BBC Voices Recordings

The Voices project provided a snapshot of the linguistic landscape of the UK at the start of the 21st century by encouraging members of the public to contribute their words and reflect on the language they use and encounter in their daily lives. The audio strand complemented an online data gathering exercise carried out by BBC Nations and Regions (see Upton and Davies forthcoming) and consisted of recordings in 303 locations involving 1,293 people. Most conversations were conducted in, and focused on, English but a small number of recordings were also made in Welsh, Scottish Gaelic, Irish, Manx and Guernsey French. 50 ‘audio gatherers’ – one from every BBC Local Radio station in England, and one each from Radio Scotland, Radio Wales, Radio Ulster, Radio Foyle, Radio Cymru, Radio Nan Gaidheal, the Asian Network and the independent Manx Radio – interviewed groups of local people between October 2004 and June 2005.

Clearly no survey, however extensive, could possibly hope to include examples of every dialect and accent, and speakers were not targeted as somehow ‘representative’ of particular speech communities. Decisions about whom to record did, however, take into account a desire to capture well established local varieties alongside English influenced by the presence of other languages and by varieties brought to this country from other parts of the world. Where possible, all members of each interview group knew each other well and came from similar geographic, social and/or ethnic backgrounds or shared the same interests and passions. Interviewers were encouraged to intervene as little as possible and most recordings took place in a location familiar to the group, whether upstairs at a local pub, at home, or at the workplace. BBC Local Radio stations in England made on average five recordings from their area; Radio Scotland made 37 recordings (including one in Berwick-upon-Tweed in England); Radio Wales made 23 recordings; and 21 recordings were made in Northern Ireland (15 by Radio Ulster and 6 by Radio Foyle). Most recordings lasted between one and two hours in total with groups generally consisting of between three and five participants.

To ensure the data would be comparable between radio stations and between speaker groups, each conversation followed the same format. Prior to a recording session each participant was given a ‘spidergram’ – a visual prompt sheet with 39 common words clustered into semantically related themes such as ‘how you feel’, ‘what you do’ and ‘getting personal’. Contributors were asked to think in advance about other words and phrases they might use for a set of control terms in each semantic field – ‘how you feel’, for instance, includes the concepts hot, cold, pleased, annoyed, tired and unwell. The Voices spidergram is a condensed version of the ‘Sense Relation Network’ sheets developed by researchers at the University of Leeds and used by a number of PhD students for extensive fieldwork in single urban locations (Llamas, Reference Llamas1999). Interviewers subsequently used this prompt sheet to initiate discussions about alternative words and expressions used by the group and to explore the contexts associated with particular variants. Participants were also encouraged to contribute words and phrases they felt might be of interest beyond the confines of the spidergram.

The recordings featured prominently in BBC Local and Nations Radio broadcast output during August 2005 and on Word4Word, a weekly slot on Radio 4 that explored the discussions in greater depth with contributions and reflections from eminent linguists. In November 2005 the audio archive, known as the BBC Voices Recordings (BL shelfmark C1190), was deposited at the British Library, ensuring it would be available to the public alongside equally prestigious linguistic audio collections, such as the Berliner Lautarchiv British and Commonwealth Recordings (BL shelfmark C1315), Survey of English Dialects (BL shelfmark C908), Survey of Anglo-Welsh Dialects (BL shelfmark C1314) and British National Corpus (BL shelfmark C898).

