Phonetic spellings in the Late Modern English dialect of the Isle of Wight (based on EDD Online )

Phonetic spelling was common practice in English dialect texts until the API (English: IPA), in 1893, launched its normative signs of transcription. Nevertheless, phonetic spellings have rarely been studied, the least in dialects. This paper ventures to trace some of the immanent norms and the sound functions of some phonetic spellings in the Isle of Wight during the Late Modern English period. The investigation is mainly based on data retrieved from EDD Online , which allows for sophisticated aggregate queries so that spelling practices can be quantified and mapped. The paper focuses on spellings unique to the island, such as double a, e , and o before another vowel, that is, in diphthongs, and u - spellings for schwa in posttonic syllables. The figures for non-unique features, such as voicing of fricatives, are normalized and allow for correlation to other counties. In sum, the Isle of Wight accent was marked by its typical “ drawl. ”


Introduction
In a recently published paper, Wolf, after a detailed review of previous literature, concluded that the study, both diachronic and synchronic, of spelling habits in dialect is an "underresearched field" (2019:192, 207).There have only been a few studies of nonstandard spelling in its own right (for example, Trudgill,  1999).They have focused on Norfolk (as did Trudgill), Scouse, Scots, and Jamaican creole.None of these contributions has examined dialects of the English south.Wolf himself underlines the significance of "dialect graphy" by using Ulster Scots as an example.
The present paper ventures to trace the immanent norms of some of the phonetic spellings used in the Isle of Wight during the Late Modern English period.The investigation, after appraising various sources and sketching the state of the art, is based on data retrieved from EDD Online (Markus, 2019a), the digital version of Joseph Wright's English Dialect Dictionary (EDD).EDD Online counts exactly 53,969 spelling variants in total and 739 variants ascribed to the Isle of Wight (I.W.).Beyond such simple counts, the online platform allows for sophisticated "aggregate" queries so that searches on spelling practices and phonetic or phonological features can be quantified, normalized, and mapped. 1 Wright's spellings of variants follow the glossotypes of the dialect glossaries and dictionaries published during the Late Modern English period to reconstruct pronunciation.Individually, these glossotypes may seem unreliable and nontransparent.When quantified, however, they allow for conclusions on some patterns of spellings as well as sounds.While Alexander J. Ellis, with his monumental work On Early English Pronunciation (1869-89), set a challenging example of what issues could be raised in a narrow analysis of phonetics, this paper is content with a broad description of some of the dialectal graphemes and phonemes relevant to the Isle of Wight.
As a follow-up paper to Markus (2022), which focused on the I.W. lexis but could trace only a moderate number of unique words in it, this study tries to show that the uniqueness of the island's dialect can mainly be found in some orthographic and, as a result, phonetic or phonemic peculiarities, some of them relics from Middle English. 2 Ellis's admirable achievement On Early English Pronunciation deserves to be examined at some length.This paper will, however, mainly deal with digraphs, conspicuous spelling practices in posttonic syllables, and the distribution of voiced word-initial fricatives.In the final conclusion and historical interpretation of the findings, the paper will contrast the longevity of accent/pronunciation with the by-and-large shorter lifespan of lexis.The Isle of Wight is used as a paradigm for other counties.

