1. Introduction

Human beings take territoriality to an extreme level. As a species, we seem very strongly predisposed towards partitioning and demarcating our spatial surroundings for the purposes of controlling access to natural resources and maintaining oversight of the activities of other individuals and groups. This apparent imperative is not enacted in an arbitrary or unconstrained way, of course: the earth’s surface is already divided into zones according to geographical terrain, habitat type, and climatic factors, and the boundaries between these zones are sometimes abrupt rather than gradual. Deserts, rivers, lakes, and mountain ranges can be significant obstacles to movement, and the oceans present barriers that have been insuperable for the great majority of human history. Within these limits, though, we have sought to carve up the earth’s land surface into larger and smaller units in ways that, in time, come to seem natural or inevitable. The processes by which borders between territories are rationalised and justified are complex and fascinating ones, combining as they do the collective histories and mythologies of tribal groups, religious and ethnic ideologies and, latterly, the political states that co-exist at the global level. Rather soon after a new state is founded, its inhabitants may come to believe that the status quo is effectively how things have always been, and that it could not tolerably be any other way (see Breuilly Reference 211Breuilly and Breuilly2013).

It seems almost self-evident that language plays a crucial part in how territories bounded by borders with their neighbours are defined. In the early nineteenth century – this side of the watershed in European political history that has been dubbed ‘The Great Divide’ (Burke Reference Burke and Breuilly2013: 21) – the emergent nation states of Europe sought to codify national languages that would be spoken everywhere that fell under the state’s control, right up to the territory’s outermost edges. A common language would, it was thought, help to unite citizens of the same state who might previously have felt more affinity with their neighbours just across the border than they did with the inhabitants of a distant capital city or remote provinces hundreds or even thousands of miles away. However, two centuries of language standardisation, linguistic prescriptivism, and the suppression of minority languages and dialects did not wipe the European map clean of state-internal linguistic variation as many of the founders of the nation states might have wished. Although a lamentably large number of European languages and dialects have been lost in recent centuries, some minority varieties have been remarkably tenacious. Others have flourished as, following the quite sudden volte face in late twentieth century national language planning policies across Europe, relegitimated ethnic identities and linguistic traditions have been given governmental endorsement and financial support, and celebration of cultural diversity is now strongly encouraged (Baldauf and Kaplan Reference Baldauf and Kaplan2006). In a significant number of cases, the areas in which minority languages and dialects are spoken straddle international borders, producing especially fruitful contexts in which to explore the dynamics of the interaction between language and identity.

It must not be supposed, having said this, that only minority languages and dialects are worth considering in such borderland contexts. It is equally interesting to consider how speakers of powerful majority languages like English make use of linguistic resources to mark their national and regional affiliations at state boundaries. The border between the Republic of Ireland and Northern Ireland, for instance, has been the focus of a number of studies (Zwickl Reference Zwickl2002; Kallen Reference Kallen, Watt and Llamas2014), as has that between the United States and Canada (Chambers Reference Chambers1995; Dollinger Reference Dollinger2012; Boberg Reference Boberg, Llamas and Watt2014). In each of these contexts, and others like them, the complex connections between linguistic behaviour, identity, and place are particularly intriguing to social scientists, not least sociolinguists. In this chapter we consider further the last of these concepts as it relates to the language and identity issues in the Scottish-English border area.

2. Place

Later in this chapter we will look more closely at aspects of how the distributions of linguistic forms in the Scottish-English borderland correlate with identity factors, and how these forms are used to encode information about place. First, though, it is worth posing the following question. What, in fact, do we mean by ‘place’? The Oxford English Dictionary defines the term in two different ways. The first set of definitions, which relate to space or location, are what we might call the ‘literal’ ones:

1. 1. a. A particular part or region of space; a physical locality, a locale; a spot, a location. Also: a region or part of the earth's surface; 10. A particular spot or area inhabited or frequented by people; a city, a town, a village.

These are, in other words, positions which do not necessarily bear any direct relation to entities found within them, but which might merit identification as a discrete position in space by virtue of their significance to human habitation or activities.

Secondly, there are somewhat more figurative definitions, which pertain to position or situation with reference to its occupation or its occupant, for example:

1. 1. a. A proper, appropriate, or natural position or spot (for a person or thing).

When we talk of people ‘being in place’, ‘finding their place’, ‘having their proper place’, ‘being well placed to do something’, or ‘having a firm place to stand’, we are not necessarily talking about their physical location. Place is about geographical positions that can be pointed to on maps or identified using GPS coordinates, to be sure, but it is also about states of mind, stances and attitudes, and the status that individuals hold within their social networks and society at large. We will focus in this chapter chiefly on place in its more concrete senses, but will not neglect the other ways in which the concept of place can inform how we interpret our linguistic data.

The notion of place has been subjected to a number of stringent theoretical treatments since classical times, and has come to preoccupy scholars across a spectrum of research disciplines including linguistics (see Auer and Schmidt Reference Auer and Schmidt2010; Auer et al. Reference Auer, Hilpert, Stukenbrock and Szmrecsanyi2013) and also encompassing cognitive and social psychology, geography, history, anthropology, town and transport planning, communications and informatics, archaeology, and numerous others (see Scheider and Janowicz Reference Scheider and Janowicz2010, Reference Scheider and Janowicz2014; Winter and Freksa Reference Winter and Freksa2012; Raubal et al. Reference Raubal, Mark and Frank2013; Richter and Winter Reference Richter and Winter2014). Scheider and Janowicz’s (Reference Scheider and Janowicz2010: 1) ‘requirements for non-reductionist accounts of place’, which build on the work of human geographers concerned with the ‘phenomenological aspect’ of place, are comprised of the following axioms:

1. 1) Places are located, but are not locations.

2. 2) Places are primary categories of human experience and social constructs.

3. 3) Places have stabilizing functions that afford insideness.

4. 4) Places have material settings (surface layouts).

In (1), we can talk of places which are not locations as such, in that they are potentially mobile (e.g. the deck of a ship, or a market which might appear only once a week and perhaps in different locations). Places must also be distinguished from locations, according to (2), in that locations can feasibly be anywhere and hence are isomorphic and arbitrary, whereas places are ‘meaningful aspects of human experience [that] involve emotional attachment and social identification’ (Scheider and Janowicz Reference Scheider and Janowicz2010: 1–2). Naming, or the association of linguistic forms with places, is a crucial component of their construction as places. Thirdly, the functions of places to ‘stabilise’ or arrest motion, as in (3), give them sets of properties that ‘fix mutual expectations among people, allowing them to meet and communicate’ (Scheider and Janowicz Reference Scheider and Janowicz2010: 2). Finally, (4) proposes that places ‘always have a concrete identifiable material form’ (Scheider and Janowicz Reference Scheider and Janowicz2010: 2). This might seem too narrow a requirement when one considers that places need not exist in any objective sense (where, e.g., are ‘the developed world’, ‘suburbia’, or ‘the street’?). However, Scheider and Janowicz extend the scope of Gibson’s (Reference Gibson1979) term surface layout to imagined places, which can be argued to share properties, if only on the conceptual level, with physically observable places.

It is worth stressing the point that even in work on place that derives from research in fields quite far removed from linguistics, matters of language are often ascribed an important, even crucial, role. Places become part of the social environment through being thought significant enough to be given names, and language is of course the principal medium by which these names and other information about places are transmitted across space and time. Language is also used as a proxy for place, in the sense that pronunciations, grammatical structures, words, and writing systems come to be associated with particular localities or regions, meaning that we can deploy linguistic resources to index non-linguistic information about our geographical provenance. Moreover, the choices that speakers make among alternative forms in their linguistic repertoires act as signals that help listeners to align speakers with relevant ingroups and outgroups. Discovering how this kind of knowledge is acquired and activated in interactions is of course one of the key objectives of sociolinguistic inquiry.

The theorisation of place and space in sociolinguistic research has progressed significantly in recent decades, thanks to the work of scholars such as Peter Auer, Joan Beal, David Britain, and Barbara Johnstone (e.g. Beal Reference Beal2006; Auer and Schmidt Reference Auer and Schmidt2010; Britain Reference Britain, Llamas and Watt2010, 2014, this volume; Johnstone Reference Johnstone, Auer and Schmidt2010, this volume; Auer et al. Reference Auer, Hilpert, Stukenbrock and Szmrecsanyi2013). Britain proposes a taxonomy of space which depends on its conceptualisation as simultaneously physical, social, and perceptual. Physical space, he argues, is objective and geometric. In this view, space is scaled linearly, such that if measured in miles or kilometres, places X and Y could be said to be the same distance apart as Y and Z. However, with respect to social distance, the gap between X and Y may be much larger than that between Y and Z. An international border may divide X and Y, for example. But even within contiguous territories, social space may present bigger divides than geography alone would predict. There are many cities around the world in which adjacent neighbourhoods may be so socioeconomically segregated (e.g. an exclusive gated community abutting a shanty town) that the inhabitants of the respective communities may never come into contact with one another. The residents of the affluent neighbourhood may have much higher levels of interaction with people from a similar suburb miles across the city. The mapping of social space onto physical space may thus be complex and distorted.

Perceptual space, finally, is a useful notion when attempting to understand how people construct their perceptions of the world around them, and how they orient to their ‘home patch’ and to places that are physically or socially removed from it. Inhabitants of a community which is well connected by transport links to a large faraway city may think of it as closer in some sense than a nearby town which is poorly connected to their own. Issues such as journey time may be a major factor in how the space is perceived, but other relevant psychological factors might include whether a state or provincial boundary intervenes, or whether there is a significant geographical barrier (for instance, a large river or mountain range) between two places. Britain (Reference Britain, Chambers, Trudgill and Schilling-Estes2002, Reference Britain, Llamas and Watt2010, Reference Britain, Chambers and Schilling2013) argues that perceptual space is constructed by people’s everyday practices, and that ‘place’ and ‘region’ emerge through the routinisation of these practices. That is, through the mundane day-to-day routines that people perform – such as travelling to work, shops, or leisure facilities – locations and areas acquire coherence in psychological space that need not map in any direct way to physical geography. Viewed in this way, place can be seen as a ‘process’ that is shaped by practice, operating within the constraints of institutional and infrastructural factors. Factors of the latter sort may force the inhabitants of a place to revise their notions of social and perceptual space: a stark example from recent British history would be the radical reconfigurations of social networks caused by the relocation of city-centre slum dwellers to peripheral housing estates during the post-war period. Perceptions of distance and proximity were also influenced by the closure of large sections of the UK’s railway network during the 1960s, and then changed again by a growth in car ownership (Vannini Reference Vannini2010).

On a subtler level, the reassignation of some towns and cities to new higher-order political units (such as counties, states) may affect how people see themselves. As Beal (Reference Beal, Llamas and Watt2010: 225) puts it, ‘the stroke of a bureaucrat’s pen can have as much effect on a speaker’s sense of place and identity as … innovations in transport infrastructure and communication’. The potential knock-on effects of this kind of restructuring of social and perceptual space on speakers’ linguistic behaviour has been investigated by Llamas (Reference Llamas2007) in the north-eastern English town of Middlesbrough. Middlesbrough was originally in the North Riding of Yorkshire but was subsequently reallocated to the County Borough of Teesside (1968), then to a new county (Cleveland, in 1974), and finally to its own unitary authority (1996). Llamas found correlations between the distribution in informants’ speech of a range of consonants and vowel variants, the political unit to which Middlesbrough belonged during individual speakers’ formative years, and the identity labels that speakers assigned to themselves, to indicate that they thought of themselves as natives of Yorkshire, Teesside or Middlesbrough.

Beal (Reference Beal, Llamas and Watt2010) reports something similar for Warrington, in north-west England, which became part of Cheshire after a redrawing of local county boundaries divided it from its original county of Lancashire. That decision met resistance among some older Warringtonians, many of whom cleave to Warrington’s historical links with the working-class ‘northern’ culture of Lancashire and resent the association of the town with the north-midland county of Cheshire, a relatively affluent part of England widely perceived to be ‘middle-class’. These perceptions may mean little to young Warringtonians, however, whose view of their hometown appears to be influenced less – perhaps not at all – by its erstwhile Lancastrian affiliation. As Beal points out, place ‘is not a given, to be taken for granted in our research designs: what appears to be a town or city delimited by boundaries on the map may actually be several different places to different groups of speakers, whose allegiance to these “places” may be indexed by linguistic variables’ (Beal Reference Beal, Llamas and Watt2010: 226).

As test sites for the investigation of how linguistic behaviour is connected to identities and perceptions of place, boundary zones of this kind could scarcely be bettered. Long-standing political divides, such as those between Britain’s historic counties, or indeed between the constituent countries of the United Kingdom, are ideal. It was for this reason that the Scottish-English border area was identified as a zone of particular sociolinguistic interest. It must also be acknowledged that the presence of the Scots language in the region weaves additional threads into the local linguistic fabric, but because English was the language spoken by all our informants as well as by our interviewer, we will not concern ourselves here with questions of the degree to which local people have competence in Scots per se.

