1 The Middle English Development of Old English y — and Lengthened y : Spelling Evidence

The ‘Great Vowel Shift’ is the term used about a set of changes in the phonetic realisation of Middle English (ME) long vowels, which took place around 1400–1750 according to the handbooks. In this shift, the non-close vowels /e:/, /ɛ:/, /a:/, /o:/, /ɔ:/ were raised one step in the vowel space, and the close vowels /i:/ and /u:/ were diphthongised (Jespersen 1909: 231 ff.; Luick 1914–40: §§479–488; Dobson 1957 passim). In the late Old English (OE) and early ME periods, changes happened to the long vowels /y:/, /ɑ:/, and /o:/, as described by e.g. Luick (1914–40: §§287, 369–370, 406) and Jordan (1968: §§39–42, 44–46, 53–54). However, these changes are not regarded as part of the ‘GVS’, because (i) they are said to have been completed before the earliest stages of the ‘GVS’ took place (the changes to /y:/ and /ɑ:/), and/or (ii) did not take place in those dialects which later contributed to the phonology of StE (the fronting of /o:/ in dialects north of the Humber). Critical voices have been raised, suggesting that the ‘GVS’ started earlier than textbooks suggest, most notably by Stockwell & Minkova (1988a, 1988b). This paper treats the ME development of OE y– and lengthened y, for convenience called ‘eME y–’, seeking to establish (a) its phonetic

In the late Old English (OE) and early ME periods, changes happened to the long vowels /y:/, /ɑ:/, and /o:/, as described by e.g. Luick (1914-40: § §287, 369-370, 406) and Jordan (1968: § §39-42, 44-46, 53-54). However, these changes are not regarded as part of the 'GVS', because (i) they are said to have been completed before the earliest stages of the 'GVS' took place (the changes to /y:/ and /ɑ:/), and/or (ii) did not take place in those dialects which later contributed to the phonology of StE (the fronting of /o:/ in dialects north of the Humber). Critical voices have been raised, suggesting that the 'GVS' started earlier than textbooks suggest, most notably by Stockwell & Minkova (1988a, 1988b. This paper treats the ME development of OE yand lengthened y, for convenience called 'eME y -', seeking to establish (a) its phonetic 1 A very early and unfinished version of this paper was read at the conference Historical Language and Literacy in the North Sea Area, Stavanger, 26-28 August 2009. I am grateful for valuable comments by Meg Laing, Roger Lass and Merja Stenroos, and for suggestions from an anonymous reviewer. Any remaining shortcomings remain my responsibility.
developments in the dialects of ME, (b) the approximate dates at which its various developments started, and (c) whether the said changes were in fact completed before the 'GVS' set in. The answers to these questions may have far-reaching consequences for our interpretation of the Shift.
However, ME spelling is not phonetic transcript, so the implied sound value can only be inferred. Traditionally, <i> and <y> for eME yare taken to indicate unrounded [i:]; <u> and <ui/uy> are believed to correspond to a retained front rounded [y:], whereas <e> and <ee> imply lowering and unrounding to [e:]. 2 When OE y --words are spelt with <i> in late OE or early ME, it seems safe to assume that such spellings do indeed indicate unrounding, especially if the modern dialect shows /aɪ/, which is the 'GVS' output of ME ı -. This assumption is strengthened if spellings with <y> for etymological ı also appear in the same ME dialects. However, it would be a mistake to view the continued use of <y> to simply represent [i:] a priori in dialects where the rounded vowel is believed (in hindsight) to have been retained. In such cases, <y> could correspond to [y:], although such an interpretation would be highly improbable if <y> also appears for etymological ı -. 3 In other words, the scribe's entire orthographical system must be taken into account before his likely pronunciation is inferred, since occasional spellings are by definition deviations from the scribe's norm, and may reveal something about his spoken system.
In those dialects where the OE yremained front and rounded, i.e. in parts of the South, and in the West (and Central) Midlands, this vowel is usually represented by <u>, <ui/uy> in ME -and not by <y> alonefrom around 1100 onwards (Wright & Wright 1928: §57). The use of <u> for this purpose was made possible when OE u -, traditionally spelt <u>, started to be spelt <ou/ow> during the ME period, due to French spelling practice (Stenbrenden 2013). Gradon (1962) cites spellings indicating late OE unrounding of the reflex of OE y, as well as conditioned rounding of the reflexes of OE i and ı -, in the SW Midlands. Forms with <y> for etymological i in a set of Exeter documents "are probably to be regarded merely as back-spellings" (1962: 66), based on the merger between OE y and i at [i], but a number of other such spellings in ten Winchester texts cannot be so dismissed. More specifically, Gradon claims that OE ı after w seems 2 Anderson (1988)  to have undergone rounding. Besides, there is evidence that OE y was unrounded before palatals even in the SW Midlands, whereas it was retained in other phonetic contexts (1962: 72).

