Buckie in the Past

The community of Buckie has its roots as a fishing settlement of the eighteenth century, and the population grew steadily from that time onwards. A fishing boom in the nineteenth century meant that the coastal settlements along the Moray Firth grew dramatically. Despite this growth, Buckie was a close-knit community, where social ties were dense and multiplex (Milroy Reference Milroy1980). In his study of north-east fisher families in Victorian Scotland, Blaikie (Reference Blaikie2002: 22) points out that while other ‘Scottish communities were sufficiently open to allow considerable migratory traffic in marriage partners and incoming families’, this was not the case in the fishing villages of the north-east coast. As a result, ‘the frequency of particular surnames within fishing communities indicates pronounced genetic concentration and reflects high levels of marital isonymy (marriage between individuals with the same surname)’. This is best demonstrated by the village of Portessie, just outside Buckie, where 41 per cent of the population were Smiths and the remaining 59 per cent shared just four other surnames (Blaikie Reference Blaikie2002: 23). Such statistics demonstrate that in these small fishing communities on the north-east coast, ‘their discrete social world maintained a hermetic family culture that brooked no incursion from the surrounding hinterland’ (Blaikie, Reference Blaikie2002: 29). Moreover, in these close-knit fishing communities, there were few class-based divisions of the type we think of in the present day. The New Statistical Account of Scotland for Rathven Parish (of which Buckie was a part), states that ‘the fishermen, as individuals, are nearly placed upon a footing of equality. Their pursuits are similar; hence their language and dealings are almost the same’ (Gordon Reference Gordon1845: 256).

Buckie in the Present Day

It is no surprise to find that Buckie has changed significantly over the last hundred years, embracing many of the innovations that have arisen over the subsequent decades. However, unlike many other rural areas in more recent history, the population of Buckie has remained stable; Figure 2.2 charts census data outputs from 1910 onwards.

2.2 Population statistics for Buckie, 1910–2011

(adapted from University of Portsmouth et al. 2016)

The stability of the population may come from economic independence: first, in the form of the fishing industry and then, with the decline of fishing, the rise of employment in the oil industry from the 1970s onwards. This employment has meant that working-age adults, and particularly the young, do not need to leave the area to find employment (although this situation might change, given the dramatic decline in North Sea oil in the past few years). Further, this stability may be indicative of social stability more generally. The 2011 Census showed that Buckie has a total population of just over 9,000, with 99.3 per cent ‘white’ and 90.6 per cent ‘white Scottish’. Of the population, 3.1 per cent were born outside the UK, and the majority of the non-Scottish population are from England (5.0 per cent). These statistics provide a snapshot of the demographics of Buckie: a small community with a tightly circumscribed ethnic base and a continuing sense of close-knit networks. It has certainly moved on from its Victorian past, but it retains a level of sociocultural cohesiveness that may be quite unusual in the twenty-first century. This is perhaps best demonstrated by the language used in the community.

The Linguistics of Buckie

The majority of speakers in Scotland speak Scots, which originally derived from Old Northumbrian English. In the present day, the term ‘Scots’ encompasses a wide spectrum of varieties, with so-called Broad Scots at one end and Scottish Standard English – a variety generally described has being ‘standard English with a Scottish accent’ – at the other. The variety used in Buckie, and in this north-east corner more generally, is known as Doric and is at the broad end of the linguistic spectrum of Scots. Despite the fact that ‘predictions of its demise continue apace, the “Doric” defiantly persists as an integral part of the region’s identity and self-image’ (McClure Reference McClure2003: 15). This statement is very much borne out in practice: step out in the streets of Buckie and you will hear a very distinctive variety of Scots, as demonstrated in (19) from Poppy, a seventy-one-year-old speaker in the community, who is telling our fieldworker, Rose, about New Year’s Eve (in Scots, ‘Hogmanay’).

1. (19)

   Rose: In that day, your father wouldna’ve gien to a pub, would he have?

   Poppy: No, your father di— Now Hogmanay the wifies got a wee sherry in a wee sherry glass. And the mannies got a whisky. That was your Hogmanay.