3. The AISEB Project

The Accent and Identity on the Scottish/English Border (AISEB) project Footnote ¹ sought to address a number of problematic questions concerning the degree to which impedance of the progress of sound changes can be attributed to identity factors and attitudes among speakers, and the part that speaker agency may play in the synchronic and diachronic distribution of phonological forms (Llamas Reference Llamas, Llamas and Watt2010; Watt et al. Reference Watt, Llamas, Johnson and Lawson2014a, Reference Watt, Llamas, Docherty, Hall, Nycz, Watt and Llamas2014b; Reference Watt, Llamas, Kendall and Fabricius2014c). Rather than attempt to survey every community along the border’s length, we opted to collect sociolinguistic data in four towns: Gretna (Scotland) and Carlisle (England) at the border’s western end, and Eyemouth (Scotland) and Berwick upon Tweed (England) at its eastern end (Figure 9.1).

Figure 9.1: Map of Scottish-English border region, showing the four fieldwork sites (labels in bold type).

Our initial hypotheses led us to expect that the speech patterns of people on the Scottish side of the border would differ markedly from those of people living just a few miles away on the English side, not least because the existing dialectological literature seemed unequivocally to suggest that the political border coincides with an entrenched linguistic divide (e.g. Ellis Reference 212Ellis1889; Zai Reference Zai1942; Kolb Reference Kolb1966; Kolb et al. Reference Kolb, Glauser, Eimer and Stamm1979; Aitken Reference Aitken and McArthur1992; Glauser Reference Glauser2000). There were also good grounds for thinking that towns at either end of the border would be sociolinguistically divergent from one another, even if they were in the same country (e.g. Gretna versus Eyemouth), owing to their geographical separation and a lack of significant contact between their inhabitants.

We also thought it important to take account of the presence of the border from the socio-psychological point of view, through probing people’s attitudes towards it. Political borders are objective boundaries insofar as they may be physically visible (fences, walls, signage, customs posts, clear-cut felling lines through forests) and/or can be pointed to on maps, and to that extent we may suppose there to be an absence of variation in how they are perceived. However, for those living in close proximity to a border, and assuming that their movement across it is not restricted, borders may acquire more subjective, symbolic qualities (Diener and Hagen Reference Diener and Hagen2010; Donnan and Wilson Reference Donnan and Wilson2010). The border may be perceived and evaluated differently by inhabitants living on either side of it, and even by people living in the same locality. In the case of the four towns investigated in the AISEB study, it seemed plausible that being close to the border might be a common factor that lent a sense of common identity to people in the four towns. Though physically at the margins of their respective nations, the towns may alternatively be seen as being at the centre of a borderland region that straddles the political divide. The border, in other words, might not be defined principally by its demarcative properties that emphasise difference, but could be seen as something that gives people in towns on either side of it a sense of affinity as ‘borderlanders’. Indeed, the region is sometimes termed ‘The Borders’, though the tendency is now to use this name just for the Scottish part of the area (‘Scottish Borders’ is the official name for the unitary authority that now subsumes the old counties of Berwickshire, Peeblesshire, Roxburghshire, and Selkirkshire). But even in a region where the border may be crossed freely, it does not necessarily follow that sharp inter-group categorisations will not emerge and persist, as theorists of borderlands such as David Newman (Reference 213Newman2006) have demonstrated. Linguistic behaviour is arguably the foremost member of the suite of cues that borderlanders can draw upon to flag their allegiances and stances. In the following section we examine one facet of the AISEB informants’ speech production, by way of illustrating this principle.

3.4 Results

Figure 9.2 shows, in the form of probability density function curves, the VOT results for the /p t k/ and /b d ɡ/ sets pooled across place of articulation within each set, for each of the four fieldwork sites. Speaker age is also represented, where the solid lines represent the data for older speakers and the dashed lines those for young speakers.

Figure 9.2: VOT values (represented by probability density functions) across the four AISEB fieldwork sites, split by speaker age group (solid lines = older speakers; dashed lines = young speakers). The pooled results for /b d ɡ/ are shown in the upper panel in each pair (‘voiced’), those for /p t k/ in the lower panel (‘voiceless’).

It is immediately clear that there are differences in each town with respect to age group, particularly for /b d ɡ/ (main effect for speaker age: t_1,1337 = 20.02, r² = 0.2589, p < 0.001).Footnote ² In the upper panel of each pair, the curve for the young group is a sharp peak centred on approximately 20 milliseconds (ms), while the corresponding curves for the older speakers in each town are bimodal, and consequently considerably flatter. The bimodality – which is not unexpected for English (see Lisker and Abramson’s [Reference Lisker and Abramson1964] results for American English) – indicates significant levels of pre-voicing, with VOT values tending to cluster between −50 and −100ms. In all four cases the other peak falls within the same positive VOT range as that found for the young speakers. The older speakers, in other words, use a combination of long-lead and short-lag VOT for /b d ɡ/, while the young speakers favour short-lag VOT almost exclusively (see the very similar results for Aberdeen English reported by Watt and Yurkova Reference Watt and Yurkova2007). It is also noticeable that for Eyemouth, unlike the other three localities, the peak corresponding to pre-voiced tokens is more prominent than that for the short-lag VOT tokens, skewing the density curve markedly leftwards.

Looking next at the pooled data for the voiceless set /p t k/, it appears that there are small but consistent differences between the age groups in each town. Again, the VOT values for the young speakers are higher overall, as we might predict if contrast between the voiced and voiceless sets is to be maintained (/p t k/ main effect for age: t_1,3313 = 16.15, r² = 0.1693, p < 0.001). As we saw for /b d ɡ/, the VOT values for older Eyemouth speakers are on average the lowest of the four communities. This finding is in line with Johnson’s (Reference Johnston and Jones1997) account of VOT in Scots-influenced varieties of English spoken in the Scottish Borders, whereby a lack of aspiration on /p t k/ correlates with a greater incidence of zero or negative VOT in the /b d ɡ/ set.

Overall, for /p t k/, the average VOT for the two Scottish varieties is lower by approximately 10ms than it is for the two English ones. Although the difference is a slight one, and there is a substantial overlap in the distributions, it nonetheless yields a significant effect for the Nation variable (t_1,3313 = 14.19, r² = 0.1693, p < 0.001). There is also an east/west split, with Eyemouth and Berwick having shorter /p t k/ VOT values than Gretna and Carlisle (effect for Coast: t_1,3313 = 7.40, r² = 0.1693, p < 0.001). Durational differences on this scale are not imperceptibly small: experimental studies have shown listeners to be sensitive to discrepancies in VOT values of as little as 10ms (Cole et al. Reference Cole, Jakimik and Cooper1978; Blumstein et al. Reference Blumstein, Myers and Rissmann2005).

Before proceeding any further, it is important to consider the results of studies that have shown a link between speaker age and VOT (Benjamin Reference Benjamin1982; Liss et al. Reference Liss, Weismer and Rosenbek1990; Ryalls et al. Reference Ryalls, Zipprer and Baldauff1997; Torre and Barlow Reference Torre and Barlow2009). Their findings reveal that, as a result of physiological changes, older speakers tend to use shorter VOT durations than younger speakers. The age effect found in the present study supports that generalisation. It does not explain, however, why the disparities in VOT values between the age groups are different across the four towns, nor why the Nation and Coast variables also appear to have an influence on VOT durations. We would suggest instead that the use of longer and shorter VOT among the AISEB speakers is related to attitudes and identity in the area, about which we say more in the following section.

4. Identity Factors

In view of its long history as a zone of conflict and contestation, and lately as a focal point for questions pertaining to issues of political autonomy – questions that culminated in the 2014 independence referendum (see, for instance, Daniel Reference Daniel2014)Footnote ³ – the Scottish-English border region is fertile ground for the investigation of social and political attitudes. The AISEB interviews were completed in 2011, prior to the announcement of the referendum but well after the establishment of the Scottish Parliament in 1999, so questions relating to the benefits or otherwise of recent and potential future constitutional changes were a reliable way of garnering opinions from interviewees.

Accompanying these relatively weighty questions were ones which sought information about participants’ attitudes towards the local area (including questions concerning the border in particular), their day-to-day routines, and how language is used in the region. Examples of the set of twenty-one identity questions are shown below.

• If you were watching a regional news programme, what places would you expect to hear news from?

• If you wanted a day out shopping, where would you go?

• How often do you cross the border?

• Do you see the border as a divide of some sort?

• Where, geographically, would you say people stop talking the same as you and start sounding different?

• Are there any pronunciations or ways of saying things that you would hear and think, that sounds really Scottish or really English?

Participants were also asked to complete tasks based upon Visual Analogue Scales (VAS; Redinger and Llamas Reference Redinger, Llamas, Watt and Llamas2014; Llamas and Watt Reference Llamas and Watt2015). In one task a VAS was used as a means of quantifying a participants’ strength of agreement with authentic statements made by people from one of the four AISEB localities – for example, ‘I’ve been abroad and everybody thinks about Britain as England so when they’re talking about Britain, they’re really talking about England’– by drawing a vertical stroke somewhere along a horizontal line representing a continuum between ‘agree’ and ‘disagree’. In another task, the one we focus on in the present chapter, participants drew multiple vertical lines across a ‘most important’ to ‘least important’ cline (the ‘Relational Analogue Scale’, RAS) to represent identity labels they felt applied to them. A list of suggested labels was shown above the RAS, tailored to the locality in question. In Gretna, for instance, the suggested labels were Borderer, British, English, European, from Gretna and Scottish (Figure 9.3).

Figure 9.3: Example of a Relational Analogue Scale (RAS) completed by an 18-year-old woman from Gretna. The positions and spacing of the labels she has chosen are quantified relative to British, the identity label available to all AISEB informants.

Informants could opt to place all, any, or none of these along the RAS cline, and were free to add labels of their own choosing. The completed RAS thus shows not just the ranking of labels along a gradient of importance, it also allows the fine-grained measurement of distances between each vertical stroke and (a) the ends of the RAS cline and (b) the other vertical strokes. The latter was achieved by scanning the hard copies of the RAS and using an on-screen measuring tool in the software package ImageJ.Footnote ⁴ Because all of the AISEB informants can be described as British (even if they choose not to attribute this label to themselves), we used the British label as the anchor point, expressing its position as a percentage based on its distance from the ‘least important’ end of the cline, and the positions of other labels relative to it (also as percentages; for the speaker shown in Figure 9.3, British was placed at 97.4%, exceeding Scottish on the importance scale by 5.7%, from Gretna by 20.6%, and Borderer by 71.2%). Where participants placed British at the ‘least important’ pole of the RAS we assigned a nominal score of 0.1% so as to distinguish these cases from those in which British was not used at all (0%).

Figure 9.4 is a summary of the RAS data relating to the relative rankings of British and Scottish (for informants in Gretna and Eyemouth) and British and English (for those in Carlisle and Berwick) among the 132 participants who used both British and one of the national labels on their RAS. The scores for individual participants are split according to age group, and the age group means are superimposed on each plot. These have been linked by a connecting line to highlight the magnitude and direction of the difference between the age groups in each of the four towns.

Figure 9.4: Relational Analogue Scale (RAS) data for the four fieldwork sites, showing individual and mean group distances (%) between the British and national (Scottish or English) labels, split by speaker age group. Individual speakers are represented by crosses, while the age group means are shown by filled circles linked by a solid line, the slope of which is an indication of size of the difference between group means. Points falling above the zero line denote a preference for ranking Scottish or English higher in importance than British.

Points falling above the zero line indicate that the individual participant preferred either Scottish or English over British. In each locality there is a spread of positive and negative values, but the greatest concentration of high positive values among the Eyemouth speakers indicates that they tended to rank Scottish considerably higher than British, and the very few points with negative values in the Eyemouth plot show that ranking British over Scottish was quite strongly disfavoured there. This is true for both the older and young groups: of the older group members, only 16.1% placed British higher than Scottish, and among the young group the figure was still lower (15.4%). The means for the older and young groups are consequently relatively high, even if the trend appears to be a downward one overall, such that Scottish is not ranked so highly above British among the young Eyemouthers as it is among their older counterparts.

Similar patterns can also be seen for the Gretna and Carlisle participants, though the average distances between British and Scottish/English are lower overall. In Carlisle, a clear age distinction is apparent, whereby the general preference for English over British among the older informants is not shared by the young group, who are evidently more likely to feel more strongly British than they do English (26.7% of the older group members ranked British higher than English, as opposed to 68% of the young group). An age-related trend is not apparent in Berwick, where speakers in both the older and young groups seem overall to be fairly equivocal about the British and English labels. Only around a third (34.8%) of the older speakers in Berwick said that English was a more important label than British to them, while the young Berwickers were split evenly (50%).

Looked at another way, it seems that on both sides of the border young people are more ready to describe themselves as British than are older ones (if older ones expressed a preference for their relevant national label), but that resistance to this trend is stronger in the two Scottish towns – and particularly Eyemouth – than it is south of the border. The proportion of informants who did not choose British at all during the RAS task was overall considerably higher in Eyemouth (17.5%) than in the other localities (Gretna = 7.5%; Carlisle = 3.6%). In Berwick it was 10.2%, but an equal proportion there chose not to use English at all either; see the sociological research of Kiely et al. (Reference Kiely, McCrone, Bechhofer and Stewart2000), who found that Berwickers tended to reject the identity labels that Kiely and his colleagues had expected them to embrace; see Section 5).

