Speech reflections in Late Modern English pauper letters from Dorset

The overall aim of this article is to show that pauper letters are a valuable, but as yet largely untapped resource for historical dialectological research. Offering a case study based on 31 poor-relief applications sent by 10 individuals to parishes in Dorset between 1742 and 1834, the article aims to identify regional variation, especially as associated with Dorset and/or the Southwest of England more generally, by comparing variant spellings and morphosyntactic usages contained in the letters with features listed in modern dialect surveys (mainly Wakelin 1986; Altendorf & Watt 2008; Wagner 2008), as well as in Dorset poet William Barnes’ Dissertation and the reconstruction of his idiolect by Burton (2013). It is possible to isolate 297 occurrences of 52 different phonological and morphosyntactic features in the pauper letters; 11 of these features are salient across the letter selection (i.e. represented by at least three paupers) and are suggestive of the provenance of the letters. The article also offers surprising findings such as the absence of the prototypically Southwestern fricative voicing, features unrecorded by modern synopses (e.g. unmarked possessive), and the presence of a feature (-ind for -ing) which had fallen out of common use in the fifteenth century.


Introduction
Dialectologists interested in regional variation in eighteenth-and nineteenth-century England are faced with an 'extremely poor data situation' (Kortmann & Wagner 2010: 290).In the absence of voice recordings as collected for the Survey of English Dialects (Orton 1962-71), for instance, scholars have consulted different kinds of written records, in particular metalinguistic comments by contemporaries from the late eighteenth century onwards (for a synopsis see Ihalainen 1994), as well as dialect writing such as dialect literature and literary dialect (e.g.Cooper 2023;Hodson 2023;Ruano-García 2023), which according to Honeybone & Maguire (2020: 3) consciously 'intends to represent a non-standard dialect in written form, at least to some degree and in some portion of a text'.However, dialectal evidence in these valuable sources is mediated through or provided by a usually well-educated writer (see also Fairman 2007c: 192;Ruano-García 2023) and does not document spontaneous, unmonitored speech.Similarly, court records as documented in the Old Bailey Corpus can also contain an element of editorial intervention (Dossena 2010: 14;Grund 2023).
In light of these constraints, Dossena (2010: 5) stresses 'the importance of studying authentic manuscripts, as it is only when we access original texts and manuscripts that we can go beyond the layers of interpretation added by later editors'.Characterised by features of orality, a fact which has long been recognised (e.g.Biber & Finegan 1989: 512), letters seem to be a promising resource for the study of regional variation.Yet even here it can be difficult to identify authentic voices relating reflections of speech.The Late Modern English period was influenced by the prescriptivist movement and the codification of a standard in pronouncing dictionaries and grammars (see Beal 2004 for an overview ;Beal 2023;Wiemann 2023), to the extent that 'all encoders [i.e.those who performed the physical act of writing], even the least schooled ones, normally attempted to imitate "standard" models, especially at the beginning and at the end of the letter' (Dossena 2010: 19; see also Gardner forthcoming on formulaic language in pauper letters).
It is precisely these 'least schooled' letter-writers which the present article focuses on, with the aim of establishing pauper letters as a source for English dialectal studies of the eighteenth and nineteenth centuries by offering a case study based on letters sent to parishes in Dorset between 1742 and 1834.The letters under investigation were collected and transcribed for the project The Language of the Labouring Poor in Late Modern England (LALP).Pauper letters are mostly applications for financial support (out-relief) written under the Old Poor Law by paupers who had moved away from their parish of legal settlement (on the term 'pauper letters' and differences between these and pauper petitions see Gardner forthcoming).A parish of legal settlement (or home parish) could be the applicant's place of birth, but settlement could also be gained elsewhere through marriage, apprenticeship, employment as an agricultural labourer for a year or rental payments exceeding £10 per year (Whyte 2004: 280).Fairman (2007a: 275) states that '[t]he language of these letters cannot be called dialect' and views their informational merit so critically (2006: 84-5) that scholars such as Kortmann & Wagner (2010: 290-1) and Wagner (2012: 936) are doubtful as to the value of pauper letters for dialectal studies.2However, Fairman himself does identify what he considers to be a small number of regional features in some letters, such as /w/ for /v/ as in werry (Fairman 2007b: 34, 38; see also Fairman 2006: 84;2007c: 191).The present case study on regional variation in Dorset intends to show that pauper letters do represent an important, largely untapped resource for the study of historical dialectology, despite some of the challenges associated with this data type.
One major challenge concerns the question of authorship and authenticity of pauper letters.The labouring poor often received only limited education, leaving elementary schooling (which became compulsory across England only in 1880) at an early age to earn wages before having received any significant training in writing (Stephens 1998: 2;Gardner 2023).An important question to ask is therefore whether the person in whose name a letter is signed and sent (i.e. the sender) is also the person who penned the letter, or whether somebody else encoded it.The 'writer-sender problem' (Nobels & van der Wal 2009;2012: 349) will be discussed in more detail in section 2 because this issue has a bearing on the localisation of regional features.Since the paupers represented by the letters under investigation no longer lived in their home parishes it is crucial to know whether the applicants wrote, or could have written, the letters themselves, or, particularly in the case of paupers living outside the Southwest, whether an encoder from their current domicile was involved.An overview of the data selection is provided in section 3.
A second challenge concerns the identification of speech reflections in pauper letters.To this end, the case study will draw, on the one hand, on a number of studies, both historical and modern, which provide lists of regional features found in Dorset and the Southwest in the nineteenth and twentieth centuries.Section 4.1 introduces these studies, including A Dissertation on the Dorset Dialect of the English Language, published by poet and philologist William Barnes in 1844, ten years after the last pauper letter investigated.Together with modern sociolinguistic work based on twentieth-century data, Barnes' dialect descriptions and the synopsis of both types of sources in Burton (2013) serve as a point of comparison for the analysis of potential regional features in pauper letters.On the other hand, the case study examines so-called occasional spellings (Wyld 1936: 65;Penzl 1957: 202;Stenbrenden 2016: 12) found in pauper letters which can be suggestive of how writers pronounced a particular word and thus indicate regional variation (Wyld 1936: 65;Matthews 1937: 4;Penzl 1957: 198).According to Penzl (1957: 198), '[t]he term "occasional spellings" suggests a minority-type of orthography which occurs together with the majority-type in identical or contemporaneous texts'.In the present article an alternative term will be used, namely 'variant spellings', which does not convey a judgement regarding frequency.This is done for two main reasons: firstly, the present study focuses on the identification of regional features only, in order to ascertain the value of pauper letters for dialectal research, and a follow-up study is required to investigate frequency patterns of these features; secondly, occasional spellings may be infrequent overall when compared with spelling practices in all available contemporary texts, but they may nevertheless represent majority-types for individual writers.On the basis of a short text extract section 4.2 illustrates how variant spellings and regional features are identified in pauper letters, while section 4.3 reflects on the robustness of dialectal data drawn from pauper letters.
Section 5 then provides an overview and analysis of the regional features which could be identified in the dataset, many of which are associated with Dorset and/or Southwestern dialects more generally.Rather than being scanned for a pre-defined 493 SPEECH REFLECTIONS IN PAUPER LETTERS set of regional features, the letters under investigation are carefully studied in their entirety to detect a wider range of regional features, with respect to both phonology and morphosyntax.This approach makes it possible to identify speech reflections otherwise unrecorded in secondary literature.Dialect lexis is not subject to investigation in this article because it appears so infrequently in pauper letters, at least in part on account of their usually specific and restricted content.The article will close in section 6 with a brief outlook.
Overall, the evidence gathered from variant spellings will allow us to draw conclusions about the presence of certain regional features, whether in individual or various writers, and show that even a single letter may reveal a noticeable number of different reflections of speech.Letters from the labouring poor, and the speech reflections they contain, help close gaps in the diachronic and geographical coverage of primary source material in current research on Late Modern English dialects, providing ample evidence for the early nineteenth century and in some cases even reaching far back into the first half of the eighteenth century.

