Home‎ > ‎

Vowel System

For a version of this article which uses IPA rather than SAMPA - as the original printed article did - see here


This article arises from a need to understand the Shetlandic vowel system for purposes first of all of orthography, and secondly for the practical and accessible representation of Shetlandic pronunciations in word lists. The first purpose equates roughly to phonology; the second requires additional phonetic information. In this context, the term 'Shetlandic' refers to the traditional language of the Shetland Islands, also known as 'The Shetland Dialect', or, along with Orcadian, as 'Insular Scots'; and the word 'dialect' is used relatively, to refer to regional dialects of Shetlandic. This article does not deal with the pronunciations of the increasing number of younger Shetlanders - mainly in the town of Lerwick - who speak only a form of Scottish Standard English. Also, in this article, the unqualified term Mainland refers to the main island of Shetland.

Academic studies have typically dealt with the Shetlandic vowel system as an aspect of some wider area of study. Jakobsen's monumental Etymological Dictionary of the Norn Language in Shetland was, as its name implies, concerned only with the Norse element of the Shetlandic vocabulary. Consequently his phonetic transcriptions are concerned mainly to identify the underlying Norse phonemes rather than to investigate how their reflexes function systemically in modern Shetlandic. The ongoing investigations of Gunnel Melchers are part of a project entitled 'The Scandinavian Element in Shetland Dialect'. Conversely, J.C. Catford's important article 'Shetland Dialect' (Shetland Folk Book, vol. 3) is primarily concerned to defend the relevance of Shetlandic in the wider context of Scots studies, particularly in extracting evidence for earlier Scots pronunciations. The limitations of the approach of the Linguistic Survey of Scotland in analysing the Shetlandic vowel system will, I hope, become evident in the course of this article.

This article is based primarily on my own (North West Burra Isle) pronunciation - represented in the LAS by Hamnavoe - and limited investigation into particular aspects of North Isles and Mainland (of Shetland) type dialects, supplemented mainly by The Linguistic Atlas of Scotland (LAS) 1.1 to 1.10, and Paul Johnston 'Regional Variation' in The Edinburgh History of the Scots Language. In order to analyse the data in LAS, I first typed the vowel representations for each section into a database in a modified form of SAMPA (1). Tables for different purposes - for example, vowel against region for one LAS section; vowel against section for one region; section against region for one or more vowels - could then be extracted at need by queries in Structured Query Language. The resulting tables were then transferred to a word processor, and converted from modified SAMPA into an IPA font by a program in the word processor's embedded programming language. The article does not deal with diphthongs.

In order to avoid confusion between different conventions, this article uses the following notations (given for reference in Table 2).

(Note: for this internet version, a modified form of SAMPA is used where the original printed article used the International Phonetic Alphabet (IPA)).

Modified SAMPA forward slash notation - e.g. /e/ - is used first of all for consonants, which are not part of the investigation; and secondly, and of vowels exclusively, for the vowel representations used in the LAS.
Modified SAMPA square bracket notation - e.g. [e] - is used for my own generalised phonetic representation of Shetland vowel sounds. It should be noted that, although these are phonetic rather than phonemic insofar as they distinguish certain allophones (2), they are nevertheless approximations for contrastive purposes and are not intended to be accurate in relation to cardinal vowels. Thus, for example, I have used [o] for a vowel sound which LAS typically represents as /O+/ or /V*/, and [I] for one which LAS often represents as /ë/. It may seem inappropriate that my less precise representation uses a notation normally associated with phonetics, while the ostensibly more accurate LAS representation uses one normally associated with phonemes. However, I shall argue that - as the introduction to LAS recognises - the LAS attempt to combine contrastive representation with a maximum of phonetic information has not always been successful in the case of Shetlandic.
Capital letters or digraphs in bold typeface - e.g. AE - represent my own categorisations of Shetlandic vowels, which are also my default spellings. Unlike my square bracket notation, this notation does not represent variation which I regard as allophonic. Rather than beg the question by referring to these as 'phonemes', however, I shall refer to them simply as 'vowels'. This notation, which reflects the current distinctions and distributions of vowels in Shetlandic irrespective of etymology, is used in preference to historical word classes (such as BEAT, BAIT, etc), and should not be confused with the LAS representation of polyphonemes, to which it is not related.
Modified SAMPA notation between backslashes - e.g. \e\ - is an 'allophonic' notation designed as a pronunciation guide, which, while essentially phonemic, distinguishes certain typical allophones by means of diacritics. This notation is described in the notes to the vowel table (Table 2) at the end of the article.