On the basis of earlier findings that have emerged from the AISEB project, we have ascertained that national identity preferences and production patterns in certain key phonological variables – notably (r) (Llamas et al. Reference Llamas, Watt and Johnson2009; Llamas Reference Llamas, Llamas and Watt2010; Watt et al. Reference Watt, Llamas, Johnson and Lawson2014a) – correlate closely in the Scottish-English border region. In the final section of this chapter, we consider how the RAS responses might relate to the VOT data from these informants’ wordlist readings.

5. Linking VOT Production to the Attitudinal Data

As mentioned in Section 4, the regression analysis carried out on the data revealed that the Nation, Coast, and Speaker Age variables all had significant effects on the distribution of VOT values. That is, VOT was shorter among older speakers than among their young counterparts in all four towns but, in line with the claims made by Johnston (Reference Johnston and Jones1997) and others, it was found to be significantly shorter overall in the Scottish localities (Gretna and Eyemouth) than in the English ones (Carlisle and Berwick). It was also shorter on the east coast (Eyemouth and Berwick) than on the west (Gretna and Carlisle). As anticipated, then, VOT in /p t k b d ɡ/ was shortest in Eyemouth and longest in Carlisle. This observation tallies with our expectations, given that in respect of several other phonological variables we have investigated – (r) and the nurse vowel, amongst others – Eyemouth speech is the most conservatively Scottish of the four AISEB varieties, while the Carlisle dialect is the most congruent with those of northern England, and indeed England generally. The Gretna and Berwick varieties each occupy an intermediate ‘hybrid’ space between these two extremes. Consequently, it is unsurprising that our Gretna speakers often say that they are thought to be English when they talk to other Scots. Similarly, it is unsurprising that Berwickers are reported by fellow Northumbrians to sound Scottish, but are described as Northumbrian or ‘Geordie’ (Newcastle) (and indisputably English), by their near-neighbours living just across the border in Eyemouth (Kiely et al. Reference Kiely, McCrone, Bechhofer and Stewart2000).

The fact that the VOT figures seem to correspond rather well with the attitudinal data reported in Section 4 confirms that there appear to be parallels between the speech patterns of the informants in our sample and the ways in which they choose to attribute identity labels to themselves. The changes we see in VOT values, which are undergoing a wholesale upward shift across the region, are reflected in the changing preferences for the British and Scottish/English identity labels, whereby young Scottish informants tend to place Scottish closer to British than older ones do, on both the west (Gretna) and east (Eyemouth) coasts. In the latter town, the tendency to rank Scottish higher than British is considerably stronger. On the English side of the border, we see in Carlisle an increase in VOT values and a sharp decrease in the preference for English over British as an identity label. In Berwick no such correlation seems to hold, however. In both age groups there seems to be a lack of any consistent tendency to rank English over British, recalling Kiely et al.’s (Reference Kiely, McCrone, Bechhofer and Stewart2000: 1.6–1.7) conclusion that, by not readily claiming English identity and by rejecting the label British altogether, their Berwick interviewees were declining to ‘play by the prevailing identity rules’ and ‘claiming, attributing, rejecting, accepting and side-stepping national identity, in ways that we had seldom or never previously encountered’. VOT values in Berwick speech are nonetheless evidently on the rise overall, to that extent Berwick speakers appear to be participating in a sound change that is in train both north and south of the border.

It would be hard to argue that there is any direct link between the trend to increase VOT in /p t k b d ɡ/ and a closing of the gap between the use of the national labels Scottish/English and British in the results of an identity questionnaire. It does seem likely, though, that this and other phonetic changes that appear to be taking place in the speech of the four AISEB localities do coincide with a reappraisal among younger people of what it means to be Scottish, English, and British. It would not be especially controversial to suggest that changes in the relative importance of identity choices might, over the course of a generation or two, have knock-on effects on pronunciation preferences, even at a level as comparatively subtle as that seen in the VOT data.

However, in spite of the similarity between the patterns found in the two Scottish towns and Carlisle, we must not lose sight of the significance of the border as a robust and persistent linguistic divide. Eyemouth and Berwick lie a mere 9 miles (15km) apart by road, and yet there remain a host of significant phonological differences between the varieties spoken there, in spite of regular and plentiful contact between the two localities. Gretna and Carlisle are about the same distance apart as Eyemouth and Berwick, and though the distinctiveness of Gretna English and Carlisle English is less marked than that between the eastern varieties, we have still been able to catalogue a series of systematic differences between them (e.g. with respect to rhoticity and the nurse vowel; see Llamas Reference Llamas, Llamas and Watt2010; Watt et al. Reference Watt, Llamas, Kendall and Fabricius2014c). Corroboration of our observations about accent similarity across the four sites has recently been provided via automated analysis of AISEB recordings, using metrics imported from the domain of speech and speaker recognition technology (Brown Reference Brown2014; Brown and Watt Reference Brown and Watt2014), on which we will report more fully in future publications.

6. Conclusions: The Border as a Place

As we have explored the AISEB corpus, we have found increasingly compelling evidence that the Scottish-English border has a multiplicity of meanings to people living in the region, and that its effects on their linguistic behaviour take many forms. The VOT data reported in this chapter are just a fragment of the complex mosaic of interconnections between language, attitudes, and identities that characterises the sociolinguistic landscape of this part of the English-speaking world. Nonetheless, they illustrate how, even at the level of slight durational differences in the pronunciation of stop consonants, we can find correspondences between speakers’ locations and how speakers orient themselves socio-psychologically towards these places, the larger geopolitical units within which they fall, and towards neighbouring communities. Though the border satisfies Scheider and Janowicz’s fourth criterion in that it has a ‘concrete identifiable material form’ (Reference Scheider and Janowicz2010: 2) manifested by the signage, flagpoles, boundary stones, and monuments that mark its presence, it is also highly porous, offering no obstacle to free movement from one side to the other. To this extent, it is more relevant in the present context to think of its significance in more abstract terms, as per those laid out in Scheider and Janowicz’s (Reference Scheider and Janowicz2010) axiom (2): the border can be thought of as a place by virtue of its effects on how people behave linguistically and non-linguistically, and how their linguistic behaviour influences and is influenced by the notion of the border as an historical, political, cultural and ideological divide. We should also take account of the third criterion – ‘places have stabilizing functions that afford insideness’ (Scheider and Janowicz Reference Scheider and Janowicz2010: 1) – in our evaluation of the extent to which the border serves on the one hand to compartmentalise people into their respective national territories and, on the other, to bring them together as ‘borderers’, even if this particular label is not one that appears to resonate very strongly among the AISEB participants. The border is thus simultaneously a place that can be conceptualised as an edge, a sharp interface between places, and as a zone or region. In some senses it is also an interstice, a narrow space between Scotland and England proper, where, as Kiely et al. (Reference Kiely, McCrone, Bechhofer and Stewart2000) found in Berwick, some of the normal rules of national identity have been suspended.

Britain’s (Reference Britain, Llamas and Watt2010) taxonomy also provides a cogent framework for the interpretation of our results. The border is fundamentally a demarcation of physical space, but on top of this basic property it also represents a bundle of divisions in social space: the boundary between two jurisdictions, between two distinct religious traditions, between two education systems (Kearney Reference Kearney2006; Torrance Reference Torrance2013), and between two linguistic continua, that of English and that of Scots (e.g. Britain Reference Britain2007). The physical and social aspects of the border respectively underpin its various meanings to local people in the domain of perceptual space, according to which the border may be seen as almost inconsequential to some borderers, but as a highly valued symbol of national distinctiveness to others.

A key lesson to be learned from AISEB and studies like it (see Watt & Llamas Reference Redinger, Llamas, Watt and Llamas2014) is that when it comes to drawing conclusions about the relationships between language and identity along national or regional borders it can at times be exceedingly difficult to make generalisations. At less than 100 miles in length, the border between Scotland and England is by most standards a short one. Yet we find that changes in the English spoken in towns at either end of it are proceeding in different ways, and that the lack of east/west symmetry appears to correlate with the ways in which the inhabitants of these towns choose to describe themselves. We believe that future research projects which attempt to treat borders all of a piece, as AISEB set out to do, are likely to yield results that are equally rich and multi-layered with respect to the evidence they provide in support of our theoretical models of the relationships between language, identity, and place.

1. Introduction

With the advent of pop and rock in the 1950s, a noted characteristic of British acts was to sing with an ‘American’ or ‘mid-Atlantic’ accent (Trudgill Reference Trudgill, Coupland and Jaworski1983: 253). However, successive waves of bands – starting with The Beatles in the 1960s (who gradually decreased the frequency of mid-Atlantic features in their singing throughout their career (Trudgill Reference Trudgill, Coupland and Jaworski1983: 258), the arrival of punk and new wave in the 1970s (Trudgill Reference Trudgill, Coupland and Jaworski1983: 261–4), and the phenomenon of Britpop, post-punk and indie music in the late 1980s and 1990s – have used an increasing number of local vernacular forms in song in the United Kingdom (Simpson Reference Simpson1999: 361–4). One recent example of a band moving towards the use of local dialect in song is the northern English indie group, Arctic Monkeys. Beal’s (Reference Beal2009) illuminating qualitative analysis of morphological, lexical and phonetic dialect forms across both speech and song lead her to conclude that the ‘Arctic Monkeys are singing in their “own” accents’ (Beal Reference Beal2009: 236). She suggests that, by doing so, the band indexes place through ‘localness’ and ‘northern-ness’ while at the same time rejecting the mainstream music industry standard of ‘mid-Atlantic’ norms (Beal Reference Beal2009: 238).

In this paper, we contribute to research on dialect in song and how it might represent ‘place’ through a variationist analysis of the indie music scene in Scotland. It has been claimed that, in the 1980s and early 1990s, Scottish indie bands showed little to no markedly Scottish features in singing.Footnote ² If they did step away from the ‘mid-Atlantic’ accent of previous decades, their linguistic variation seemed to be a nod towards the indie scene of the south of England, with a rather generic Standard British English accent. In more recent years, however, it has been suggested that, along with the changing political context of Scotland, a rise in the use of Scots in contemporary music can be observed.Footnote ³ Specifically, ‘talk of independence is giving Scottish culture a boost – especially in the thriving music scene. The new bands are inspired by the country's traditional music, and they sing in their own accents too’.Footnote ⁴ To date, however, no empirical analysis of the use of Scots in song has been undertaken, thus the linguistic outcomes of this changing political context remain unknown. In the current analysis, we address this gap by targeting the speech and song of two frontmen from contemporary indie bands in Glasgow: James from The Twilight Sad, and Craig from The Unwinding Hours. Both singers come from Scotland and both speak with an identifiably Scots accent. However, one of them, James, is noted to also sing in a broad Scottish accent, as demonstrated in the quotes from reviews and interviews:

1. 1. a. ‘… James Graham’s shaggy Glaswegian accent … ’. Footnote ⁵

   2. b. ‘… James Graham, whose broad Scots brogue is instantly recognisable in a musical landscape of bogus Yankee twangs.’Footnote ⁶

   3. c. ‘… wryly sardonic lyrics muttered in Scots-burr … ’.Footnote ⁷

James himself is well aware of using Scots in song: ‘[My accent is] too thick for them, … if you don’t like the Scottish accent you won’t like it … if someone’s writing about me, it’s just always the Scottish brogue, which I am fed up of reading about. (laughter) Or Groundskeeper Willie, from The Simpsons.’Footnote ⁸ No such descriptions are found for Craig, and he himself says: ‘I don’t think I sing in a Scottish accent. … I feel completely separate from that. I don’t feel any pressure to sing … in a Scottish accent. … it’s just a natural thing. It’s just the voice that came out.’

Thus we have two very different overt stances on the use of Scots by these two performers, but how does this correlate with their actual use? Is it really the case that James has a ‘thick brogue’ in song? And is Craig’s singing accent ‘just a natural thing?’ To answer these questions, we target an iconic stereotype of Scots, postvocalic /r/. Through quantitative and qualitative analysis, we investigate how the singers’ use of the different variants across speech and song may index their attitudes and concepts of place and social identity (Beal Reference Beal2009: 236–8).

We first begin by reviewing the literature on use of dialect in song and the current situation in Scotland. We then provide a brief overview of postvocalic /r/ followed by the data and methodology for the present study. Finally, we turn to the quantitative analysis and discuss the results in the context of the qualitative findings regarding the singers’ attitudes.

2. Sociolinguistic Variation in Song

Whilst no research on the sociolinguistics of singing in the Scottish context has been conducted, a number of studies look at accents in song with regard to place (e.g. New Zealand - see Gibson Reference Gibson2008) or enregisterment with a specific music genre (e.g. Hip-hop – see Clarke and Hiscock Reference Clarke and Hiscock2009). We consider the following four studies of paramount importance for this research.

One of the earliest and most influential sociolinguistic studies on language variation in song is by Trudgill (Reference Trudgill, Coupland and Jaworski1983). He notes the adoption of General American (GA) variants by British bands, and explains this in the context of ‘Acts of Identity’ (Le Page and Tabouret-Keller Reference Le Page and Tabouret-Keller1985), where speakers’ linguistic behaviour is motivated by the wish to resemble as closely as possible that of the group or groups with which they wish to identify. As Trudgill (Reference Trudgill, Coupland and Jaworski1983: 254) states ‘Americans have dominated the field, and cultural domination leads to imitation: it is appropriate to sound like an American when performing what is predominantly an American activity; and one attempts to model one’s singing style on that of those who do it best and who one admires most.’