British Library dialect resources and services

The BL has had a number of significant successes in recent years in developing access to its unique collections of sound recordings of authentic regional speech. The extraordinary popularity of BL Sounds: Accents and Dialects (http://sounds.bl.uk/Accents-and-dialects), an online dialect archive first published in 2004, and of Sounds Familiar? (http://www.bl.uk/soundsfamiliar), an interactive, educational website celebrating and exploring the diversity of British accents and dialects, confirms the academic value of, and enormous public interest in, this type of content. Both resources achieved impressive press and media coverage at launch and are widely incorporated into school and HE teaching syllabuses in the UK and overseas. Above all user figures remain consistently high – in 2011, for instance, Sounds Familiar? attracted 292,000 individual visitors, maintaining its position by a considerable margin as the BL's most widely used learning resource every year since 2007. The success of the two resources derives from the unprecedented access they give to important, and above all reliable, sources. The Survey of English Dialects (SED), internationally acknowledged as the definitive study of dialects in England, features prominently on both sites. However, before the Library provided online access to the survey's sound recordings, the SED was largely only known to, and appreciated by, academics with an understanding of phonetics and historical linguistics. The public response to the web resources shows that specialist and non-specialist audiences alike can relate much more immediately to the rich variety of voices captured in a sound recording than in a written description. Moreover, academics have access for the first time to audio data to complement the meticulously recorded SED published output.

The BL Sounds website was re-launched in 2012 with additional content and enhanced functionality designed to improve the user experience. Recordings on the site can be browsed by location and by date and fully searchable content summaries ensure there are numerous ways of locating appropriate content. Users can now additionally create and save their own play-list(s), tag recordings and insert notes in individual audio files. This will allow users to build their own linguistic corpora to support their teaching and/or research. However, untranscribed sound files present considerable challenges for researchers wishing to interrogate audio data more closely. Anyone wishing, for example, to compile evidence of the distribution of individual linguistic features has little choice but to audit several hundred hours of recordings in the hope of locating useful material. VoUK was conceived to address this challenge and funding was secured from The Leverhulme Trust for three years from 2009 to 2012. The intention is to expose the linguistic content of the BBC Voices Recordings to meet current and future sociolinguistic research enquiries for a wide variety of audiences, but also to create a template that can be applied retrospectively to existing BL sound collections and to future acquisitions. To achieve this we will create a searchable dialect dataset deriving initially from detailed linguistic descriptions of a selection of BBC Voices Recordings. The data extracted will not only contribute to our understanding of contemporary varieties of spoken English in the UK, but will enable the creation of an online discovery and access resource allowing linguists to identify relevant content.

Voices of the UK

VoUK follows on from previous BL studies by Robinson (Reference Robinson2004; Reference Robinson2007) by extracting detailed linguistic information from individual BBC Voices Recordings to create an inventory of lexical, phonological and grammatical phenomena that occur in each recording. The style of presentation intentionally allows both academic linguists and non-specialist audiences to interpret the content. In spring 2012 worldwide access to a catalogue of the BBC Voices Recordings, including fully searchable summaries, was made available at http://cadensa.bl.uk/cgi-bin/webcat and the full set of 283 recordings in English (and Scots/Ulster Scots) was published at http://sounds.bl.uk/. Onsite users at the BL can access sound files directly from the catalogue and remote users can stream audio files from the BL Sounds site. Time stamps included in the summaries allow listeners to locate the appropriate point in a recording quickly and accurately. Individual VoUK linguistic descriptions – one for each group conversation – will be attached as PDF documents to the corresponding audio file at BL Sounds. These will be released in batches once edited to ensure consistency, with a complete set available from summer 2013. A discovery and access functionality to enable users to browse recordings by linguistic criteria will be implemented once the editorial process is complete. This will allow researchers to find recordings according to a pre-determined set of linguistic variables, e.g. grammatical features such as were-generalisation (e.g. she were tired) or phonological phenomena such as TH-fronting (e.g. I fink so). The methodology for phonological and grammatical description is described fully in Upton and Davies (forthcoming); here we focus on the lexical component of the VoUK descriptions (see appendix 1 for a sample lexical description).

