her speech. Unlike Mr Ritchie, she has close realisations of /æ/ as [ε ] and /e / as [e], while he typically has more open variants. She also lacks the distinction between gate and chain, found in Mr Ritchie’s speech. In Arrowtown as a whole, we find variable rhoticity, variable h-dropping, variable tapping of intervocalic /t /, and variable use of clear (vs. dark) /l/ in syllable-coda positions. What is happening here is a kind of ‘shopping-basket’ effect: in situations where there is no stable adult model, children are able to choose from a wider variety of adult models than otherwise. Choice would seem to be highly individual, provided the features are heard sufficiently frequently among the adults to be noticed by the children (what Trudgill 2004 calls the threshold rider). This is not to say that the choices are entirely random. Trudgill argues that though it is impossible to predict which form any individual will select from those available, the proportions of variants present in the accents of the children, taken as a whole, derive in a probabilistic manner from, and will therefore reflect – subject to the threshold rider – the proportions of the same variants present in the different varieties spoken by their parents’ generation taken as a whole. In the Norwegian Arctic territory of Spitsbergen, which has a highly transient population, Mæhlum (1992) finds children likewise adopting very original combinations of features. Nevertheless, she concludes that there are a number of linguistic tendencies around which the children cluster, by which they signal various kinds of regional and other group allegiances – or a strategy of neutrality. In a similar fashion, ethnographic methods in the Milton Keynes study show children signalling social identities such as these, as we shall see shortly. The Stage II speakers in the ONZE corpus also show a great deal of intraindividual variation. For example, Mr Riddle, who was born in Palmerston in 1860, shows an astonishing degree of variability: 1. /æ/ as in trap can be either [ε ] or [a] (but not [æ]) 2. /i / as in fleece varies between the short monophthong [i] typical of Scots and a long diphthong [ə i] typical of southern England – and individual lexical items can occur with both pronunciations 3. /e i/ and /əυ / as in face and goat alternate between Scottish-sounding monophthongal pronunciations with [e] and [o] and very un-Scottish pronunciations with the wide diphthongs [æi] and [ɐυ ] 4. similarly, /a i/ as in price alternates between a typical Scottish diphthong [ε i] and an open central monophthong [a ], half-way between cardinal 4 and 5 in quality, whose provenance is not entirely clear, but could be Lancashire The most likely explanation would appear to be that Mr Riddle grew up in a community providing both English English and Scottish English models, and that for some phonological features he acquired both variants. Mr Riddle is by P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. May 27, 2005 15:13 The birth of new dialects 211 no means the only speaker to display this kind of behaviour. Acoustic analysis, too, shows unusually large envelopes of phonetic space for vowel realisations for a number of informants (Gordon et al. 2004). Care does have to be taken at this point, it is true, since some such variability may be the result of accommodation to the speech of younger New Zealanders during the ONZE informants’ long lifetimes. A parallel case is the speech of adult rural migrants in the Norwegian city of Bergen, whose speech shows similar intra-individual variation as part of their long-term accommodation to the city dialect (Kerswill 1994: 148). Nevertheless, the amount of variability in the speech of many individual speakers is very striking indeed and strongly suggests that idiolects formed in dialect mixture situations may be much more variable than idiolects formed in stable speech communities. This kind of variability is paralleled by Stage II speakers in Spitsbergen (Mæhlum 1992) and, if Reitan’s (1928) report is correct, in Bardu, too. 3.1.2 Milton Keynes The Milton Keynes children generally show much less variability than the New Zealand informants. This is because the town lies between two regions that have been identified as the most dialectally levelled in the country, and most of the in-migrants also come from these regions (see Kerswill and Williams 2000: 80–81; Edwards 1993: 215–216). However, careful phonetic analysis and quantification can reveal quite detailed patterns. We begin by considering the variable (ou), which refers to the realisation of the offset of the vowel /əυ / as in goat, which is currently being fronted in the southeast of England. The parents of the children originate from various parts of