State of the art
Phonetic spelling was a common practice in written texts of spoken English or English with an affinity to the spoken word (as in private letters) until the late nineteenth century. 3Writers of correspondence, lexicographers, glossarists, orthoepists, language teachers, and novelists often or occasionally practiced these suggestive and seemingly subjective spellings in the eighteenth and nineteenth centuries.In 1886, however, the Association Phonétique Internationale (API; English: IPA, for International Phonetic Association) was founded, and two years later, its English phonetic alphabet was published, still commonly used today.A cursory look into Walker's Pronouncing Dictionary (1791/ 1823) and into Ellis's survey of phonotypical systems prospering from the 1840s to 1870s (Ellis, 1869-89: IV, 1183) tells us how much the English language had been in need of a normalized phonetic transcription.The used symbols of phonetic spellings are practically incomprehensible to uninitiated readers, 4 and Walker's introduction, by its mere length, indicates the enormous difficulties of explaining English sound symbols in relation to their orthography.
Ellis's comprehensive achievement On Early English Pronunciation (1869-89) is symptomatic of the hopelessly chaotic situation.While admirably competent in the diverse linguistic domains at issue-historical linguistics, dialectology, phonetics/phonology/prosody, comparative linguistics-even Ellis failed to present a coherent and systematic dialect phonology in the first four parts of his work, published in two volumes of 1,432 pages from 1869-1874.Several reasons for this initial failure are close at hand: (1) The English language was (and today still is) particularly challenging when it comes to the differences between pronunciation and orthography, with many spelling survivals from Middle English; (2) the diversity of the traditional English dialects was extreme; (3) several spelling habits and coding systems expressing speech sounds were available in Ellis's time, from Orm's tentative spelling reform in the twelfth century to A. Melville Bell's Visible Speech of 1867.Faced with these insufficient coding systems, Ellis invented one of his own, which he called Palaeotypes ("old types").It widely worked with normal Latin letters, except that they partly came in digraphs or trigraphs (for one phoneme), and were partly inverted, italicized, capitalized, or complemented by all kinds of signs. 5In sum, they are almost impossible to memorize.Given that nineteenth-century dialect glossaries were written in more or less standardized "glossotypes," Ellis also tried to provide a survey of these (I, 13-16; II, 565-80, 614-15).Like the palaeotypes (always rendered in parentheses), the glossotypes, always in italics, come in a confusing number of variants, the more since Ellis also includes spelling habits in other European languages than English.Instead of trying to provide an overview here, I shall refer to individual glossotypes later in this paper.
The keys for the basic letters and signs of these two codes may seem unacceptable in our time of API.However, they reflect the subtlety and theoretical ambition of Ellis's concepts concerning not only sounds and spellings, but also phonotactic features such as diaeresis, word and sentence accent, and pitch.It is almost tragic that Ellis did not have the API alphabet at his disposal yet and had to make use of the complicated letter and sign constructions of his. 6s a result of this, and also owing to the mixture of problemoriented phonetic description with historical criticism of contemporary and previous "pronouncing prophets" (IV:1218), Parts I to IV of Ellis's magnum opus have, up to now, not been used by phonologists as much as they otherwise might have been. 7owever, some fifteen years after the publication of the four parts of his book, Ellis, in 1889, came back with his (previously envisaged) Part V, entitled The Existing Phonology of English Dialects (EPED) and augmented by a long chapter Preliminary Matter (V, 1-88).EPED, unlike the preceding parts, is the applied analysis of the sound inventory and spelling practices attributable to the English, Scottish, and Welsh counties, or areas smaller or larger than counties.
The value of this substantial description of dialect-specific pronunciation in the UK of the late nineteenth century has been controversial in recent dialectology.Maguire (2003:1-2) refers back to Anderson (1977) and Shorrocks (1991) as examples of positive appraisals, and to Wright (1892), Dieth (1946), and Wakelin  (1972) as negative ones.Maguire himself, in 2003 and again in  2012, also strongly advocates the exploitation of Ellis (1889), rather than of Wright's Dialect Grammar (EDG, 1905), on the grounds of Ellis's "large number of narrowly defined locations in the EPED" (2012:91) and because "most (perhaps all) of Wright's data for the north-east is derived from Ellis (1889)" (2003:1).In (2012:91), Maguire repeats the downranking of Wright's achievement in the English Dialect Grammar (the Dictionary is not even mentioned): Wright's data for parts of Northumberland and of Durham "are re-transcribed examples from the EPED." There is, however, no factual evidence for this complaint about Wright's borrowing from Ellis. 8 The English Dialect Dictionary (EDD), in its online version (2019a), allows for finding all headwords that are, wholly or partially, ascribed to either Northumberland or county Durham (Maguire's "north-east"): 7,549 matches.Only 163 of these headwords also contain one or more references to Ellis with, overall, 376 such references.Of the just over 300 references to south Durham, which is one of the county sections that Maguire (2012:91)  specifically selected to attest to Wright's indebtedness to Ellis, 295 are attributed to "J.E.D.," which stands for a Miss J. E. Dent. 9In other words, Maguire's complaint about Wright's "incorporation" of Ellis's data proves unjustified.Ellis's role may be slightly different in the case of Wright's EDG, but, then, the question is how Maguire could possibly know about Ellis's share in the EDG, given that no electronic version of the text was available.Even assuming that Maguire studied the EDG with sufficient care, its contents are bound to be less comprehensive in a one-volume grammar than in the Dictionary.
Another great admirer of Ellis's work is MacMahon (1998, see 390-2).His substantial survey article "Phonology," in the Cambridge History of the English Language dedicated to 1776 to 1997 (= vol.IV), in its first part on the historical sources, repeatedly refers to features of non-RP pronunciation in line with his sources, but these sources, as he states himself, tended to be polemically critical of each other, RP-oriented, opposed to the factual variability of English pronunciation, and/or driven by the attempt to stabilize the pronunciation of the English language (see MacMahon,  1998:375-96).MacMahon's paper, in its subsequent (descriptive) part on vowels and consonants, is mainly based on "educated" speech.No coherent picture of the "vulgar" speech is provided.
Ellis, for his part, arranged the matter of his EPED no longer by speech sounds, as before in Parts I to IV and as later practiced by MacMahon, but by dialect areas, proceeding in 42 local categories from England's south "up" to the Shetlands and Orkneys ("Insular Lowlands").This basic structure of the book allows for an easy examination of dialect-specific features.The Isle of Wight is mostly seen as part of, and undistinguished from, Hampshire (1889: V,  viii, 41, 96-7).Though the book as a whole uses material from 1,145 places and 811 informants (1889: xvii), there is only one paragraph on I.W.-specific data.Ellis reports that he had two informants, one for the north of the island, a Rev. C.E. Seaman, and a certain Mr. Titmouse, schoolmaster of Shorewell, for the south.Ellis then comments that he had "some difficulty in interpreting some of Mr. Seaman's spellings" and, therefore, preferred to "confine myself to giving those words which Mr. Titmouse has re-spelled" (1889: V, 107).However, he has then to admit that his main informant, Mr. Titmouse, was a native of another county and had previously been a schoolmaster in Somerset.
As these details show, Ellis is very scrupulous in documenting his informants, by name, profession, and location.For I.W., he explicitly lists Northwood, Shorwell, and "whole Isle" as the places of dialectal information (1889: V, 92).Generally, and in principle, Ellis could therefore draw higher-resolution pictures of dialectal distribution than any phonetician before him.It is also an asset of Ellis's approach that he often took care of local areas smaller than counties.However, given the poor evidence provided by sources in concrete cases such as I.W., it does not come as a surprise that the final output about dialect features is occasionally disappointing.
It is, at least for I.W. Various features are mentioned with reference to I.W. jointly with Hampshire or even the wider area of the "eastern Mid-Southern dialect" (see Ellis, 1889: V, 91-2, 96).Ellis generally works with Classified Word Lists (CWL), but in the case of I.W., only one single sample, the place name Shorwell, with its rhotic pronunciation (:shɔRˈl), is given (Ellis, 1889: V, 107); as a detailed phonetic feature only the diphthong (ᴔ 0 i) is mentioned, which "may be (Ǝ 0 i)," as "conjecturally" palaeotyped by Ellis himself. 10Other characteristics, such as the voicing of fricatives and the shift of thr-to dr-in thresher/dresher, are mentioned by Ellis on occasion (EPED:107), but not decidedly and uniquely attributed to I.W.
In sum, Ellis's EPED, apart from its demanding inaccessibility to readers, contains a few observations about I.W., but they are scarce and eclectic, nonquantified and apparently nonunique to the island.By comparison, Joseph Wright, in his EDD, did not personify and localize his informants as much as Ellis did, and instead he also very much relied on written sources.He used the spellings that he found in these written sources less critically than Ellis by just passing them on to his readers, particularly in the form of variants.Given that these spellings are not as erratic and untrustworthy as it may seem at first sight, I will, in the following, begin with spellings, rather than with sounds.One of the most striking spelling characteristics on I.W. is the use of <aa>.