When giving word examples from the LAS, I use their standard English spellings. My own examples are given in a Shetlandic spelling, whether or not this coincides with that of standard English, with translations in brackets where necessary. Most of the non-English examples I have given - e.g. caib, peig, plag, oag, buil, roed - are words I have heard and used in speech since I was a child, with a few - such as leid, loumie, inbuis - which I have heard more recently; and I am thus familiar with their pronunciation in my dialect. The only exceptions are houb (lagoon) and Houbie (a place name) where I have had to rely on pronunciations from other areas.

For the purposes of this article, it is appropriate to make some general comments about the LAS data for Shetland.

Firstly, the regions of Shetland represented in the LAS are strikingly peripheral. There are two from the North Isles (North Yell and Fetlar); three from the far West (Papa Stour, Walls and Foula); three from the South (Dunrossness, Scousburgh and Fair Isle) and two (Hamnavoe and Oxna - now uninhabited) from the small group of islands to the South and West of Scalloway. The entire North, East and Central Mainland is completely unrepresented, as are some of the more distinctive dialects such as that of Whalsay.

Secondly, the list of words used is not always well suited to eliciting the characteristics of Shetland phonology. For example, in Section 1 (vowels + /d/) the OA [o:] vowel is attested only in a few regions where there are particular exceptions to the general merger of this vowel with the O [O] vowel. The inclusion of words such as froad (froth) and affroad (dissuade), and the proper names Mowat ([mo:@d] in my pronunciation) and Dodie would have elicited this vowel in more regions. Other, less straightforward, examples will emerge in the course of the article.

Thirdly, I am surprised by the apparent differences between my own pronunciations and those given for Hamnavoe - about half a mile away from where I grew up - which I perceive as being almost identical to my own. There are certain phonetic features of the informant's pronunciation - mostly diphthongisations - which I am unable to hear in my own pronunciation. More seriously, there are several cases - some of which will be discussed in the article - where words which I pronounce differently are grouped together under the same phoneme. Even though I have questioned people who live within the village of Hamnavoe I have not been able to explain all these anomalies satisfactorily.


I first became aware of the importance of vowel length in Shetlandic when I asked my father the meaning of a word I had read in a poem, which was spelt loomi, and which I pronounced [lumi] to rhyme with roomy. My father had no idea what it was until I explained the context; then he said, 'Oh! Du means loumie!' [lu:mi]. Though my father knew the word perfectly well, I had pronounced it with a short rather than a long 'oo' sound and it had therefore been incomprehensible out of context.

Many similar problems with the written representation of vowel length in Shetlandic are owing to graphemic rather than phonological factors. There would be no way, for example, to predict from the orthography whether the word which the Shetland Dictionary spells inbös has a long or a short [ø] sound. This problem, however, is related to the fact that the final consonant in inbös is /s/, not /z/ (which is often written <s>, as in English), and is thus easily explained in terms of the Scottish Vowel Length Rule. It is possible that the length of the vowel in loumie as opposed to that in roomy could be explained in morphological terms. However, there are a number of words which, in my own pronunciation and generally within Shetlandic, have long/short vowel contrasts in similar environments:

Short Long
back [bak] baak [ba:k] (balk)
baet [bet] (beat) bait [be:@t]
got [gOt] goat [go:@t]
need [nid] meid [mi:d] (landmark)
buil[bøl] (resting place)
guid [gød]
broel [brø:l] (bellow)
(speak nonsense)
room [rum] loumie [lu:mi] (oil slick)

Catford identifies three such long vowels - AI as in bait, AA as in faat (fault) and OA as in boat - giving his maximum Scots twelve-vowel system. As his inventory was based on monosyllables closed by /t/, however, he does not remark upon the contrastive long vowels which are noted in LAS in such words as meid (landmark), leid (diligence) and broel (bellow).