the British Isles, and would therefore be expected to show a range of pronunciations for this vowel, both from the southeast and elsewhere. In order to see whether any focusing among the children has occurred, we can compare the fronting score for the parent/caregiver recorded for each child (all but one of whom were women) with that of the child. The variable has the following values: (ou) – 0: [o ], [oυ ] (ou) – 1: [əυ ], [əυ ] (ou) – 2: [ə y] (ou) – 3: [ə i] score: 0 score: 1 score: 2 score: 3 (northern and Scottish realisation) (older Buckinghamshire and London) (fronting) (fronting and unrounding) An index score was calculated for each speaker, on a scale from 0 to 3, in interview style. One of the hypotheses of the Milton Keynes project was that the 4-year-old children would be measurably closer in their speech to their caregiver than are either the 8- or the 12-year-olds. Figure 7.1 shows the correlation of the 4-year-olds’ index scores with those of their caregivers. Taking the caregivers’ scores first, we note that they cover a very wide range. Four of the 16 have scores close to 0, indicating high-back rounded pronunciations characteristic P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. 212 May 27, 2005 15:13 Macrosociolinguistic motivations of convergence and divergence 2.5 Children's (ou) indices 2 1.5 girls age 4 boys age 4 1 0.5 0 0 0.5 1 1.5 Caregivers’ (ou) indices 2 2.5 Fig. 7.1 Correlation of 4-year-old children’s and caregivers’ (ou) indices. of the north of England and Scotland. The remaining 12 are all from the south of England, and show different degrees of fronting. Like the adults, the children fall into two groups: those using high-back northern variants; and those favouring southern diphthongs. However, all the children are Milton Keynes-born, so we have here a case of some young children acquiring their parents’ dialect, while others have either not acquired it or have already accommodated to southeastern speech before the time of the interview. In fact, we have direct evidence of this type of accent mobility in this age group: one of the two boys at bottom left of the figure, the offspring of Scottish parents, was using a mainstream southeastern accent by the time he was recorded for a second time 18 months later. For these 4-year-olds, the choice between a high-back variant and a central or fronted diphthong is a binary one. Some follow their parents; others turn away from them. However, there is a further, more subtle, pattern in the data. Among the 12 children who have southeastern parents (represented by the large cluster in the centre and top right of the figure), and therefore do not have a gross binary choice to make, there is a strong positive correlation with the degree of fronting of their caregivers, with an r2 of .3551 (Pearson). This suggests, of course, that these children match their parents’ quality for this vowel very closely. The 2 children with non-southern caregivers, at top left, have made the binary choice away from their parents’ pronunciation. Interestingly, they have then actually accentuated the difference by going for quite a fronted vowel. P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. May 27, 2005 15:13 The birth of new dialects 213 2.5 Children's (ou) index 2 1.5 girls age 12 1 boys age 12 girls age 8 boys age 8 0.5 0 0 0.5 1 1.5 Caregivers (ou) index 2 2.5 Fig. 7.2 Correlation of (ou) indices for 8- and 12-year-olds and caregivers. There is a great deal of diversity among the 4-year-olds. Yet we cannot say that these Stage II speakers are following the same track as the ONZE Stage II informants, because they are at a stage in their socialisation where the parental model will still be very strong. However, we can examine the older children to see if the pattern has changed. Figure 7.2 shows the same information for the remainder of the children, divided by age and sex. This time, once the 3 children from northern families have been removed, there is an almost complete absence of correlation (r2 = .0532). By the age of 8, the children are no longer affected at all by their parents’ vowel articulations. Not only is there greater homogeneity but there is also focusing on a different norm even from that of the southeastern caregivers. Figure 7.3 shows the data for all female child subjects along with only those caregivers who are from the southeast (all of whom are female). This is to gain maximum comparability across the age groups. As can be seen, all the children have, on average, a higher fronting score than the caregivers. Of the children, it is the 4-year-olds whose score is closest to that of the adults. This suggests that, as the children grow older, they settle on a new, (in this case) partly external norm. In other words, they change their habitual vowel realisations. This interpretation was argued for strongly in Kerswill and Williams (2000: 107). A re-examination of the data reveals that, in fact, there is a more complex pattern. The effect illustrated in figure 7.3 is actually rather slight, and is affected P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. 