How to find phonetic spellings in the EDD: Example <aa>
The default mode for finding spellings in the EDD, such as <aa>, via EDD Online, is searching for strings within headwords. 11"Unique" items, that is, those that are attested to no other area than the selected one, are retrieved when the button ONLY, instead of the alternative buttons OR or AND, is switched on in the interface of EDD Online.For example, if one types in aa in the search box and activates the filter for the selection of I.W. under dialects/ English counties, the query delivers five matches-too few for drawing conclusions (see Figure 1).
The screenshot of Figure 1 is meant to familiarize the reader somewhat with the basic search mode of the EDD Online interface.The deficient evidence does not only follow from the small number of findings for the Isle of Wight.A search for <aa>-spellings in all English counties (click on "England" on top of the county list) would provide 56 matches with 101 county references involved.These are again numbers that do not promise to be statistically significant.The result on <aa>-spellings in headwords, as shown in Figure 1, is, therefore, at best suggestive of the spelling practices at issue.The reason for this by-and-large poor result is clear: Wright's headword spellings are compromises, meant to somehow "cover" the specific spellings in the areas that the entry as a whole refers to.To get the real dialectal spellings, we must leave the headwords behind and direct our attention to the contents of the entries or articles, for example, to the often itemized variants.
Figure 2 shows the beginning of the result list of a query for <aa> as part of variants.The result is now 55 variants for I.W. alone, of overall 1,004 variants for all counties worldwide (Figure 2).The numbers of counties ascribed to these variants may vary from case to case: the references to the I.W. amount to 65, with 1,438 county references in total (see Figure 3). 12he map in Figure 3 has automatically turned the quantities counted on the left (after normalization) into colours. 13It clearly shows that <aa>-spelling was not unique to the Isle of Wight but also known in other areas of the UK and in nearly all English counties.The white patches in Scotland, Wales, and Ireland are irrelevant in our context, given that Wright often ascribed dialect items not to the counties there but to larger areas (such as Ulster), which I have not activated in our search in order to keep things simple.
However, the main point concerning the map of Figure 3 is that the I.W. (in brown) clearly diverts from the counties directly neighboring it on the mainland (in blue).Instead, there is an affinity to Berkshire and Wiltshire, Cornwall, and most of the English counties of the north (not to mention Scotland, Wales, and Ireland).The reasons for this nonconcentric distribution of <aa> will be discussed in the next section, as the focus of the present section has been introducing the reader to the functioning of the search engine of EDD Online.All we can now provisionally say about the motivation for <aa> in I.W. is that the spelling cannot have come directly across The Solent, the strait that separates the island from the mainland, but that its raison d'être lies deeper.