   Rose: That was it.

   Poppy: That— there was nothin’ in atween.

   Rose: No.

   Poppy: A sherry or a whisky, that was fit— That was it and you used to hae oot a wee spread. Maybe a bit o’ shortbread or a sponge that your mother had made. You used to sit and take the bells in. Ken this, fan I think aboot it, abody was happy.

   Rose: Aye.

   Poppy: We’d all be whooching it up and obviously you n— you tried the sherry, like, fan you was a bairn.

   Rose: Aye.

   Poppy: If you were caught, God help you.

   Rose: Oh me, Poppy!

This short extract is full of forms from older stages of Scots and spans lexical, phonological and morphosyntactic variation. The extract includes lexical items such as ken for know, wee for small and wifie for woman, the negative clitic na for n’t (as in your father wouldna’ve gien to a pub) and the prefix a- for be- (in there was nothin’ in atween). To add to historical retention, there is innovation: the isolated nature of the north-east led to the development of a number of features that are not found elsewhere in Scotland, as exemplified in the use of fan for when (in fan I think aboot it) and its cohort of other wh- question words (including fit for what and fa for who). Thus retention and innovation have gone hand in hand in providing a highly differentiated variety at the lexical, phonetic and grammatical levels of language use.

The linguistics of this community hold a mirror to the larger context of development over the past hundred years, and Buckie’s past is imprinted in its linguistic present. At the same time, Buckie is not completely immune to change. While many new so-called ‘diffused forms’, such as th-fronting and /l/ vocalisation (e.g. Kerswill Reference Kerswill, Britain and Cheshire2003) remain absent in the dialect, despite widespread use elsewhere, other supralocal forms, such as the be like quotative in (20) from thirteen-year-old Griff, and glottal replacement, as in example (21) from twelve-year-old Gillian, have spread rapidly in this community:

1. (20)

   Griff: I asked Mum ‘Fit did you do without a mobile phone?’ and she was like ‘Well I play my— eh I played Space Invaders’ or something. I was like ‘Mam, get with the times’.

1. (21)

   Gillian: Well I ken how to do i[ʔ] on the iPhone bu[ʔ] I wouldna ken how to do it— like tell you fae tha[ʔ] … but there’s usually like a li[ʔ]le bu[ʔ]on in your se[ʔ]ings and then it’s like your Facebook contacts but then you can like change i[ʔ].

Against the backdrop of high mobility, urban growth and loose networks, this rural area remains largely monoethnic, with a stable population and very close-knit ties. In many ways, this community is quite unrepresentative of twenty-first century life. However, we argue that such a community provides an excellent test site for examining the mechanisms involved in the acquisition of structured variation at the earliest stages exactly because of its circumscribed make-up. As demonstrated in Chapter 1, we are still very much at the early stages in establishing how and when children acquire variation. Buckie thus provides an appropriately controlled research site for examining the processes underlying sociolinguistic development in these early stages.

We turn now to the data used in the present study, beginning with the key corpus of caregiver/child speech.

Recruiting Participants

The first task of recruitment was greatly facilitated by the fact that the first author, Jennifer, is from Buckie. Most of her family still live there, as did her mother, Moira, who agreed to be fieldworker in the crucial phase of data collection. Moira’s involvement meant direct access to this tight-knit community, which circumvented the need for ‘community gatekeepers’ or ‘brokers’ (e.g. Schilling-Estes Reference Schilling-Estes, Bayley and Lucas2007). Moreover, Moira was an extremely well-known figure in the community, which meant that the potential difficulty of, for example, ‘cold-calling’ was circumvented. Using the criteria for selection noted below, she identified potential participants and contacted them initially by telephone. She then made a follow-up face-to-face visit to each caregiver to talk them through what the recordings would involve. This approach proved to be invaluable in persuading caregivers to take part, with 91 per cent of those approached agreeing to participate in this study.