His quantitative analysis of /t/ and /r/ reveals a more nuanced view of the British/American split: the use of these features by The Beatles and the Rolling Stones shows that they move from a predominantly ‘mid-Atlantic’ to a more British use over a six-year period from 1963 to 1969. Trudgill (Reference Trudgill, Coupland and Jaworski1983: 253) claims that this is further evidence of British singers modifying and manipulating their singing accent towards a culturally dominant linguistic model: at first towards the American features of the globally successful US artists of the early 1960s, but later towards a British-based use as a result of the growing popularity and power of the UK music scene. In his words, ‘British pop music acquired a validity of its own, and this has been reflected in linguistic behaviour’ (Trudgill Reference Trudgill, Coupland and Jaworski1983: 261). Moreover, Trudgill (Reference Trudgill, Coupland and Jaworski1983: 262) goes on to show that the advent of punk in the mid-1970s brought an increase in markedly British, and particularly working-class Cockney, features, with the singers adapting to the linguistic model of their targeted audience – the urban working-class youth.

In a follow up to Trudgill’s study, Simpson (Reference Simpson1999: 343) notes that ‘pop and rock singers, when singing, often use accents which are noticeably different from those used in their ordinary speech styles’. He suggests that the perceived homogeneity resulting from globalisation in the music industry has led to a breakdown into different styles, with artists trying ‘to carve out their identity by searching for some generic label that marks them out as different or unique’ (Simpson Reference Simpson1999: 362). Crucially, he argues that in examining the use of accent and dialect in song ‘such a study also needs to take into account those aspects of the wider sociopolitical and cultural context which act as determinants on particular singing styles’ (Simpson Reference Simpson1999: 364).

In his study on British singing styles, Morrissey (Reference Morrissey, Locker and Strässler2008) extends the linguistic discussion to include phonological considerations, deliberating factors such as ‘musical genre, song topics and cultural considerations, as well as … performance’ (Morrissey Reference Morrissey, Locker and Strässler2008: 194) and ‘singability’ – the variants’ ability to carry sound (Morrissey Reference Morrissey, Locker and Strässler2008: 213) in sung performance. He points out that singing is largely unidirectional, lacking an interpersonal interaction between singer and audience (Morrissey Reference Morrissey, Locker and Strässler2008: 195). Nonetheless, he draws upon Bell’s (Reference Bell1984) ‘Audience Design’ framework to distinguish between Outgroup Referee Design (the singers’ wish to identify with a linguistic group they do not belong to, but consider prestigious) and Ingroup Referee Design (their interaction with an outgroup by referencing their own – possibly absent – ingroup’s style features) (Bell Reference Bell1984: 197–8).

The latest analysis of dialect use in song comes from Beal (Reference Beal2009), who takes a language-ideological approach by examining the linguistic behaviour of the lead singer of Yorkshire indie band Arctic Monkeys. She compares his linguistic behaviour in song with his speech in a radio interview. Beal comes to the conclusion that the original ‘mid-Atlantic’ features adopted in the past have become enregistered as appropriate for the performance of mainstream pop. In other words: they have become socially recognised as indexical (Agha Reference Agha2005: 38) of a pop singing accent. Singers who diverge from this model express their rejection of it (Beal Reference Beal2009: 238). Moreover, Beal argues that the Artic Monkeys’ Sheffield singing accent is not ‘merely a default’ accent resulting from said rejection of the American model (Beal Reference Beal2009: 237). It is, in fact, a consciously made ‘positive choice’ in favour of specific traditional and modern local features indexing ‘localness’, independence and authenticity, and thus a case of enregisterment not only of their hometown variety, but also with attributes of British indie music (Beal Reference Beal2009: 238). Singing is a highly performed act – a fact which complicates ‘the links between sociolinguistic practice and social meaning, [but] can also expose those links quite strikingly and make them available for critical reassessment’ (Coupland Reference Coupland2007: 171). Hence song is not merely a simple analogue of speech, but a metalinguistically constructed act, which constitutes a most interesting medium for the study of enregisterment.

These studies demonstrate that the accent a singer or band chooses to portray in song is highly context-dependent, and includes both convergence with and divergence from outgroups and ingroups (Morrissey Reference Morrissey, Locker and Strässler2008: 197–8) in the pull between mainstream and local musical identities. In order to contextualise the present study, we next describe the indie music scene in the United Kingdom and Scotland in particular, and consider how it is situated within the current social, political and linguistic environment.

3. Data and Methodology

3.2 Postvocalic /r/ in Scots

Scots is stereotypically described as a rhotic accent (Wells Reference Wells1982: 10–1) and the historical record indicates that /r/ was once an apical tap [ɾ] and often a trill [r] (Grant Reference Grant1914, Johnston Reference Johnston and Jones1997). However since the nineteenth century, derhoticisation in working-class speech has been noted, with the increasing use of approximant forms of / r/ as in (2a–c) and the traditional stereotypical trill (2d) increasingly circumscribed to the speech of adult broad Scots speakers, particularly in rural areas (Johnston Reference Johnston and Jones1997: 511; Macafee Reference Macafee1983: 32; Scobbie et al. Reference Scobbie, Gordeeva and Matthews2006: 4). The following examples demonstrate these features:

1. 1. a. ‘And then sometimes, it’s just- it’s wei[ɻ]d, it just clicks like that.’ (James: speech)

   2. b. ‘I’d move from the drums to pick up the guita[ɻ].’ (Craig: speech)

   3. c. ‘It’s the dance that I will neve[ə˞] share.’ (James: song)Footnote ¹⁷

   4. d. ‘The cu[r]tains closed again.’ (James: song)Footnote ¹⁸

This change has led to a sociophonetic continuum in the realisation of postvocalic /r/ from weakly to strongly rhotic as demonstrated in Table 10.1 (adapted from Lawson et al. Reference Lawson, Scobbie, Stuart-Smith and Lawson2014: 61).

Table 10.1: Continuum of use for postvocalic /r/.

Weakly Rhotic					Strongly Rhotic
no /r/	derhotic	alveolar	retroflex	schwarFootnote 19	tap	trill

For urban Scots, /r/-vocalisation is becoming increasingly common in working-class speech (Johnston Reference Johnston and Jones1997: 511; Romaine Reference Romaine and Trudgill1978; Macafee Reference Macafee1983: 32; Stuart-Smith et al. Reference Stuart-Smith, Trouvain and Barry2007: 241). However, in Glasgow, a complex class split has emerged in use of /r/: working-class speech is characterised by increasing derhoticisation, with speakers using barely audible pharyngealised or plain vowel-sounds, while middle-class speakers, on the other hand, appear to be becoming more and more /r/-ful, using strongly rhotic postvocalic realisations such as [ɻ] (Lawson et al. Reference Lawson, Stuart-Smith and Scobbie2011: 257; Lennon Reference Lennon2011; Stuart-Smith et al. Reference Stuart-Smith, Lawson, Scobbie, Chiara and Calamai2014).

Furthermore, speakers may have the ability to ‘style-drift’ from one end of the continuum to the other (Stuart-Smith Reference Stuart-Smith, Corbett, McClure and Stuart-Smith2003), thus this particular variable may be available to do sociolinguistic work. To answer the question of how James and Craig make use of this continuum, we now turn to the analysis of the data at hand.

4. Results

Table 10.2 shows the overall distribution of the seven variants in the dataset. It reveals that the non-rhotic and derhoticised variants together make up the majority of forms at over 60% of the data, with the alveolar and retroflex accounting for 10% and 3% respectively. The most interesting result from Table 10.2 is the lack of strongly rhotic variants in this Scottish context: schwar is used only 3% of the time and, perhaps not surprisingly for an urban dialect, there is an almost total lack of the traditional Scots trill, with only two tokens. However, the outlier in this continuum from weakly to strongly rhotic is the use of the tap, at a full 21%.

Table 10.2: Overall distribution of variants.

	Weakly rhotic				Strongly rhotic
	no /r/	derhotic	alveolar	retroflex	schwar	tap	trill
Ns	144	225	61	19	22	125	2
%	24.1	37.6	10.2	3.2	3.7	20.9	0.3

Closer investigation of the individual speakers may shed further light on this use. Table 10.3 shows the results for James and Craig separately.

Table 10.3 reveals a key difference across these individuals: James shows a polar use between the weakly and strongly rhotic variants of the continuum (compare 57% no /r/ and derhotics to 35% taps), whereas Craig is situated at the weakly rhotic end of the spectrum (with a full 90% no /r/, derhotics and alveolar approximants). In other words, James and Craig have very different patterns of variant use.

Due to the marginal use of the retroflex, schwar and trill variants, these are removed from subsequent analysis. We further reconfigure the data to look at the crucial effect of speech versus singing in the deployment of these variants. Table 10.4 shows the results, repeated graphically in Figure 10.1.

Figure 10.1: Postvocalic /r/ by speaker across song and speech.

Table 10.4 and Figure 10.1 illustrate noteworthy differences between the speakers and styles. Craig shows a similar distribution for variants in song and speech, with a high level of alveolar approximants and derhoticised variants across both. James on the other hand, shows a different pattern; in song he uses much higher rates of the tap compared to his speech.

Closer examination of James’s use of the tap can be gleaned from the phonetic context in which it occurs, specifically the following phonetic environment. Previous research on Scottish English has shown that phonetic environment influences which variants are used and when (Romaine Reference Romaine and Trudgill1978; Macafee Reference Macafee1983; Stuart-Smith et al. Reference Stuart-Smith, Trouvain and Barry2007; Schützler Reference 232Schützler2010). For example, Romaine (Reference Romaine and Trudgill1978) finds that the loss of postvocalic /r/ is most likely to appear in prepausal contexts, and [ɹ] is most frequent when followed by a consonant-initial word. Linking contexts, namely a following vowel, are most likely with [ɾ]. The least likely context for the tap is prepausal. In contrast, Schützler (Reference 232Schützler2010) also finds that word-final prevocalic and prepausal contexts are most likely to appear with rhotic variants.Footnote ³¹

In this analysis, we divide the data into 4 different categories: coda cluster (5a–b), coda vowel (5c–d), coda consonant (5e–f), and prepausal contexts, represented by coda # (5g–h).

1. 1. a. ‘There’s a girl in the crowd … ’ (James: song)Footnote ³²

   2. b. ‘Sin and holy words, I know.’ (Craig: song)Footnote ³³

   3. c. ‘Put it in your eye.’ (James: song)Footnote ³⁴

   4. d. ‘You’re insecure and bored.’ (Craig: song)Footnote ³⁵

   5. e. ‘And there’s your sister with her answer and she’s always right.’ (James: song)Footnote ³⁶

   6. f. ‘Cause your cracked dry lips … ’ (Craig: song)Footnote ³⁷

   7. g. ‘And you’ll go nowhere#’ (James: song)Footnote ³⁸

   8. h. ‘And I wish that I could see you there#’ (Craig: song)Footnote ³⁹

Table 10.5 shows the results. Alveolar variants are removed due to the small number (N = 10).

With the exception of coda #, all the linguistic contexts show a heightened use of the tap variant: coda consonant at 50% in song versus 13% in speech, coda vowel at 92% versus 60% and coda clusters at 35% versus 22%. In other words, the linguistic contexts behave much the same with respect to variant use.Footnote ⁴⁰

5. Discussion

The examination of the social and linguistic constraints on use reveal a number of findings:

• Overall, there is a high rate of the variants at the weakly, rather than the strongly rhotic end of our continuum. This is despite postvocalic /r/ being a classic stereotype of Scots.

• James and Craig exhibit different rates of use of the variants: the majority of Craig’s variants are from the weakly rhotic end of the spectrum, whereas James uses variants from both ends of the continuum.

• Craig does not differentiate song and speech in the use of postvocalic /r/: he is generally at the non-rhotic end of the continuum in both. James’s results on the other hand point to a higher rate of the rhotic tap in song when compared to speech.

• James uses the higher rates of taps in particular linguistic contexts only in song: predominantly in coda vowel and coda consonant environments, and, to a lesser extent, in coda clusters.

How can these results be explained? Despite postvocalic /r/ being a stereotype of Scots, the overall distribution of forms shows that the majority of variants appear at the weakly rhotic end of the scale. Moreover, the most iconic variant – the trill – appears only twice. This initial result aligns fully with reports that postvocalic /r/ is receding in the Central Belt, at least in working-class speech (see Stuart-Smith Reference Stuart-Smith, Foulkes and Docherty1999, Reference Stuart-Smith, Corbett, McClure and Stuart-Smith2003, Reference Stuart-Smith, Llamas, Mullany and Stockwell2006, Reference Stuart-Smith, Trouvain and Barry2007; Stuart-Smith et al. Reference Stuart-Smith, Trouvain and Barry2007, 2010, Reference Stuart-Smith, Lawson, Scobbie, Chiara and Calamai2014). Given this context of change, especially where derhoticisation is ‘most advanced in the more populous western conurbation’ (Stuart-Smith et al. Reference Stuart-Smith, Lawson, Scobbie, Chiara and Calamai2014), these initial results are not surprising.