Double-vowel spellings on the Isle of Wight and their distribution
Ellis (1889: V, 96) was certainly right in referring to the great towns of Winchester, Southampton, and Portsmouth to "have acted seriously on the dialect, which however crops up again in Wi." [i.e., I.W.].This is, however, too general a statement about the dialect of I.W.The evidence of Figure 3, with Hampshire being part of a wider navy-or light-blue area, requires further arguments that take into account I.W. alongside the other areas marked by the brown patches on the map of Figure 3.
The map shows that the north of England was the most coherent area of <aa>-distribution.Given that the double spelling was meant to express phonemic length, 14 the dialectal distribution is connected with the general rounding from /ā/ to /ǭ/ south of the Humber from the twelfth to the fourteenth century (Pinsker  1963,  §37).In this Southumbrian area, only new /ā/ could avoid the rounding, such as in new loanwords from French, or when an originally short /a/ was lengthened before voiceless spirants in the seventeenth century (Pinsker, 1963,  §59: in the south of England). 15This would explain the patchy distribution on the map of Figure 3 concerning the south.In any case, the <aa>-spellings for /ā/ are interpretable as survivals from Middle English (see Markus, 2001) and thus, in the south, an example of extreme conservativism, typical of rural and relic areas such as the English southwest.This would explain the intense use of double <a>-spellings in Cornwall, in the south of Wales and in Irish County Wexford-all these were former English colonies where English had been imposed on the Celtic inhabitants and then tended to be relatively conservative ("colonial lag").
The <aa>-spellings look strange from a present point of view, but in the standard of the English language other vowel doublets, <ee> and <oo>, are of course quite normal for expressing phonemic length, with old /e:/ and /o:/ long ago raised to /i:/ and / u:/, respectively (as in fleet and root).The <aa>-spellings in dialect are nothing but the adherence to an old spelling in a niche of the phonological system.
Why was /a:/ in a "niche"?The reason is that it was, in the south and in the standard, secondary; it had, as it were, come in by the backdoor.Actually, there were several backdoors.I mentioned two of the reasons, among them loanwords from French, why /a:/ had to be expressed in the south even after the Southumbrian rounding.The spelling gaaigement, an aphaeretic form for "engagement," may serve as an example.According to the OED, it came into the English language in the seventeenth century (see under engagement).The /a:/ of the stressed syllable was expressed by the duplication of spelling.Other newly long vowels in the south were the originally short ones that were lengthened, again after the Southumbrian rounding, in open syllables of disyllabic words.The common abbreviation for this sound change is MEOSL (Middle English Open-Syllable Lengthening).The rule only affected nonhigh vowels /a/, /e/, and /o/, and only in open stressed syllables of disyllabic words.The word snake, in I.W. also written snaake, is on the list of our I.W.-retrievals in Figure 2 and is a good example of MEOSL. 16nother "backdoor" by which originally short stem vowels could be lengthened is the loss of liquids or nasals (l, r, m, and n) before another consonant (as in half), resulting in svarabhakti (insertion of a vowel, here /u/) and, finally, the lengthening of the stem vowel (see Pinsker, 1963,  §53,1 and 55,2).The headword HAULM (for "straw," cf.German Halm) may be mentioned as an example.It occurs as haum in some counties and as haam in others, among them the I.W.
Yet another chance for short vowels late to be secondarily lengthened was their being part of glides or diphthongs.The spelling gaaigement, mentioned earlier, with its additional <i>, suggests the diphthongs, here /ai/, to be the causal factor.I will, however, discuss this point of double spellings in glides in the next section, where the focus is on the additional spelling elements (the <i> in the case of gaaigement).
The complexity of the development of vowels in the history of English dialects cannot be fathomed out here in view of the reasons for digraph-spellings.It seems, however, fair to draw two conclusions from our observations so far: (1) the reasons for digraphs lie in the past of the sounds and spellings concerned; and (2) there is no single cause.I have detailed a few cases why <aa> was meant to express a long vowel /ā/: Middle English original ā; ā from ă owing to MEOSL; and lengthened ā before liquids and nasals (i.e., before weak consonants), which were eventually given up in favour of the quantity of the preceding vowel (the HAULM type). 17Going through the list of the 55 <aa>-spellings in the I.W., as pasted from the interface of EDD Online in Table 1, confirms the role of these types (Table 1).
Thus, MEOSL is involved in raames, traade, and wraathy.And the type HAULM is confirmed by baak (BALK) and baam (BALM).However, the list of Table 1 also tells us that there are other reasons beyond the ones I have so far distinguished.
One of the additional factors is an original /w/ that affected the preceding stem vowel, as in claa ('claw'), craa ('craw'), daa ('daw'), draa ('draw'), gaak ('gawk'), laa ('law'), and others.The /w/, preserved in the spelling of the English standard, was vowelized at some stage in the history of English (see Pinsker, 1963:  §40), thus producing a diphthong /au/.A parallel development can be observed with /a/ followed by <y>, which, for its part, originated from a plosive /g/ or fricative /ʒ/.Again, the result was, earlier or later, a diphthong, as in Standard English day (cf.Gothic dags), which is listed in the I.W.-variants as daay.
Diphthongs were very much a "playground" of variation in the history of English.So was the expression of the length of vowels by double spellings.Accordingly, practically all spelling habits commented on so far, while found in I.W., cannot be ascribed to it as unique forms.However, when it comes to the expression of the length of the first elements in diphthongs, the Isle of Wight plays quite a unique role of its own, as the next section will show.

