2.1 British Library Sound Recordings of Vernacular Speech

The British Library (BL) has in its sound archives recordings that document spoken English over a period of more than 100 years. The recordings range from ‘performance’, such as speeches, literary productions, audio books, public talks and lectures to more ‘naturalistic’ speech contained in radio and television broadcasts, oral history interviews and linguistic surveys. These unique collections are an especially rich resource for researchers interested in varieties of British English and are used by diverse audiences from academic linguists, language teachers and students to researchers and practitioners in the creative industries such as actors, voice coaches, authors, script-writers, journalists and the broadcast media. A number of recent initiatives have enabled the British Library to extend its resources and services by acquiring existing collections and creating new content, while simultaneously developing remote access to selected material. This chapter provides an overview of a collection of historic sound recordings – the Berliner Lautarchiv British and Commonwealth Recordings (BL shelfmark: C1315) – and evaluates their relevance to contemporary research by comparing them with data from the Survey of English Dialects and similar authoritative linguistic sources.

2.2 The Berliner Lautarchiv British and Commonwealth Recordings

The Berliner Lautarchiv British and Commonwealth Recordings (BLBCR) is a collection of digital audio transfers from shellac disc recordings made between 1916 and 1938 by Wilhelm Doegen, Director of the Lautabteilung an der Preußischen Staatsbibliothek (Prussian State Library Sound Department) in Berlin and its predecessor the Königlich Preußische Phonographische Kommission (Royal Prussian Phonographic Commission). It consists of 821 digital copies of recordings that feature speakers from the British Isles and Commonwealth nations and constitutes a subset of the extensive Berliner Lautarchiv at the Humboldt-Universität in Berlin (publicus.culture.hu-berlin.de/lautarchiv). The BLBCR audio files fall into three categories: recordings in English of British and Irish Prisoners of War (POWs) held in captivity on German soil between 1916 and 1918, recordings in a variety of indigenous languages of British colonial troops made in the same circumstances, and later recordings made by Doegen, including a set of recordings in Irish made during fieldwork in Ireland in the 1920s and 1930s (see Hickey Reference Hickey2011: 98–99). The recordings of British and Irish POWs represent the earliest known collection of sound recordings of vernacular English speech and are the focus of this chapter.

2.3 Wilhelm Doegen and Alois Brandl

Wilhelm Doegen (1877–1967) was a graduate of modern languages at the Friedrich-Wilhelms-Universität in Berlin who developed an interest in the use of phonetics in language teaching as a result of meeting British phonetician Henry Sweet while studying at Oxford in 1899. After graduating, he began publishing several volumes of teaching materials with the Odeon Recording Company in Berlin (Doegen Reference Doegen1909). Following the outbreak of World War I Doegen became increasingly intrigued by the opportunity afforded by the presence on German soil of unprecedented numbers of peoples from all over the world in the form of POWs. Enlisting the support of academics working in such diverse fields as philology, ethnomusicology and anthropology, Doegen and philosopher/psychologist Carl Stumpf (1848–1936) proposed the creation of a commission to coordinate a comprehensive recording programme. The Königlich Preußische Phonographische Kommission was established in October 1915, with Stumpf as its Chairman and Doegen responsible for implementing the ambitious recording programme. The intention was to capture native German dialects, the voices of famous people, and languages, music and songs from all over the world. Between 1915 and 1918, 2,677 recordings were made in 70 POW camps across Germany: 1,022 recorded on wax cylinder by ethnomusicologist Georg Schünemann (1884–1945) and 1,650 on shellac disc by Doegen and his associates (Mahrenholz Reference Mahrenholz, Dendooven and Chielens2008).

While Doegen was responsible for technical production, eminent German linguists assumed responsibility for identifying and recording individual language groups. Alois Brandl (1855–1940), Austrian-born Professor of Anglistik at the Friedrich-Wilhelms-Universität, Münster, and President of the Deutsche Shakespeare-Gesellschaft (German Shakespeare Society), directed the recording of English-speaking POWs. Like Doegen, Brandl had met Henry Sweet in London, receiving weekly tutorials during the winter of 1879–1880 to improve his English pronunciation. As he describes in his autobiography (Brandl Reference Brandl1936: 138), Sweet's instruction involved exercises in articulation, transcription and accurate reproduction of transcribed speech, a process which he acknowledges informed his approach to documenting the recordings made in the POW camps (Doegen Reference Doegen1925: 367). After the war, Doegen secured a permanent home for the recordings in the Lautabteilung an der Preußischen Staatsbibliothek, founded in 1920. He subsequently resumed a productive collaboration with distinguished British linguists including Daniel Jones (1881–1967), Arthur Lloyd James (1884–1943) and Ida Ward (1880–1949), with whom he published a series of commercial recordings on the Odeon label, such as Examples of English Intonation (Jones Reference Jones1921). He also assisted Ferdinand Wrede (1863–1934), director of the Deutscher Sprachatlas (Wrede et al. Reference Wrede1927–1956), with sound recordings for the survey of German dialects instigated by Georg Wenker (1852–1911).