A more nuanced picture emerges when the results are divided across the two singers, James and Craig, who show different patterns of use. Attitudes have been shown to play a defining role in linguistic behaviour (see, e.g., Garrett Reference Garrett2010 for an overview of research). Comments made by Craig and James during the sociolinguistic interview with the first author regarding their attitudes towards the use of Scots may provide clues to these patterns of use.

In reflecting on the use of Scots in song over the past few years, Craig thinks that ‘singing in an American accent was going too far, … straying way too far from your roots’. At the same time, however, ‘you didn’t have to sing in a particular Scottish brogue, whereas now, it’s become a very popular stance … It’s become more apparent in recent years, but when I was coming up … the Scottish voice wasn’t as strong.’ Despite this observed increase in the use of Scots in song, Craig states ‘I don’t think I sing in a Scottish accent’ and ‘I don’t think about an accent at all. It’s what just comes out’. In comparing Craig’s data across the two different media, a clear picture emerges: his use of postvocalic /r/ is the same across speech and song. Thus, in spite of Craig declaring that he does not sing in the Scottish accent he uses in speech, at least for this variable, Craig is not using a ‘mid-Atlantic’ accent or a stereotypically Scottish accent, but instead is singing in his own accent. A plausible explanation for him not being aware of how similar his use of /r/ is throughout song and speech is that postvocalic /r/ is actually not enregistered as Scottish for Craig – or if it is, he associates it with a type of Scottishness he does not relate to.

In contrast to Craig, James states that he sings is his ‘natural way of speaking’, and that ‘dialect was the only way to go about it’ in his music career. Moreover, he has been under no pressure to change his accent by his record label, and indeed his Scottish accent in song has ‘been a kinda unique selling point about the band’. This may explain why the no /r/ and derhotics of his speech are substituted by taps in song, suggesting that this variant indexes Scottishness for James. In other words, his sense of place.

In the projection of place, however, James has an overt concern with ‘authenticity’ of voice, as revealed through the interview data in the use of Scots in song:

It may appear from the heightened use of taps that James is ‘trying too hard’. However, the linguistic constraints reveal that he increases his use of taps across all linguistic contexts,Footnote ⁴¹ rather than on a more idiosyncratic level that we might expect from, for example, hyperdialectal use (Trudgill Reference Trudgill1986: 75). Moreover, he uses a variant extant in his speech – the tap – rather than turning to the even more stereotypical ‘Groundskeeper Willie’ trill of the media comment. This suggests that he is doing what ‘comes kinda naturally’ in song – using taps as in his speech – but just at slightly heightened rates. These heightened rates could relate to the level of performance; while Morrissey emphasises the likeliness of non-rhoticity in favour of ‘singability’ (Morrissey Reference Morrissey, Locker and Strässler2008: 211), the high level of performance and theatricality in song might lead to James increasingly using the tap to ‘style’ and ‘stage’ his Scottish identity (Coupland Reference Coupland2007: 25).

6. Conclusion

Beal (Reference Beal2009: 224) describes indexicality as the process ‘whereby linguistic features become associated with social categories and can then be used to do social work’. This utilisation of language variation works particularly well in song, since verbal performance has the power ‘to transform social structures’ (Bauman Reference Bauman1975: 305). As discussed in the Introduction, it has been suggested that there has been a rise in the use of Scots in contemporary music over the last twenty years in Scotland, and this small-scale study set out to investigate the linguistic validity of this claim. The study of the speech and song of the frontmen from two Scottish indie bands demonstrates that ‘indexical meaning is created and reinforced in local practices in which different people participate in different ways, if at all’ (Johnstone and Kiesling Reference Johnstone and Kiesling2008: 6). Hence we cannot expect indexicalities ‘to have unique social meanings, even in the same socio-cultural settings’, but rather to ‘become amenable to being discussed, argued over and renegotiated, metalinguistically’ (Coupland Reference Coupland2007: 23). Our analysis of the use of postvocalic /r/ in the participants’ speech and singing has shown striking differences in use, especially in the highly performed context of the latter. A number of social factors might have contributed to this divergence in singing accents, such as the singers’ age gap and the difference in influences from the music industry and peers when first embarking on their music career. However, it has become apparent that what is and is not enregistered for a particular community member may also play a key role in the projection of place and social identity in an evolving music industry.

1. Introduction

Recent studies have demonstrated the need to place speakers’ own evaluations of local and linguistic identity at the forefront of interpretations of phonological variation (Dyer Reference Dyer2002; Llamas Reference Llamas2007; Llamas and Watt, Chapter 9 in this volume). This approach provides a more nuanced understanding of the links between a speaker’s orientation to place and their linguistic practice, highlighting the relationship between locally salient identities and patterns of phonological change.

This paper will focus upon phonological variation in the dialect of Royston, an ex-mining community in the Metropolitan Borough of Barnsley in South Yorkshire. The findings will question assumptions regarding the inevitability of levelling and mutual convergence in situations of dialect contact, by demonstrating how three successive generations have resisted convergence to pan-regional phonological norms. In recent studies of dialect contact, levelling is taken as ‘something of a “given’’ (Kerswill Reference Kerswill2002: 187); an anticipated consequence of this process is the gradual disappearance of linguistically marked and minority variants (Britain and Trudgill Reference Britain and Trudgill1999: 246). Kerswill makes the distinction between levelling and geographical diffusion, the former he confines to relatively compact geographical areas, and the latter he describes as a process ‘by which features spread out from a populous and economically and culturally dominant centre’ (Reference Kerswill2002: 187–8). Kerswill (Reference Kerswill2002: 188) also asserts that features which spread via geographical diffusion are likely to be acquired in cities and towns before infiltrating the dialects of the rural areas ‘in between’. However, both levelling and geographical diffusion depend upon mutual linguistic convergence and, to date, relatively little sociolinguistic research has focused upon speech communities which diverge from rather than accommodate to surrounding varieties. Research by Watson (Reference Watson2006: 55) begins to explore the phenomenon of ‘phonological resistance’ establishing that aspects of Liverpool English are diverging from supralocal phonological norms. However, the factors which motivate a speech community to resist the diffusion of supralocal or pan-regional phonological norms need further exploration.

This paper will examine both production and attitudinal data in order to show that greater levels of dialect contact do not necessarily lead to the rejection or suppression of minority variants and greater linguistic homogeneity. Furthermore, it will explore the ways in which attempts to assert a distinct local identity can result in the retention and deliberate exaggeration of demographically and linguistically marked forms (see Johnstone et al. Reference Johnstone, Andrus and Danielson2006: 92). The study will focus upon speakers’ metalinguistic commentary in order to make sense of locally salient identities (Dyer Reference Dyer2002; Llamas Reference Llamas2007). It will also consider how these identities of place interact with linguistic production, recognising that ideology is key to any interpretation of speakers’ linguistic choices (Eckert Reference Eckert2008: 456).

This paper will reinforce the need to take an ethnographic approach to the collection and analysis of data. It will also demonstrate the importance of an understanding of local historical contexts, social experiences and tensions which may impact upon the use and perception of linguistic variables. Silverstein (Reference Silverstein2003) makes the connection between language and the construction of identity, emphasising the indexical link between linguistic forms and social value judgements. His concept of an indexical order illustrates that social or indexical values are not inert, and he asserts that an effective indexical analysis must acknowledge the ‘duplex’ nature of language use, namely that it is both ‘pragmatic’ and ‘ideologically informed’ (Reference Silverstein2003: 227). In line with Silverstein’s notion of the fluidity of indexical values, work by Britain and Trudgill (Reference Britain and Trudgill1999) demonstrates that the social meaning of variables can be adapted or reallocated, thus revealing complex and multifaceted aspects of local and linguistic identity. Viewing identity as a ‘socio-cultural phenomenon’ (Bucholtz and Hall Reference Bucholtz and Hall2005: 585), the Royston study links acts of linguistic identity to the ideological perceptions which form via membership of a speech community, in order to demonstrate how phonological resistance can act as a powerful indicator of a speaker’s sense of place.

2. The Research Location

The town of Barnsley is situated in South Yorkshire , approximately 16 miles north of the city of Sheffield, and approximately 21 miles south of the city of Leeds. Figure 11.1 shows the location of Barnsley relative to the north of England. The Metropolitan Borough of Barnsley was formed as a result of the Local Government Act 1972 and is made up of twenty-one smaller regions, or wards. The borough has a total population of 231,221, which equates to approximately 4.38 per cent of the total Yorkshire region. Barnsley, once famous for its mining industry, has no working pits remaining; wholesale and retail trades now account for the majority of employment (Barnsley Council Online, Census 2011).

Figure 11.1: Location of Barnsley relative to the north of England.Footnote ²

The urban township of Royston lies on the northernmost boundary of the Barnsley borough, a boundary which also serves to divide South from West Yorkshire. Royston has a current population of 10,728 (approximately 4.64 per cent of the borough of Barnsley as a whole) and is situated approximately 4.3 miles from the centre of Barnsley, and 7.2 miles from the centre of Wakefield in West Yorkshire. However, a journey through Royston is not the most efficient route between the two centres. The major roads linking Barnsley and Wakefield are the M1 and the A61; Royston lies well to the east of these routes, and the B6132, which links Barnsley and Wakefield via Royston, is a comparatively indirect and meandering route (see Figure 11.2).

Figure 11.2: Location of Royston in relation to the Metropolitan Boroughs of Barnsley and Wakefield.Footnote ³

The township of Royston is relatively isolated from other settlements as recent descriptions of the area confirm: ‘To the north and west are extensive areas of countryside and to the south open land separates the settlement from Athersley and Carlton in Urban Barnsley’ (Unitary Development Plan, Barnsley Metropolitan Borough Council 1995: 3). This geographical isolation is, however, being eroded. Within the last decade a rapid programme of house building on the peripheries of Royston has brought the township nearer to surrounding settlements, most notably to Carlton. The potential loss of Royston’s isolated status is not viewed positively by older residents, as can be seen from the comment in (1) below:

1. (1) OM5: ‘Carlton is getting nearer and nearer; it won’t be long until it joins up with Royston and we’ll have lost our identity altogether’

2.1 Employment and Industry

By the nineteenth century Royston had developed as an area rich in market gardens; as recently as 1893 there were at least six commercial market gardens and eight farms in the village (Elliott Reference Elliott2000: 4). However the nature of employment in Royston changed rapidly following the opening of Monckton Colliery in 1876. This large new colliery was situated on the edge of the township attracting large numbers of migrant workers who would initially lodge with local residents, but later settle in Royston following a period of extensive house building. As shown in Figure 11.3, census records for the first three quarters of the nineteenth centuryFootnote ⁴ reveal that the population of Royston fluctuated slightly in the decades preceding the opening of the colliery, but the total never peaked above 676. However, in the forty-year period following the opening of Monckton, Royston’s population rose rapidly to over 6,000, an increase of over nine times the total prior to the establishment of the colliery.

Figure 11.3: Population totals for Royston 1801 to 1911 (UK Census data).

The majority of this incoming workforce came from areas of the Black Country.Footnote ⁵ As Figure 11.4 shows, the Black Country lies approximately 100 miles south of Royston and is described by Asprey (Reference Asprey2007) as: ‘unusual in that it is not an area delimited by political, physical, or economic boundaries’ (Reference Asprey2007: 78). Despite the disputed boundaries of the region, the Black Country is generally thought to encompass parts of the counties of Staffordshire, Warwickshire and North Worcestershire (Asprey Reference Asprey2007). The name Black Country derives from the Industrial Revolution and denotes the impact of pollution on the general landscape of the region. The Black Country had embraced industrialisation rapidly and by the late nineteenth century many of the collieries in the region had been exhausted of resources leaving a workforce skilled in deep seam mining and coke production, yet in desperate need of employment (Barnsby Reference 255Barnsby1971).

Figure 11.4: Location of the Black Country in relation to Barnsley and Royston.Footnote ⁶

The 1901 census for Royston gives a sense of the proportions of incoming workers. The figures for the whole of Royston show that 31 per cent of household heads came from the Black Country compared with 12 per cent from Derbyshire and Nottinghamshire, with the latter providing the second largest portion of the influx (UK Census data 1901). In areas of Royston where most incoming workers settled, still known today as ‘Little Staffordshires’, the figures rise to 38 per cent from the Black Country and 13 per cent from Derbyshire and Nottinghamshire (UK Census data 1901). It is clear that the majority of long distance migration came from the Black Country, but with significant numbers from Derbyshire and Nottinghamshire. Yet Cave (Reference Cave2001:15) finds that only the Black Country/Staffordshire influx gains any perceived status in terms of the local character and dialect of the township:

Many people interviewed for this study referred to Royston as ‘Little Staffs’. A quarter of those interviewed from Royston claimed Staffordshire or Black Country ancestry, and informants interviewed from neighbouring localities, such as Darfield, Havercroft, and Central Barnsley, confirmed that they believed Royston was full of ‘Staffordshire folk’ with a distinctive regional speech pattern.