Spelling practices in the context of diphthongs on the Isle of Wight
Figure 4 shows that there is no one-to-one relationship between variant form and dialect attribution in the case of HAULM, mentioned earlier.
On the one hand, haam is not I.W.-specific; on the other hand, another variant ham is equally attributed to the I.W.By contrast, the <aa>-spellings being part of, or historically resulting from, diphthongs are remarkably often only ascribed to the I.W.Here are the first two entries of this kind on our list (Figure 5).
Again, only the linguistic historical background can bring some light into this seemingly confusing distribution.As is well-known (see Pinsker, 1963: §55, 16,1 and 62, A1), old diphthongs, as in Middle English day ['dai], were monophthongised in the English standard to [e:] (Great Vowel Shift) and again diphthongized in the nineteenth century to [eɪ].In the "vulgar" dialects of the south (Pinsker: "Vulgärsprache"), the two elements of the diphthongs drifted further apart than in the standard in terms of their height of tongue: the [e] of the diphthong was lowered to [a], whereas [ɪ] was raised to [i] (see Pinsker, 1963: §62, A2).We do not exactly know whether the first element [a] was also lengthened on this occasion or whether and to what extent it had remained long when it resulted from the original long monophthong [e:].In any case, <aa>-spelling can be taken as the attempt of I.W. writers of dialect texts to mark the special quality of the diphthong in their pronunciation.The grapheme <aa> as such was, as we saw in Figure 3, widespread in the British Isles and has historical roots.However, the use of this grapheme for a markedly emphatic or extended pronunciation of [a] in the [ai]-diphthong is a characteristic orthographic feature that is fairly unique to the I.W.Why Berkshire, unlike Hampshire, shared this feature in a handful of cases, and whether the noticeable pronunciation of the diphthong may be classified as a Cockneyish vulgarism-these are questions that this paper, given its narrow limits, cannot answer.A survey concerning group relationships between English county dialects by Ihalainen  (1994:211), based on Halliwell's Dictionary of Archaic and Provincial Words (10th ed., 1881), classifies Berkshire to belong to the midland as well as the south-western dialect area.Wakelin (1986:1), quoted by Ihalainen, called Berkshire, in the light of mid-twentieth-century evidence, a "marginal area."There is no reason then to assume that the dialects of the Isle of Wight and of Berkshire were generally connected, and that this is the reason for similar phonetic or spelling habits.

Other spelling characteristics concerning diphthongs
The In this case, <y> was not simply used as a marker of the length of the preceding vowels.For this, there were other means.As was suggested by <aa>, digraph vowel spellings were commonly used for expressing length, and digraph spelling also worked for /e:/ and /o:/.However, the vowels concerned by these double spellings discussed earlier were, in nearly all cases, originally open vowels (as the modern English spellings <ea> and <oa> respectively suggest).The group of long vowels now under discussion, that is, those marked by an additional <y>, owe their length to MEOSL, the Middle English Open Syllable Lengthening, and the resulting long vowels were all closed vowels.We must go back to the thirteenth century, when words of the type blade (to use the first sample of the list above) were pronounced the way they are spelled now (['blade]) but were changed to ['bla:de] (in historical notation: blăde > blāde).The rule affected nonhigh vowels a, e, and o and only in open stressed syllables of disyllabic words.The word examples quoted in our list completely agree with these conditions.
The further development of the blade-type words is wellknown: weakening and, finally, loss of the unstressed second syllable, strengthening of the stem-syllable by first raising the vowel to /e:/, and then making it a diphthong /ei/.This is the development applying in the English standard.Dialects, of course, went their own way.
Which brings us back to the Isle of Wight.Our sample illustrating <y>-spelling, bleyad for BLADE, reflects the loss of the original second (unstressed) syllable just mentioned.The orthography without the final -e is a clear case of phonetic spelling.It seems fair to assume that the stem syllables of the words concerned likewise have spellings that imitated the pronunciation.The letter <y> generally has a double function in English of expressing either the fricative "consonant" /j/ or the vowel /i/.Given that /i/ is the highest vowel and that /j/ is one of the the lowest consonants (in terms of consonantal strength), the two sounds nearly collapse.In concrete terms, the word bleyad had either a triphthongal pronunciation /'bleiad/ or a newly disyllabic one /'blejad/.In my opinion, the second option is the more likely one, because y-insertion before a vowel was a general phenomenon in English dialects (see Markus,  2011) and here simply shared by the I.W. Accordingly, the headwords EARLY, EARNEST, EARTH, and EAT can be retrieved for the I.W. in the variants yearly, yearnst, yeath, and yett, respectively.While this word-initial yod-formation was common in the nineteenth century, 18 the same process in the middle of diphthongs traced in this section was unique to the I.W.It resulted in turning the diphthong to be disyllabic and is based on the generalization of the rule affecting words of the early-type to apply not only wordinitially but also syllable-initially.
To sum up this section, we have found a systematic use of <y> as a unique spelling marker in <e>-initial diphthongs, just as the previous section had delivered an I.W.-specific spelling marker mainly in <a>-initial diphthongs.Both characteristics testify to a specific realization of diphthongs on the island.Phonotactically, the two processes observed are connected.While the more radical sound shift, yod-insertion, as in bleyad (for BLADE) and yearly (for EARLY), only affected (original) diphthongs with the midvowel /e/ as their first element, the low-tongue vowel /a/, as in daay, only made it to being lengthened.
The main reason for this general practice seems to be so-called "reverse spelling." 19In line with the common development of the phoneme /u/ to schwa (/ə/) in the "Second Vowel Shift" (sixteenth century, see Pinsker 1963, §56, 1), the traditional <u>-spelling was used unetymologically, that is, also for schwa sounds that did not result from previous /u/.The rule of the Second Vowel Shift affected the stem vowels of words such as but, plum, and cudgel, where the vowels continued to be opened or lowered from /ə/ to /ʌ/ (according to Pinsker 1963 in the eighteenth century), but where the old <u>-spellings survived.By contrast, the <u>spellings in our I.W. words just quoted concern the unstressed syllables of (mostly disyllabic) words.Why were these posttonic vowels in need of being marked?The answer is that they, from Late Middle English onwards, were affected by the tendency toward reduction or even omission, above all when the following consonant was a liquid or nasal. 20Looking at our selection of <u>-spelling examples again, it turns out that they are nearly all cases of word-final liquids or nasals. 21The answer to our question of the motivation of <u>-spellings is this: the words were probably marked by reverse spellings (<u>) for /ə/ because this schwa was on the point of being completely omitted by the nineteenth century.The entry-headword of our sample devvul, with its seventeen variants, shows this erosion of the unstressed syllable very clearly (Figure 6).
One can see that several variants (2, 5, 7, 8, 15, 16) indicate a monosyllabic pronunciation, which is also confirmed by Wright's phonetic transcriptions.My conclusion: in contrast to this obvious trend, I.W. spellings in posttonic syllables (followed by liquids or nasals) mark the second syllable to be still articulated.One may interpret both the marked pronunciation and the marker <u>, with its reverse spelling background, as resulting from linguistic conservativism.