We used judgement sampling (e.g. Chambers Reference Chambers2003: 44), in this recruitment phase, as we had a checklist of predefined criteria for the participants. In order to control the sample as much as possible, we started with a highly elaborated list of criteria for inclusion in the study, including, for example, no older siblings. However, as with most studies of this type, we had to remove a number of constraints in order to recruit enough participants.

We focused in the end on four main criteria:

1. (1) both parents must have been born and raised in the area

2. (2) no referrals for speech or language therapy

3. (3) the mother was the main caregiver (i.e. there was no substantial time with extended family, childminders, etc.)

4. (4) no child in formal, state-run nursery education.

These criteria were imposed because they have implications for the ambient language to which the child is exposed and how this might impact on language production. For example, previous studies have found that parental input impacts a child’s ability to acquire a particular dialect (e.g. Payne Reference Payne1976; Kerswill & Williams Reference Kerswill, Williams, Jones and Esch2002), while the more formal language used in institutional settings such as formal education may also have an impact (e.g. Romaine Reference Romaine1984; Patterson Reference Patterson1992; Chevrot et al. Reference Chevrot, Beaud and Varga2000).

We also wanted to capture children’s vernacular use at a crucial stage of language development, and ‘the 3- to 4-year old level is a critical period for the acquisition of dialectal norms of the speech community’ (Roberts & Labov Reference Roberts and Labov1995: 110). In addition, while most studies of language acquisition start where language starts (i.e. in the first months, when the child moves from babbling to the one-word phase), we needed data that would allow us to investigate a range of variables, from individual phonemes to use of whole clauses. For this reason, we targeted children between the ages of 2;6 and 4;6, with the final sample ranging from 2;10 to 4;2. Data collection ran from December 2003 to June 2004.

Table 2.1 shows the caregivers and children in this sample.

The Recordings

Studies of first language acquisition are often longitudinal, consisting of a number of recordings over a set period of time. In this study, we chose to gather cross-sectional data – recordings of each speaker pair across one time period. In doing so, our aim was to gain a ‘snapshot’ of caregiver/child interaction across a range of ages. Further, in much of the previous research of preschooler language, the researcher has been present during the recordings. Crucial to our endeavour was obtaining data that replicated as far as possible the everyday interactions of caregiver and child. Having the researcher present, even though in this case they are from the community, would highly compromise this endeavour, as issues of the Observer’s Paradox (Labov Reference Labov1972) are even more acute where preschool children are concerned; the presence of an ‘unknown’ observer in close proximity may result in a child being ‘speechless’. This was clearly demonstrated in cases where the first author visited some participants to check on progress: despite her attempts to engage in conversation, the lips of the child were clamped tightly shut, only to be opened to utter I‘m nae speakin’ to you, or some other equally withering response. This meant that we needed to come up with a data collection technique that would allow us to capture more felicitously the everyday speech patterns between parent and child. To do this, we provided the caregivers and children with lightweight recording equipment (Sony MZ-R700 MiniDiscs and Sony ECM-T145 lapel microphones) and asked them to record themselves in a range of situations over a one-week period. Each caregiver and child had a separate recorder and lapel mic, which meant that they were not constrained to sit in one place for an hour’s recording or to be in close proximity to each other at all times. The lightweight recorders were placed in the child’s trouser or coat pocket, and the lapel mic was placed under clothes to ensure that it stayed in place.

We asked that the recorded interactions total ten hours. This may seem like quite a lot, given that sociolinguistic research typically aims for between one and two hours of recorded speech during a sociolinguistic interview (Labov Reference Labov, Baugh and Sherzer1984), but simply put, kids talk less. Roberts (Reference Roberts, Chambers, Schilling-Estes and Trudgill2002: 336), for example, estimated that eight to fourteen hours of young children’s speech is required to collect data comparable to that found in one to two hours of adult speech. We were not prescriptive as to what activities the caregivers and their children should engage in, and we stressed that they should not do anything ‘special’ for the recordings, but merely carry on their everyday lives. This, of course, is partly a fiction – in everyday life it is unusual to be hooked up to a recorder – but what we were aiming for was the closest possible approximation to caregiver/child interaction and the language that arises from this.