Monckton Colliery reached its zenith in terms of production and employment in the early decades of the twentieth century. The colliery eventually closed in December 1966 and the shops and amenities that had flourished in Royston providing facilities for the community of mineworkers and their families went into rapid decline. Today the majority of Royston residents have to travel beyond the township to find work, and wholesale and retail trades now employ the greatest proportion of Royston’s working population at 17.6 per cent, followed by manufacturing at 14.8 per cent (Barnsley Council Online, Census 2011).

3. The Data

3.3 Methodology

The word lists comprised 105 tokens in total; these included thirty FACE and thirty GOAT tokens. However, as previous studies have highlighted the difficulties of separating the vocalic portion from preceding or following approximants (Ferragne and Pellegrino Reference Ferragne and Pellegrino2010), tokens with these phonetic environments were eliminated. This reduced the total to fifteen FACE and fifteen GOAT tokens per speaker, retaining vowels with preceding/following obstruents and nasals.

Initial auditory analysis and visual inspection of the corresponding spectrogram was undertaken in Praat (Boersma and Weenink Reference Boersma and Weenink2008) to determine the qualities of the FACE and GOAT vowels produced by speakers in the three datasets, and to code tokens as diphthongal or monophthongal realisations. This analysis was then tested via the extraction and measurement of formant data. To minimise the effects of coarticulation, measurements were extracted for F1 and F2 at the 25 per cent and 75 per cent points into the vowel segment using a script in Praat. Each speaker’s tokens of FLEECE, TRAP and GOOSE were also measured and used as reference vowels for the purposes of normalisation. The data was normalised using the Watt and Fabricius modified method (see Fabricius et al. Reference Fabricius, Watt and Johnson2009; Flynn Reference Flynn2011) in NORM (Kendall and Thomas Reference Kendall and Thomas2007).

In order to quantify the degree of movement throughout the vocalic portion of each FACE and GOAT token, the Euclidean distance between F1 and F2 at 25 per cent and 75 per cent was calculated using a script in R (R Core Team 2013) and the measurements were converted to Equivalent Rectangular Bandwidth (ERB) in a move to model human perception of frequency (see Moore and Carter, Chapter 12 in this volume). The resulting data provides a degree of diphthongisation for each speaker’s FACE and GOAT tokens (see Haddican et al. Reference 256Haddican, Foulkes, Hughes and Richards2013); the mean value of these totals for each participant is represented in Figures 11.6–11.8. Means lower than 1.5 ERB are classed as monophthongal as this matches most closely the auditory and visual analysis of the word list recordings. The horizontal line on each graph represents this division between monophthongal and diphthongal realisations.

Figure 11.6: Degree of diphthongisation in ERB of FACE and GOAT tokens for Royston speakers.

Figure 11.7: Degree of diphthongisation in ERB of FACE and GOAT tokens for Barnsley speakers.

Figure 11.8: Degree of diphthongisation in ERB of FACE and GOAT for Wakefield speakers.

The following section will examine these data in order to establish whether metalinguistic claims regarding the Royston dialect are evident in linguistic practice. It will illustrate the degree of diphthongisation of FACE and GOAT tokens produced by Royston speakers in comparison to that found in the adjacent speech communities of Barnsley and Wakefield.

4. Results

5. Discussion

Acoustic analysis of production data reveals that diphthongisation of FACE and GOAT vowels in the Royston dialect is not receding. The younger generation of Royston speakers show greater phonological resistance to supralocal norms than the middle and older generations of speakers in the township. A small minority of the Barnsley and Wakefield corpora diphthongise FACE or GOAT and yet there is only scant evidence of a resemblance between the vowel qualities of the Royston and Wakefield FACE diphthongs. This would indicate that FACE and GOAT variants in the Royston dialect are not influenced by dialect contact with these adjacent speech communities. There is, however, a match between current variants of FACE and GOAT in the Royston dialect and forms that are receding in the dialects of the Black Country and Derby. As the Black Country, Nottinghamshire and Derbyshire comprised the largest proportion of long-distance migration to Royston in the late nineteenth and early twentieth centuries, this could indicate that the regionally distinctive Royston variants have been influenced by historical dialect contact with these varieties.

Looking at the dialect of York in North Yorkshire, Haddican et al. (Reference 256Haddican, Foulkes, Hughes and Richards2013) found that respondents in their study were keen to retain the distinction between northern monophthongal and southern diphthongal realisations of FACE and GOAT, and that patterns of FACE and GOAT monophthongisation were ‘strongly shaped by identities of place’ (Haddican et al. Reference 256Haddican, Foulkes, Hughes and Richards2013: 373). Prior studies have established that monophthongal forms of FACE and GOAT are emblematic of northern varieties; however, the Royston production data presented in this paper runs counter to this evidence showing that three generations of speakers in the township of Royston have resisted levelling to pan-regional norms.

The following section will examine attitudinal data in order to understand more about Royston speakers’ orientation to place, and to consider why the Royston dialect has resisted the regional diffusion of long monophthongal forms of FACE and GOAT evident across other Yorkshire varieties.

6. Conclusion

This chapter provides a closer study of a place in between; it explores patterns of phonological divergence which suggest that speakers are actively resisting participation in the geographical diffusion of supralocal norms. A multifaceted approach has been undertaken to make sense of local and linguistic identity creation in the dialect of Royston, within broader patterns of regional dialect variation and change. This has shown the benefits of combining phonetic analysis with knowledge of locally significant events and a speaker’s sense of place.

The Royston study demonstrates the resilience of the diphthongal Royston forms. Referring to Labov’s study of Martha’s Vineyard (1963), Eckert (Reference Eckert2008: 454) observes that linguistic variables can be employed as part of a ‘local ideological struggle’. In the Royston study, the nature of that ideological struggle changes according to generational perspective. Both generations of Royston speakers are using FACE and GOAT variants in order to construct social meaning. Older speakers clearly see the FACE and GOAT forms as indicative of the township’s linguistic heritage employing them as part of a conscious resistance to linguistic and social change. Younger speakers, however, do not associate the FACE and GOAT variants with heritage and tradition, they associate them with the unique character of the township. There is clear evidence of reallocation in terms of the ways in which the two generations assign labels to their linguistic practice. The Royston variants have, over a period of three generations, resisted levelling to supralocal norms and part of this resilience can be attributed to a process of ‘socio-stylistic reallocation’ (Britain and Trudgill Reference Britain and Trudgill1999: 247), whereby new social values have been assigned to the Royston FACE and GOAT variants.

The findings from Royston clearly challenge the idea that dialect levelling is a given in situations of dialect contact. There is no evidence to show that Royston speakers, spanning three generations, are accommodating towards adjacent speech communities; in fact the reverse is true, and younger Royston speakers are intensifying phonological resistance to supralocal norms. The sociolinguistic situation in existence at the beginning of the twentieth century, which likely gave rise to the Royston variants of FACE and GOAT, no longer exists, and yet successive generations of speakers have maintained the distinctive forms as a consequence of Royston’s unique demographic, political and geographical history. In order to make sense of such linguistic phenomena, it is crucial that interpretations of the history, geography and the general narrative of place come from the speakers themselves. Assumptions of place imposed from outside can distort the link between linguistic forms and the construction of local identities. The Royston study shows that patterns of dialect reallocation, resistance and divergence are symbiotically linked to speakers’ interpretations of the physical and ideological landscape in which they are situated.

1. Introduction

Discussions of language and place sometimes give the impression that the ‘local’ is a clearly defined and recognisable entity. This is often the case when a language feature is recognised as belonging to and defining the social characteristics of a particular place. However, as Eckert (Reference Eckert and Fought2004: 109) has noted, ‘the community is a contested entity that is differentially constructed in the practices and in the speech of different factions, as well as individuals.’ This can give rise to competing and contended ideas about what it means to be ‘local’ (Johnstone Reference Johnstone and Fought2004: 71).

This paper explores how life trajectory and gender interact to affect how individuals from the same small island community use language to index ‘local’. In particular, we show how ‘human geographies, particularly the built and social environments we inhabit, create varying opportunities for individuals and social groups’ (Laws Reference Laws, Jones, Nast and Roberts1997: 48). These ‘varying opportunities’ result in the construction of alternative local identity types. In particular, we demonstrate that defining place is no straightforward endeavour when ‘place’ takes on specific meanings linked to alternative life trajectories, and that this is true no matter how small the community studied. As Britain (Reference Britain, Auer and Erich Schmidt2009a: this volume) has argued, rural areas have been particularly ‘fetishised’ as homogeneous and uncontested entities (and this is despite the fact that the types of diversity examined in urban spaces may be considered to be relatively simplistic – see Britain Reference Britain, Al-Wer and de Jong2009b: 228). By focusing on a small, rural, island community, our work will demonstrate that ‘individual speakers distinguish themselves linguistically no matter what type of community they live in’ (Schreier Reference Schreier2006: 27). We show that what matters is not the size or type of community, but the necessity for individuals to index distinct styles and identities within a particular social space.

2. The Research Location

The Isles of Scilly are a group of islands, situated approximately 28 miles off the south-west coast of Cornwall, England. They consist of numerous islands, but only five of these are inhabited. Of these, St. Mary’s is the largest, both in terms of physical size and population. The islands as a whole had a population of 2,203 in the last census (Office for National Statistics 2011), of which 75 per cent lived on St. Mary’s. All of the speakers considered in our paper come from St. Mary’s.

Figure 12.1 shows the location of the islands relative to the south-west English mainland. It also indicates the mainland departure points for the islands. The islands are accessible by air and sea. Light aircrafts fly between St. Mary’s and Land’s End or Newquay all year round (weather-permitting), and the trip takes between 15 and 25 minutes. The islands also have a passenger ferry service, which operates between Penzance and St. Mary’s from Easter to October. At all other times, the only scheduled boat is the freight service, which serves the islands three times a week year-round.

Figure 12.1: Location of the Isles of Scilly relative to the south-west of England.Footnote ¹

The islands’ major industry is tourism, which accounts for over 85 per cent of its income (The Isles of Scilly Council 2005: 14). The tourism trade is long established, but was particularly bolstered by the arrival of the railway to Penzance in 1867, and the replacement of sailing vessels by a steamer in 1859. The islands also have a reasonably robust farming industry (which benefits from favourable weather conditions), and some small family fishing businesses.

The islands have a particularly interesting history of governance, having been leased from the Crown and managed by a series of governors between 1571 and 1920. This makes them quite distinct from the nearest Cornish mainland – a fact of which Scillonians are very proud. This can be seen in (1), a quotation from Scillonian in exile, Frank Banfield, writing in the Gentleman’s Magazine in 1888.

Banfield (Reference Banfield and Urban1888: 54) goes on to claim that Scilly’s ‘correct speech’ is a consequence of the influence of the first governor, Sir Francis Godolphin, and his descendants, who ‘impressed their own correcter locution and more Eastern English of inheritance and education upon the population.’ The link that Banfield makes between education and ‘correct speech’ (by which we assume he means something akin to ‘Standard English’), and the ideology that Scillonians are better educated and more well-spoken than mainlanders, can be found in metalinguistic commentary about the variety across history, as shown in (2) and (3) below.

It is certainly true that education has featured highly in Scillonian life. Education was made compulsory on Scilly before it was on the mainland (Mothersole Reference Mothersole1914: 48). Also, before the islands built their own secondary school in 1966, wealthy and/or especially bright children were sent away to selective and fee-paying boarding school between the ages of 11 and 16. These education patterns undoubtedly provided ideological pressure supporting the use of Standard forms of English on Scilly. However, given that type of education also served to segregate the population, it does not necessarily follow that there was a consistent effect of ideology on language use across the population. That is to say, whilst the historical metalinguistic commentary implies a uniform orientation to standard forms on the islands, in the absence of substantial linguistic work (Scilly has been neglected in dialect surveys, although see Thomas Reference Thomas1979), there is no way to confirm whether these perceptions reflect actual patterns of language use across the community.

In order to explore the extent to which islanders orient to Standard English forms, we examine how a cross-section of Scillonians make use of two especially salient markers of Standard English, the vowels in the TRAP and BATH lexical sets. The next section explains the suitability of the TRAP and BATH lexical sets for this study and outlines how we undertook our analysis.

3. The Data

The vowels in the TRAP and BATH lexical sets (Wells Reference Wells1982) are particularly useful variables for our purposes because they carry a good deal of ideological baggage in English English. In particular, Mugglestone (Reference Mugglestone2003: 78) refers to the pronunciation of the BATH vowel as ‘a salient feature of “talking proper”’. Most historical linguists agree that, prior to the seventeenth century, words in both of these lexical sets were pronounced with the same low front vowel (Lass Reference Lass1976: chapter 4; MacMahon Reference MacMahon and Romaine1998; Beal Reference Beal1999: 105–11; Reference Beal2004: 139; Piercy Reference Piercy2010: 9–24). Gradually, in Standard English English (StEE), this vowel raised for some and lengthened for all before the voiceless fricatives /f, θ, s/, and before nasal clusters. However, it is important to note that this change did not proceed to completion across every lexical item, as Piercy (Reference Piercy2010: 17–18) has observed. Nonetheless, despite being ‘a half completed sound change’ (Wells Reference Wells1982: 233), the outcome was a TRAP/BATH split, based on duration, with vowels in the BATH lexical set having longer duration than those in the TRAP lexical set. Then, very gradually, between the eighteenth and twentieth centuries, the BATH vowels also retracted, so TRAP and BATH eventually came to differ by vowel quality and duration.