2.4 The 1916‒1918 Recording Programme

From Brandl's own account of the POW recording programme (Doegen Reference Doegen1925: 362–375), it is clear he viewed the opportunity as a natural extension in regrettable circumstances of work he had begun before the war. He recalls several visits he and his students – undergraduate trainee teachers – made in the early part of the twentieth century to carry out fieldwork in locations across England and Scotland, expressing particular delight in observing dialect verse and songs. In common with many of his contemporaries, Brandl considered urban speech somewhat diluted and less worthy of linguistic study than rural dialects, a bias that continued well into the twentieth century as demonstrated some forty years later by the essentially rural focus of the Survey of English Dialects (Orton et al. Reference Orton1962–1971). The process for selecting informants was assisted by the German military authorities who made available documentation they held on all British and Irish internees, including essential biographical details such as age, home and occupation in civilian life. Access to this data enabled Brandl to compile an initial sample of POWs that met his criteria: ‘Stammte er aus der gebildeten Schicht, aus einer größeren Stadt oder aus dem weiten Cockneygebiet des Südostens, so wurde sein Zettel von vornherein bei Seite gelegt … Bauernburschen aus abgelegener Gegend, Fischer aus kleinen Häfen, Schafhirten und namentlich halbe Analphabeten waren zumeist gesucht’ (‘anyone from the educated class, from larger towns or from the wider Cockney area of the South East was rejected from the outset … farmhands from remote locations, fishermen from small fishing villages, shepherds and especially semi-literates were prioritised’ – my translation). Brandl's dismissal of speech in the south-east of England contrasted with his admiration for dialect speech in Lancashire and Scotland which he felt showed greater stability and continuity and explains the prominence within the collection of speakers from the north of England and Scotland.

Having identified a long list of potential candidates, Brandl then arranged visits to individual camps to interview speakers and assess their suitability for inclusion in the survey. Presented to Brandl in groups and asked simply to state occupation and describe briefly where they lived was considered sufficient to eliminate ‘eine große Zahl von Sprechern, indem sich nach den ersten Worten ergab, dass sie niemals einen Dialekt sprachen oder ihn gründlich abgelegt hatten’ (‘a large number of speakers who, within a few words, revealed they had never spoken dialect or had abandoned it completely’ – my translation). This rather superficial assessment by modern standards enabled Brandl to select his ‘gutes Material’. Those chosen were then made aware there was no compulsion to participate in the study, but evidently few declined. Sensitive to the controversial recording circumstances and keen to avoid contentious subjects, Brandl then spent some time interviewing individual speakers, commenting that most soon relaxed and seemed to enjoy both his interest in their speech and the opportunity for diversion from camp life. Discussions ranged from parents to school and working life, with Brandl encouraging speakers to volunteer local expressions, which he observed were particularly forthcoming when speakers reminisced about their childhood. Unfortunately, this is the only record of these spontaneous conversations, which served, for Brandl, as final confirmation of the authenticity of an informant's dialect and, for the POWs, as reassurance of Brandl's purely academic intentions prior to agreeing to make a sound recording. The content of the subsequent sound recordings varies and includes reading passages, word lists and recitals of songs and/or folk tales. The most frequently recorded text is a recital of the Parable of the Prodigal Son (Luke 8.11–32) in the speaker's own dialect. Clearly a popular device with linguists at the time, this passage also served as the principal ‘specimen’ in the Linguistic Survey of India (Grierson Reference Grierson1903–1928) chosen by its director, Sir George Grierson (1851–1941), as it contained ‘the three personal pronouns, most of the cases found in the declension of nouns, and the present, past and future tenses of the verb’ (Grierson Reference Grierson1927: 18). Although Brandl makes no reference to this precedent, he also commends the passage as containing several ‘Scheideformen’ (‘distinguishing forms’) that enable comparison between dialects.

The original shellac discs are held at the Berliner Lautarchiv along with documentation created at the time of recording. Biographical details of informants (e.g. date of birth, religion, occupation in civilian life, schooling, level of literacy) are presented on a ‘Personalbogen’, most of which are complete and survive intact. It is difficult to ascertain the version of the parable used for the recordings, but Brandl describes how a text was ‘vorbereitet’ (‘prepared’), thereby implying some editing of the original. Speakers were given the text in advance and encouraged to adapt it to reflect their own local speech forms, making it even more challenging to identify the source text. These handwritten adaptations survive in most cases and were used as prompt sheets for the actual recording. They do not correspond exactly with what appears on the discs as, presumably, speakers deviated from the script due to the pressure – and indeed novelty – of being recorded. These orthographic transcripts are accompanied by phonetic transcriptions made by Brandl using a notation scheme which he acknowledges derives primarily from Henry Sweet, although he stresses the discs themselves are the more accurate record of a given dialect. The International Phonetic Association was established in 1888, so Brandl's transcriptions are an example of practice at a time when the International Phonetic Alphabet was not yet universally applied. As such, they are of considerable academic interest and the subject of a forthcoming doctoral thesis by Valentina Tarantelli at the University of Sheffield. The handwritten transcripts are equally intriguing as they capture laymen's attempts to modify English orthography to convey vernacular forms and localised pronunciation.

It has been surprisingly difficult to find any reference to the discs or to Brandl's descriptions in the years immediately after the recordings were made. There are 20 dialects featured in Englische Dialekte (Brandl Reference Brandl1926) and a series of discs and pamphlets reviewed in The Review of English Studies (1931: 372–373) but few other leads exist. From 1949 onwards the discs were located in the German Democratic Republic and hence, perhaps understandably, disappeared from view. ‘Re-discovered’ and digitised in the 1990s, digital audio files were acquired by the BL in 2007 along with photocopies of the accompanying documentation. Given the extraordinary circumstances in which they were created and the centenary in 2014 of the outbreak of hostilities, there has been considerable interest in the collection in recent years. The recordings featured in the Radio 4 broadcast Barbed Wire Ballads (2005), and were the subject of a BBC 4 documentary, How the Edwardians Spoke (2007), which traced descendants of three speakers to ‘re-unite’ families with the recordings for the first time and an exhibition to commemorate the contribution made by Sikh soldiers to World War I, Empire, Faith and War: The Sikhs and World War One (2014) included several recordings with Indian POWs.