The Challenges of Data Collection with Young Children

As with all data collection, a number of problems arose, exacerbated by the fact that this procedure had not been tried and tested to any extent in previous research. The first problem was time: although we aimed to have each participant complete the recordings within one week, in reality most pairs took longer, with an average of two weeks’ collection time. The majority of pairs completed all ten hours of recording, but this was not without its issues, as we will discuss.

The next problem was that some children simply did not want to wear the recording equipment, as shown in (22) and (23):

A range of creative tactics was employed by the caregivers to attempt to persuade the children to wear the equipment, from general cajoling to more overt bribes, as shown in example (24):

Once the recording equipment was in place, another problem arose. We had suggested that the recorders be carried in a trousers or coat pocket, but we had not foreseen the weight of the recorder as an issue (25–26):

The microphones were also dropped, pulled and even eaten (27):

It has to be said however that many children positively loved the novelty of being recorded, as demonstrated in (28–30):

A more serious problem arose with regard to the quality of the data. Our request that the participants carry on as normally as possible resulted in some of the recordings not being of high enough quality for fine-grained linguistic analysis. For example, a walk along the cliffs in this north-east corner of Scotland in high winds meant that the voices were all but obscured by the inclement weather. Some of the bath-time recordings were similarly compromised, with the splish-splash of water drowning out conversation. In the end, for each caregiver/child pair, we selected five hours of speech in which the quality was good enough from the ten hours available in order that all speakers had the same amount of data in terms of length of recording.

In sum, caregivers and children go for walks in the wind, take baths, eat baked beans. This is the language of the home, not of the laboratory, and this methodology decisively captures the diverse linguistic contexts of use that arise naturally between caregiver and child in everyday interaction. Our next task was to transcribe the data.

Transcription

To process the data, we created time-aligned transcripts of each sound file using CLAN (MacWhinney Reference MacWhinney1991), a computer programme designed to facilitate the transcription of child language data (Figure 2.3).

2.3 Screenshot of transcription in CLAN

A strict transcription protocol was followed for all recordings, which laid out the conventions for representation of the speech data. We did not attempt to represent at this first pass every nuance of variation found in the data, as this would effectively mean carrying out multiple analyses at the same time as attempting the already onerous task of transcription. Instead the transcription protocol represented faithfully the local lexical and morphosyntactic differences found in each pair’s interactions (e.g. in Figure 2.3, how for why), but used standard English orthography for more fine-grained phonological differences. For example, the lexical item sweetie in the transcription in Figure 2.3 appears as both swee[t]ie and swee[ʔ]ie in the actual recording. The same is true of -ing forms; many of these appeared with the alveolar realisation, but all were represented with -ing in orthography. To avoid full analysis in this initial stage, we did not annotate such variable phonological realisations. The utterances were broken down into finite clauses, and these ‘first pass’ transcripts were further annotated in subsequent analyses where necessary.

These protocols were followed diligently in the transcription phase, but we adapted them to aid the reader in interpreting the interactions from here onwards, with key glosses provided in the section that follows.

Comparison Across Community, Caregivers and Children

As indicated in Chapter 1, CDS has been shown to be qualitatively different from adult-to-adult speech, with modifications at the lexical, syntactic and phonetic levels including higher pitch, slower speech rate and simple sentence structure (e.g. Snow Reference Snow, Fletcher and MacWhinney1995). We also showed that CDS might differ with respect to variation: caregivers use fewer local forms in CDS when compared to adult norms (e.g. Foulkes et al. Reference Foulkes, Docherty and Watt1999, Reference Foulkes, Docherty and Watt2005; Roberts Reference Roberts, Chambers, Schilling-Estes and Trudgill2002). Put very simply, what we might expect is that when adults talk to each other, they would make a lot of use of aye, r[u:]nd and be[ʔ]er, but when caregivers talk to children, they would make more use of yes and r[ʌʉ]nd and be[t]er. This is said to arise from concern on the part of caregivers, and in particular female caregivers, with ‘teaching’ their children the ‘proper’ language. At the same time, not all communities studied showed this effect (e.g. Van Hofwegen & Stob Reference Van Hofwegen and Stob2011), nor did all variables (Smith et al. Reference Smith, Durham and Fortune2007, Reference Smith2013). In order to test caregiver input vis-à-vis community norms, across each variable we compare the rates of use of particular features across community and caregivers (schematised in Figure 2.4).