Authorship and authenticity of pauper letters from Dorset
The following describes different strategies of how the authorship and authenticity of pauper letters can be assessed.As mentioned earlier, this is particularly relevant for paupers who no longer lived in Dorset or even the Southwest, for an encoder with a different dialectal background may not have represented the applicant's variety in their writing.All examples in this section are taken from letters sent to parishes in Dorset which were transcribed as part of the LALP project as faithfully to the original as possible in terms of spelling and self-corrections (see also Auer et al. 2022;Gardner et al. 2022;Gardner 2023), and the focus is placed on those letters which will also be analysed in section 5.
Text-internal evidence in pauper letters pointing towards authorship is generally rare (Fairman 2007c: 169).While third-party encoders are known to have been involved on occasion in the writing of poor-relief applications, 'professional writers … were apparently only very rarely resorted to' (Sokoll 2001: 65).In consequence, earlier studies conclude that pauper letters are largely authentic representations of the language of the labouring poor in that 'people either wrote their letters themselves or had them written by someone who was close to them' (Sokoll 2001: 65; see also King 2019: 36) and therefore came from a similar background.These appraisals are supported by evidence from the letters sampled for the present article, as shown in the following.
For the determination of authorship, three different kinds of clues are instrumental for proving that a letter is autographical (see also Gardner forthcoming).Firstly, the production circumstances may be described in a letter in such a way as to leave no doubt concerning the writing skills of the applicant.Ann Weakford, for instance, writes in (1) about the physical difficulties she experiences when trying to hold her pen:

494
A NN E -CHRISTINE GARDNER (1) I myself been down in the fefver this eight wicks and could case hold my penn to rite this to you I am so weak (Ann Weakford, Frome, 5 November 1789, DO/BF/6)3 A second type of clue is provided by persons who witness the applicant's handwriting by adding their name to the letter in their own hand.In (2) Augustine Morgan introduces Joseph Barratt as 'witnes', who acts as letter-carrier for Morgan and whose name also appears in an earlier letter (DO/BF/20).
(2) witnes our ands Augustne Morgan Joseph Barratt (Augustine Morgan, Beaminster, 9 November 1805, DO/BF/21) A third type of evidence comes from additional archival sources such as marriage registers, where married applicants either signed their name, if they could write, or added a mark like a cross instead of their name, if they could not.Regrettably, this type of evidence requires human resources outside the scope of the LALP project, and it only applies to paupers who actually got married.For Augustine Morgan, however, Fairman (2007c: 179) was able to locate the relevant entry noting the marriage between Morgan and Sarah Dike in Beaminster on 20 November 1785, where Morgan signed his name and his wife added a cross.Both Morgan's and Weakford's letters are clearly autographical.Unfortunately, these two are the only applicants writing to Dorset parishes where such incontrovertible proof exists.Nevertheless, there is also an indirect clue which suggests that the sender of a letter was also its encoder.Charls Ann Green, for instance, sent eight letters from London to her home parish Wimborne between 1818 and 1826, which are all in the same less experienced hand characterised from a non-linguistic perspective by uneven lines and inkflow (see also Auer et al. 2023).King (2019: 37) argues that in these circumstances the letters are probably autographical since '[i]t is implausible that the poor would have found the same person to write for them over these periods and on such a sustained basis' (see also Sokoll 2001: 64).
In contrast to Green's case, when several hands are involved in a series of letters by the same sender, the letters are regarded as non-autographical.That the sender could not write themselves is particularly evident when the handwriting of the signature also differs across the series.In the case of John Bartlett, who sent four letters from Poole to Beaminster in 1834, two different hands encoded two letters each and in each set the signature matches the handwriting of the body of the letters.What is important to note, however, is that all four letters are nevertheless authentic representations of lower-class writing.Bartlett's encoders were not professionals with experienced handwriting, but probably from his social circle, and had relatively limited schooling as shown by the irregularity of the lines and uneven height of characters (how authentic letters can be identified non-linguistically is described in more detail in Gardner (forthcoming), which also offers linguistic evidence in support of this procedure).That said, the first encoder (DO/BE/2, 5) seems to have received more training than the second (DO/BE/ 3, 4) where orthography conforms to a much higher degree to conventions.Two different encoders also prepared letters for Susannah Fuller (DO/SM/3, 1 May 1811; DO/SM/4, undated), with the first offering somewhat neater handwriting than the second.
The level of training evidenced in the handwriting can also be used as an indirect clue for smaller sets of letters or individual applications.Although it is impossible to determine, without additional evidence, whether a letter is autographical or not, it can be regarded as authentic if the handwriting betrays limited training (e.g.uneven lines, graph formation or inkflow; see also Gardner forthcoming).This criterion applies to two letters in the same hand sent by James Summers from Fort Monckton to Corfe Castle between 1783 and 1785 (DO/CC/1, 2), two letters sent by an unnamed woman from Poole to Wimborne (DO/WM/14, 26 February 1827; DO/WM/15, undated), and one letter sent by James Headen from Plymouth to Blandford Forum on 2 June 1810 (DO/BF/26).Finally, an interesting case is presented by a married couple where both husband (DO/BF/1, 27 January 1742) and wife (DO/BF/4, 6 August 174X) each sent a letter from Bristol to Blandford Forum.Both letters are written by different, untrained hands, the wife signing in her husband's name, which was not unusual (Fairman 2007c: 169).Considering that the letters do not contain any indication that either spouse was absent during the time of writing (for instance, the husband might have been travelling across the country, i.e. been 'on tramp', to search for work; see Gardner et al. 2022), it seems feasible that they each wrote their own letter themselves.
Identifying regional variation in pauper letters may assist us in determining authenticity.Gardner et al. (2022) show that the regional features observable in Green's (probably autographical) letters point towards a linguistic anchor in the Southwest, meaning that the encoder of her letters must hail from this larger dialect area.With Green living in London with her family during the eight-year period in which the letters were sent, the encoder can in all likelihood only be Green herself or someone from her familybut since she repeatedly writes about her husband and his injured hand as well as her three children, it is highly plausible that Green wrote the letters herself.Isolating regional features typically found in Dorset, or the Southwest in general, in a wider range of pauper letters as will be done in section 5 will contribute to our understanding of regional variation in the Late Modern English period and help us gather more information concerning the encoders of these letters.

Data
The LALP corpus (under construction) currently contains 71 letters (13,270 words) by members of the lower ranks with a home parish in Dorset.For reasons of space, the focus in this article lies on all letters classified as autographical (Morgan, Green, Weakford), as well as a selection of letters regarded as authentic, i.e. representative of lower-class writing, which are particularly interesting with respect to the speech reflections they contain.Most letters sampled for the LALP project and this study date

496
A NN E -CHRISTINE GARDNER from the period 1795 to 1834, with 1795 seeing a number of modifications to the law and 1834 marking the introduction of the significantly different New Poor Law.In the correspondence from this period, in particular, applicants emerge as strategic letter-writers negotiating with their home parish for out-parish relief (Sokoll 2008;King 2019).However, a small number of poor-relief applications from the eighteenth century survive as well.The uneven historical survival of source material is reflected in the diachronic distribution of the letters selected for the present study in table 1.
In total, 31 letters by ten individuals (4,803 words) will be investigated, all of whom had a parish of legal settlement in the eastern half of Dorset but had moved away (see table 1).Only Morgan, Bartlett and the unnamed woman from Poole still had a domicile in Dorset (on migration patterns of paupers see Gardner et al. 2022).In the case of Fuller and Bartlett it will be interesting to see whether the letters of the different hands show similar speech reflections or differ from each other.If congruences are found, this could indicate that a particular feature is salient in the communityor that the writer was influenced by the applicant during dictation.However, a recent study (Gardner forthcoming) on formulaic sequences in non-autographical letters could find no evidence of any linguistic influence of applicants on the encoders of their letters.
To conclude, all letters discussed in the previous section and sampled for the present study are either autographical or at least authentic, i.e. representative of lower-class writing, and will be analysed in section 5 with respect to reflections of speech.

Identifying regional variation in pauper letters
For the identification of dialect features in pauper letters this article draws, firstly, on both modern and historical sources describing regional variation in the Southwest and, secondly, on textual evidence, which includes the study of variant spellings.These two approaches will be described and illustrated in more detail in sections 4.1 and 4.2, respectively.The importance of combining the analysis of variant spellings with other pertinent evidence, such as that drawn from sociolinguistic studies, is underlined by Stenbrenden (2016: 30-1), who also provides a synopsis of earlier scholarship on this matter.Section 4.3 discusses issues concerning the distinctiveness, frequency and salience of regional features identified in pauper letters, which are relevant for the analysis in section 5.