However, this change did not proceed to completion in all varieties of English English. Most notably for our purposes, the traditional varieties of Cornish English (the mainland variety closest to the Isles of Scilly) have TRAP/BATH patterns that were fossilised at the point during which English was introduced to this area following the loss of the Cornish language (Wakelin Reference Wakelin1975: 1986). Whilst there is some variation in this region with regard to the precise vowel quality of these vowels (reflecting the phased introduction of English westwards along the peninsula), in any given location, the vowel quality is the same for both lexical sets. In traditional varieties of Cornish English, then, the TRAP/BATH split is marked only by duration, with BATH vowels being longer than TRAP vowels.

Although the historical metalinguistic commentary emphases dialect contact with StEE speakers via the elite networks provided by the islands’ governors, there has been on-going Cornish immigration into Scilly over time. For instance, in the 1901 census, almost half (48.3 per cent) of Scilly’s 571 migrants were from Cornwall (data from The Isles of Scilly Museum 2007). This opens up different possibilities for our Scillonian speakers. Given that, historically, some of them had access to education in boarding schools on the mainland (and Cheshire and Trudgill Reference Cheshire, Trudgill, Cheshire, Edwards, Munstermann and Weltens1989: 95, amongst others, consider boarding schools to be one of the main places in which Standard English is cultivated), we might expect mainland-educated Scillonians to show a more StEE-like pattern than Scilly-educated Scillonians. Furthermore, given that sociolinguistic studies have consistently shown that women lead men of the same socioeconomic group in the use of standard language features (so much so, that Holmes [Reference Holmes and Coates1997] refers to this as a ‘sociolinguistic universal’), we might expect mainland-educated women to show the most StEE-like pattern in our dataset.

To examine these patterns, we analysed a sample of data from a group of Scillonians born before the islands’ own secondary school opened in 1966. This data was obtained from interviews from the Isles of Scilly Museum’s Oral History Archive. This archive contains recordings made by local people interviewing other local people. The archive recordings date from the 1970s through to the present day. The purpose of the archive is to record island history, as told by the individuals who experienced it. Interviewees were selected by museum volunteers on the basis of their status as a ‘Scillonian character’ (a vague definition, but one which includes consideration of island heritage and community roles). Table 12.1 provides information on the participants used in our analysis.

In order to test the quality of the TRAP and BATH vowels used by these speakers, we extracted formant data, sampling formant tracks every 5ms through each vowel, with LPC order set to appropriate values for each speaker. This allowed us to obtain the median value for each formant in the vowel. We focus on F1 and F2 in our analysis, given that F1 has been found to correlate with vowel height, and F2 with how front or back a vowel is (see Ladefoged Reference Ladefoged1982, amongst others). We also measured the duration of each vowel, which we transformed into a logarithmic domain to account for the fact that hearers seem to perceive durations as ratios rather than as absolute amounts. This kind of transformation makes sense statistically because logarithmically transformed durations also approximate the normal distribution more closely. We also transformed our formants into the domain of the equivalent rectangular bandwidth (Glasberg and Moore Reference Glasberg and Moore1990), as a step towards speaker normalisation. For clarity of presentation, axis notation is provided in our figures in Hertz for spectral data and milliseconds for temporal data.

We coded for a number of phonological environment factors, namely: the position of the syllable in question within each utterance (initial, final or somewhere in the middle of the utterance), the number of syllables in the word, whether the rhyme of the syllable was open or closed, the manner of articulation, the voicing of the following consonant, and whether the syllable seemed to be carrying sentence stress. We also coded for lexical versus grammatical words. In order to focus on the social patterns in our data, we do not deal with these phonological environments in this paper – only to note that the social patterns we discuss below stayed robust when linguistic factors were included in statistical models of the data (see Moore and Carter [Reference Moore and Carter2015] for a more comprehensive discussion of the linguistic predictors of TRAP and BATH in this dataset). In the next section we describe the outcome of our analysis.

4. Results

Figure 12.2 shows a series of density plots of our raw formant data, according to education type and gender. Density plots are like contours on a map. Peaks show areas of greater density where formant measurements cluster. TRAP vowels are shown in grey and BATH vowels in black. Figure 12.2 reveals variation amongst our Scillonian speakers, and suggests effects of both education type and gender. As expected, the mainland-educated speakers seem to have a TRAP/BATH pattern that is more like present-day StEE. Their TRAP and BATH vowels are more distinct from one another, and they appear to have more of an F2 difference between these lexical sets; that is to say, the BATH vowels seem to be further back than the TRAP vowels. On the other hand, the Scilly-educated speakers’ plots show much more similarity between TRAP and BATH, with the two vowel clusters largely overlapped. With regard to education type, these plots suggest that mainland boarding school education may indeed correlate with more StEE-like patterns of TRAP and BATH in our sample of Scillonian speakers.

Figure 12.2: Density plot showing raw formant data for TRAP and BATH, according to education type (horizontal axis) and gender (vertical axis).

The effects of gender are less easy to deduce from these plots. Whilst it seems that the Scilly-educated women have a more StEE-like pattern than the Scilly-educated men (the centre of their BATH density plot has a lower F2 value than that of their TRAP density plot, whereas the distinction is much less clear for the Scilly-educated men), it is not clear that the mainland-educated women are leading the mainland-educated men in having the most StEE-like pattern. The mainland-educated women’s TRAP/BATH split does not appear to be any more extreme than that of the mainland-educated men – indeed, a closer look at Figure 12.2 suggests that it may be less so.Footnote ² Furthermore, there appears to be less differentiation in the patterning of the women overall in our Scillonian sample. That is to say, Figure 12.2 suggests that it is men who define the envelope of variation in this community: the mainland-educated men seem to have more of a distinct TRAP/BATH quality split than any other group, and the Scilly-educated men seem to have a less distinct TRAP/BATH quality split than anyone else.

In order to test the patterns suggested by the raw data in Figure 12.2, we ran random forest variable importance measures on the data. Strobl et al. (Reference Strobl, Boulesteix, Kneib, Augustin and Zeileis2008; Reference Strobl, Hothorn and Zeileis2009) provide an explanation of this type of modelling. Put simply, this technique models the relative importance of predictor variables in explaining the TRAP/BATH split in our data. This modelling allowed us to add duration as a predictor variable alongside F1 and F2. The results of this analysis are shown in Figure 12.3, which shows the variable importance plots predicting whether a vowel is TRAP or BATH for the data from each of the speaker groups listed in Table 12.1. Black columns indicate that a factor is a significant predictor and, the larger the bar, the more significant a factor is.

Figure 12.3: Variable importance plots predicting the relative importance of F1, F2 and duration in the TRAP/BATH split for each speaker group.

Figure 12.3 shows that F1 is not very important for predicting whether a vowel is TRAP or BATH, although it does have a small effect on the data of all of our speaker groups, with the exception of the mainland-educated females (the F1 bar here is insignificant, but the bar is so small that it is not possible to see its colour clearly in the figure). On the other hand, F2 and duration are clear predictors of whether a vowel is TRAP or BATH for all of our groups. However, these two factors do not pattern in the same ways across our speaker set. F2 is the most significant factor predicting whether a vowel is TRAP or BATH for everyone except the Scilly-educated men. For the mainland-educated males and females, and the Scilly-educated females, duration matters, but it is less important than vowel quality in differentiating TRAP and BATH, and it is much less important than F2 for the mainland-educated men. On the other hand, this pattern is reversed for the Scilly-educated men; duration is the most significant factor in predicting whether a vowel is TRAP or BATH for this group. F2 also matters for these speakers, but it is less important than duration.

In sum, the results shown in Figure 12.3 seem to support the patterns evident in the raw data. However, they elaborate on these results in two key ways. Firstly, Figure 12.3 suggests that the Scillonian men have oppositional patterns of language use. Their patterns of F2 and their patterns of duration work to create distinction – where one group uses more of one of these factors to mark the TRAP/BATH split, the other group use more of the other. Secondly, and in contrast to the men, the two groups of Scillonian women are much less different from one another – both groups seem to be moving towards a present-day standard-like norm, irrespective of education type, and there is much less differentiation in their linguistic patterning, as shown in Figure 12.3. We attempt to explain these patterns in the next section.

5. Discussion: The Social Meanings of TRAP and BATH

The results in Section 4 suggest that women and mainland-educated men use more StEE-like forms and Scilly-educated men use more Cornish-like (regional vernacular) forms. Previous research has suggested that speakers use vernacular forms to index local identity. This was shown in Labov’s (Reference Labov1963) Martha’s Vineyard study, and has been demonstrated in subsequent work many times since then (see, for instance, Schilling-Estes’ [Reference Schilling-Estes1998] research in Ocracoke; Johnstone et al.’s [Reference Johnstone, Andrus and Danielson2006] work on Pittsburgh; and Wong and Hall-Lew’s [Reference 280Wong and Hall-Lew2014] San Francisco study). Other research has suggested that increased use of standard language forms reflects orientation away from the local and towards less traditional or more global language norms. For instance, Holmquist’s (Reference Holmquist1985: 199) study of a rural Spanish village discovered increased use of Castilian Spanish being driven by ‘a general turning away from things rural’. These findings echo those in Gal’s (Reference Gal1978) study of a rural Austrian village a decade earlier, where increased use of German (the more prestigious code in this bilingual context), was most frequent amongst those who oriented away from the community’s traditional peasant roles and values. In both Holmquist’s and Gal’s studies, it was women who drove language change, given that, socially and economically, they suffered most in the rural economies of the locations studied. On the other hand, the men who controlled the local rural economy resisted language change the most.

It is tempting to apply these findings to the Scillonian context: we might assume that (i) the Scilly-educated men are resisting StEE-like forms to reflect and construct their stake in the local economy, whereas (ii) women and mainland-educated men are driving change towards the standard as a reflection of their orientation towards island-external prestige norms. Whilst the first part of this explanation may hold true, the second part may require further elaboration. Firstly, Scilly’s unique sociocultural history suggests that it may be problematic to assume that the StEE-like patterns found on the island reflect a straightforward orientation away from the islands. Secondly, Scillonian women use StEE-like forms, but, unlike the women in Holmquist’s and Gal’s studies, it is not straightforwardly the case that they have the most advanced use of these forms on Scilly.

With regard to the first of these points, we refer back to the discussion in Section 2. This described a history of metalinguistic commentary linking Scillonian English and StEE. As can be seen in the quotations in (1), (2) and (3), this discourse typically juxtaposes Scilly with Cornwall, and Scillonian English with Cornish varieties of English. In this commentary, speaking in a style which more closely approximates StEE is a defining characteristic of being Scillonian according to the historical record. Consequently, one could view the use of this style as a claim regarding ‘inheritance and education’ (Banfield Reference Banfield and Urban1888: 54) and, as such, as a claim for a particular kind of Scillonian authenticity. In this sense, use of StEE-like forms may not solely indicate orientation to island-external or global language norms; it may also indicate orientation to a particular (and historically substantiated) Scillonian social type. That this is possible is further substantiated by the fact that the context of the interview itself focuses our mainland-educated speakers upon what they consider to be the important aspects of Scillonian life. That is to say, their discourse in the interview is focused on all matters Scillonian, not island-external concerns.

On the other hand, the oppositional behaviour of the Scilly-educated men suggests that they are orienting to a quite different Scillonian identity type, which is associated with different forms of status. This can be seen if we compare the roles and responsibilities undertaken by our male speakers over their lifetimes, and the dominant topics that these speakers discussed in their interviews. This information is given in Table 12.2, which suggests two distinct life trajectories for men according to education type.

Whereas the men who were educated on the mainland tended to end up being managers of their own businesses and taking on authoritative positions in the community, such as being councillors or magistrates, the men educated on Scilly tend to find employment in seafaring activities or local trades. The differences in roles and responsibilities are also highlighted in the topics the men discuss in their interviews. As noted in Section 3, the purpose of the oral history interviews was to reminisce about Scillonian life. Consequently, the topics discussed in the interviews offer a window on how our interviewees characterise Scillonian status and construct it in their talk. In line with their employment, the most frequent topic of conversation for the mainland-educated males is their management responsibilities. These might be responsibilities to do with their own businesses (Victor and Jim managed their own farm and also had various management roles in the islands transport company, and Ted managed a butcher’s shop and a hotel), or authoritative roles such as being local councillors or magistrates. The Scilly-educated men, on the other hand, are more likely to talk about local history (that is events and occurrences on the islands) linked to their employment on the islands or at sea. For instance, Stan ran an inter-island boat service during the wars (navigating the treacherous seas around the islands in difficult conditions) and he tells stories about this time, and Luke was a fisherman and served as the lifeboat coxswain for many years, hence most of his interview is concerned with stories about lifeboat rescues.

The two different life trajectories experienced by male Scillonians prioritise different types of knowledge. Whereas the mainland-educated men achieve institutional status following their mainland schooling, there is evidence that at least some of the Scilly-educated men are strongly resistant to the idea that education gained elsewhere is more important than practical, local knowledge. In particular, the archive materials include evidence of some islanders expressing concern about too much intervention from outside the islands. This can be seen in the transcript in (4), which comes from a news report featuring a Scilly-educated Scillonian, Luke, warning against outside influences on the development of a new runway on the islands.