The length distinctions noted by Catford are easily explained in terms of distinctions in older Scots, and thus fall within the normal scope of Scots dialectology. Of the remaining distinctions, most of those before /d/ are presumably owing to the change of /D/ to /d/ in most dialects of Shetlandic (e.g. meid [mi:d] from an original meith [mi:D]) where the SVLR-conditioned long vowel has been retained even where the fricative has become a stop. In words of Norse etymology, the long vowel often seems to derive directly from Old Norse - so e.g. roug [ru:g] (heap), O.N. hrúga; leid [li:d] (diligence), O.N. hlít; neib [ni:b] (to hang the head, nod off), O.N. hnípa; loumie (oil slick), O.N. ljómi (brightness, radiance). (3) The fact that Old Norse intervocalic and final /p/, /t/, and /k/ usually became /b/, /d/ and /g/ respectively in Shetland Norn (Jakobsen vol. 1 p. lv) may account, along with the SVLR and the shift of /D/ to /d/, for the fact that long [i:], [u:] and [ø:] occur only before voiced consonants and are therefore not evident in an analysis based on syllables closed by /t/. Occasionally the preservation of Old Norse vowel length leads to minimal pairs, such as screed [scrid] (n: a swarm), O.N. scriðr; and screid [scri:d] (v: to swarm), O.N. skríða (4).

While the LAS records contrastive [i:] and [ø:] before /d/, it does not identify [u:] as a separate phoneme (unless the [ju:] sound which occurs in words like feud, huge is regarded as such.) This may be because it is not easy to find words in current spoken Shetlandic with long [u:] before /d/, and although Jakobsen records many such words it is not easy to establish that these would be contrastive in the same dialect with a word such as lood [lud] (loud). Johnston's comment ( p. 475) that this vowel obeys Aitken's Law strictly is concerned only with reflexes of the historical OUT class. Long [u:] does, however, occur in words of Norse origin like oub [u:b] (whimper); roug [ru:g] (heap); toug [tu:g] (tussock); and, though contrastive short [u] does not seem to occur before final /b/ and /g/ - presumably because of the rarity of this combination in SSE - the length of the vowel is similar, compared to that in words where short [u] does occur, such as lood, to the length of the [i:] in meid compared to that of the [i] in need. To put it another way, I, as a native speaker, perceive the vowel in oub, toug, etc. as being long in the same way that I perceive the vowels in meid and roed [rø:d] (to speak nonsense) as being long; and pronunciations of oub and toug as [ub] and [tug] would sound just as wrong as if I were to reverse the length of the vowel sounds in words such as buil [bøl] (resting-place) and broel [brø:l], or need [nid] and meid [mi:d]. Both long and short [u:] occur before /m/, as in loumie [lu:mi] (oil-slick), cf. room [rum] (and perhaps before /b/ in e.g. houb (lagoon); Houbie - a place name - as opposed to ruby, though I am not familiar with these words in my own region.) As my mispronunciation of the written loomi shows, such words seem to have unpredictable vowel length irrespective of SVLR criteria. (5)


'Soft Mutation' is my own term for a change in quality which affects certain Shetlandic vowels when they occur before certain consonants, and for the sake of brevity rather than technical accuracy I shall refer to the range of allophones which occur typically before voiceless consonants as 'hard' and those which occur typically before voiced consonants as 'soft'. In my dialect, this mutation affects eight vowels and occurs almost regularly before voiced stops, though the data from the LAS suggests that there is considerable local variation as to which vowels are affected in which environments. This characteristic may correspond, in part at least, to what Johnston describes as the Insular Scots Clockwise Vowel Shift ( p. 485; 11.2.6. p. 447) - certainly the general trend is for vowel sounds to be raised and/or fronted before voiced consonants - but Johnston does not define this term sufficiently to make its scope clear. In my analysis, Soft Mutation is a general tendency in Shetlandic phonology which underlies the further diversification of realisation which may occur in more specific environments, and also lies behind certain phonemic mergers which occur only under certain conditions. It could be said that the phonetic realisations of affected vowels before certain consonants are not simply allophones of the phoneme which occur before specific consonants, but specific realisations in their turn of a general tendency for certain vowels to undergo a systemic shift before certain types of consonant.