Dialect descriptions from the nineteenth and twentieth centuries
The main nineteenth-century source consulted for the present investigation is William Barnes' Dissertation (1844), which was published only ten years after the latest letter investigated.Not only did Barnes write dialect poems (Burton 2013(Burton , 2017a/b/b), but also treatises on his native dialect, which are very valuable for historical dialect studies.His Dissertation was prefaced to his first collection of poems, a Glossary followed in 1847, and a grammar in 1868 entitled A Grammar and Glossary of the Dorset Dialect with the History, Outspreading, and Bearings of South-Western English.Incidentally, many 497 SPEECH REFLECTIONS IN PAUPER LETTERS words from Barnes' Glossary were included in the English Dialect Dictionary (Burton & Ruthven 2013: 399;Wright 1898Wright -1905)).In his Dissertation, Barnes identifies the location of his own regional variety: 'The dialect in which he writes is spoken in its greatest purity in the villages and hamlets of the secluded and beautiful Vale of Blackmore' (Burton & Ruthven 2013: 19).Barnes also draws attention to the fact that he writes in a rural dialect which is not spoken by the poor living in towns: 'He needs not observe that in the towns the poor commonly speak a mixed jargon, violating the canons of the pure dialect as well as those of English' (Burton & Ruthven 2013: 19).
From the perspective of the pauper letters under investigation the final observation is very poignant.Following Barnes, the paupers are unlikely to speak a 'pure dialect', having migrated within and outside Dorset.Consequently, one cannot expect to find all dialect features in their letters which the poet describes, even if they were born in the Vale of Blackmore.It is also important to bear in mind that the variety Barnes describes is not necessarily representative of this area, but may merely be his own idiolect.As Burton & Ruthven (2013: lxviii) note, this idiolect 'was likely to have been modified by schooling by the time Barnes grew up speaking it'.The poet was born into a working-class family and through self-education eventually became a respected school master with a command of several languages and a bachelor's degree from Cambridge, giving lectures into his late seventies (Wrigley 2004).
Despite these caveats, Barnes' works are still helpful in that they document selected regional features which in his view are the most salient.Two examples are given to illustrate how Barnes describes regional features.One feature mentioned by Barnes in Barnes divides vowels and diphthongs into close and open sounds, with four long and short sounds each (Burton & Ruthven 2013: 9).The sounds referred to in the previous quotation correspond to [eɪ] ('third long close sound'), [ɑ:] ('first open sound') and [i:] ('first close sound') in IPA.The features used for illustration are two of several which Barnes marks typographically in his poems, elisions being marked with an apostrophe and the pronunciation of the FACE subset with a circumflex.This can be seen, for instance, in the following line from Martin's Tide, written for the feast day of Saint Martin: 'An' scores o' tricks have we a-plây'd' (Burton 2013: 296).
Barnes' Dissertation usefully supplements major dialect studies which do not go as far back in time.For the identification of Southwestern features this article also draws on Burton (2013), who published editions of Barnes' poems together with a detailed reconstruction of the likely sound of his dialect, as well as Wakelin (1986), Wells (1982), Ihalainen (1994), Altendorf & Watt (2008) and Wagner (2008).All these studies at least in part rely on data from the Survey of English Dialects.Its findings on Dorset (Orton & Wakelin 1967a, 1967b, 1968) provide additional insights, as does Wright's English Dialect Grammar (1905), but these additional sources are consulted for isolated features only.Ellis' On Early English Pronunciation (1889) is the earliest dialectal survey approaching modern principles collecting data from young and old speakers c.1865-88 (Wagner 2012: 923) and deserves more detailed treatment elsewhere.