2.4 Comparison between community and caregiver

The variationist literature suggests that there is a statistically significant correlation between caregiver and child rates of use across a range of variables (e.g. Kerswill & Williams Reference Kerswill and Williams2005; Patterson Reference Patterson1992; Nardy et al. Reference Aurélie, Chevrot and Barbu2013). In other words, if a caregiver uses lots of glottal stops, then their child will too. We test this in these data by comparing overall rates of caregiver input to child output (Figure 2.5).

2.5 Comparison between caregiver and child

These analyses provide a triangularisation between community, caregiver and child (Figure 2.6) that will allow us to compare and contrast what is the same, what is different and why across these three different groups.

2.6 Comparison of community, caregiver and child

We also test for the correlation between caregiver input and child output, both in overall distributions and across caregiver/child individual pairs (Figure 2.7).

2.7 Individual caregiver/child pairs

Individual vs Group Norms

In sociolinguistic research, variation is generally treated as a community-based norm, with each individual conforming to a pattern of use. However, there may be ‘outliers’ who do not conform in this way (e.g. Guy Reference Guy and Labov1980), oddballs ‘whose speech seems completely anomalous’ (Chambers Reference Chambers2003: 93) and who are swimming against the tide (Figure 2.8). As Britain (Reference Britain, Britain and Cheshire2003) points out, it is important not to ignore such speakers if a full picture of the variation is to be gained. Thus, we will also make note of individual differences in use across the different variables.

2.8 The individual swimming against the tide of group norms

Age of Child

Research has shown that caregivers decrease the use of specific CDS forms, such as vowel lengthening and exaggerated pitch, as a child matures (e.g. Huttenlocher et al. Reference Huttenlocher, Vasilyeva, Waterfall, Vevea and Hedges2007; Kaye Reference Kaye1980; Phillips Reference Phillips1973). Similar conclusions are reached in the study of variation, where caregiver talk ‘gradually becomes more similar in character to that of the interadult mode as the children get older’ (Foulkes et al. Reference Foulkes, Docherty and Watt2005: 201), i.e. there are higher frequencies of the standard variant with caregivers in interaction with very young children, but these are replaced by increasing frequencies of the non-standard variant as the child gets older (Figure 2.9). Thus, for example, there may be lots of yes with younger children but more aye with older children.

2.9 The effect of children’s age on use of variants

In order to test this in our data, we investigate rates of use of the local form from (caregivers of) the youngest child to (caregivers of) the oldest in order to gain insights into the driving force behind CDS across the different stages of acquisition and its possible effects on the child’s language. If use is conditioned by age, we might expect to see lower rates of the local forms with the (caregivers of) younger children, as in the stylised graph in Figure 2.10. If use is not governed by age, we would expect to see a flat line from youngest to oldest, as in the stylised graph in Figure 2.11. Alternatively, as per previous research (Smith et al. Reference Smith, Durham and Fortune2007, Reference Smith, Durham and Richards2013), the caregiver/child pairs may fall into two main groups (Figure 2.12).

2.10 Stylised graph showing age differences

2.11 Stylised graph showing no age differences

2.12 Stylised graph showing split between (caregivers of) younger and older children

We can also use an age comparison to examine a key question raised in Chapter 1: do children acquire one variant first, and if so, which one – the standard or the local form? Or do they acquire both variants at once? An optimal way to address that question is through longitudinal data, where a child’s linguistic development is recorded over a set period of time (e.g. twelve months or two years). Our data in this study is not longitudinal but cross-sectional: the children are compared at a single point in time. However, because we have children of different ages in our sample, we can use the cross-sectional data as a proxy for maturational change. Thus, if the variable patterns as in Figure 2.12, we can hypothesise that one variant is acquired first and another added to that as the child grows older (as in Figure 2.13). If the variable patterns as in Figure 2.11, we can hypothesise that both variants are acquired at the same time (Figure 2.14).