Voiced fricatives
One of the most striking spelling features typical of the Isle of Wight, alongside other counties, is the voicing of word-initial fricatives, attested, for example, by <z>-spelling for standard <s>. 22he feature is well-known.Some people from Somerset today still refer to their county as ['zaməzet].Ihalainen (1994:214) classified this feature as a south-western one.My analysis with EDD Online by-and-large confirms this ascription, but the normalized frequencies present a more precise picture, in particular as far as the I.W. and differences between the various fricatives are concerned.
The search for a word-initial <z> instead of <s> in variants of all English counties delivers 350 results in EDD Online, with 55 ascriptions to the I.W. 23 One of the output headwords is SADDLE, which Wright found in one of his I.W. sources as zaddle.
As we are not interested in the headwords as such but in the numbers of dialect ascriptions and the mapping of these numbers after normalization, I directly switch to the sorting mode column-3counted (Figure 7). 24he column-3-counted mode of the sorting box provides the quantification of the tokens and types of the dialectal references.The map in Figure 7 clearly shows the epicenter of word-initial voicing in the case of /s/.Yorkshire and Durham in the north could be interpreted as exceptions, with only one occurrence in each case. 25As far as I.W. is concerned, the map of Figure 7 shows that the island, unlike adjoining Hampshire, shares the southwest's relatively high frequencies, namely ≥ 90 references per 10,000.Berkshire (again Berkshire!) and Gloucestershire, that is, counties bordering the southwest, are also members of the same statistical group (brown colour!).Cornwall, probably owing to its non-Anglo-Saxon background, noticeably deviates from the frequencies of the high figures of the southwest.
In addition to the /s/-/z/ alternation, voicing of the voiceless fricatives also includes /v/ versus /f/ and /ð/ versus /θ/.The maps for these two oppositions (Figures 8 and 9) show marked similarities with the map in Figure 7 in that the southwest of England is again clearly prominent in the use of voiced word-initial fricatives.
The word-initial labiodental fricative, in its voiced version /v/, occurs in 32 I.W.variants, which are ascribed to the island 37 times.The first three samples of the list of variants are van (FAN), varm (FARM), and vather (FATHER).Cornwall is now in the same (highest) statistical group as the rest of the southwest and the I.W.However, in the west Midlands, the southeast (in particular: Sussex), and in the  area north of London, voicing of voiceless /f/ was also known, but its frequencies, according to the map, peter out there.
As regards the dental fricatives [ð] (so-called eth) and [θ] (Wright uses [þ] for the voiceless version), three factors disturb the clear ascription of the forms to dialect areas.First, LModE and Present-day English <th>-spelling does not mark the feature [± voiced].Second, since the fourteenth century, <th> in word-initial position had been voiced in the south of England since late Old English, as well as, more generally, in words unstressed within clauses and in syllables unstressed within words (see Pinsker, 1963: §83,1-3), as in that, the, and knowledge.Third, the EDD does not closely correlate the two phonetic transcriptions [ð] and [θ] to dialect areas. 26For example, in the entry on THICK ('this'), some sixty areas are listed, with dozens of variants, both followed by only three transcriptions with the symbol for eth ([ðik; ðek, ðɐk.]).The dialect areas are mostly southern counties, so that the voicing seems to be meant to apply to all of them.But then, there are also a few <d>-spellings in the list of variants (e.g., dik for Dorsetshire).Though we, therefore, do not exactly know which transcriptions belong to which counties and variants, a tentative analysis is feasible.Figure 9 provides the result.
The map of Figure 9 is based on a search for word-initial [ð] combined with all English counties, no matter whether the corresponding sound in the English standard was voiced or voiceless.This method seems justified on the assumption that Wright mainly provided the voiced transcriptions for the counties previously listed.Our result of Figure 9, however, is also statistically questionable as far as the I.W. is concerned.While the overall number of ascriptions to counties is 833, the Isle of Wight has only twelve references.One should therefore not take the map at its face value in all its details.Focusing on the only 53 headwords concerned, we find that most of them are the very pronouns, mentioned above, that, under lack of stress in clauses, had long before the nineteenth century developed a voiced version (as in the, that, etc.).There are only a few content words in our retrieval list, such as THETCH [ðetʃ, ðātʃ.]('vetch'), THWART [ðēt.]('cross'), and THWARTLE [ðə ¯•tl.]('cross-plough').
Despite these imponderable factors, the distribution of /ð/ in word-initial position confirms the predominant role of words with voiced word-initial fricatives in the southwest.As regards the I.W., the map clearly marks it to belong to this block of southwestern England.