Analysing variant spellings in pauper letters
Paupers applying for out-relief had usually received only limited schooling, as mentioned sections 1 and 2, and this could significantly impact their spelling.As Allen (2015: 211) notes: Some non-standard spellings suggest that the writers had only ever heard the word spoken and had not seen, or did not remember seeing it written or printed, so their spelling was a representation of the sound of the word as they pronounced it, based on their understanding of the orthographic system of the language.
In other words, variant spellings can be indicative of 'subconscious interference from speech habits' (Stenbrenden 2016: 33; original emphasis).An excerpt (see figure 1) from one of the most prolific letter-writers in the sample, Augustine Morgan, will be used in the following to illustrate (a) the wealth of speech reflections which can be found in pauper letters, and (b) which type of variant spellings are excluded from analysis.In the transcription in (3), variant spellings which are considered to represent regional features are highlighted in bold and numbered with Arabic numerals; those which are excluded from further analysis are in italics and numbered with Roman numerals.
(3) S r I hope that you will Sind Types of variant spellings highlighted in (3) which are not considered in the dialectal analysis include doubled consonants as in 'Gett ' [i] or alternative renditions of vowels [iii, iv] where the spellings do not (necessarily) reflect a pronunciation differing from what is expected.For instance, <ou> used for /u:/ in 'dou' instead of <o> can represent this vowel in e.g.ghoul, and <ey> for /ɪ/ in 'aney' instead of <y> reflects this sound in e.g.money.
Likewise excluded from consideration are words spelt as two orthographic units instead of one, such as 'a bel' [ii], as well as capitalisation (the numerous examples in (3) include e.g.'Sind', 'Som' and 'Money'), which do not impact the pronunciation of the individual sounds.Lastly, an interesting case is presented by 'hopet ' [v], where the final consonant is written with a <t> instead of a <d>, reflecting the assimilation process to the preceding voiceless consonant [p] in speech.Since this does not represent regional usage, it is not included in the present study, but could be of interest to researchers investigating connected speech in the past.
4.3 On the distinctiveness, frequency and salience of regional features identified Some features like multiple negation, but also h-insertion, th-fronting and g-dropping, are also socially stratified, at least in modern times, and multiple negation in particular can be considered so universal that it 'does not exhibit regional contrasts or a clear-cut regional distribution' (Wagner 2012: 931).The remaining features are listed specifically in the dialect descriptions focused on Dorset and/or the Southwest mentioned in section 4.1 (for references to specific features see sections 5.1-5.3).However, this of course does not preclude features from being common in other regions of England as well, as is the case, for instance, with elision of final /d, t/ (Wyld 1936: 303).As indicated previously, this article does not intend to make claims on whether certain features are unique to Dorset and/or the Southwest, but rather aims to underline the richness of dialectal data which can be drawn from pauper letters, using Dorset as an example, and to establish such texts as valuable sources for the historical study of regional variation.Additional data, also from the surrounding counties and across England, is necessary when attempting a more fine-grained localisation of dialect features.
Nevertheless, it is important to critically discuss the reliability of the data gathered from the letters, particularly when it is based on variant spellings.Considering that paupers usually did not receive much education, how reliable are they, or rather their spellings, as witnesses of speech reflections?For instance, how reliable is Morgan's spelling <o> in Sonday as an indicator for the backing of the STRUT vowel, when he uses the same <o>-spelling for STRUT in Som for 'some' (see example (3) in section 4.2)? 9 In order to achieve a degree of robustness concerning the identification of regional variation, the analysis relies on the notion of salience.A distinction is made between more prominent, or salient, features which are found in letters from at least three different paupers, and less salient features which are attested by only one or two writers.In addition, for each feature the number of observed instances in all letters by all paupers will be provided as well.Comparing modern and historical dialect data, Kortmann & Wagner (2010: 284) observe that 'high text frequency is not a necessary prerequisite for salience' and that 'low overall frequencies of … comparatively rare features … do not necessarily stand in the way of conclusions drawn from regional distributions'.In the context of the pauper letter data this means that low-salience features (such as those listed in table 2) are worth recording and should not be disregarded as 'accidental spellings'.Furthermore, a feature that is not shared by many writers may still be dominant in the speech of an individual (e.g.THOUGHT ≈ GOOSE for Morgan in table 3).However, it lies outside the scope of this study to identify all instances where a dialectal feature could potentially have occurred, i.e.where a variant spelling might have been found, but was not employed.A follow-up study based on a larger dataset is required to examine frequency patterns and to what extent variant spellings may be linked to specific lexical items.The merit of such an investigation will be briefly illustrated in section 5.4.

Reflections of speech in pauper letters from Dorset
This section provides an inventory of speech reflections observable in selected pauper letters with links to Dorset, beginning with consonants (section 5.1), followed by vowels and diphthongs (section 5.2), and morphosyntax (section 5.3).Features deemed prominent or salient in the data, i.e. those which are attested in the writings of at least three different paupers (see section 4.3), are discussed in more detail than features only attested in the letters of one or two paupers.Section 5.4 provides an overview of the features identified and discusses the reliability of and challenges surrounding data from pauper letters for historical dialectological research.

Phonology: consonants
Southwestern dialects are traditionally rhotic (Burton 2013: 582) and speakers may even insert /r/ after vowels when it is not expected etymologically (Wells 1982: 341-3;Altendorf & Watt 2008: 214, 218).This kind of hyper-rhoticity is alluded to by Barnes ( §27 in Burton & Ruthven 2013: 10) when he notes that in words like pillow the ending becomes -er.In writing, rhoticity can only more safely be assumed through the presence of hyper-rhoticity.Six of the ten individuals under investigation show r-insertion with spellings such as 'wagers' (for wages; Summers and Headen), 'Deart' (dead; Unnamed woman), 'Susaner' (Susanna; Fuller Hand 2) and 'Elizar' (Eliza; Bartlett Hand 2), as well as 'a torll' (at all; Green) and 'Laſtorll St' (Laystall St; Green).The examples range from Summers writing in 1785 to Bartlett in 1834, suggesting that this feature remains salient over time, as confirmed by modern data.Only one of these examples could be an instance of intrusive r, i.e. 'Elizar', which is followed by a word with an initial vowel, 'is'; in all other cases r appears before a consonant.
Similarly, the elision of final /d/ after an alveolar nasal mentioned by Barnes (section 4) is a ubiquitous feature, and according to Wells (1982: 344) extends to /t/ as well as other contexts.Elision of /d, t/ is attested in letters by five paupers sent between 1742 and 1827, for instance in 'presan' for present (E.Chappell), 'an' for and (3x) and 'derick' for direct (Chappell's wife), 'eldes' for eldest (Weakford), 'han' for hand, 'Sho' for showed and 'nex' for next (3x) (Morgan), as well as 'an' for and besides 'husban' for husband (4x) (Unnamed woman).
Five consonant features appear in the letters which are not necessarily regionally restrictive, but (also) social variants, certainly from a modern perspective.The most frequent are h-dropping (15 instances by five writers), e.g.'ands' for hands or 'is' for his (Morgan), and hypercorrect h-insertion (40 instances by four writers), e.g.'ham' for am (Morgan, Headen, Fuller Hand 1, Green) or 'honley' for only and 'hower' for our (Headen).Wakelin (1986: 31) states that Dorset belongs to an area where /h/ was traditionally retained.The higher proportion of h-insertion could speak in favour of this, but the presence of h-dropping suggests that social variation concerning this stigmatised feature was already at play in the late eighteenth and early nineteenth centuries.
G-dropping with preceding KIT lowering (see also section 5.2.1) is present in three writers who must have pronounced /ɪŋ/ as /ən/ (Burton 2013: 534; Barnes Diss.§42 in Burton & Ruthven 2013: 12) as relayed by spellings such as 'Shilens' (Morgan) and 'shillns' (Headen) for shilling, as well as 'Writen' for writing (Unnamed woman).Contrary to Burton and Barnes, however, g-dropping with concomitant KIT lowering is evidently not restricted to present participles and verbal nouns and is therefore not treated here as a morphosyntactic feature, but as two separate phonological features.G-dropping without KIT lowering is also attested, for instance in 'thins' for things (Morgan) and 'shillin' (Unnamed woman).Two rare examples of th-fronting are provided by the pauper letters, in 'Boof' (Morgan) and 'bofe' (Unnamed woman) for both.Finally, writing 'triful' for trifle, Green may have pronounced a short [ʊ] before /l/, a feature which serves as one criterion in Trudgill's modern dialect classification (Ihalainen 1994: 255).
Voicing of initial fricatives, one of the most well-known features of the Southwest besides rhoticity, is not attested at all through orthography in the letters sent to Dorset parishes.