2.13 Variants acquired sequentially

2.14 Variants acquired simultaneously

Style-Shifting

As discussed in Chapter 1, acquiring a language demands much more than knowledge of its words, sounds and sentences. It also requires knowledge of how to talk to different addressees on different occasions, in different settings and on different topics. Although this is often characterised as a mapping of discrete voices for different contexts, the sociolinguistic literature demonstrates that the situation is much more fluid, with subtle but dynamic adjustments to a speaker’s style across a range of situations. Despite this complexity, adolescents and adults systemically style-shift quite seamlessly, and the emergence of adult-like patterns of style-shifting may appear in the preschool years. In these data, we wanted to specifically test Labov’s observation that ‘linguistic variation is transmitted to children as stylistic differentiation on the formal/informal dimension’ (Reference Labov2001: 437). To do this, we divided the recordings into different stylistic contexts that arise naturally from the caregiver’s multifaceted roles in daily interaction with her child: teaching and discipline versus routine, play and intimacy.

Rather than coding piecemeal for individual variables, where the wider context may be obscured, we coded every single utterance in the corpus for a particular context. This meant that the different contexts were not restricted to categorisation based on individual words, sentences or speaker turns, but instead covered the full context of use, which could include one or two interactions in a specific situation or multiple interactions across varying contextual fields. This is exemplified in the following interaction (56), where Max (2;11) and his caregiver, Alice, move from a teaching context, to discipline, to play, to routine and back to discipline in a short space of time.

Not only was this task extremely time-consuming, but how an utterance should be categorised was not at all times black and white. While some were very clear, others were not, as we will discuss.

Routine was fairly straightforward and included the rituals of everyday life such as eating breakfast (57), shopping (58) and getting dressed (59).

It was much trickier to code for teaching and play, as these are often one and the same in caregiver/child interaction, teaching through play being a key component of the early years (e.g. O’Reilly and Bornstein Reference O’Reilly, Bornstein, Bornstein and O’Reilly1993). For this reason, we created a strict protocol for how to code the data. Contexts were coded as teaching only in cases where the caregivers were overtly teaching the children, for example, asking them what colour something was (60) or about left and right (61):

Teaching segments often demonstrate some degree of help from the caregivers, such as ‘sounding out’ the start of the word, as in (62):

Segments where a book was being described either by the caregiver or by the child were also coded as teaching. Although in the former cases, it was mainly the child who was asking the questions, the manner of description made it clear that the caregiver was to some extent teaching the child new words (63):

Questions from the child that involved requests for information on issues of, for example, daily routine (64) were not counted as teaching, however.

Contexts were categorised as play when they involved not only actual games (65) but also more general play, as in (66). They also included imaginary play (67):

Contexts were coded as discipline when the caregiver was scolding their child or admonishing them to behave better, as in (68) and (69). This involved a lot of the negative imperative forms ‘don’t’ and ‘dona’. The child’s response was also considered to belong to the context of discipline.

Intimacy (70) made up the final context of use.

In analysing these different situational contexts, if there is style-shifting we expect to see a pattern such as Figure 2.15, where there are higher rates of the local form in the informal contexts of play, routine and intimacy and lower rates in teaching and discipline. A pattern as in Figure 2.16, on the other hand, would suggest that there is little to differentiate the different contexts of use (i.e. no style-shifting according to situational context). In cases where we do not have enough tokens to divide the data in this way, we will collapse play/routine/intimacyFootnote ⁵ and teaching/discipline into two broad categories: formal and informal, as in Figure 2.17.