214 May 27, 2005 15:13 Macrosociolinguistic motivations of convergence and divergence Table 7.1 Distribution of three variants of (ou) across sample (%) (children: elicitation tasks; adults: interviews) (from Kerswill and Williams 2000: 93) 4-year-olds 8-year-olds 12-year-olds Caregivers [ə i],[ɐ i] [ə y],[ɐ y] [əυ ], [əυ], [ɐυ ], [ɐυ] (also [oυ ] and [o]) 13.5 12.9 3.0 3.5 30.2 53.6 68.6 37.3 55.7 33.3 28.2 60.0 Age effect among children: p < .001 (MANOVA) All children vs. adults: p < .001 (t-test) 1.8 1.6 1.4 girls age 12 1.2 1 girls age 8 0.8 girls age 4 0.6 southeastern caregivers 0.4 0.2 0 Fig. 7.3 (ou) index scores for all girls and southeastern caregivers. by the fact that fully front and unrounded vowels get a higher weighting in the index than their rounded counterparts. The argument in the earlier paper was based instead on a binary division between front variants ([ə i], [ə y], [ɐ i], [ɐ y]) and central or back variants ([əυ ], [əυ ], [oυ ], and [o ]). This showed a considerable shift away from the central/back variants of the adults, and a strong tendency for the older children to lead the way. Data collected from 14-yearolds in 1996 suggest that children continue to increase their fronting of this vowel into their teens, albeit slightly (Cheshire et al. 1999). Table 7.1 shows the distribution of three variants on which this argument was based. Two patterns become apparent, in addition to the clear move away from central/back variants by the older children. The first is that it is not the most P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. May 27, 2005 The birth of new dialects 15:13 215 front and unrounded offsets that are in the ascendancy, but variants with fronted and rounded offsets, that is, not the most ‘shifted’ variant (that is, away from the older central/back southeastern diphthong). Second, and this is especially relevant to our argument in this chapter, it is clear that the youngest children exhibit rather greater variation than the older children. Not only do some still use their northern parents’ variants but others go for vowels with unrounded offsets, overshooting the use of this variant by the older children. This inter-individual variation is likely to be greater than in established communities because, as in New Zealand, the range of adult models is greater. However, compared to New Zealand, something quite different is going on here. Focusing seems to have taken place, at least for this vowel, during Stage II. One can suggest a number of reasons for this. First, the linguistic differences involved are less in Milton Keynes than in any of the New Zealand localities studied by ONZE. Second, we are dealing with children who, perhaps, already have a stable (or stabilising) norm to aspire to, not from their parents but from older children and teenagers. We should note that a whole cohort of schoolchildren had already grown up in Milton Keynes in the twenty-four years between designation and the recordings. There may be a contrast between this late twentieth-century new town and many of the earlier cases studied in that there is universal education in quite large schools, with the attendant opportunities for the development of norms. And, finally, it should be remembered that Milton Keynes is not by any means isolated; in this, it differs sharply from all the other settlements we have discussed. (ou)-fronting is, in fact, an external norm, probably (though we can’t be sure) being spread from London by diffusion. We have noted it for Reading, where instrumental analysis of the shift has been carried out (Kerswill and Williams to appear). As a new town with (by definition) a mobile population, Milton Keynes is probably extremely receptive to such influences. Added to this is the unstable dialect situation there (relative to other towns), which will favour the adoption and development of new features. In terms of both regional dialect levelling and the adoption of innovations, Milton Keynes may be ahead of neighbouring towns, prefiguring later regional trends in a similar way to the new dialects elsewhere. Yet there are two ways in which Milton Keynes shows the characteristics of new-dialect formation. The first is that the children indeed show one expected aspect of Stage II speakers, and this is in the amount of inter-individual variability. Children who are socially isolated from other children