Recognition of soft mutation as a general tendency provides, along with vowel length, a framework for looking at Shetlandic vowel phonology as a whole rather than as an aspect of wider analyses. For example, it is well recognised that the A vowel is typically raised before /d/, /b/, and usually /g/ - in LAS typically to /æ/ or /a+/ before /d/, as opposed to /a/ before /t/. Johnston, analysing Shetlandic vowels from the perspective of historical Scots phonemes, describes this (p. 485) in analytical terms as applying to the A phoneme alone without relating it directly to similar characteristics of other vowels. Catford (p. 72) notes that diphthongisation of certain vowels may indicate former palatalisation in the following consonant, and attributes to this (p. 74) the well known merger of the AE vowel with E or EE in different types of Shetlandic dialects - e.g. haed [hed] (head) to 'hed' [hEId] in the Central Mainland and to 'heed' [hid] in Yell (see Table 1). From the point of view of this article, both of these phenomena are aspects of a general phenomenon of soft mutation. In the rest of the article I shall specify the hard or soft mutation of a vowel where required by suffixing a lower case letter to my capital notation - so AEh = hard AE, AEs = soft AE, etc.

According to the LAS data, this phenomenon does not occur as regularly in all forms of Shetlandic as it does in my own dialect. Whereas this is no doubt true to some extent, it is also evident that the approach of the LSS is not well suited to collecting the necessary information. The introduction to the LAS itself states that the practice of giving voiced and voiceless counterparts of following consonants in the same section has not always been technically successful, and this would certainly seem to be true for Shetlandic, leading in some cases to the grouping together of what I perceive as different vowel phonemes.

The tables derived from the LAS data suggest that, in spite of the impression of phonetic accuracy given in the representation of phonemes, where voiced and voiceless following consonants are combined in the same section words have been grouped under what is perceived - presumably by the investigators, not by the informant - as being a single phoneme, and represented by the phonetic value for one or other of the soft or hard mutations. Thus, for example, whereas all but two of the ten localities investigated exhibit a raising of the A vowel before /d/ (Section 0) as opposed to /t/ (Section 1), only three localities are recorded as exhibiting this characteristic before voiced as opposed to voiceless velars, and only two before voiced as opposed to voiceless labials, where the voiced and voiceless consonants are not automatically differentiated by section. However, as Johnston states (p. 485), A raising occurs typically before /d/ and /b/, and often also before /g/. This suggests that raising of A before voiced stops has been 'levelled' in the LAS analysis according to perception of the underlying phoneme.

In most cases this levelling is phonologically justified, but it is not applied consistently throughout the localities under investigation. In one case - Fair Isle - we have the following allocations (LAS spellings and notation):

/E;/ bag, bang, long, neck, tack, wrong
/a;/ back, bank, break, drank, make, sank, shake, take

Here the opposite of phonemic levelling - or a mistaken levelling owing to a failure to identify the effects of soft mutation - has occurred. In my own accent, there is little phonetic difference between As as in bad and Eh as in set - although I would write them as [æ] and [E] respectively, they probably occupy overlapping phonetic ranges. It is obvious to me from the data that the Fair Isle pronunciation is similar, in that As, as in bag, bang, long and wrong, is pronounced the same as Eh in neck and tack, and that bag, bang, long and wrong should be grouped phonemically with back rather than with neck; so:

1. E - Eh [E] neck, tack (6)
2. AE - (a) AEh
[a] back, bank, break, drank, make, sank, shake, take
(b) AEs
[E] bag, bang, long, wrong

This may be what Johnston means when he says (p. 485) that the raising of A can lead to some phonemic overlap with E allophones. However, because [E] representing As and [E] representing Eh do not occur in the same phonetic environment they can never be confused, and there is therefore no real phonemic overlap.