2.5 The Survey of English Dialects

The Survey of English Dialects (SED) was a groundbreaking nationwide survey of the vernacular speech of England, undertaken by researchers based at the University of Leeds under the direction of Harold Orton (1898–1975). As the SED is well known to linguists and dialectologists, only a brief summary is provided here, but Fees (Reference Fees1991) gives a more detailed description of the survey and its legacy. From 1950 to 1961, a team of fieldworkers collected data in 313 mostly rural localities, initially in the form of transcribed responses to a questionnaire containing over 1,300 items, meticulously recorded and subsequently published in the Survey of English Dialects: The Basic Material, Vols. I–IV (Orton et al. 1962–1971). The informants were mostly farm labourers, predominantly male and generally over 65 years old and thus, like the BLBCR informants, born in the second half of the nineteenth century. Advances in audio technology during the 1950s made it increasingly possible, and indeed desirable, to record informal conversations on site, so several localities were revisited to make sound recordings with original contributors or replacements with similar profiles. The recordings vary in length and quality, but generally consist of ten to twenty minutes of unscripted, spontaneous discussions of working life, domestic routine and local custom. The audio archive complements the survey's published output and includes recordings from 288 SED localities, additional recordings from Orton's pre-SED Northumbrian corpus and several pilot recordings from non-SED localities. The original open-reel tapes and gramophone discs are held at the Brotherton Library at the University of Leeds, and digital copies are available at the BL (BL shelfmark: C908).

Until recently these two collections of early sound recordings of British dialects were inaccessible to all but a handful of academics. In 2011 the complete set of 66 recordings of British and Irish POWs reading the Parable of the Prodigal Son were made available on the Sounds website (sounds.bl.uk) alongside extracts from 288 SED recordings uploaded in 2005. Audio clips from both collections were included in the BL's Evolving English exhibition (2010–2011) and on the accompanying Voices of the UK audio CD (British Library 2010). The Library's Sound and Moving Image catalogue (cadensa.bl.uk) gives complete details of both collections. To assess the validity of the BLBCR recordings for present-day linguistic researchers and dialect enthusiasts, I now consider a sample of lexical, phonological and grammatical variables contained within the Parable text and compare the variants supplied by one POW with SED published data and audio files to determine how these two sets of early sound recordings corroborate existing knowledge and/or allow us to re-evaluate previous studies.

2.6 The Dialects of Shelley and Skelmanthorpe

The following is a comparison of the BLBCR recording of John Townend (BL shelfmark: C1315/1/813) with SED data recorded in Skelmanthorpe (SED ref: 6Y31). John Townend was recorded in Güstrow POW camp on 3 July 1917. Born on 1 January 1882 in Shelley in the West Riding of Yorkshire, he attended the village school, then lived and worked locally as a steam engine driver before enlisting. Shelley is situated in present-day West Yorkshire, approximately six miles south-east of Huddersfield and one mile north-west of Skelmanthorpe. The two SED informants for Skelmanthorpe were aged seventy-five and seventy-six when interviewed in October 1952 and gave their occupations as weaver/miner/gravedigger/farm labourer and farmer/smallholder respectively. One of the informants, Wibsey Dyson, also made a sound recording on 20 October 1952 (BL shelfmark: C908/48 C1). Brandl's desire to give each POW creative licence to modify the text means the content of the Parable inevitably varies from speaker to speaker. Nonetheless, there is considerable overlap with items contained in the SED questionnaire, and in most cases it is possible to map up to 100 lexical items with a corresponding entry in the relevant volume of the SED Basic Material (Orton et al. Reference Orton1962–1971). Table 2.1 compares the realisation of a selection of variables that occur in the recording of John Townend with corresponding data from Skelmanthorpe presented in the SED (Orton and Halliday Reference Orton and Halliday1962–1963). The variables are restricted to full content words as function words, such as pronouns and prepositions, frequently show contrasting strong and weak forms, and are discussed separately below.

The phonetic transcriptions in Table 2.1 column 2 are based on an auditory assessment of the sound recording of John Townend. To avoid potential influence from SED data, the recording was transcribed prior to consulting the published responses for Skelmanthorpe. Multiple entries in columns 2 and 3 show alternative forms recorded in each source. In column 3 these either reflect pronunciations contained within an informant's response to other prompts in the SED questionnaire or derive from the SED ‘Incidental Material’ (Orton Reference Orton1962: 17) – tokens that occurred during spontaneous discussions that arose naturally between fieldworker and informant. Table 2.1 shows a remarkably close correspondence between the forms supplied by John Townend and the SED informants’ responses recorded in Skelmanthorpe. Of the thirty-three variables under scrutiny, twenty-four contain an identical match and six differ only in minor phonetic detail as itemised below:

1. 1. a. always shows deletion of word medial /w/ for both speakers but John Townend favours a voiced final consonant.

   2. b. clothes shares the same dialectal centring diphthong (more below), albeit the SED entry shows the word initial cluster /kl/ was realised as [tl], a phenomenon noted in several SED localities in the north and Midlands.

   3. c. closes [= ‘fields’] shares a dialectal diphthong (more below), but the SED entry indicates a more open onset but a common schwa vowel for the plural <-es> morpheme, a distinctive feature of the accent in this area then and now.

   4. d. give is realised as northern dialect/archaic gie (see, e.g., Wright Reference Wright1898–1905), but John Townend uses a lax vowel where SED shows a tense vowel, a subtle distinction in all probability prompted by contrasting phonetic environment: for John Townend gie occurs pre-consonantally (gie me [gɪ mɪ] that part of your goods that belongs me) whereas for the SED informant it precedes a vowel (gie it me [giː ɪt mɪi]).

   5. e. naught [= ‘nothing’] differs in phonetic quality in that the SED form shows a lowered onset for the diphthong but crucially the word belongs in the thought set (Wells Reference Wells1982) for both speakers – not the mouth set (Wells Reference Wells1982) as elsewhere in the north – and, like analogous words such as daughter and brought, is realised with a localised diphthong still typical of this part of West Yorkshire.

   6. f. very differs only in John Townend's use of an intervocalic tapped /r/, a feature conspicuously absent from the published SED data in Yorkshire but present in many SED sound recordings, including Skelmanthorpe (more below).