The BBC Voices Recordings differ from the vast majority of recent sociolinguistic surveys in a number of ways. Firstly the involvement of BBC Nations and Regions was essential in facilitating fieldwork over a greater geographic area than any previous study. Secondly the prominence afforded to lexical data collection contrasts markedly with the main focus of most post-SED studies in the UK. Finally, an important aspect of the Voices initiative as a whole is public engagement – both in terms of participation in the survey itself, but also in terms of enthusiasm for sharing and acquiring knowledge about language. The British Library performs a unique role in supporting Knowledge Transfer between HE research and the general public and VoUK was intentionally designed to disseminate this primary research data to the widest possible audience. Lexical variation seems to feature prominently in public discourse about linguistic variation, as demonstrated by the continued demand for popular print publications in the form of slang and dialect glossaries and by the proliferation of online platforms such as Urban Dictionary (http://www.urbandictionary.com/). A quick online search shows, for instance, that Countryside Books currently publish a ‘regional dialect series’ that runs to 15 books, while Abson Books publish 21 items in their ‘language glossaries’ series, with many titles appearing in revised editions and enjoying multiple print runs. It is fundamental for our purposes to appreciate that the vast majority of these popular publications and online forums are presented in the form of word lists or ‘dictionaries’ with entries occasionally blurring the distinction maintained by linguists between lexical, phonological and morpho-syntactic phenomena. In many cases items included in these accounts more accurately reflect local pronunciation patterns or grammatical features with authors modifying spelling conventions to convey these localised or colloquial forms. Typical of such entries are deef [= ‘deaf’] or dusta [= ‘are you’] – the former representing broad Cumbrian dialect pronunciation, the latter capturing the use in some northern dialects of a verb phrase with second person familiar pronoun ‘dost thou’ (Barker, Reference Barker2007: 17–18). It is also instructive to note that press interest in the VoUK project has focused predominantly on lexical matters, as illustrated by an ‘A to Z of regional English’ (G2, 27 March Reference Jeffries2009: 12) compiled by a Guardian journalist from a number of sources, including early VoUK descriptive work. Such evidence of popular consumption of linguistic data was sufficiently compelling to encourage us to present our lexical descriptions using standard orthography where possible. We hope this will ensure that the lexical data can be used by a variety of audiences, including schools and teachers, learners of English as a foreign language and the general public.

The lexical data we record falls into two categories: responses provided by participants to the 39 target words on the spidergram prompt sheet and lexical items that occur spontaneously during the course of subsequent discussions. For the sake of clarity we refer to the former as ‘elicited lexis’ and the latter we categorise as ‘spontaneous lexis’. Our treatment of the two types differs for reasons outlined below, but our overriding concern has been a desire to record the data as neutrally as possible. Rather than contribute to linguistic theory VoUK's primary objective is to present linguistic data with minimal analysis to allow researchers to locate content and interpret, evaluate or manipulate it for their own purposes. Despite our desire to remain theory-neutral any written representation of speech data inevitably requires choices to be made. The following account of our descriptive methodology explains the reasoning behind the judgements we made on issues from spelling conventions to debates about whether to categorise a particular variant as lexical, phonological or morpho-syntactical variation or – our default option – a combination of all three. Other challenges we encountered include whether to merge closely related forms into a single entry (e.g. chucking, chucking down, chucking it down, chucking it = TO RAIN HEAVILY); the equivalence (or not) of multi-word expressions elicited; and how to present responses in their discourse context (e.g. humour or deliberate use of antonyms).

Elicited lexis

We initially considered adopting SED conventions for presenting the elicited lexis, i.e. a close phonetic transcription of every response, but this approach was rejected for several reasons. Firstly, SED fieldworkers had limited access to sound recording technology so hand written transcriptions were the most appropriate solution for recording responses to the survey questionnaire for interpretation and analysis at a later date. Due to the prohibitive cost of equipment and inaccessibility of most SED localities it was only possible in the 1950s to make brief sound recordings on site and most actually took place some time after the original fieldwork. A small number of recordings offer tantalising glimpses of the SED questionnaire in practice, such as the recording made by Stanley Ellis in Eggleston, County Durham (BL shelfmark C908/4 C3), but most contain equally rich complementary data in the form of spontaneous reflections on working life or discussions of local farming practices. Much though we applaud the extraordinary foresight of the SED researchers in making their entire raw data available in the Survey of English Dialects Basic Materials (Orton et al., Reference Orton1962–1971) the recording of entries in IPA transcription is a considerable obstacle for a general audience, albeit much valued by linguists. Modern technology offers us the opportunity to make the data available in a variety of formats to suit different audiences. As the BBC Voices Recordings audio files are available online we felt that linguists would be able to create transcriptions to suit their own purposes should they wish, so we were keen to use standard orthography in our lexical descriptions to suit a wider readership.