A rather more complex case occurs in the following group from the LAS data on the Hamnavoe dialect:

/ë/ bird, earth, fir, firth, her, hoarse, kirk, kirn, our, their, word, work (verb)

In my pronunciation - which I should expect in this case to be systemically identical to that of the informant - these words have two different vowel sounds, roughly as follows:

[I] bird, earth, hoarse, their, word
[3] fir, firth, her, kirk, kirn, our, work (verb)

Presumably, these differing pronunciations have been grouped under /ë/ (my [I]) according to the aforementioned practice of phoneme levelling. What has apparently been missed here, however, is the fact that where a vowel is followed by /r/, /l/ /m/ or /n/ plus a stop, it is the second consonant in the cluster which determines whether the vowel is realised as its soft or hard allophone. (7) Bird and earth have the same, or a very similar, vowel sound; but in bird it represents Is, determined by the /d/ after the /r/, and in earth it represents AEh, determined by the /t/ (not /T/ in Shetlandic) after the /r/. The words should therefore be grouped phonemically as follows:

1. AE - AEh [I] earth, hoarse
2. I - (a) Is
[I] bird, their, (8) word
[3] fir, firth, her, kirk, kirn, our, work

While voiced stops have this effect when second in a cluster, other voiced consonants may not - kirn (churn) has Ih, though kin would have Is. Again, although Is and AEh probably occupy overlapping phonetic ranges they do not occur in the same environments and therefore there cannot be any phonemic confusion.

This is true generally in Shetlandic. Although, when I am speaking Shetlandic, I pronounce bad with approximately the same vowel sound that I use for bed when I am speaking English, there is no possibility of confusion with bed when I am speaking Shetlandic, because in Shetlandic I pronounce bed differently from both my English pronunciation of bed and my Shetlandic pronunciation of bad; so (very roughly):

bat badset (9) bed
My English[a] [a] [E][E]
My Shetlandic [a] [E][E] [EI]

There is therefore a domino effect by which most of the vowel system shifts before voiced consonants, and phonemic integrity is maintained. (10)

In certain cases, however, soft mutation does lead to phonemic merger. The best known is the merger of AEs, as in e.g. haed (head), hael (whole) with Es - as in hell - in Mainland type dialects, and with EE - as in heel - in North Isles type dialects (including, paradoxically, Fair Isle), with the result that North Isles and Mainland type dialects lose a phoneme in these environments as compared to dialects like my own which distinguish hael from both hell and heel (Table 1; Catford p. 74; cf. Mather, 1964, p. 44). There are other cases. I can hear no difference in my own pronunciation between the Us vowel in words like bud, gub (lather), dug (dog) and the UI vowel in words like cuit (ankle), guid (good), which does not undergo soft mutation. This perception is not echoed by LAS with respect to the Hamnavoe dialect, but in Dunrossness and Scousburgh Us and UI words are grouped together under /Ö-/ in certain sections.

There are many apparent discrepancies in the LAS data which seem to be due either to a failure to identify soft mutation, or to differences of approach in allocating its results. Often the data from Fair Isle shows this; for example:

Vowel + my pronunciation Eh [E] Es [EI] Ah [a] As [æ] AA [A:]
LAS example lengthbend wantbound (v.pt) band
Hamnavoe E;+^IE;+^I a+ a+Ä;
Fair Isle æ e; a;æ A;

In this case, the Hamnavoe data uses the phonetic transcription for the soft allophones of both E and A (in bend and bound) as the phonemic index for the phoneme as a whole, although the hard allophones (in length and want) are in fact pronounced quite differently. The Fair Isle data, on the other hand, has not identified the soft and hard allophones - determined by the consonant after the /n/ - as the same phoneme, but has grouped Eh with As because of their similarity in sound, as with bag and neck above. This peculiarity of the Fair Isle data often means that it alone provides phonetic information on soft mutation.