Of the three remaining variables, two show morphonological differences and one contrasts phonetically. The SED past forms begun and come or comed are clearly more dialectally marked than the counterparts John Townend uses: began and came are Standard English past forms, albeit his pronunciation of came [keːm] contains a distinctly ‘northern’ monophthong. The pronunciation of home shows predictable H-dropping (more below) for both speakers, but the SED informant uses the same dialectal centring diphthong noted for clothes above, whereas John Townend uses a checked vowel with a /w/ onglide. Although this phenomenon is not recorded in the SED for Skelmanthorpe at home (VIII.5.2), it is noted in several nearby localities, such as Golcar (SED ref: 6Y29) and Holmbridge (SED ref: 6Y30) and the response [wʊts] recorded for Skelmanthorpe at oats (II.5.1) confirms this onglide as a genuine local feature.

A smaller subset of the same variables occurs in the SED sound recording from Skelmanthorpe, confirming the same close correspondence. The passages below are extracted from the sound recording of Wibsey Dyson published at sounds.bl.uk and include IPA transcriptions of lexical items that match entries in Table 2.1.

1. 1. a. I were very [vaɾɪ] oft first there if there were a funeral on

   2. b. ‘well’ I said ‘I never seen a ghost in my life but I [inaudible] and I'd look at one’ [wʊn]

   3. c. don't make [mak] no mistake [mɪstak] when I got nearly up to it I heard [jəd] it were saying summat

   4. d. a woman'd comed [kʊmd] out in her nightdress and and and were seeking her cat

   5. e. I used to get up about half past four and set off to my work [waːk] I used to leave the house [æəs] at half past five and I'd to be at s… at pit at six o'clock and we worked [wʊkt] while two [tʊuː]

   6. f. before I got to my work [waːk] I'd gotten a right sweat on, you know, and then I'd to wade nearly knee-deep in water and I'd to work [wʊk] in water all the day through

   7. g. no wonder at folk dying when I reckon it up prematurely conditions that they had to work [wʊk] under in them days you were seldom ever come home [ʊəm] dry you're always [ɔːlɪz] wet to the skin and besides that you had naught [nɔʊt] for it

There is further confirmation here of a direct match with John Townend's speech in the case of house, make, one, naught [= ‘nothing’], take and two. Furthermore, Wibsey Dyson uses a final voiced consonant in always as noted for John Townend in Table 2.1 and indeed a tapped /r/ in very (more below). H-dropping is confirmed on heard and home: heard contains the /j/ onglide also used by John Townend, but a different vowel; John Townend's characteristic /w/ onglide in home is, however, again not evident here. Wibsey Dyson confirms his preference for past tense comed, but shares with John Townend a distinction between the noun work [waːk] and verbal work [wʊk~wɔk~wɒk] with a back rounded vowel of varying degrees of openness.

2.7 The Phonology of the Dialects of Shelley and Skelmanthorpe

Thus far this analysis has been restricted to comparing individual lexical items with direct counterparts in each source. To test the value of the BLBCR recording further I now offer a more comprehensive analysis. Despite pre-dating by 70 years the system now widely adopted for comparing accents of English, most recordings of the Parable text include at least one member of most lexical sets (Wells Reference Wells1982). In the case of John Townend only the choice and north sets are completely unrepresented, although some sets have only a small number of tokens. It is, however, possible to give an account of the general characteristics of John Townend's accent and compare it with SED data from Skelmanthorpe. The following description focuses on five vowel sets considered salient in this dialect – bath, strut, face, goat and goose – with additional observations on two significant consonantal features and two important connected speech processes. Table 2.2 indicates John Townend's realisation of the selected phonological variables and includes all the tokens in his recital of the Parable for each variable. As with Table 2.1, only full content words are included here: function words are discussed in the section below addressing morphonological phenomena.

In common with most speakers in this part of Yorkshire then and now, the quality of John Townend's bath and strut vowels shows no evidence of the trap-bath and strut-foot splits that characterise southern English accents. His pronunciation of words in the trap set (e.g. back, angry with [a]) and foot set (e.g. goods, put with [ʊ]) confirms this. Table 2.2 shows he is consistent in using [ʊ] with strut and only deviates from [a] with bath on dancing [dɒnsɪn]. Unfortunately, dance was not included in the SED questionnaire, but there is evidence of dialects in which words with orthographic <-an-> are realised with [ɒ], although the data locate the relevant isogloss some distance to the west of Shelley (Upton and Widdowson Reference Upton and Widdowson2006: 16–17). Furthermore, man occurs six times in the recording consistently with [a] and SED entries at man (VIII.1.6) and hand (VI.7.1) show no sign of [ɒ] in Skelmanthorpe, although [mɒn] is recorded in nearby Heptonstall (SED ref: 6Y21). I can cite anecdotal evidence of this pronunciation in Barnsley (personal experience from attending a football match at Oakwell c. 2003 when a local fan within earshot repeatedly taunted the referee with the observation: you want glasses, man [mɒn]). Perhaps, however, the English Dialect Dictionary entry for dancing (Wright Reference Wright1898–1905) offers most by way of clarification, as a citation records a spelling for West Yorkshire that corroborates John Townend's pronunciation: nivver let noan at lasses gooa to donsin [= ‘never let any of the girls go dancing’].