Short vowels
One feature is particularly dominant in the pauper letters, that is the lowering of the KIT vowel to /ε/ or /ə/ in some words (33 instances), which is also present in Barnes' poems (Burton 2013: 534).In later data KIT lowering is found by Wakelin only sporadically, 'especially in Cornwall, Devon and Somerset, among older speakers ' (1986: 21), which indicates that the feature is recessive.However, it is still recorded for Dorset by the Survey of English Dialects, for instance for the pronunciation of shilling (Orton & Wakelin 1968: 894).This word also appears regularly in the pauper letters, by nature of the negotiation of financial support contained therein, typically in connection with g-dropping (see section 5.1); spellings include 'Shilens', 'Shelens' and 'Shelons' (Morgan) with single or double KIT lowering.Further examples of this feature are 'Cheldon' (Morgan) and 'Cheldren' (Bartlett Hand 2) for children, 'famelly' (Weakford) and 'Famely' (Green, Bartlett Hand 2) for family, 'hende r ' for hinder (Summers), 'lettell' for little and 'tell' for till (Weakford), and 'paresh' for parish (Morgan).Overall, KIT lowering can be found in seven of ten paupers from 1783 to 1834, demonstrating the continued presence of this feature.
The raising of the DRESS vowel is also mentioned as an 'occasional variant' by Burton (2013: 535) and as recessive by Wakelin (1986: 21).It is evidenced by three paupers between 1783 and c.1827 in 'git' (get; Summers), 'Litt' and 'Sind' (let and send; Morgan), and 'will' (well; Unnamed woman).A backed version of STRUT is also attested by three paupers between 1789 and 1821: 'ombel(l)' with h-dropping for humble (Weakford), 'Homble' for humble (2x) and 'oſs' for us (Green), and 21 instances occur in Morgan's letters, among them 'hop' (up) with h-insertion, 'most' (must), 'Som' (sum) and 'Sonday' (Sunday).While this is not documented as a feature by Barnes, Burton (2013) or Wakelin (1986), Wright (1905: 628, 660) does list backed pronunciations for eastern Dorset, for instance in summer, sun, Sunday and up.Perhaps this was not felt salient enough to be listed by the other sources, but it does appear to be a bona fide feature of Dorset.
Eight further features are suggested by the spellings of only one or two writers each (table 2).The value of such low-salience features is discussed in section 4.2, and further attestations may be found in additional data sources.

Long vowels
Only one long-vowel feature is prominent in the pauper letters, namely the shortening of the FLEECE vowel.Variant spellings attest to this in four individuals between 1789 and c.1827.The only word affected by this vowel shortening (19x) is week(s), typically spelt 'wick(s)' (Weakford, Morgan, Headen, Unnamed woman).FLEECE shortening is variable and not pervasive (Burton 2013: 544, 545-6), Wakelin (1986: 25)

Diphthongs
As with long vowels, only one diphthong feature can be observed in more than two writers, i.e. the monophthongisation of the FACE set also mentioned by Barnes (see section 4.1).Morgan displays this in 'Lebour' (labour), as well as 'Tread' (2x) and 'Tred' for (trade).Monophthongisation is also suggested by Green's rendering of Laystall St as 'Laſtorll St' and they as 'tha', which the unnamed woman spells 'the'.Two further features are attested by one pauper each.Firstly, Morgan has a spelling that suggests a monophthong in the GOAT set (Wakelin 1986: 28), as shown by 'Boof' (both).Secondly, Fuller (Hand 2) seems to pronounce the PRICE set with the CHOICE diphthong (Wakelin 1986: 27) in 'a bloige' (obliged).