It seems that the environments required for raising of A are different from those required for raising of AA. In both Hamnavoe and Fair Isle there is a three-way distinction between bend, bound and band, and whereas A is raised to As before /nd/ in bound (the phoneme cannot be E, because bend is quite distinct) band has the length and quality of vowel expected of AAh, rather than that expected of AAs [æ:], which occurs in Hamnavoe in words such as cause [kæ:z] and false [fæ:z]. This variation in the mutation of different vowels in the same environment is common before /m/, /n/ and /l/ - for example, before final /l/ I pronounce A as As but U as Uh, e.g. bal [bæl] (throw), but full [fol]. It is apparently the raising of A but not AA before final /l/ which creates the following distinctions in my pronunciation:

Es bell [bEIl]
As bal [bæl] - to throw
AAh baal
[bA:l] - eyeball
AAh baa
[bA:] - ball

The distinction between soft and hard allophones is here reinforced by the fact that the final /l/ in bell and bal is palatalised - as typically after front vowels - whereas in baal it is velarised (cf. Catford p. 72).

Many local shibboleths depend on this variation in the mutation of particular vowels under certain circumstances. The variation between soft and hard U in the word yun (yon) is well known - I say [jon] with Uh, whereas speakers from many other parts use the fronted vowel ([ø~8]) typical of Us. In this case my pronunciation is irregular, because I say e.g. bun, fun and gun with the fronted (soft) vowel sound. A well known shibboleth of my own Burra dialect is the pronunciation of A as [E] before /S/ and /tS/ (LAS 1.3. sect 9), often caricatured by other Shetlanders as the 'wesh', 'metch' syndrome. At first sight this seems like a transfer of A to E; but it is more probable that, uniquely in Burra, As occurs before /S/ and /tS/. This is supported by the fact that E, as in fresh and flesh, becomes [EI] - the sound expected of Es - in Burra before /S/ and /tS/, preserving the distinction of A from E in this environment.

In my pronunciation, a following syllable - e.g. /@r/ - can affect mutation. Thus dim [dIm], with Is, but simmer, [s3m@r] (summer) with Ih; gun [gøn] with Us but scunner [skon@r] with Uh. When the ending is a separate morpheme, however, the mutation is not affected; so I say dimmer [dIm@r], with Is, gunner [gøn@r] with Us.

In regions where AA has a soft allophone, [æ:], the pronunciation in final position seems to be indeterminate. Thus I make a distinction between caa [kA:] (to drive) and caa [kæ:] (to call), and also between faa [fA:] (1. to fall, 2. offal) and faa [fæ:] (to be obliged); but speakers from other regions do not necessarily make such a distinction. Similarly, before /r/, I distinguish war [wA:r] (1. war 2. seaweed) from waar [wæ:r] (worse), and I pronounce car, laar (breath of wind), etc. to rhyme with war, but tar, bar, far, mar (to confuse), etc. to rhyme with waar. In some other regions, however, car rhymes with waar.

Johnston ( p.485) regards the [æ:] sounds in words like this as variants of /Er/, developed from Older Scots fer, etc, rather than as an exception to the hand darkening to which he attributes the occurrence of AAh before /nd/ and /l r/, and while this accords with the quality (though not the length) of Eh in words like ert (direction) mert (cow for beef), it seems also possible that either (a) there is here a distinction between AAh in laar, war, etc. and As in tar, bar, etc. as apparently with AAh in baal (eyeball), paal (pawl) and As in bal (throw), wal (well), and that the regional variations of words like car are due to differences of distribution; or (b) that without a determining following stop as in wart [wart] and hard [hæ:rd], the mutations before /r/ are indeterminate, as they seem to be in the final position; or that there is a combination of these factors.