The data in Table 2.2 for face and goat show several variants. Local speech in much of West Yorkshire today is characterised by monopthongal realisations in these sets, with [eː ~ ɛː] and [oː ~ ɔː] respectively. John Townend clearly favours a monophthong for a small set of words in the goat set here and for the majority of the face set, but also uses broader dialectal variants. As noted in Table 2.1 [ʊə] is recorded for clothes and [ɔɪ] for closes [= ‘fields’]. All these variants are also present in the sound recording made with SED informant Wibsey Dyson:

1. 1. a. we'd a grave [gɾeːv] to dig and my mate [meːt] had gotten there before me

   2. b. I says, ‘sithee, Bill what's yond?’ he says, ‘it's a ghost [gʊəst] it's a ghost’ [gʊəst] he says

   3. c. in them days [deːz] there no [nɔː] there were no [nɔː] mechanism in the pit hardly everyth… every every bit of coal [kɔɪɫ] were hand-gotten

   4. d. before I got to my work I'd gotten a right sweat on, you know, [jənɔː] and then I'd to wade [weːd] nearly knee-deep in water

   5. e. no [nɔː] wonder at folk [fʊək] dying when I reckon it up prematurely conditions that they had to work under in them days [deːz] you were seldom ever come home [ʊəm] dry

The diphthongal goat variants appear to be recessive in present-day Yorkshire dialect, although Stoddart reports don't with [ʊə] among older speakers in nearby Sheffield (Stoddart et al. Reference Stoddart, Upton, Widdowson, Foulkes and Docherty1999: 74) and [ɔɪ] survives locally if only in fossilised phrases like fish-’oil (= ‘fish and chip shop’; see, e.g., Kellett Reference Kellett1994: 61) and putwoodintoil (= ‘shut the door’; see, e.g., Mitchell Reference Mitchell1987: 20) – expressions used affectionately and/or humorously even by speakers who otherwise consciously avoid overtly dialectal features. The checked variant go with [ʊ] is confirmed in the SED (VIII.6.1) and in the sound recording with Wibsey Dyson: I started a-working at a shilling in the day shilling in the day for going [gʊɪn] down the pit. Unlike the diphthongs, this pronunciation remains a distinctive feature of the present-day local accent and is reported by Stoddart for Sheffield with go, goes and going (Stoddart et al. Reference Stoddart, Upton, Widdowson, Foulkes and Docherty1999: 74). Rather surprisingly, given its potential for variation, over was not included in the SED questionnaire, but it does occur in the SED sound recording with Wibsey Dyson, pronounced [ɔʊə]. The use of a checked vowel, however, is certainly possible locally nowadays in a restricted set of words in the goat set including broke(n), froze(n), only, open and spoke(n). Of the other face variants, make and take with [a], always with [ɪ] for the final vowel and great with [ɛ] are included in Table 2.1. More recent BL sound recordings confirm always is still widely pronounced with [ɪ] in a number of British dialects, while in this part of the North and Midlands make and take appear to have joined items within the face set that can surface with [ɛ]: words such as ain't, came, break, gave, great (particularly in the collocation great big), laid, made, make, making, say, take(n) and taking. The final variant noted for face – again with [ɪə] – is not apparent in the SED published data or sound recording for Skelmanthorpe. However, again was elicited in a number of SED localities in response to the prompt beside (IX.2.5), albeit not in Skelmanthorpe. John Townend uses again in the Parable text as a preposition meaning ‘against’, mirroring those SED informants who supplied it as a variant of the preposition beside. The SED entry [əgɪən] for Thornhill (SED ref: 6Y26), the closest locality to Skelmanthorpe, offers confirmation of this as a genuine local feature.

The narrow diphthong with an extended offset noted for most items in the goose set in Table 2.2 is confirmed by the published SED data, including at two (VII.1.2) and shoe (VI.14.22) as shown in Table 2.1. This realisation remains distinctive of present-day speech locally and is also apparent in the SED sound recording with Wibsey Dyson, although not always with such distinctive length:

1. 1. a. I were very oft first there if there were a funeral [fjʊunɹəl] on

   2. b. I used [ʊust]to get up about half past four and set off to my work I used [jʊust]to leave the house at half past five and I'd to be at s… at pit at six o'clock and we worked while two [tʊuː]

   3. c. and then I'd to wade nearly knee-deep in water and I'd to work in water all the day through [èɾʊuː]

John Townend's pronunciation of soon as [sʊɪn] is replicated in the SED Skelmanthorpe entry for moon as [mʊɪn] (VII.6.3) and indeed occurs in the sound recording made with Wibsey Dyson: I says, ‘art thou late Bill or I'm soon [sʊɪn] or I'm soon [sʊɪn] and thou late?’

Table 2.2 includes data relating to two consonantal features: initial /h/ and the realisation of /r/. John Townend's almost categorical deletion of initial /h/ is unsurprising as H-dropping is a well-known feature of most vernacular accents in England. SED data shows it was a particularly productive process in West Yorkshire dialects, as confirmed by several entries for Skelmanthorpe including head (VI.1.1), hedgehog (IV.5.5) and help (V.8.13) and in the corresponding sound recording: if it'd a gone ‘sh’ and waved it's hands [andz] I should've very likely either had [ad] a fit or taken my hook [ʊuːk]. John Townend's realisation of /r/ varies between an alveolar approximant [ɹ] and, as noted previously, a tapped [ɾ]. This is one of the more intriguing elements of the recording as, unlike other features discussed here, it differs quite significantly from SED published data and thus perhaps offers new insights into the dialect of the period. I have always considered this a distinctive feature of present-day speech in this part of West Yorkshire and hear it in the speech of family in Castleford and Pontefract, and so I have consequently always been intrigued by its absence from SED descriptions. As Table 2.2 shows, John is certainly not consistent in his use, and we have insufficient data to present a definitive view of the distribution of each variant, but there are nonetheless observable tendencies. Some urban dialects of the north-west – notably Manchester and Liverpool – are characterised by almost categorical use of tapped /r/. John, however, uses both available variants word initially (ran with [ɹ] versus ring with [ɾ]) and within identical consonant strings (e.g. brass with [ɹ] but bring with [ɾ], begrudged with [ɹ] and angry with [ɾ]), but consistently uses an alveolar tap intervocalically in very, merry and arose. We might expect to find some trace, therefore, of this in the SED published data at entries like carrot (V.7.18), porridge (V.7.1) and very (VIII.3.2). It is difficult to speculate why this does not occur more systematically in the SED Basic Material (Orton et al. Reference Orton1962–1971), but it is certainly audible in several relevant SED sound recordings, including in Carleton (BL shelfmark: C908/9 C4) and Holmbridge (BL shelfmark: C908/47 C6) and, most importantly, in Skelmanthorpe: I were very [vaɾɪ] of first there if there were a funeral [fjʊunɹəl] on and this particular morning […] we'd a funeral [fjʊunɹəl] to we'd a grave [gɾeːv] to dig. It is also unquestionably characteristic of present-day speech in West Yorkshire as demonstrated by more recent BL recordings in nearby Golcar (BL shelfmark: C900/08618 – I used to go to the Methodist Chapel but my friends [fɹɛndz] went to the Baptist Chapel so I actually did a swap and went to the Baptist my parents [pɛːɾənts] were quite happy for me to do that) and Castleford (BL shelfmark: C1190/19/01 – I can remember [ɹɪmɛmbə] five ton of n … nuts being pulped on a warehouse [wɛːɾaʊs] floor and these nuts had to be turned every month) which confirm a similar distribution to John Townend and Wibsey Dyson, i.e. R-tapping is more likely to appear intervocalically.