are likely to rely heavily on their parents as a model for their vernacular as compared to children who are involved in peer groups. In a new town at Stage II, this means that, compared to established towns, there will be a greater presence, among children, of features from dialects elsewhere. In an established town, isolated children will simply sound ‘old fashioned’, not as if they come from somewhere else. We can again use the (ou) data as an illustration. Kerswill and Williams P1: FXS/ABE P2: FXS 0521806879c07.xml 216 CU1837B-Auer et al. May 27, 2005 15:13 Macrosociolinguistic motivations of convergence and divergence (2000: 93–94) argue that individuals who are well integrated into peer groups have high degrees of (ou)-fronting, and that this is part of the mechanism for the establishment of this as a norm. A small number of children in the sample were, however, virtually excluded from participation in peer groups, for a variety of reasons. These children reflected their parents’ speech more clearly, and consequently had accent features from elsewhere. The clearest example of this was a 12-year-old, the son of London parents, whose (ou)-fronting score was low, at 1.4, exactly matching that of his mother. He also used London monophthongal variants of mouth, typically [æ ], which, as we shall see shortly, is not at all characteristic of Milton Keynes children. Like the ‘lames’ in Labov’s (1972b) study of teenage gangs in New York City, this boy was not mentioned as a friend by any of the other children. For the ONZE data, it is probably unrealistic to infer anything about the quality of the speakers’ relationships in early childhood; yet we must assume that the effects noted in Milton Keynes – namely, varying relationships with family and peers mirrored in the degree of use of innovative linguistic features – must have pertained in those communities, too. This means that the ‘shoppingbasket’ model alluded to earlier, whereby the selection of variants by children was seen as a random matter, cannot have been the whole story. Factors such as those just discussed must also have contributed to the inter-individual variation. More generally, they may also help to determine which dialectal ‘strategy’ a young speaker will home in on, in Mæhlum’s (1992) terms. As alluded to before, there is a second way in which the status of Milton Keynes as a town with a new dialect can be confirmed. A mark of new dialects is the absence of any local, stable model, implying a loss of continuity across generations. The language, with its dialect features, is not transmitted in the ‘normal’, intergenerational sense identified by Thomason and Kaufman, as we saw in section 1.3. The clearest example of this is the Milton Keynes pronunciation of the vowel /aυ / as in mouth. Table 7.2 shows the distribution of a number of variants of this vowel, which appears to be converging on a Received Pronunciation-like [aυ ], moving away from local pronunciations such as [ε i] and [ευ ]. The child data given here are from recordings of 14-year-olds conducted in 1996. This table gives an apparent-time snapshot of four generations of the area, with the new town being established between the ‘elderly’ and ‘women’s’ generations. In Milton Keynes, there appear to be three stages in the development of this vowel: first, a period of stability in which [ευ ] and [ε i] predominated (generations 1 and 2), followed at the height of the Milton Keynes settlement in the 1970s by a period of greater heterogeneity in which [æυ ], the form favoured by the majority of the in-migrants (represented here by the women in generation 3), was dominant. A ‘re-focusing’ finally began with the second-generation migrants (today’s children, generation 4), who are rejecting [æυ ] and settling P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. May 27, 2005 15:13 The birth of new dialects 217 Table 7.2 Percentage use of variants of /aυ / (mouth), Milton Keynes working class, interview style (adapted from Kerswill 2002: 697) Generation Stage in new-dialect formation 1 pre-Stage I 2 3 pre-Stage I Stage I 4 Stage II Speakers [ευ ] Survey of English Dialects (SED) informants, 1950–1960sa Elderly (2f, 2m) Women age 25–40 (n = 48) Girls age 14 (n = 8) Boys age 14 (n = 8) √ [ε i] [ε ] [a