It is worth stressing that the SED Basic Materials do, of course, include sets of responses in conventional orthography, listed alphabetically at the head of each item in the SED questionnaire, such as the set for LEFT-HANDED (Orton and Halliday, Reference Orton and Halliday1962: VI.7.13). Unfortunately familiarity with IPA phonetics is required to establish the correspondence between individual items in this orthographic list and the phonetically transcribed responses arranged by locality in the main text so that only a linguist would be able to distinguish, for instance, between informants who supplied cow-pawed or kay-pawed. However, the SED orthographic entries have been extremely helpful in enabling us to transcribe lexical items encountered in the BBC Voices Recordings for which we have found no record in other sources – responses such as cowey elicited in Kerseley, Warwickshire (BL shelfmark C1190/01/04) and keg-handed recorded in Heanor, Derbyshire (BL shelfmark C1190/12/04), for example. We also adapt SED conventions for presenting additional information that record an informant's intuitions about the responses they supplied. The SED entry for YAWNING (Orton and Halliday, Reference Orton and Halliday1962: VI.13.4) shows typical supporting information, which includes contextual detail such as [‘older’] or [rare]; suggestions of additional meanings as in [also = staring]; or related phrases like [gape like a cuckoo]. As was intended, the BBC journalists who guided the Voices conversations often probe participants for further clarification. Where this additional reflective information is linguistically revealing we include it. In maintaining this link with existing data and with previous dialect research, we hope to ensure a high degree of consistency and continuity in terms of methodology and practice, and to enable linguistic comparisons to be made between BBC Voices Recordings and with earlier surveys and other contemporary studies.

For each recording we list all the responses to each of the items on the spidergram. Some variables – notably KIT OF TOOLS and TO PLAY (A GAME) – are occasionally overlooked and we often hear an interviewer justify this by commenting that he or she has struggled to elicit responses from previous contributors. In such cases we flag in our descriptions that this item was ‘not discussed’ to avoid the implication that no alternatives were known within the group. Most interviewers do allow contributors sufficient opportunity to ensure the whole set of variables is discussed, but inevitably some items prompt more animated discussion than others. We record each response in italics in the form given by the speaker and in the order in which it appears within the conversation. A verb might therefore appear as a lemma (e.g. belt = TO HIT HARD) or in its inflected form (e.g. belted), depending on how it is supplied by an individual speaker. Supplementary information, where relevant, follows individual entries in brackets following SED practice. The following examples, extracted from descriptions of four groups' responses to the stimulus UNATTRACTIVE, illustrate this approach:

A speaker's actual utterance(s) are recorded in italics and the order reflects the sequence in which they were supplied in the conversation. Additional information is included in brackets immediately after the variant(s) to which it applies. The entries above therefore show that while ugly is considered more broadly applicable by all groups, a speaker in Aston Cantlow suggests the collocation ugly bugger is more likely to refer to a male. There is consensus among three groups here that dog, on the other hand, is commonly restricted to female referents. In the case of the group in Aston Cantlow this is explicit in the feminine antecedent in she's a dog and for the speakers in Salford it also applies to mess. In the description for Salford we include both minger and minging – the former presumably a noun and the latter an adjective – thus replicating faithfully the forms in which they were given. Similarly skanker, skank and skanky are all included as although they clearly share the same root they differ structurally. The additional information here also records that these terms are not necessarily directly synonymous with UNATTRACTIVE, rather they convey the additional sense of someone considered by this group to be negligent in regard to their own personal appearance. Other information included in brackets sheds further light on who might use a particular lexical item or in what social context it might occur. The phrase face like a dag, for instance, is used by the mother of a speaker in Mawla, while ‘modern’ here denotes a word or phrase only recently acquired or recognised by a group, such as minger and he wasn't in the pretty queue when the looks were being given out in Maldon. A speaker in Aston Cantlow also supplies the similar expression back of the queue when the looks were given out and comments that this is a light-hearted expression associated with “banter”. We italicise entries in double quotation marks within brackets to indicate a speaker's actual words when offering observations on a given usage – as also in the case of “scrubber” in Salford.