Another intriguing pronunciation noted in Table 2.2 is John Townend's word final devoicing of fed and had in the utterances the husks that he fed the swine [fɛt∩ʔ swɑːn] with and his father saw him and had pity [aʔ pɪtɪ] and ran to meet him. These are examples of Yorkshire assimilation (Wells Reference Wells1982: 366), a term that describes a localised process whereby a voiced consonant assimilates to a following voiceless consonant. In certain high-frequency collocations this can occur in many accents of English (e.g. has to [has tu], have to [haf tu], had to [hat tu] and used to [juːst tu]) but is less restricted for many speakers in West Yorkshire, for whom e.g. goldfish [gɔːltfɪʃ] or big fish [bɪk fɪʃ] are typical pronunciations. This phenomenon is popularly associated almost exclusively with the city of Bradford, stereotyped locally as [bɹatfəd] or [bɹaʔfəd], but evidence from modern BL recordings confirms Wells's impression of a much wider distribution across most of present-day West and South Yorkshire and extending into East and North Yorkshire too. In the case of fed, here this process is possible despite it preceding a definite article with an underlying voiced onset /ð/, as the local preference for definite article reduction (more below) allows speakers to substitute a voiceless /t/ (or its allophonic equivalent [ʔ]) for the, thereby creating the required environment for Yorkshire assimilation. It is difficult to locate entries that show evidence of Yorkshire assimilation in the SED Basic Material (Orton et al. Reference Orton1962–1971), but there are examples in SED sound recordings in, for example, Sheffield (BL shelfmark: C908/48 C2 – if he's a good crane driver [gʊʔ kɹɛːn dɾɑɪvə] and I've had one or two good ’uns uh he'll be able to shout up to the crane man, you know, which way to go).

The final entry in Table 2.2, categorised as liaison, captures a local process that distinguishes some speakers in West Yorkshire and other parts of the central north from accents elsewhere in England. Many speakers with non-rhotic accents, including RP, tend in connected speech to insert a /r/ sound across certain morpheme and word boundaries where an open syllable meets a syllable with a vowel onset, such as do you want Fanta or Tango? [fantəɹɔːtaŋgəʊ] or drawing [dɹɔːɹɪŋ]. This so-called intrusive /r/ is thought to result from the loss of historic postvocalic /r/ (Wells Reference Wells1982: 222–227). Present-day speech in West Yorkshire is non-rhotic, albeit relatively recently in the extreme west as shown by evidence of r-coloured vowels in SED localities such as Golcar (Orton and Halliday Reference Orton1962: 36) but not Skelmanthorpe. Thus intrusive /r/ is allowed in most environments, but appears to be resisted locally in the environment <-awing> and <-aw#> + V, where many speakers favour either [ɸ] or insert a sandhi /w/. Both alternatives are illustrated in the recording with John Townend:

1. 1. a. he saw others [sɔː ʊðəz] at their meals but had naught hissen to eat

   2. b. his father saw him [sɔːʷɪm] and had pity and ran to meet him

More recent examples can be found in several BL audio files, including in Huddersfield (BL shelfmark: C900/08561 – ‘we were at the back of uh of the rows in assembly and all we were doing were footsie like this and she saw us [sɔːʷʊz] and had us out and we each had a stroke of the cane on the back of us hand’).

2.8 The Lexis and Grammar of the Dialects of Shelley and Skelmanthorpe

We now turn to a description of significant lexical and morphonological features that occur in the BLBCR recording with John Townend. Table 2.3 lists a set of features and gives the variant(s) demonstrated by John Townend and the relevant token(s) within the Parable text.

Of the five lexical items included in Table 2.3, gie (= ‘to give’) closes (= ‘field’), naught (= ‘nothing’) and brass (= ‘money’) are all also captured in the SED Basic Material (Orton et al. Reference Orton1962–1971) at IX.8.2, I.1.1, VIII.8.4 and VII.8.7 respectively. Although yet (= ‘still, now as formerly’) was not an SED prompt word and does not occur in the sound recording with Wibsey Dyson, its use in this sense by the SED speaker recorded in nearby Thornhill (BL shelfmark: C908/9 C3) confirms it as a local variant: aye, well I always did a clean job and I do yet.