Discussion
The extracted LAEME material shows a variety of spellings for eME y -: <i>, <y>, <e>, <ee>, <eo>, <ey>, <u>, <ui>, <uy>, <yu>, <ou>. Again, it must be stressed that spellings cannot simply be interpreted as transcriptions of sounds. However, interaction between written norms and spoken systems must be assumed, resulting in hyper-adaptations, back spellings, and the like, and when the material is systematised, patterns emerge. Most LAEME sources show a mixture of spellings for etymological ywhich seem to contradict each other in terms of their implied sound value. A case in point is the text with index number 1300, whose language has been localised to Suffolk and dated to the second half of the twelfth century: it has dominant <i> (indicating unrounding), a secondary variant <u> (implying a retained front rounded vowel), and minor variants <ui> (implying retained [y:]) and <e>, <eo> (implying lowered and unrounded [e:]). Thus, it is difficult to draw any definite conclusions from the material. Nevertheless, the following observations can be made.
Unrounding of OE y -/y to [i:] started in late OE and is indicated in source texts whose language has been localised to Essex, Suffolk and perhaps Hampshire from the late twelfth century; in sources localised to Oxfordshire, Kent, Northamptonshire and Worcestershire from the early thirteenth century; in texts localised to Cumberland, Cheshire, Somerset and Surrey from the mid-thirteenth century; in sources localised to Lincolnshire, Norfolk, Devon, Gloucestershire and Herefordshire from the late thirteenth century; and in texts localised to Ely, Huntingdonshire and the North Riding of Yorkshire from the early fourteenth century. Thus, unrounding seems to have started in the South-East and South-West, and to have spread northwards, which goes against the assumption that the unrounding originated in the North (Jordan 1968: §41). However, the paucity of ME texts from northern England from the early ME period precludes any definite conclusion regarding the locus of this change.
Lowering and unrounding to [e:] is indicated in sources whose language has been localised to Essex and Suffolk from the late twelfth century; in texts localised to Kent from the early thirteenth century; in texts localised to Somerset and dated to 1240; in sources localised to Gloucestershire and Wiltshire from the second half of the thirteenth century; and in a text whose language has been localised to Lincolnshire from the early fourteenth century. Hence, eME y -> [e:] seems to have started in the South-East (Kent, Essex, Suffolk), but also to have taken place independently barely a half-century later in the South-West. Forms with <e> are dominant in sources whose language has been localised to Kent (the texts with index nos. 8, with a secondary variant <i>, and 142, with minor variants <éé> and <ie>), Essex (no. 160), Gloucestershire (no. 161), Somerset (no. 156, with <y> co-varying with <e>), and Lincolnshire (no. 169, also with <y> co-varying with <e>).
Retained [y:] is indicated in sources whose language has been localised to Berkshire, Essex, Suffolk and Worcestershire from the latter half of the twelfth century; in texts from Northamptonshire, Herefordshire and Shropshire from the early thirteenth century; in texts localised to Cheshire, Gloucestershire, Wiltshire and Surrey from the midto-late thirteenth century; and in sources from Oxfordshire, Ely and Huntingdonshire from the late thirteenth or early fourteenth century.
The <u> spellings from Berkshire, Essex, Suffolk and Surrey are early, but they seem to suggest that Wright & Wright (1928: §57) may be wrong in stating that the reflex of OE yhad become [e:] in Essex and Suffolk in the late OE period; <e> forms do indeed occur in Essex (text nos. 4, 64, 1200) and Suffolk (text no. 1300), but they are not dominant. Sussex is poorly represented in the early ME material, but text no. 67 (1200-50), shows <i>, not <e>, for eME y -. Surprisingly, <u>-type spellings also linger on in the East (Ely, Huntingdonshire) as late as the early fourteenth century, although the <u> forms here are minor variants.
Regarding retained [y:], the LAEME material seems to also run counter to Wright & Wright's explicit claim concerning the development of yin Wiltshire: dominant <u> in text no. 280 (1250-74) suggests that yhad not been unrounded in Wiltshire in late OE, but remained rounded. The same text shows dominant <ou> and <u> for the reflex of eME u -, and interestingly shows one <ou> for the reflex of eME yas well, which indicates a rounded vowel. Lass & Laing (2005)  ] is no more than the traditional account, but Lass & Laing's claim that it merged with the reflex of eME u in the SW Midlands certainly needs closer examination. Lass & Laing use material from LAEME texts 277 (Worcestershire), 272 (Shropshire), and 280 (Wiltshire) to back up their claims, which is why spellings for eME yin these three texts must be investigated in some detail.