] [æυ ] [aυ ] 63.2 0 25.6 0 9.8 11.7 0 17.2 1.2 38.6 0 31.5 0 0 0 0 0 0 5.9 12.3 4.7 3.8 88.8 83.1 a Informants recorded in Stewkley, a village 5 km from present-day Milton Keynes (Orton et al. 1968) on [aυ ], probably influenced (in this case) by the fact that this form is spreading in the region. Starting with generation 2, there is a marked discontinuity in the scores between each succeeding generation, shown particularly by the total absence of the older forms in the speech of the women and children. This reflects, we argue, the lack of social continuity in this town, where most children have parents as well as grandparents originating elsewhere. That we really are dealing with koineisation and not just regional dialect levelling by diffusion is demonstrated by data from Reading, a long-established town some 80 kilometres south of Milton Keynes. Here, we see a similar shift towards [aυ ]. In Reading, however, the old form [ε i] is still used by a number of children, thus demonstrating the effect of the demographic continuity there (Kerswill 2002a: 697). 3.2 The transition to Stage III The relationship between Stage II and Stage III is, on a linguistic level, on the whole very well understood. To summarise, there will be further (probably accelerated) levelling, as the number of variants of a variable reduces from several in Stages I and II to just one, or two or more if reallocation has taken place. Which features survive is largely dependent on demographic factors, as well as on the factor of markedness (Trudgill et al. 2000: 311–312). Trudgill (2004) also discusses the contribution to new-dialect formation of drift, a phenomenon first discussed by Sapir (1921): this is the propensity of a language to change in a certain ways. Thus, related languages, such as English and P1: FXS/ABE P2: FXS 0521806879c07.xml 218 CU1837B-Auer et al. May 27, 2005 15:13 Macrosociolinguistic motivations of convergence and divergence German, or British English and New Zealand English, may show similar, but independent (that is, not contact-induced) changes. However, the social mechanism of dialect transmission that is entailed by the rapid focusing that takes place between Stages II and III is less well understood. We offer some thoughts here. Clearly, some general factors come into play, as we have seen. The main one, already discussed, is the degree to which children are able to interact socially with other children. This will be important in determining whether a child will grow up with a mainly parent-based variety, or will start the gradual mutual accommodation process with other children we have witnessed in Milton Keynes. As we have seen, the ONZE informants show both inter-individual and intra-individual variability, and we argue that this in all essential respects reflects their childhood speech as well – at least from the stage in their socialisation when they began to interact outside the home. The Milton Keynes data notwithstanding, we have no evidence of the childhood speech of these or other such individuals. The Milton Keynes children do, however, provide a pointer in that they can be shown to be moving away from parental norms towards new norms as they grow older. But they show none of the variability of the ONZE speakers – with two exceptions: first, nonintegrated children’s speech will be oriented to varieties in another community (such as London); and second, children with distinctively non-southeastern home accents, such those originating in Scotland, will shift dialect as they begin to interact with other children, and may retain a distinct ‘home’ variety for use with family members. It may well be the case that many of the New Zealand speakers would, likewise, have had a ‘home’ and a ‘playground’ code, and that these would have differed more than they do in a community with cross-generational continuity. A remarkable fact is the degree of homogeneity in New Zealand English, achieved at least in part, it seems, by 1900 (Woods 2000: 138). Given the degree of variability among the ONZE informants, and the absence of easy communication routes until well after this time, this is very surprising. We argue that a reason for this is that approximately the same mix of forms, in similar demographic proportions, underlay most New Zealand settlements (Trudgill 2004). Since similar mixes formed the input to both Australian and South African English, it is not surprising that rather similar varieties ensued. Yet this alone cannot account for the extreme uniformity of New Zealand English. Trudgill (2004) agrees with Bernard (1969), who argues, in the case of Australia, for geographical mobility and the fact that Australians in different parts of the country kept in touch with one another, mainly though sea travel from one port to another, as accounting for the uniformity. For nineteenth-century New Zealand, Britain (forthcoming) similarly cites high levels of mobility and transience, and suggests that these factors ‘led to the emergence of an atomistic society freeing people both from subservience and from the need to conform that tight-knit P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. May 27, 2005 The birth of new dialects 15:13 219 local communities often