The use of second person <th-> pronouns noted in Table 2.3 is widely associated with Yorkshire dialect and John Townend's near-categorical use of thou [ða] as a subject form, thee [ðiː, ðɪ] as an object form and thy [ðɪ] and thine [ðɑːn] as possessives mirror SED responses in Skelmanthorpe at IX.9.9, VI.14.2, IX.8.6 and V1.5.17 respectively. Recent surveys and present-day BL sound recordings suggest this feature is now recessive, although it survives among older speakers in some areas and, as with many dialect variants, remains available as an identity marker or to be used for effect in particular contexts – witness the use by Leeds band Kaiser Chiefs in their 2004 single ‘I Predict A Riot’: watching the people get lairy it's not very pretty I tell thee. It is intriguing that John Townend uses the more widespread/standard possessive your [jʊə] initially, which is recorded alongside thy in the SED Skelmanthorpe data (IX.8.5). An editorial note for Skelmanthorpe (IX.9.5) indicates that the informant judged thou to be ‘older’ than you, so perhaps John Townend's use of <th-> forms here might be performed. As the original source text for the Parable probably contained <th-> forms, this might also have influenced John Townend to retain them as indicative of Shelley dialect. Certainly, the regularity with which <th-> forms occur in the SED Basic Material (Orton et al. Reference Orton1962–1971) is not reflected in the SED sound recordings, although there are several possible explanations for this. Firstly, as interviewees, the SED informants seldom ask questions of the fieldworkers, and secondly, as local familiar forms, the <th-> variants might not have been deemed appropriate by informants speaking to fieldworkers whom they presumably considered strangers or social superiors (despite the warm and lasting friendships that often developed during fieldwork). Where <th-> forms do occur in SED sound recordings, they generally arise in reported speech, as is the case in the sound recording with Wibsey Dyson (I says, ‘art thou late Bill or I'm soon or I'm soon and thou late?’, I says, ‘sithee, Bill, what's yond?’ and she says, ‘hast thou seen my cat?’;).

Table 2.3 records three non-standard past verb forms captured in the Parable recording. John Townend's consistent use of third person singular were remains a distinctive feature of speech locally as confirmed by numerous contemporary recordings, and not surprisingly this is the form recorded by SED fieldwork in Skelmanthorpe (VIII.9.5), but what is striking here is John's alternation between a strong and weak variant of were:

1. 1. a. there were a [wɒɾə] man who had two sons

   2. b. when he were [wɒ] yet a great way off his father saw him

   3. c. now his elder son were [wə] working in the closes

   4. d. then he were angry [wəɾaŋgɾɪ] and wouldn't go in

   5. e. this thy brother were [wə] dead and is alive again he were [wə] lost and is found

This distinction was noted by Wright (Reference Wright1892: 161) but is not apparent in SED published data, as the questionnaire only elicited unstressed past ‘be’ (VIII.9.5), something Petyt considers an oversight (Reference Petyt1985: 194) as he too regarded this contrast as a noteworthy feature of West Yorkshire dialect. In the SED's defence, no questionnaire can possibly provide comprehensive coverage of every aspect of variation, but the extensive supplementary data provided by the SED audio files and ‘Incidental Material’ (Orton Reference Orton1962: 17) plugs many gaps, and the recording in Skelmanthorpe does in fact capture the two forms very clearly:

1. 1. a. when I got nearly up to it I heard it were [wɒ] saying summat

   2. b. when I got a bit gainer [= ‘nearer’] I could hear what it were [wɒ] saying

   3. c. a woman'd comed out in her nightdress and and and were [wə] seeking her cat

   4. d. in them days there no there were [wə] no mechanism in the pit hardly everyth every every bit of coal were hand-gotten [wəɾandgɒʔn̩]

Recent BL recordings in South Elmsall (BL shelfmark: C900/08505 – and it were [wɒ] I just forget how it were [wɒ] worded and it and it were [wə] ‘True Labour’ or ‘Socialist’ or whatever) and Castleford (BL shelfmark: C1190/19/01 – my eldest lad didn't want to do anything else all he wanted to do were [wɒ] come into the business which he did at the age of sixteen but let me say this he'd been working since he were eight [wəɾ ɛɪt]) confirm that this distinction continues to exist for some speakers locally.

The SED questionnaire did not elicit past forms of either ‘get’ or ‘tell’, but the fieldworker's notebook for Skelmanthorpe held in the Brotherton Library at the University of Leeds (LAVC/SED/2/2/6/31: 114) records telled [tɛld] in the ‘Incidental Material’ (Orton Reference Orton1962: 17), and gotten occurs repeatedly in the sound recording with Wibsey Dyson: we'd a grave to dig and my mate had gotten [gɒʔn̩] there before me; before I got to my work I'd gotten [gɒʔn̩] a right sweat on, you know and in them days there no there were no mechanism in the pit hardly everyth … every bit of coal were hand-gotten [andgɒʔn̩]. Both forms remain present in dialects in the north of England, although telled is arguably increasingly rare.

The final two features noted in Table 2.3 relate to morphonological processes applied in connected speech to the definite article and certain prepositions. Definite article reduction (DAR) – a contracted pronunciation of the word the – is a distinctive and relatively stable feature of speech throughout Yorkshire and some neighbouring counties. Often rendered in popular representations as if to imply people say t’ book or even omit the article altogether, DAR is in fact a complex phonetic process with speakers realising ‘the’ using a variety of allophones of /t/, most commonly a glottal stop, or, in certain phonetic environments, very occasionally /ʔ∩è/. John Townend frequently uses a glottal stop in several grammatical contexts – preceding a subject noun (not many days after the young man [ʔ jʊŋ man] gathered all his belongings together), before an object noun (the husks that he fed the swine [fɛt∩ʔ swɑːn] with) and in combination with a preposition (a great famine came over the country [ɒvəʔ kʊntɹɪ]). This articulation occurs frequently in the SED recording in Skelmanthorpe (e.g. and I were in setting the key [sɛtɪnt∩ʔ kɛɪ] into the lock [ɪntəʔ lɒk] and I turned me round) and persists as the most common realisation today. The fricative variant John Townend uses in the phrases with the husks and to the house is worth further scrutiny. Unfortunately, there are no examples in the Parable text of ‘the’ preceding a noun with vowel onset, but as noted above, John Townend deletes initial /h/ meaning both husks and house effectively have vowel onsets thus prompting his use of the variant /ʔ∩è/. The definite article was consistently recorded as [t] during SED fieldwork in Skelmanthorpe, even preceding a vowel at in the oven (V.6.6) and after H-dropping at the heat (VI.13.6), although [è] is noted for both these prompts in nearby Holmbridge. There is also one example of a similar pronunciation in the sound recording with Wibsey Dyson (I used to leave the house [tèæəs] at half past five and I'd to be at s … at pit at six o'clock and we worked while two). As noted above, this articulation is increasingly rare and does not feature in many present-day BL recordings, although there are isolated examples, such as in a recording from Leeds in exactly the same phonetic environment (BL shelfmark: C900/08627 – oh the steel works [ʔ stiːlwəːks] were massive the Hunslet steelworks [èʊnslɪʔ stiːlwəːks] it were a massive place).