In (4), Luke stresses that islanders like him have local knowledge and know what is best for the islands. This is articulated by positioning himself in opposition to mainland development companies (ll.1–2), and by opposing the ways in which mainland coastal locations have developed (l.2–5). Luke stresses Scilly’s uniqueness (l.8–10), insinuating that it requires people like him to protect the islands from outside influence. Whilst Luke does not mention education in this extract, the implication is clear: knowledge about what is best for the islands comes from having a situated understanding of Scilly, and this is not necessarily achieved via formal education.

The behaviour of the men in our sample suggests that male Scillonians have adapted variants of TRAP and BATH to construct oppositional local identity types: one concerned with education and aspiration, and one concerned with local island knowledge. Britain and Trudgill (Reference Britain and Trudgill1999: 247–8) refer to this kind of adaptation as ‘socio-stylistic reallocation’. It is evidenced here by the oppositional TRAP/BATH patterns in the male data and (i) the differences in the roles and responsibilities of the two groups of Scillonian men, (ii) the likely differences in social practices that such activities entail, (iii) the matters with which the men are concerned in the topics of their interviews and (iv) in other evidence available in the archive, such as the extract in (4).

Turning now to our female speakers, we see that there are far fewer opportunities for women to gain explicit status within the community. Table 12.3 provides information about the roles and responsibilities undertaken by our female speakers over their lifetimes, and the topics that these speakers discussed in their interviews. Comparing the two groups of women reveals that, whilst Scilly-educated women tended to have more jobs than mainland-educated women, both groups of women undertook employment in the flower and farming industries, irrespective of their educational background or the other activities in which they were engaged. In practice, this meant that women were recruited to work in their husband’s or wider family’s main lines of work, by tying and packing flowers, in addition to undertaking all of the domestic responsibilities of the home, and any other form of employment they undertook. Furthermore, unlike the men, these women did not have roles and responsibilities in the community that brought them into conflict with each other. In fact, as their list of roles and responsibilities suggest, they were more often working alongside each other sustaining island industries.

This is not to say that these two groups of women are entirely homogeneous. The topics shown in Table 12.3 reveal some differences between the groups. Like their male counterparts, Scilly-educated women talk about local history. However, they do not talk about their roles or responsibilities in relation to this history as the Scilly-educated men do. Instead they tend to talk about people, that is, island folk and their families. Sometimes this is in relation to a specific historical event; for instance, Tess was asked by her interviewer to focus on the period around the World Wars. Whilst mainland-educated women also focus on local history, their emphasis is on horticultural matters – they either talk about their own interests in gardening, or they talk about the flower farming industries managed by their husbands or fathers. For instance, Margaret talks about the varieties of flowers farmed on the farm managed by her husband, and Elizabeth talks about the way flowers were packed on the farm managed by her father. These slight differences in topic reflect that these two groups of women have different histories, concerns and priorities.

Nonetheless, Table 12.3 suggests neither group of women is acknowledged as having visible institutional roles in this island community. This is not to say that they do not contribute significantly to community life, but that their topics are not driven by their roles and responsibilities, because their roles are not acknowledged historically in the way that those occupied by men are. Women may sustain many of the island industries but none of these women were councillors, or chairs of local businesses; none of them served on the lifeboat crew or ran local boat services. This suggests that the type of education that women had gave them different life experiences, but it did little to affect how they are recognised within the community.

This is exemplified in the extract in (5), which provided the title of this chapter. It is taken from a story told by one of the Scilly-educated women, Tess, who is in conversation with another Scilly-educated woman, Gloria. They are talking about farmworkers’ wives having to tie flowers for their husbands’ employers.

The extract outlines the roles of women in the farming industry. Women are expected to work in the same industries as men, although this is often characterised as ‘supporting’ their menfolk, rather than perceived as work in its own right. Women may be unable to refuse to take on this work (ll.13–15), no matter what other responsibilities they have (ll.16–23). Notably, however, women are not allowed to be ‘farmers’ (l.31) – no matter how well educated they are, or if their deceased relative had served as a manager (ll.33–34).

The ‘resilience of the Scillonian woman’ is an ideological trope which has endured across time. The extract in (6) is taken from a 2014 interview with two young Scillonian women who were also discussing the role of women in the farming industry.

Much like the description in (5), in (6), Kate describes a situation in which women take on multiple jobs, including skilled work such as nursing. Interestingly, Kate does not perceive there to be inequity between men and women (l.31), despite defining women’s roles in farming as ‘helping’ (l.26), rather than defining women as ‘farmers’, or workers in their own right. This provides further evidence to support the observation that education or training have few practical consequences on how hard or how much a woman is required to work, or how much acknowledgement she receives for this work. Whereas men get to define their Scillonian status on the basis of their education or their local knowledge, Kate characterises the true Scillonian woman as someone who ‘get[s] in with everything’ (l.33), despite receiving little explicit status in return for her labour.

Several sociolinguistic studies have demonstrated how women’s language use may be conditioned by their access to particular ‘linguistic markets’ (Bourdieu Reference Bourdieu1977; Woolard Reference Woolard1985) and their ability to gain status within those markets (Milroy Reference 279Milroy1980; Eckert Reference Eckert2000). The market in which the Scilly-educated men operate is largely inaccessible to women by virtue of the sharp gender segregation in many of the activities undertaken by the Scilly-educated men. For instance, women do not engage in boating and fishing. These activities entail engagement with the tourist industry (tourists are taken on fishing trips and on pleasure trips from St. Mary’s to Scilly’s other islands), and they also bring Scilly-educated men in contact with other seafaring men from Cornwall, who use Scilly as a port. It is possible that contact around these shared practices may serve to reinforce, or at least support, the Scilly-educated males’ Cornish English-like TRAP and BATH patterns. Dubois and Horvath (Reference 278Dubois and Horvath2000) observe a similar pattern of language use in the Cajun community where young men (but not young women) were found to ‘recycle’ traditional Cajun English features which were previously in decline. Dubois and Horvath (Reference 278Dubois and Horvath2000: 306) attribute this ‘to the fact that the Cajun Renaissance has largely affected the sphere of traditional male activities, such as boating, fishing and hunting, and the display of Cajun culture associated with tourism’. Of course, this is not to say that the Cornish-like variants of TRAP and BATH are ‘male’ in any direct sense. Rather, they have acquired social meanings that are linked to characteristics that are more commonly associated with men than with women. That is to say, their social meanings are associated with styles of speakers (personae) and social stances which occur more frequently in the discourse of Scilly-educated Scillonian men.Footnote ⁴

Many of the activities undertaken by the mainland-educated males are also largely inaccessible to women. In this generation, women are not councillors, managers or magistrates. However, the kinds of social meanings linked to these activities (‘educated’, ‘discerning’) index the characteristics of the historically dominant Scillonian identity type, as articulated in the metalinguistic commentary on the islands. This might explain why we find both groups of Scillonian women using relatively more standard-like variants of TRAP and BATH, irrespective of their social and educational background. StEE-like variants of TRAP and BATH index a Scillonian identity that is also in line with ideological expectations about female behaviour.

The ideologically loaded nature of TRAP and BATH may also help to explain why women are more constrained in their ability to use these particular linguistic forms to signal intra-gender differentiation. In Section 3, we discussed how the backing of the BATH vowel in StEE happened very gradually, and may not have been entirely complete until the beginning of the twentieth century – that is, at the very time that our speakers were born and acquiring their language variation. As Beal (Reference Beal2004: 141) has observed, when this change was in progress ‘those who aspired to “correct” pronunciation in England had to steer a very narrow course, avoiding both the “broad” [ɑ:] and the “mincing” [ӕ]’. Tracing pronunciation guidance over this period, Mugglestone (Reference Mugglestone2003: 78) notes that these pressures were particularly strong on women, for whom a less than careful enunciation may have led to their speech being perceived as ‘“inaccurate”, “vulgar” or indeed “uneducated”’. Perhaps unsurprisingly, then, both Ellis (Reference Ellis1869 iv: 1152) and Ripman (Reference Ripman1906: 55), writing at the turn of the twentieth century, describe an ‘intermediate’ sound as more typical of female speakers.

Figures 12.2 and 12.3 suggest that our female speakers have a TRAP/BATH pattern that is situated in between that of our two groups of male speakers; that is to say, they have an intermediate pattern for these vowels. What we see for the Scillonian female speakers, then, may be sensitivity to the historical ideological pressures on TRAP/BATH usage, given the incredible salience of these particular vowels. That is to say, the female speakers’ patterns reflect an orientation to a conservative and old-fashioned standard norm of pronunciation – one where TRAP has a quality distinct from BATH, but not too distinct as to risk being misinterpreted.

A counter argument to this might be that the behaviour of the mainland-educated men indicates that the uncertainty about the social meaning of backed BATH variants has passed for this generation of speakers. However, this would assume that both men and women are responding to the same linguistic norms, and that they are in the same stage of language change. It is not necessarily the case that something that is presumed ‘vulgar’ in female practice, necessarily carries the same evaluation when observed in male practice. As Eckert (Reference Eckert1989) has argued, whilst sociolinguists have tended to evaluate language use in terms of oppositional gender categories, the effects of gender on language variation are not necessarily consistent across gender groups. This is likely to be particular true in communities, such as the one considered here, that demonstrate gender segregation in social and cultural roles.

It is important to stress that, although women behave more homogeneously than men with regard to their use of these variables, we do not wish to imply, as Jespersen (Reference Jespersen1922: 258) once did, that women are simply more linguistically alike than men. The nature of the variables we have analysed in this chapter is key. It is probable (and indeed likely, as our current research on other variables in this community is suggesting) that women use other, less ideologically loaded, linguistic features to mark intra-gender differentiation. With regards to their use of TRAP and BATH, female Scillonians’ linguistic behaviour is constrained by the following factors: (i) the peculiar (and especially salient) status of TRAP and BATH in English, (ii) the link these variables have with ‘education’ and associated characteristics, (iii) the indexical links between ‘education’ and the historically dominant delineation of the ‘Scillonian’ and, finally, (iv) the ideological constraints on women to conform to this ‘educated’ social type, and avoid linguistic behaviours that could be negatively evaluated.

6. Conclusion

In this chapter, we have explored how life trajectory and gender interact in a small island community to affect how men and women use variables of the TRAP and BATH lexical sets. Our results showed that men defined the envelope of variation for these vowels. This was explained as a consequence of different educational experiences, which resulted in divergent life trajectories and the construction of two oppositional Scillonian identity types. The women in the community were found to show less differentiation in their linguistic variation for these vowels. Women’s linguistic behaviour was explained as a consequence of constraints on women’s sociolinguistic behaviour, irrespective of education types, limitations on the kind of Scillonian identities available to them, and ideological pressures to conform to gendered expectations about language use.

In the course of our analysis, we have questioned the assumption implicit in much sociolinguistic work that use of standard language features straightforwardly reflects orientation to norms external to the local community being studied. The complicated social history of the Isles of Scilly provides a context in which standard language forms take on local meanings linked to local identity types. This is not to say that the StEE-like variants of TRAP and BATH found on Scilly do not carry traces of social meanings available beyond the islands – indeed, our analysis has shown that the ideological baggage associated with variants of these vowels affects how they are used by different groups of speakers. Our point is that their use on Scilly expands and adapts their precise indexical values to reflect the social relations and histories of the islanders. In the context of their use on Scilly, the ‘local “palimpsest”’ effect (Lass Reference Lass1976: 268–9) on these variants means that they are not just StEE-like forms, they are Scillonian forms.

Our study has suggested that there are multiple ways to index a Scillonian identity. These different identity types reflect different experiences of the local environment and different entitlements to local space and resources. Whilst it is tempting to find ways to ‘distil’ the essence of a particular location, our research suggests that place takes on a range of meanings linked to alternative life trajectories. We have found this to be true even in a very small, rural, island location. The Isles of Scilly meet many of the criteria of an isolated community (see Montgomery Reference Montgomery2000; Schilling-Estes Reference Schilling-Estes2002: 65; Wolfram and Schilling-Estes Reference Wolfram and Schilling-Estes2006: 174–8); they are geographically remote, they exhibit historical continuity in their population, and they are relatively autonomous from the mainland in terms of their governance. Nonetheless, they exhibit sociolinguistic heterogeneity, linked to different claims about Scillonian identity. Geographical isolation may be less important than how open or exocentric community members are, as Andersen (Reference Andersen and Fisiak1988) has observed. Schilling-Estes (Reference Schilling-Estes2002: 77) has observed that the need to mark out very local distinctions may actually mean that some geographically isolated communities actually support heterogeneity better than less isolated communities. This seems to be the case in Smith and Durham’s (Reference Smith and Durham2011) study, which shows divergent language use in one very tight-knit community in Shetland, and our data also suggests that this could be true. However, it may just be that research on geographically isolated communities has tended to more closely interrogate the social criteria by which speakers are identified, given that these communities less easily conform to the hierarchical social models used in sociolinguistic work. Perhaps all communities are more heterogeneous than has been assumed, and we are still in the process of findings ways to adequately conceptualise how individuals inhabit space and create meaning in relation to it.