The extracted LAEME material for text 272 shows dominant <u>, and a secondary variant <v> for eME u -; and dominant <u>, and secondary <i>, <e> for eME y -. Likewise, text 277 shows dominant <u> and secondary variants <ou>, <v>, <o> for eME u -; and dominant <u>, and minor variants <i>, <eo> for eME y -. Text 280 shows dominant <ou>, <u>, and minor variants <v>, <o>, <ow> for eME u -; and dominant <u>, and minor variants <i>, <ou>, <eo> for eME y -. Lass & Laing also claim that there are no instances of <y> for eME yin the SW Midlands. Close inspection of all LAEME source texts localised to the W Midlands reveals that there are, but only for WHY, in text nos. 246 and 1100 from Herefordshire, 2002 from Gloucestershire, and 1600 from Oxfordshire. Table 1 provides a complete list of all LAEME texts whose language has been localised to the W Midlands, and their spellings for eME yand u -. It seems to be true that many W Midlands texts show <u> for both eME u and y -, but most of them also show different secondary and minor spellings co-varying for each reflex. For instance, <ou/ow>, <o>, <uu>, <v>, <ov>, <w> are not infrequent as non-dominant variants for eME u -, whereas such spellings are rare for eME y -. For eME y -, non-dominant spellings such as <ui/uy>, <e>, <eo>, <i> are more frequent. In some W Midlands texts, the two reflexes appear to be kept apart; in these, the spellings suggest unrounding (and sometimes lowering) of eME y -. Such sources are no. 232 (Oxfordshire, 1175-1224), no. 189 (Herefordshire, 1200-24), no. 273 (Herefordshire, 1225-49), as well as no. 161 (Gloucestershire), no. 248 (Herefordshire), and no. 3 (Worcestershire; all 1275-99). Again, most of the W Midlands sources show a mixture of spellings which often contradict each other in terms of their implied sound value. LAEME spellings for lengthened OE y may prove helpful. Most of the source texts whose language has been localised to the W Midlands show dominant <u> for the reflexes of OE y in lengthening contexts, though quite a few show minor <i>, particularly for OE yht, and particularly towards the later period. In other words, lengthened OE y seems to have remained rounded in most of the W Midlands in the ME period.  ((u, v, o)) ‡ Suggested dates for source texts follow LAEME: the first number refers to century; 'a' and 'b' refer to the first half and the second half of the century, respectively; and '1' and '2' refer to the first and second quarter of each half-century. † Use of round brackets to enclose non-dominant spellings follows LALME practice: single brackets enclose secondary variants, and double brackets enclose minor variants. * An asterisk indicates that the text has <i> or <y> for why and/or OE bysen only. Exceptions are found in the following sources: no. 260 (Shropshire, 1200-24) has dominant <ey> for lengthened OE y other than yht (for which <uht>, <uh> are found); no. 158 (Gloucestershire, 1250-74) has <e> as well as <u>; no. 161 (Gloucestershire, 1275-99) shows only <ey> for lengthened OE y; no. 10 (Gloucestershire, 1275-1324) has <u> and <ei>. All of the preceding forms may point to lowering and unrounding of yto [e:]. However, generally speaking, the reflex of OE yht appears to be in the process of merging with the reflex of OE iht.
Thus, examination of spellings from all LAEME sources localised to the W Midlands corroborates at least two of Lass & Laing's claims: (a) there is no "neat geographical tri-partition for /y/", and (b) not "only are the symbol-to-sound mappings more multiplex than is suggested, but there is a strong element of lexical specificity in the set of reflexes" (Lass & Laing 2005: 281). In other words, certain lexical items seem to be spelt in certain ways, which supports the theory of change by lexical diffusion (Phillips 2006a(Phillips , 2006b: sound changes seem to start in certain phonetic contexts, and/or in very frequent words, whence they spread from context to context, and from frequent to less frequent words. Their third claim, that there are no "particular spellings uniquely associated with OE /y(:)/" (ibid.), is not entirely correct, as <ui/uy> seem to be used for the reflex of eME yonly. 5 It is their assumption that the reflex of eME yfell in with the reflex of eME u in the SW Midlands which is the most difficult to prove. ME spellings seem to lend support to their view: Table 1 shows clearly that <u> remains dominant for the reflexes of both eME yand u in the SW Midlands up until the last quarter of the thirteenth century, although <ou/ow> slowly take over as the dominant spellings for eME u from c. 1250. A very few examples of <ou> for the reflex of eME yappear to support the merger hypothesis also.
Evidence from the modern dialects may settle the issue: if presentday SW Midland accents show a merger between the reflexes of eME u and y -, the ME spellings may be taken to indicate just that; if not, the same spellings simply show a lack of distinct representations on the orthographic level, but not phonetic or phonological merger. The Survey of English Dialects (Orton & Dieth 1962) yields material for the lexical items listed in the Appendix for the W Midlands, and for Wiltshire and Berkshire. This phonological material from the modern dialects implies that there was no merger between the reflexes of eME yand u in the W Midlands, since generally eME yis reflected as a number of diphthongs with unrounded off-glides, and eME u is reflected as a series of diphthongs with rounded off-glides. However, exceptions are found in some localities in Cheshire and Staffordshire: That is, in these localities, there is sometimes a diphthong with an unrounded off-glide for the stressed vowel of cows/kyes (OE cy -), mouse, house, mouth, drought, thousand, clouds (all with eME u -), as well as for dry, hide, fire, wright, why (all with eME y -). Interestingly, these have converged on the 'normal' reflex for eME yand not for eME u -. Thus, the ME spellings from the SW Midlands likely indicate merger of spellings, but not of sounds, as otherwise it would have been impossible for the reflexes of the two sounds to be distinguished again later on an etymologically correct basis. The only way in which merged sounds could unmerge would be if two different systems co-existed, one of which kept the reflexes apart, and the distinction was re-introduced into the system in which merger had taken place. But such a scenario remains speculative and unlikely.