2.15 Stylised graph showing use of local variant in different situational contexts (style-shifting)

2.16 Stylised graph showing use of local variant in different situational contexts (no style-shifting)

2.17 Stylised graph showing use of local variant in different situational contexts (formal vs informal)

Gender

Chapter 1 details findings on gender differences in caregiver speech with respect to variation, with, for example, higher rates of the non-standard form used with boys when compared to girls (Foulkes et al. Reference Foulkes, Docherty and Watt2005). Gender differences in child speech are not so clear cut, as also discussed in Chapter 1, one hypothesis being that gender differences do not arise until children move into the wider community (e.g. Roberts Reference Roberts1997). In these data, we test for potential gender effects in both caregiver and child speech across the range of variables that we examine (Figure 2.18).

2.18 Comparison of girls and boys

Linguistic Constraints

In addition to social constraints, a number of linguistic constraints will also be tested within each linguistic variable. This will include, where relevant, phonetic context, subject type and lexical collocation. Such analyses also relate to the ‘complexity of the conditioning’ (Kerswill Reference Kerswill1996: 199) of a variable that might influence when linguistic constraints are acquired. To date, studies have largely concentrated on one variable in isolation, and thus systematic comparison across a range of variables has not been possible. This means that it has not been possible to test, for example, Chambers’ (Reference Chambers1992) principles of second dialect acquisition, according to which lexical replacements are acquired faster than phonetic or phonological rules and further, with phonological variables, simple rules are acquired earlier and faster than complex rules. In this study, we contribute to the question of how different types of variables are acquired by comparing multiple variables in order to establish which types are acquired at what stage of development (Figure 2.19).

2.19 Order of acquisition of different types of variable

The Interaction of Linguistic and Social Constraints

Roberts (Reference Roberts1994: 30) points out that most studies of acquisition tend to ‘focus on one type of constraint or the other’. However, variation is the result of multiple competing effects in the ambient language, as in Figure 2.20. By testing both linguistic and social constraints, we may be able to contribute to the debate on order of acquisition (e.g. Labov Reference Labov1989). Simply put, do children learn social and stylistic constraints first, or grammatical ones or both at the same time (Figure 2.21)?

2.20 Potential influences on the acquisition of variation

2.21 Order of acquisition of social and linguistic constraints

Developmental vs Sociolinguistic Variability

Before embarking on the analyses, we need to consider one more key issue in the investigation of the acquisition of variation: how to distinguish between variation arising from developmental processes and variation that is part of community norms (e.g. Roberts Reference Roberts and Ball2005; Tagliamonte & Molfenter Reference Tagliamonte and Molfenter2007; Miller Reference Miller, Donaher and Katz2015).

In (71), Luke is telling his mam, Molly, about a particularly exciting accident that he and his granny witnessed some time ago in:

In this extract, Luke uses an array of irregular verbs forms, including seen and see for standard saw, goed for went and three different forms for the preterite of take: took, tooken, ticken. Luke also uses yet another form in the mix: taen for preterit took (72):

An important part of the analysis is to distinguish which of these forms are bona fide sociolinguistic variables and which are the result of Luke’s linguistic journey to adult-like attainment of forms. One way to distinguish these is to see what the adult speakers in the community are doing. In examining the data from the ambient language in the adult corpus, we find that goed for went is never heard in adult speech in this community. Instead, this is a very common overgeneralisation used by children in the early stages of language acquisition (e.g. Pinker Reference Pinker1994: 273). Seen for saw, on the other hand, is also a productive form in adult speech in the community, as demonstrated in the examples from the caregivers in our corpus in (73–75). In other words, it is a sociolinguistic variable.

And what of the verb take? Luke’s use of tooken and ticken are both non-target-like forms. In Buckie, you would not hear an adult use these. Taen on the other hand, is part of community grammar, and it can be used in both past participle (76–77) and preterite (78–79) contexts:

In most cases, what is part of community grammar and what is developmental is quite straightforward: if we find it in the adult data, it is a sociolinguistic variable; if we do not find it, it is likely that it is part of developmental processes. However, some cases are less straightforward, particularly those cases where a form is both developmental and also part of community grammar. For example, a child is known to go through a stage where 3rd person singular -s on lexical verbs is sometimes present and sometimes not, as in (80–81) from Harris and Wexler (Reference Harris, Wexler and Clahsen1996: 11).