engender’. As Britain points out, this was a society with relatively weak social network ties, precisely the sorts of ties that are the breeding ground for rapid supralocal linguistic change. Trudgill also adds diffusion as an explanation: geographical diffusion of linguistic innovations from one place to another would have continued in the normal way during the new-dialect formation period and would have helped, aided by high mobility, to strengthen the reduction of geographical and social differences. In any case, we can see, even in the ONZE data, very clear changes in speech correlating with the speaker’s date of birth. In every case, we see an increase in the use of modern New Zealand English features, and a decrease in features that have died out. An example is so-called happy-tensing,5 the use of a peripheral vowel [i] in some final unstressed syllables (such as that in happy), replacing [i]. The feature, which is now universal in New Zealand, is actually absent from the speech of the oldest informants. It then makes an appearance, and subsequently increases. The percentage of speakers who have happy-tensing by decade of birth in the ONZE corpus is as follows: 1850–1859 1860–1869 1870–1879 1880–1889 0% 25% 48% 43% This strongly suggests that happy-tensing did not arrive in New Zealand from Britain at all (British varieties, too, have shown an increase in the use of the feature over the same period), but started life independently in New Zealand as an example of drift. It is speakers at the transition between Stage II and III who are gradually adopting it, showing both inter- and intra-individual variation as they do so. Rather than being rapid, we see the process happening over at least forty years. Work on the loss of rhoticity in New Zealand English shows a similar gradualness, and that the process started early in the settlement history (Hay and Sudbury 2002). Hay and Sudbury’s results show considerable variability already among the oldest ONZE speakers (born in the 1860s and 1870s), and that the loss of rhoticity was categorical in informants born after about 1910. From the point of view of discovering the geographical spread of a fully focused variety in New Zealand, one result in particular is revealing. This is that, of the two factors ‘North Island vs. South Island’ and ‘year of birth’, the former is statistically the stronger predictor. This strongly suggests that focusing was geographically gradual, and (we suggest) that the spread of the loss of rhoticity may, in fact, have been by geographical diffusion; as is well known, rhoticity is a recessive feature that is now restricted to the south of the South Island. 5 This term is taken from Wells (1982). P1: FXS/ABE P2: FXS 0521806879c07.xml CU1837B-Auer et al. May 27, 2005 15:13 220 Macrosociolinguistic motivations of convergence and divergence 4 Conclusion New-dialect formation is in many ways the quintessence of dialect convergence: all the processes described in this chapter can be found in one or more of the other scenarios for dialect contact described in this book. It also has the potential (not always realised) to be a very rapid process, observable using standard techniques from social dialectology. This enables researchers to apply the full range of sociolinguistic techniques, from the quantitative (Labovian) method, through various (ethnographic and quantitative) means of gaining social-psychological information, to examining the contribution of different age groups to dialect convergence in a new community. Over a longer time perspective and using archive material, it is possible to compare the outcomes of different cases of new-dialect formation, drawing linguistic conclusions that shed light on language change generally. As such, the study of new dialects is a contribution to historical linguistics. New-dialect formation is related, in particular, to creolisation (cf. Hinskens, Auer, and Kerswill, this volume), in that it implies a lack of generational continuity at the community level. This is an example of dialect divergence, which can also be seen at the family level as the children, through interaction with peers outside the home, begin to form a new dialect. New-dialect formation is the product of linguistic accommodation between speakers, both in the immediate context of conversations and in changes in people’s long-term speech habits. We show (Trudgill 2004) that the form of English in New Zealand is highly predictable from knowledge of where in the English-speaking world the immigrants came from, and in what proportions. This does not tell us whether the accommodation was actual (pattern-matching) or more in line with the identityprojection model; in the end, this does not seem to matter for the outcome.