Morphosyntax
The most common Southwestern morphosyntactic feature found in the pauper letters, shared by six writers and with 11 occurrences, is universal -s (Ihalainen 1994: 214).
The unmarked possessive is a feature similarly attested during the whole period and shared by five writers; one example is 'my Chldren Ages' in (6).This feature is not recorded for Dorset in the main dialect surveys listed in section 4.1; however, in their grammar based on the Survey of English Dialects Upton, Parry & Widdowson (1994: 483) note that the possessive singular with a zero ending is found 'mainly in the Green Wakelin (1986: 26) northern dialects' and that examples from other counties are rare.Considering its diachronic persistence in the pauper letters from Dorset, this feature should be investigated further.
Uninflected do (4x) and have (3x) can be observed in the writings of four paupers.According to Wagner (2008: 433), do and have remain uninflected only when used as auxiliaries, which is confirmed by the pauper data in (7) for do, and in (8) for have: (7) he dont troble for much (Ann Weakford, Frome, 5 November 1789, DO/BF/6) (8) we had a very Sudent discharge which have been very hurtfull for the money I had Saved (James Summers, Fort Monckton, 18 November 1783, DO/CC/1) In a similar way, one pauper does not inflect the full verbs come (9) and make for the third-person singular.The consulted dialect synopses do not mention this, but the uninflected third-person singular present tense is documented as a feature in the linguistic description of a recording of Sid Hodder (b.1878), a retired farm worker from Portesham, Dorset, who was interviewed in 1956 as part of the Survey of English Dialects (provided online in the database Sounds by the British Library).12(9) this come with our humble seruis to you (E.Chappell's wife, Bristol, 6 August 1740s?, DO/BF/4) Pronoun exchange is a well-known feature of Southwestern dialects (Wakelin 1986: 34;Ihalainen 1994: 214;Wagner 2008: 420-3) and occurs four times in three different pauper letters sent between the 1740s and 1810.In ( 10) and ( 11) we and he are used as objects, respectively, whereas in (12) her is used with subject function: (10) to desire of you to to send we some relife (E.Chappel's wife, Bristol, 6 August 174X, DO/BF/4) (11) you will never have he there no more (Ann Weakford, Frome, 5 November 1789, DO/BF/6) (12) hear forst Me to see […] I Recved 2 lawers letters from M r Georg Moor of a dept that hear Crontracted (James Headen, Plymouth, 2 June 1810, DO/BF/26) These examples are reminiscent of the findings on pronoun exchange presented by Ruano-García ( 2023).Yet while two paupers use subject for object forms and only one writer object pronouns in subject function, there is not enough data to suggest that the former type of pronoun exchange is more frequent than the latter.Example (10) predates the first known metalinguistic comment on pronoun exchange from 1777-8 by at least one generation (Ruano-García 2023), but the use of pronoun forms as subjects was already attested in Old English, and subject forms with object function can be found from the Early Modern English period onwards (Hernández 2011: 95).That pronoun exchange is not present in pauper letters after 1810 could be an early sign of the feature being recessive (cp.Wagner 2008: 423

on modern data).
There are some interesting known morphosyntactic features occurring with low salience and frequency in the pauper letters, listed in table 4.Among the features in table 4 it is noteworthy that the indefinite article a before a vowel only occurs in the earliest letters by the Chappells in the 1740s, a feature the couple shared.It is also only Edward Chappell who shows another very traditional feature, namely ye as an object pronoun.A further traditional feature, which must have been so rare in the eighteenth and nineteenth century that it is not noted in the modern synopses or in Barnes' Dissertation, is the use of -ind for -ing, which is attested in a letter by James Summers from 1785.This feature occurs twice in ( 13), but interestingly in the second instance the writer corrects '-ind' to the expected '-ing', which suggests that Summers must have been aware of common usage patterns.Following Lass (2006: 81), Southern -ind was already being superseded by -ing during the fifteenth century.Example ( 13) thus provides a surprisingly late attestation of the older inflection ending, which merits further study.Lastly, the pauper letters also contain morphosyntactic features which are so universal that it is 'impossible to pinpoint their regional basis' (Wagner 2008: 437).Among these features range those which were targeted as non-standard by the prescriptivist movement and have received scholarly treatment in this context (e.g.Tieken-Boon van Ostade 2011, 2014).As mentioned in section 4.2, we find multiple negation in the pauper letters, not only in Morgan's writing (7x; see, for instance, (3)), but also in letters by Weakford