Summary and conclusion: The Isle of Wight in a historical context
This paper was primarily concerned with spellings of vowels: in sections 2-4 with those of stressed stem vowels of words, and in section 5 with the conspicuous <u>-spellings in posttonic syllables.The last section 8 dealt with the voicing of fricatives, that is, shifts of consonants.There are many other orthographic and phonetic features that could be found to apply to I.W., such as h-dropping, hypercorrect h-insertion, evidence of glottal stops, fricatives changed to homorganic stops, etc.I have disregarded these features because they were apparently not unique to I.W. and, in some cases, proved to be most infrequent, as in the case of drinstead of thr-(droat for throat).
The empirical study, based on data as retrieved from EDD Online, focused on the Isle of Wight, using two types of evidence.On the one hand, we could characterize the I.W. by affiliating it with relative frequencies of features that proved not to be unique to the island, but could be ascribed to it in comparison with other counties and with a remarkably high statistical significance.Such features were the use of <aa>, in some cases also of <ee> and <oo>, for expressing the length of the vowels /a/, /e/, and /o/, respectively.The well-known substitution of voiced fricatives for voiceless ones in word-initial (i.e., usually pretonic) position was observed to be a common phenomenon in England's southwest and also widespread in some western counties and the I.W., but, strangely enough, not in Hampshire.This special role of Hampshire may result from this county's urbanization, with the big cities affecting the dialect of this county as a whole.
On the other hand (Section 2), the paper traced some features practically unique to the Isle of Wight: the use of <aa> for the first element in diphthongs and when the stem syllable was closed by a liquid of nasal, or, in a smaller number of cases, by the half-vowels <w> and <y> (/w/ and /j/).The function of the digraph in such cases was certainly to mark the first element of the diphthong to be long.Since the standard had shifted the diphthong at issue, as in day, to either the monophthong /e:/ or further to the diphthong /ei/ by the nineteenth century, the use of <aa> for /a:/ is a relic of former centuries, at least in terms of spelling.Whether the pronunciation /a:i/ was also a survival from ME or the result of a new radical development of the diphthong (see Pinsker, 1963: §62, A, on Cockney), can be left undecided here; I tend to favor the option of the survival.
The "drawl" that must have resulted from the lengthening of /a/ in diphthongs is enhanced by another unique feature concerning the spelling of diphthongs.Section 4 provided evidence of a systematic insertion of <y> in formerly disyllabic words of the blade-type, so that [ble:d], supposedly the standard form of the word around 1800, was, then and later, pronounced ['blejəd] (spelt <bleyad>), thus preserving the old disyllabicity.Our distinction of different types of words in line with the historical development of their stem vowels was necessary to explain why the stem vowel /a/, spelled <a>, ended up lengthened in some cases and gave rise to disyllabicity in other cases.
In any case, the dialect accent resulting from the two features is that I.W. was marked by "an emphasis on longer vowels," as an anonymous Wikipedia author describes the present-day dialect on the Isle of Wight. 27The same author also says that the accent "is similar to the West Country drawl heard in south-western England."I hope that this paper has presented concrete features of evidence for this "drawl." 28n additional feature that contributes to the same drawling accent was traced in Section 5, concerned with the emphasis on posttonic syllables by replacing an unstressed /ə/ with a stronger articulation, expressed by the grapheme for the high-tongue vowel, <u>.The general trend in standard British English pronunciation over the last centuries has been to reduce or even omit unstressed syllables (see, e.g., the pronunciation of words such as medicine and Worcester).Not so on the Isle of Wight.Not only stem vowels but also unstressed-syllable nuclei were accentedly articulated.In sum, the accent worked either quantitatively (length), qualitatively (prolonged diphthongs, <u> in posttonic syllables), or phonotactically (syllable-increase by yod-formation). 29hese features of a "drawl" are in full agreement with how MacArthur (1992:325) defined the term.Drawl is "speech in which words are drawn out, especially prolonging vowels and final syllables."MacArthur's reference to the origin of drawling speech in vagabonds' cant nourishes the suspicion that the Isle of Wight's former role as a center of smuggling and piracy might be one of the factors why our analysis of I.W. variants has provided evidential details of this part of "vulgar" speech. 30Be that as it may, what we do know for sure is that some of the spelling characteristics of I.W. reveal phonetic and phonotactic features of drawling that were, during the time covered by the EDD, unique to the island, and moreover, that they result from spelling habits and sound shifts going back by centuries and partly as far as Middle English. 31 This tenacity of an English county dialect, paired with a remarkable autonomy, does not come as a surprise in the domains of phonology and orthography.Unlike sounds, words tend to migrate and, particularly since the flourishing of I.W. tourism in the nineteenth century, certainly crossed The Solvent easily, thus undermining the island's autonomous dialect in lexis.By contrast, features of accent are more tenacious and long-lived and could, therefore, more traceably survive in Late Modern English.It is for this reason that the dialect of the Isle of Wight, when it comes to its unique features, has to be inferred from the confusing spelling practices and sound-shift rules extant in the past of the English language.
12 As we are now interested in comparing the Isle of Wight with other counties, the former ONLY-button was replaced with the OR-button.This is one of the reasons why we have received so much higher output.13 The given number for the Isle of Wight is related to the total figure 1,438 in terms of percentage, per mil or per 10,000, and it is also related to the overall role of each of these counties in the EDD.Coloured maps of the kind of Figure 3 are a subtype of "choropleth maps"-the term implies a multitude of areas.14 Both Bell (1867) and Ellis (1869-89:1171, 1178) used <aa> for /a:/ on the basis of historical evidence (see the overview in Ellis, 1869-89: II, 566).This does, of course, not mean that the quality of the vowel remained unchanged.15 Examples are staff, glass, ask, path.16 The further development of the snake-type words is well-known: weakening and, finally, loss of the unstressed vowel of the second syllable, compensatory strengthening of the stem syllable by first raising the vowel to /e:/ and then changing it to a diphthong /ei/.This is the development in the English standard.17 There is a similarity to the well-known pronunciation of words ending in -l in African American Vernacular Englishes, as in [fuʊ] for Standard [ful].18 See Markus (2011).Ihalainen (1994:213), however, somewhat one-sidedly ascribes it to the north.19 Ellis (1869-89: III, ix-x) mentions the "obscure sounds which actually replace unaccented vowels."20 Liquids (l, r) and nasals (m, n) are, according to phonotactic theory, "weak" consonants so that they tend to merge with the preceding vowels and even syllables.See Vennemann (1988:9).For the details of the rule concerning unstressed syllables see Pinsker, 1963:  §69, A3b, B3c and B4.21 Ellis's wordlist, mentioned in a previous footnote, consists of words whose posttonic vowel is likewise followed by a liquid or nasal in almost all cases.22 Ellis (1889: V, 107) says that initial /z/ is "not frequent" but "a tendency," and that there is also the tendency to use initial /v/ for /f/.23 One needs a "last-result" query for this: first search for all headwords beginning with <s>; then activate the last-result button; under dialect areas tick all English counties and type in z* in the search box.24 The term column implies that the information of the retrieval window is immanently arranged in the form of a table in which, in the present query, column 1 lists the headwords, column 2 the variants, and column 3 the county references.The headwords are always column 1.The definition of the other columns depends on the query at issue.25 On checking these cases, I have to admit that the Yorkshire occurrence is based on an erroneous reference of the dialect tag.The correct referencing of the variants was extremely difficult for the Innsbruck programming team to achieve so that occasional mistakes cannot be excluded.The team is presently trying to have this bug eliminated.26 This is why the interface of EDD Online in the filter phonetic contains a warning to the user: "note: incorrect ref. to dialect areas."27 https://www.cs.mcgill.ca/∼rwest/wikispeedia/wpcd/wp/i/Isle_of_Wight.htm (accessed November 20, 2020).28 The I.W. recording by the Survey of English Dialects (SED) in 1959 in the linguistic description of the spoken text features the deviant pronunciation of vowels and diphthongs, but the phonetic symbols are distorted.The other features mentioned are mostly nonunique to the I.W. (e.g.voicing and h-dropping).The description of the spoken text is, however, very useful when it comes to grammar (https://sounds.bl.uk/related-content/TEXTS/ 021T-C0908X0032XX-0200A1.pdf).29 I am using the term accent, similar to Horn/Lehnert (1954:51), as a cover term for different dimensions of sounds.30 "Topographically the Isle of Wight always has been a smuggler's delight." (http://www.wightfarmholidays.co.uk/iow/smuggling-and-piracy-on-the-isleof-wight/) (accessed November 21, 2020).31 Ellis (1869-89:IV, 1249) pointed out the "comparative tenacity" of dialect pronunciation "as against the received pronunciation, which [ : : : ] has been and is still continually altering."