Finally, it would seem that phonemic levelling in LAS has led, in some cases, to a confusion of long and short vowels. The Hamnavoe data groups the following words (Section 5):

/ë/ apple, babe, cheap, cream

In my pronunciation these are approximately apple [e:@pl], babe [be:b], cheap [tSIp], cream [krem]. In my estimation only the respective differences between apple and babe, and that between cheap and cream, are due to soft mutation. The difference between apple and cheap and that between babe and cream are length differences, and are potentially contrastive, e.g. (with I words for comparison):

Hard Soft
Short I[3] rip, hip (omit) [I] rib, crim (clear throat)
Short AE[I] raep (rope; towel rail), cheap [e] laem (crockery), cream
Long AI[e@] rape, apple, frame [e:] caib (thole pin), babe

The fact that it is soft AE which occurs before /m/ in laem, but hard AI in frame, is similar to the occurrence of As in bal (throw) but AAh in baal (eyeball), and this type of distinction is frequent before /l/, /m/ and /n/. I am not aware of an occurrence of AE before /b/ to contrast with AIs in words like babe, caib (wooden thole pin), but this contrast occurs in my dialect before /d/ and /g/ - e.g. haed [hed] (head), maid [me:d]; paeg [peg] (peg), stravaig [str@ve:g]; and the length of the vowel in babe is that of maid rather than that of haed. Again, phonetic similarity between AEs and AIh has led to beg and egg (AE words) being grouped with ache, bake, etc. (AI words) in LAS 1.3. sect 5.

A contributing factor here may be the tendency of some Burra speakers to 'drawl' vowel sounds. The Hamnavoe data shows no difference between the length of the vowel in mad, bad, and that in fraud, Maud. But there is a distinction in my speech - bad [bæd], with As, fraud [fræ:d] with AAs - and the failure to distinguish e.g. the vowel in cream from that in babe may be due to a drawling tendency on the part of the informant. A possible criterion is that, whereas [bæ:d] (with lengthened vowel) is reminiscent of the pronunciation in some accents, especially perhaps when emphatic, [fræd] (with shortened vowel) simply sounds wrong and would probably be misunderstood out of context. (Fred is [frEId].)

The fact that soft mutation is not considered in the selection of the LAS word list means that the data is inadequate for understanding its effects on Shetlandic phonology even apart from phoneme levelling. For the Hamnavoe data, apple and babe are the only words in Section 5 which I would expect to have the AI vowel, and nieve is the only word in Section 4 which I would expect to have the E vowel. In the latter case the word has correctly been grouped alone. In cases where the mutation is dependant on the second following consonant the problem is even more acute. In Section 8 the only Ah word is want and the only Uh word dunt - the data for Fair Isle, true to form, groups both of these alone - and there are no words such as lint, tint, mint, which would have elicited Ih before /nt/.

Where the LAS data elicits sufficient information, it shows some regional variation in soft mutation. A comparison of Sections 0 (+/t/) and 1 (+/d/) shows that raising of As as opposed to Ah is recorded for eight out of ten of the districts represented. The ratios of fronting and/or raising for Es (9/10) and AEs (5/10) as opposed to their 'hard' counterparts are complicated by the merger of Es with AEs in Mainland-type dialects. Ratios of 7/10 are recorded for AIs, AAs and Us, and 5/10 for Is. (11) Only one district - North Yell - shows a raising of Os: no difference is shown for Hamnavoe, though fronting of Os as opposed to Oh is regular before voiced stops in my own pronunciation. A rounded front vowel /œ*-/, which corresponds to my own pronunciation of Os, is recorded for Hamnavoe only before /l/ (Section 7), though Scousburgh exhibits a similar vowel before both /l/ and /n/ (Section 8), and Dunrossness before /l/, /n/ and /b/ (Section 5).