Finally, we turn to the prepositions listed in Table 2.3. In the case of in, of and with, John Townend produces a contracted form in which he deletes the final consonant. A contracted form for of is common in vernacular English worldwide and is, not surprisingly, well documented for Skelmanthorpe (e.g. at III.13.3) and in the sound recording (e.g. every bit of [ə] coal were hand-gotten) and numerous present-day recordings. The contracted form of with is also common in a number of varieties of British English and certainly persists locally. John's articulation with a lax vowel preconsonantally is recorded in the SED for Skelmanthorpe (IV.8.6) and in the corresponding sound recording (with [wɪ] fear and trembling I went my way towards this here […] apparition in white). The deletion of the nasal consonant of in is probably less common nowadays, but it is certainly evident in the SED (V.6.6) and occurs frequently in the recording with Wibsey Dyson:

1. 1. a. with fear and trembling I went my way towards this here […] apparition in [ɪ] white

   2. b. I'd to wade nearly knee-deep in [ɪ] water and I'd to work in [ɪ] water all the day through

   3. c. I started a-working at a shilling in the [ɪʔ] day shilling in the [ɪʔ] day for going down the pit

Finally John shows a predictable contrast between a weak vowel for the preposition to preceding a consonant (near to [tə] the house) and a stronger vowel preceding a vowel (he arose and came to his [tu ɪz] father), but also a more marked pronunciation (and the servant said to him [ti ɪm] ‘thy brother has come home’;) that is perhaps nowadays more readily associated with speakers further north, especially Scotland. There is no record of this realisation in the Skelmanthorpe data, but there are several examples in SED sound recordings in Yorkshire, although, interestingly, only in localities in the North and East Ridings with the closest example to Shelley in Rillington (BL shelfmark: C908/46 C8 – when he got to the [tɪt] far end of the field as he was mowing across there was a lad leading a Galloway over the field). However, a single token in a BBC Voices Recording in Castleford (no, not today, love, that's all gone but uh like Andrew's just said to [tɪ] you you'd have more walnuts and Brazil-nuts than you would filberts because people (they're more popular) didn't care for a lot of filberts) perhaps adds some validity to John Townend's use.

2.9 The Value of British Library Archival Sound Recordings to Present-Day Audiences

As noted above, we rely principally on surviving documentation and Brandl's testimony to determine whether the BLBCR recordings are faithful records of the dialects they represent. Clearly an element of performance was encouraged, making it difficult to judge whether individual features capture an informant's typical usage or reflect forms imitative of broader or older dialect speech back home. We might perhaps suspect an artificially inflated number of dialectal forms – lexical variants, non-standard grammatical constructions and archaic or localised pronunciations – chosen and rehearsed in advance, rather than indicative of widespread usage. Whether individual utterances are contrived or intuitive, one might nonetheless accept them as valid evidence of a speaker's active and passive knowledge of his local dialect, but treat them with caution in terms of authenticity. However, a detailed analysis of the recording with John Townend bears comparison with other sources for which we have greater provenance and confirms the collection as a rich resource for linguists wishing to investigate early twentieth-century English dialects. As an internationally acclaimed survey, the SED is rightly acknowledged as a reliable source of dialect data, but the analysis here of the recording with Wibsey Dyson should hopefully also raise awareness of the availability and value of SED sound recordings.

The recordings of John Townend and Wibsey Dyson offer linguists insights into phenomena unrecorded in the SED published volumes but captured in contemporary written accounts – as demonstrated most notably here in the case of John Townend's and Wibsey Dyson's realisation of /r/ and third person were. Online access to a further 65 BLBCR recordings and 287 SED audio files at bl.uk/sounds therefore provides an essential teaching and learning resource for re-examining other less well-documented features of British dialects. A similar exercise as undertaken here would be informative for many other varieties of English. Equally importantly, audio collections are open to interpretation by a wider range of audiences in that non-specialists can respond much more readily to a sound recording than to a purely written description of speech. To the inexperienced researcher, academic descriptions of broad twentieth-century dialect, such as those contained in the SED, can be extremely daunting, as they inevitably require familiarity with IPA transcription. Hearing the actual voices of real speakers will not only support academic enquiry, but also open up respected linguistic data sets for the first time to teachers, students, actors and the general public. The BL is committed to supporting knowledge transfer between higher education research and the general public and has enjoyed considerable success in meeting growing public demand for knowledge about English accents and dialects by presenting and interpreting its linguistic audio content to new audiences. The Voices of the UK project (Robinson et al. Reference Robinson, Herring, Gilbert, Upton and Davies2013) will develop access further by creating detailed linguistic descriptions of the dialect content at bl.uk/sounds, with the ultimate aim of making the sound recordings browsable by linguistic criteria. This will mean that in future students unfamiliar with phenomena such as, for instance, the dialectal goat diphthongs discussed here for West Yorkshire will be able to locate and audit authentic spoken examples to assist them with interpreting and evaluating more challenging dialect data. Finally, one should simply appreciate the unique value of these audio archives as the earliest known sound recordings of the speech of ‘ordinary’ folk.