Discussion
In the 31 pauper letters investigated a total of 297 instances of 52 different types of speech reflections could be observed.These do not necessarily occur only in Dorset, but may be common in many other varieties as well, and two were not listed as dialect features in modern synopses.Eleven regional features are salient in the letters, meaning that they are represented by at least three of the ten paupers included in the selection.Listing the features by salience ('Number of paupers') rather than frequency ('Number of occurrences'), table 5 contains three morphosyntactic, consonant and short vowel features each (one of each making up the top three), as well as one feature each relating to long vowels and diphthongs.With the exception of pronoun exchange (discussed in section 5.3) the features are attested into the 1820s and 1830s, which suggests their prominence in the earlier nineteenth century and possibly beyond.Although the number of 11 salient features might appear small, the presence of these features in many letters helps link them linguistically to Dorset according to the sociolinguistic studies consulted, or at least the Southwest more generally.It has to be noted that only the phonological features in table 5 are localisable to this county.The morphosyntactic features, in contrast, are more widely associated with the Southwest in general.Among the morphosyntactic features in table 5 pronoun exchange alone seems to have a more specific regional distribution which could help distinguish eastern from western localities within the Southwest (e.g.Wagner 2008: 422-3;Ruano-García 2023), but this feature is too infrequent in the selected pauper letters to allow a more fine-grained analysis.A final point of observation: of the prototypically Southwestern features (hyper-)rhoticity, fricative voicing and pronoun exchange, only the first is represented among the most salient, and fricative voicing is not revealed through spellings in the selected letters at all.Five of the salient features, marked with an asterisk in table 5, are mentioned in Barnes' Dissertation.However, two of them (with the asterisk in parentheses) are alluded to with reference to more restricted contexts than observed in the pauper letter data: in Barnes, KIT lowering is coupled with g-dropping concerning the inflectional ending -ing, while hyper-rhoticity only appears in words ending in <ow>.Three additional features mentioned by Barnes are evidenced in the pauper material with low frequencies only: omission of -v-, backing of the DRESS vowel, and the verbal prefix a-.As explained in section 4.1, the analysis of regional features in the pauper letters was not expected to match Barnes' depiction of the sound of the Vale of Blackmore.However, the eight features found in the pauper letters which are mentioned in Barnes' Dissertation do support the regional anchoring of the pauper letters to Dorset.Some additional 509 SPEECH REFLECTIONS IN PAUPER LETTERS morphosyntactic features may be discussed in his grammar published in 1868, which needs to be consulted in follow-up studies.
Overall, the 41 low-salience (and low-frequency) features paint a picture of diversity and contribute to the expression of the paupers' idiolects.For reasons of space, these features will not be listed again here, but in summary they consist of five consonant, eight short vowel, six long vowel, two diphthong and nine morphosyntactic features; three features (unmarked possessive, uninflected full verb other than do or have, -ind for -ing) were not recorded by the modern dialect studies consulted for the present article, and may have been too infrequent to be included in them.
Studying the distribution of all regional and universal features among the paupers (table 6) is revealing on several levels.A total of 297 instances of speech reflections in a corpus of 4,803 words amounts to a normalised frequency of 6.18 occurrences per 100 words.Five paupers exceed this figure, most notably Augustine Morgan (10.99),followed by the unnamed woman (8.49) and Ann Weakford (6.86), as well as the Chappells (6.21 and 6.41 for husband and wife, respectively).The first three are represented prominently in the analysis of speech reflections in sections 5.1-5.3.
As mentioned in section 1, dialect lexis was not investigated extensively in this article on account of its low frequency of occurrence.However, the three instances of regional lexis observable in the letter selection occur in the writings of the three paupers who have the highest frequency of dialect features overall: bide 'stay' (attested in the entire Southwest, Wakelin 1986: 39) is used by Morgan andWeakford, andafeard 'afraid' (attested throughout England, Wright 1898-1905) by the unnamed woman with the spelling 'Afared'.
Considered diachronically, the data in table 6 does not reveal any clear tendency towards a reduction in dialect features over time, which might have been expected considering the increased educational opportunities for the lower classes particularly in the nineteenth century (e.g.Stephens 1998: 10-1;Gardner 2023, forthcoming).Rather, there are noticeable differences in feature frequency between the paupers independent of time, and these are likely to be a reflection of the varying degrees of education of the individuals.The more education a pauper received, i.e. training in reading, spelling and writing, the more likely it is for spellings to be more standard-like and the less likely it is for variant spellings to emerge.This becomes particularly apparent when comparing the two different (authentic) hands writing for John Bartlett.The better-trained Hand 1 has no variant spellings which point towards reflections of regional speech, whereas Hand 2 with its more limited training does.This also suggests that the pauper's speech is unlikely to have influenced the encoders, otherwise the regional patterns would be more congruous.The same applies to the two authentic hands encoding letters for Fuller.Both hands reveal regional features through their spelling, but not the same ones.In this context it is interesting to consider another set of related letters, that of husband and wife from the 1740s.In their letters the Chappells share only two dialect features, namely the generally prominent elision of final /d, t/ and the low-salience feature of the indefinite article a appearing before a vowel.This example serves to illustrate the idiolectal diversity revealed in the letters and reminds us that the absence of a feature in writing does not mean that this feature was not present in speech.After all, pauper letters are not dialect literature and, considering the limited schooling the labouring poor usually received, are unlikely to represent conscious and consistent efforts to 'write in dialect'.
As noted in section 4.3, investigating the frequency of occurrence of regional features in further detail unfortunately lies outside the scope of this article.That it is an avenue worth pursuing will be briefly illustrated on the basis of two higher-salience features.Owing to the specific content of the pauper letters, some words occur with noticeable frequency in the corpus and also appear as examples in the analysis of variant spellings (4) I and my family is Starveing (Edward Chappell, Bristol, 27 January 1742, DO/BF/1) (5) I hoes two pou nd (Charls Ann Green, London, 6 December 1826, DO/WM/13) (6) you Say that I Must Send my Chldren Ages and what Thay Earns and what I Earns my Self (John Bartlett, Hand 2, Poole, 2 April 1834, DO/BE/3)

( 13 )
Should be payind you […] I Sopose you think because I have had bad look for not havein [^g OVERWRITES d^] work Last year (James Summers, Fort Monckton, 17 March 1785, DO/CC/2)

Table 1 .
Letter selection (chronological order) 4 Dissertation is the elision of word-final d: 'd, after n, as in an', and; boun', bound; groun', ground; roun', round; soun', sound; is commonly thrown out, as it is after l: as in veel, for field' ( §30 inBurton & Ruthven 2013: 11).The local pronunciation of a subset of the FACE 5 set serves as a second example:The diphthongs ai or ay and ei or ey, the third long close sound as in May, hay, maid, paid, vein, neighbour, prey, are sounded,like the Greek ai,the a or e the first open sound as a in father and the i or y as ee as the first close sound.The author has marked the a of diphthongs so sounded with a circumflex; asMây, hây, mâid, pâid, vâin, nâighbour, prây.( §22 in Burton  & Ruthven 2013: 10) Morphosyntactic features can usually be isolated without having to examine variant spellings; examples in (3) are multiple negation ('I Cant Gett no Bread'), and universal -s where a singular verb is used for a plural subject ('Boof my Laigs is').

Table 2 .
stating that in the Southwest it is 'sometimes found before a final voiceless' consonant with keep, sheep and week as examples.All remaining long-vowel features appear in individual writers only (table 3).Low-salience features in pauper letters (short vowels)

Table 3 .
Low-salience features in pauper letters (long vowels)

Table 4 .
Low-salience features in pauper letters (morphosyntax) Headen and Green.Green is unique in having unmarked plurals for measurement nouns (5) and a flat adverb in 'Mendes very Slow' (see table 4), while Weakford has the double comparative 'more softer'.Finally, the past tense form is used as a past participle by two paupers, in 'had a went' (Weakford, see table 4) and 'I have rote this letter' (Unnamed woman).

Table 5 .
Salient dialectal features in pauper letters

Table 6 .
Distribution of features in letter selection