Figure 1 .
Figure 1.Search for <aa> in Isle of Wight headwords.

Figure 2 .
Figure 2. Searching for <aa>-spellings in variants ascribed to the I.W. alongside other counties.

Figure 3 .
Figure 3. Quantified references of search string <aa> to I.W. in comparison to other counties, with survey map opened.

Figure 4 .
Figure 4. Variants of HAULM, with forms of the I.W. highlighted.

Figure 5 .
Figure 5. Two examples of <aa>-spellings unique to the Isle of Wight.

Figure 6 .
Figure 6.Variants of DEVIL in EDD Online (number 6 is the form for the I.W.).

Figure 7 .
Figure7.Word-initial <z> spelling for <s> in all English counties quantified and mapped.

Figure 8 .
Figure 8. Map of /v/ instead of /f/ in word-initial position.Figure 9. Map for /ð/ in word-initial position.

Figure 9 .
Figure 8. Map of /v/ instead of /f/ in word-initial position.Figure 9. Map for /ð/ in word-initial position.

Table 1 .
List of 55 I.W.-variants with <aa> list of the 739 Isle of Wight variants produced by EDD Online reveals another idiosyncrasy of spelling: the use of unetymological <y> in stressed syllables.We are talking about forms such as bleyad for BLADE, bleyar for BLARE, geyenge for GENGE ('depth of soil'), leyace for LACE, leyadul for LADLE, meyat for MATE, preyat for PRATE, pleyagy for PLAGUEY 'annoying'), ruineyat for RUINATE, seyav for SAVE, steyal for STALE, steyev for STAVE, treyad for TRADE, and so on.The function of <y> as part of the trigraphs <eya> and <eye> is again connected with the vowel shifts of Modern English, in this case, with the shift of ē (in the spelling sequence <a : : : e>, as in blade) to a diphthong, first [ēi] and then [ę ¯i] (seePinsker, 1963: §61,1).This shift of the nineteenth century, as Pinsker ( §62, A.2) suggests and as was mentioned earlier, was carried on in the "vulgar speech" of the English south; we all know it from Cockney [ai] for Standard[ei]in lady, day etc.So what is, in the face of this vowel shift in stressed syllables, the possible function of the <y>-spellings?