Cheshire
Interestingly, even in many LAEME source texts with dominant <u> for eME y -, there is only <i> for the lexemes WHY and OE bysen 'example'. This seems to suggest that if the unrounding of ystarted in any one word, that word is most likely WHY, although it should be noted that there was an OE variant whie. 6

Summary of early ME material
The material from LAEME suggests the following developments for the reflex of eME y -: 1. It was unrounded to [i:] across the country, and this process seems to have started in the (South-)East and the W Midlands.

It was lowered and unrounded to [e:] in Essex, Suffolk and
Kent in the late twelfth or early thirteenth century; in Somerset, Gloucestershire and Wiltshire in the latter half of the thirteenth 6 Likewise, if the lowering and unrounding of yto [e:] started in a specific word, it may have been in the OE word þyster 'dark, gloomy' and cognates, since there are frequent <e> and <eo> for these lexemes even in texts with dominant or exclusive <u>, <ui/uy> for all other words with eME y -. However, OE þyster and cognates also appear with OE ie, eo, so ME <e>, <ie> etc. for these probably go back to forms which did not have OE y -, since they show a very distinct pattern in the extracted material. For this reason, ME spellings for OE þyster and cognates have not been included in Table 1. century; and perhaps in Lincolnshire in the early fourteenth century. 7 3. It remained as [y:] in the W Midlands and parts of the South-West, as well as in Ely and Huntingdonshire (at least as a minor variant) in the late thirteenth or early fourteenth century.
With regard to Lass & Laing's hypothesis concerning the phonetic nature of the ME reflex of OE y -, the modern material does not support their claim that the reflex of eME yhad merged with the reflex of eME u in the ME dialects of the W Midlands, even though <u> is the dominant spelling for both in ME in the area in question.

The later ME material
The LAEME material for eME yneeds to be tied up with the later material from SMED and LALME. Conclusions based on analysis of this material are briefly summarised below.   The development of eME ybefore palatals appears to have been different from that in other contexts: in this environment, ywas unrounded to [i:] in Hampshire and Dorset, and also in Berkshire, Wiltshire and Somerset. There are traces of such a development in Surrey also, but not in Sussex or Kent. Finally, Devon shows a tendency to unrounding regardless of phonetic context. Ek (1972) investigates the ME development of OE eo and yin the South-East, using onomastic material which partly overlaps with that of SMED, although much of Ek's material is earlier. His conclusions differ somewhat from Kristensson's, particularly regarding the extent of the [e:]-area. However, Kitson (1998: 170) concludes that since Ek's material is earlier, "what the two investigations show between them is a retreat of the e-reflex in favour of the u-reflex as well as, further north, the i-reflex". 8 In other words, Ek's and Kristensson's data demonstrate change in progress.
The LALME material suggests that the reflexes of eME yhad been unrounded to [i:] in the East and the North in the late ME period. Retained [y:] is implied by spellings from the West and the South, and from the West Riding of Yorkshire, whereas <e>-type spellings linger on in most of the East, and in parts of the West (Gloucestershire, Worcestershire, Oxfordshire), the South-West (Devon, Dorset, Hampshire, Wiltshire) and the South-East (Kent, Sussex, Surrey). That is, <i>-type and <e>-type forms co-vary in the East, but whether these systematically correspond to [i:] and [e:] is an open question.