We also record responses that result from association with the prompt word, but that are not intended by the speaker(s) as synonyms. Speakers often suggest antonyms, offer idiomatic usage or partake in amusing word play inspired by the discussion of a particular notion word. As this is potentially of interest we include it within the list of variants, marking it by the use of italics within double quotation marks. The example in Aston Cantlow, “I wouldn't fancy yours”, is typical of this kind of entry. Clearly this phrase could not simply replace the adjective ‘unattractive’ in any meaningful statement, but it nonetheless captures a speech act these speakers associate with discourse surrounding the notion of attractiveness. It is, we suspect, routinely associated by native speakers of a certain generation with superficial or flippant discussions among friends of the relative attractiveness of potential romantic partners. Interestingly the form “don't fancy yours much” is captured in the definitions of several entries (e.g. minger) submitted to Urban Dictionary. We feel there is sufficient interest in this kind of additional pragmatic detail to justify its inclusion in our lexical descriptions. As can be seen from the entry for beaten with the ugly stick in Aston Cantlow, we also indicate whether a response is unsolicited or results from an interviewer probing for a group's active use or passive knowledge of particular variants. The following list of common conventions is used as shorthand and gives a sense of the range of additional contextual information recorded:

As can be seen from the examples provided above, we follow established spelling conventions where available and attach a superscript symbol to an entry to indicate the source of any precedent. Table 1 below shows sources consulted thus far for spelling guidance. This list will expand to include additional sources as the editorial process continues and we have already identified a number of works we know will be of assistance, such as The Lore of the Playground (Roud, Reference Roud2010) and Fantabulosa (Baker, Reference Baker2002).

Table 1:

Previous lexical datasets consulted to establish spelling precedent

We consult the Oxford English Dictionary (online edition, OED) in the first instance. If a lexical item has an entry in the OED we adopt this spelling and leave the entry unmarked in our description. If we are unable to find a suitable entry in the OED we turn to other authoritative dictionaries and specialist glossaries as necessary, annotating the entry with the corresponding symbol in table 1. Where no previous written evidence has been found or legitimate alternatives exist, decisions regarding appropriate spelling are made on a case by case basis. The following examples from descriptions of three groups responding to the prompt DRUNK illustrate this methodology:

The superscript triangular symbol (^∆) that follows bladdered and tied one on indicates both expressions feature in Dalzell & Victor (Reference Dalzell and Victor2006); the superscript lozenge symbol (^◊) indicates rattled is recorded in Green (Reference Green2010); and the superscript pentagonal symbol (^⌂) records entries such as sealed, smelt the barmaid's apron, stooled and jelly-legged for which spelling is arguably predictable but for which we have found no previous evidence of use in this sense. The phrases blotto as a bob-owler and had a crooked elbow elicited in Dudley are particularly worthy of further explanation here. The term blotto is recorded in the OED while bob-owler is recorded in Wright (Reference Wright1898–1905) as a West Midlands term for ‘any large moth’ – hence the latter is indexed with the superscript circular symbol (^○). However, as we have found no record of the phrase blotto as a bob-owler for ‘drunk’ we also attach (^⌂). This use of multiple symbols is rare, but always explained more fully within individual descriptions to avoid confusion. In the case of the description for Dudley the meaning of bob-owler is included within the spontaneous lexis for that recording. The phrase had a crooked elbow, on the other hand, contains recognisable individual elements for which spelling is not in dispute, but the phrase itself is not recorded in any of the sources we consulted. The inclusion of elbow-crooker meaning ‘drunk’ in Dalzell & Victor (Reference Dalzell and Victor2006) is deemed sufficiently close to merit a citation, but we sense additional information included in a footnote is needed to assist researchers in locating the relevant entry. The response smelt the barmaid's apron [apən] also recorded in Dudley includes a phonetic transcription in IPA as the pronunciation is neither immediately apparent nor easy to record appropriately in the corresponding phonological description. As noted earlier, speakers occasionally make an impromptu remark in response to a prompt word, as in the case of Robin Hood and his Merry Men in Mawla. This comment appears to be an example of playful word association triggered by the previous response – merry – the use of which has become fossilised in popular references to the folk legendary figure of Robin Hood and his band of outlaws. As such responses are potentially of interest we include them, but indicate by the use of double quotation marks that they are not intended as direct variants of the stimulus word itself. The description of an interview in Southall here also records the use of a Punjabi word. It has not been possible to transcribe or provide transliterations of many items in languages other than English, but we hope by including this with a time stamp we will be able to crowd-source suggestions for appropriate entries once descriptions are published online. Finally, we try to avoid unnecessary footnotes, but one entry here illustrates the occasional need for further clarification. In the case of hissed in Mawla, we suspect native speakers would recognise a possible euphemism for ‘pissed’. This inference is made even more plausible by a speaker subsequently suggesting hissed as a newt with its obvious similarity to pissed as a newt which is sufficiently well-established to merit inclusion in the OED.

Spontaneous lexis

Additional lexical items that occur spontaneously during the conversations are recorded alphabetically and cited within context with a time stamp to enable users to locate the point in the audio file where the token occurs. As is normal in group conversation many of the lexical items occur in passages that feature multiple speakers, interruptions, hesitations, asides, changes of direction and incomplete sentences. We reproduce as faithfully as possible every element of the cited speech act in which the word or phrase under scrutiny occurs. We use brackets within the citation to indicate a change of speaker but otherwise keep punctuation to a minimum. The extracts from four descriptions below illustrate the kind of incidental lexical material included, and our style of presentation:

As can be seen from the selection above, we include words or expressions we sense are not part of mainstream usage. This covers potentially unfamiliar terms relating to speaker interest or expertise, such as sporting references (bodyline in Lewes and stack in Milton Keynes); words relating to particular industries, such as agriculture (swap or swap-hook in Lewes); cultural and leisure pursuits (tipteers in Lewes, doctor's orders in Driffield and grebo in Milton Keynes); and items which might be considered geographically restricted (forkin-robin in Driffield and dae, gan and hersel in Coxhoe), age-specific (court in Driffield and baddest in Milton Keynes) or possibly both (else in Driffield, dead in Lewes, well in Milton Keynes and by in Coxhoe). As with the elicited lexis, we cite previous evidence of the data we extract. Thus forkin-robin, gan, hersel and tipteers have entries in Wright (Reference Wright1898–1905), while baddest is recorded in Dalzell & Victor (Reference Dalzell and Victor2006) and dae has an entry in the Dictionary of the Scots Language (online edition). We are also mindful of a worldwide audience and thus define entries (e.g. doctor's orders, bodyline and grebo) that might cause problems for non-native speakers and/or users unfamiliar with English-speaking, particularly British, cultural references.