1. (80)

   1. a. It only write in the pad.

   2. b. My finger hurts.

   (Eve, 2;0 file 14)

1. (81)

   1. a. Patsy need a screw.

   2. b. This goes in here.

   (Peter 3;3; file 8)

However, variation in 3rd person marking in adult speech, where -s can be absent, is amply attested, particularly in AAVE, as demonstrated in (82) from Cukor-Avila and Bailey (Reference Cukor-Avila, Bailey and Lanehart2015: 189).

1. (82) She bring ’em in every mornin’.

A similar example comes from finite be deletion. Children are noted to go through a stage where both copula and auxiliary be are absent in utterances, as in (83) from Vainikka (Reference Vainikka1993: 272):

1. (83)

   I ø popping balloons

   (Nina, 2;0 file 5)

Such forms start to disappear from around twenty-eight months (Brown Reference Brown1973). However, finite be deletion is also well-attested in adult data in some communities, as demonstrated in (84) from Labov (Reference Labov1969: 717).

1. 1. a. He ø fast in everything he do.

   2. b. We ø on tape.

The possible entanglement between variation arising from stage of development and that of community norms is not restricted to morphosyntax. Phonetic forms, too, may present problems in analysis. For example, coronal stop deletion, or more commonly (-t,d) deletion, in word-final consonant clusters (85) is amply attested in the variationist literature for adult speech (e.g. Bayley Reference Bayley1994; Fasold Reference Fasold1972; Guy Reference Guy and Labov1980; Tagliamonte & Temple Reference Tagliamonte and Temple2005).

1. (85)

   We was walking down Micklegate and we grabbed him and grabØ this lad as well.

   (Tagliamonte & Temple Reference Tagliamonte and Temple2005: 281)

At the same time, consonant cluster deletion is described as ‘the most common and longest lasting stage’ (Shriberg & Kwiatkowski Reference Shriberg and Kwlatkowski1980: 138) in first language acquisition, where word-final consonant clusters with /t/ or /d/ may not be fully acquired until around 4;0 (e.g. McLeod et al. Reference McLeod, van Doorn and Reed2001b; Stoel-Gammon Reference Stoel-Gammon1987). Thus, deciphering whether the child’s deletion of /t/ or /d/ in coda clusters as in (84) is the result of developmental or community norms may be problematic (see also Smith et al. Reference Smith, Durham and Fortune2009).

A number of researchers have tackled this potential problem head on, and through careful analysis have been able to disentangle the competing influences – developmental vs community variation. For example, Green (Reference Green2011: 39) examines finite be deletion in the speech of African-American children. The results led her to conclude that children start with no be in non-past contexts but slowly realise from the input that these can occur variably. In other words, this is an insertion, rather than a deletion process in the acquisition of this variable. She thus foregoes the description of deletion, as her data shows ‘no evidence that the BE form was ever present’ (Green Reference Green2011: 45). Miller (Reference Miller2013) demonstrates, through a longitudinal comparative study of Nina and Sarah from the Brown (Reference Brown1973) corpus, that non-agreeing don’t in the speech of these children is the same in the initial stages due to developmental processes, but Sarah’s continuing use in the later stages (86) is due to caregiver input, where her mother and father (87) also used non-agreeing don’t.

1. 1. a.

      She don’t get no one to play with.

      (Sarah 147; 4;10)

   2. b.

      It don’t have no wings!

      (Sarah 117; 4;07)

1. (87)

   She don’t have no pennies.

   (Sarah’s father)

Beyond disentangling developmental vs community-based variation, such analyses may have implications for first language acquisition research more broadly (e.g. Miller Reference Miller2013). While such analyses are beyond the scope of the current study, it is important that we note potential overlap between developmental and systematic adult-like variation in examining the data at hand.

We have now provided a qualitative, descriptive account of the data for analysis. Before we turn to the variationist analyses of individual variables, we first provide an overall view of the corpus as a whole.