As noted earlier we try to restrict the use of footnotes, but two entries here illustrate instances where we feel they are justified. Most entries above capture lexical items that occur absolutely naturally, such as the Driffield speaker's use of ‘to court’ in I'd been courting for quite a while actually because uh Audrey who's me wife we used to go to school together. There are, however, examples here that are not strictly ‘spontaneous’, rather they are supplied as illustrative of a particular speaker or speech community or are associated with language use in particular domains. In many cases this can be deduced from the citation, as in (like the way you fall over as well there's different, like, words for that) yeah, like ‘stacked’ (yeah) (‘spank’) (‘slam’) in Milton Keynes. The expression ‘stacked’ here is offered in response to a question from the interviewer about language use among skateboarders that is not widely understood beyond that community of practice. The speaker's words include sufficient contextual information here to ascertain that ‘stacked’ is provided by way of illustration of skateboarding terminology. Where this is not sufficiently clear we use footnotes to indicate that an entry is ‘performed’ or provided in imitation of other speakers, as in the example here of dead supplied by a speaker in Lewes who is quoting her mother. We also use footnotes to provide clarification if a quoted passage contains references to people or events that might impede understanding for some audiences. Only dedicated sports fans (and probably only those of a certain generation and/or from cricketing nations) are likely to recognise Harold Larwood, for instance, in the extract from Lewes above. Hence footnotes contain brief details of historical and contemporary figures, fictional characters, commercial organisations, public institutions, television programmes, films, music and so on. Some references (e.g. ‘The Beatles’ or ‘Adidas’) might seem superfluous to the vast majority of audiences in 2012, but by including it we hope both to future-proof the data, but also to ensure users worldwide can appreciate the fine distinctions made by speakers in their linguistic choices.

VoUK descriptions occasionally include multiple entries of the same linguistic phenomenon. For instance, in the extract from Driffield above we transcribe the speaker's pronunciation of the first person singular possessive pronoun with a comparatively weak vowel as ‘me’:

This spelling captures a feature that occurs in many varieties of English in the UK and implies a merger of the first person singular object and possessive pronoun. We are aware, however, that it is equally valid to interpret this phenomenon as a phonetic process so we include ‘my’ as a special case within our phonological description of the PRICE lexical set (Wells, Reference Wells1982) for this recording:

and include ‘possessive me’ within our grammatical description:

Likewise the entry for dae recorded in Coxhoe we also repeat within our description of the GOOSE lexical set (Wells Reference Wells1982) for this recording:

while dead (from Lewes), well, baddest (both from Milton Keynes) hersel and by (both from Coxhoe) we include in the grammatical descriptions for the respective recording:

Many of the lexical items collected are attested in the reference works cited above, but a surprisingly high proportion of the words remain elusive. For these entries we use our own judgment based on the ‘default’ spelling patterns of British English. There are examples of lexical items that are transparent in terms of spelling but represent private code used within a family or group of friends such as ‘fully-handed’ here:

This expression captures the concept perfectly and is in fact grammatically acceptable, but does not reflect idiomatic usage. We choose to hyphenate the two components by analogy with words like ‘left-handed’ and ‘right-handed’ (cf. jelly-legged in Mawla above). In other cases we have had to create our own spelling for entries such as namna, munger and gargy. These forms represent childhood attempts at pronouncing variants for GRANDMOTHER that have subsequently been adopted and used by a family in Ingatestone, Essex (BL shelfmark C1190/02/02). There are also examples of words formed from productive morphological processes such as mama-ji and mummy-ji [= MOTHER] created by adding the Hindi-Urdu honorific suffix <-ji> to English morphemes and used by English-Punjabi bilinguals in Southall (BL shelfmark C1190/03/05). Potential performance errors or slips of the tongue such as battlecock and shuttledore [= ‘shuttlecock and battledore’] in Heanor, Derbyshire (BL shelfmark C1190/12/04) are recorded as closely as possible, as are nonce-formations such as the spoonerism supplied by a landlord in Mansfield (BL shelfmark C1190/26/04) to demonstrate his customary way of encouraging customers to leave his pub: “he's pissed he's not getting no more tell him to fup up and suck off!”. Finally, we occasionally adopt established transliteration conventions of other languages as in the case of kappre [= CLOTHES in Punjabi], which is used by a Punjabi-English bilingual speaker in Southall (BL shelfmark C1190/03/05) during a discussion of code-switching: and they used to say “we are going for shopping to buy some ‘kappre’ for wedding”.