Another feature worthy of mention is the tendency for the vowels E, O and A to be lengthened before voiceless fricatives, in words such as greth (urine), loff (loaf) and bath (cf. Johnston p. 471; p. 479). However, these vowels are sometimes also lengthened before stops - examples from my dialect are slock [slO:k], (slake); body [bœ:di] (person - where [œ] is the realisation of Os; cf. boady [bo:di], body), and gret [grE:t] (great). The last word, which might be expected to have [E:] in regions where this is the regular realisation of the AIh vowel, also has it irregularly in some regions like my own where AIh is realised more like [e@], and in some places - such as Papa Stour - this leads to its being grouped alone in the LAS data. In some cases these lengthenings can lead to distinctions such as that between great [grE:t] and gret [grEt] (wept); but as a whole the phenomenon is sporadic and can scarcely be regarded as phonologically significant. (12) However, in areas where length is the only difference between A and AA it may cause a merger before voiceless fricatives between A in words like bath, haff (ocean) and AA in words like traath (troth, in truth), which are pronounced differently in other regions. Other vowels are not lengthened before voiceless fricatives - e.g. ruif [røf] (roof); rooth [ruT] (rowing block).


For practical purposes, these characteristics imply that it is useful to distinguish six long vowels orthographically, as the length of the vowel cannot be reliably deduced from environmental criteria. Soft mutation, on the other hand, is usually predictable according to phonetic environment and region, and need not be represented orthographically. For a pronunciation guide in word lists, however, a pure phonemic representation which does not indicate the tendency to soft mutation seems inadequate, while a more phonetic representation, which would distinguish the allophones by different symbols, would be too region-specific and would disguise the identity of the phoneme. My 'dictionary' notation, described in the notes to the Vowel Table (Table 2) and marked in this article by backslashes, seeks to meet these requirements.


My reasons for investigating the Shetlandic vowel system are essentially practical, and I am not in a position to assess conclusions derived from historical sources. Nevertheless, from the perspective suggested in this article and in view of the limitations of the LAS data, some of the statements about the Shetlandic vowel system given in the literature seem surprising.

Johnston's comment ( p. 477, cf. p. 498), that 'in some central Shetlandic, CUT diphthongises to [O+u] = LOUP before all velars, not just /g/ (Mather and Speitel 1986: 3)' might suggest that pronunciations such as 'dowg' (dog) and 'lowk' (luck) are characteristic of central Shetlandic. Firstly, however, reference to the LAS shows that this characteristic is recorded only for Hamnavoe and Oxna, which are both from the small group of islands West of Scalloway (though with the [u], which Johnston gives in full, superscripted). Secondly, from the group of words given with this pronunciation from Hamnavoe - bunk, bog, dog, luck, lug, rug, shrunk, (13) sprung, sunk, tongue - only tongue and sprung may have a slight suggestion of diphthongisation in my pronunciation - certainly spellings like towng occur in Shetland literature, though I think principally from the North rather than the Central Mainland. I can detect no diphthongisation in my pronunciation of words like lug and luck, which have the Us [ø~8] and Uh [ø] vowels respectively. It would seem that the impression given of a diphthong before all velars is here owing to phonemic levelling in LAS, by which the phonetic realisation of one allophone (here perhaps +/N/) has been used as the index for the phoneme as a whole. Thirdly, the equation with the LOUP class is surprising, as in this section of LAS (Section 6, + velars) the data for Hamnavoe correctly shows that most LOUP class words, such as folk, howk and yolk, merge with e.g. box, cloak, clock, rock, soak, sock under /Q*+/ (my O [O]) as generally in Shetlandic before /k/. In most environments, LOUP words - including loup itself - tend to have the OO [u] vowel (Johnston p. 498).

Johnston does not apparently relate the merger of AEs with EE in North Isles type dialects to that of AEs with Es in Mainland type dialects, attributing it directly to clean- and stone- (and in outlying varieties heal- and tale-) raising ( p. 457). However, Johnston's description of these conditions (11.4.3. p. 457) apparently accounts for raising only before /n/, /l/ and /m/, whereas, as Catford and LAS show, the transfer occurs also before voiced stops. EGG class words, which follow the same pattern (Table 1) Johnston explains in terms of a reverse process, considering that the North Isles type [i] pronunciation is original ( p.471, cf. p. 460).

From the perspective suggested in this article, which is not however concerned with historical causes, the mergers of AEs with EE and Es are seen rather as the result of a systemic change which occurs before most voiced consonants, including stops. There are no obvious instances of AE before /b/; but as EGG class words tend to have the AEs rather than the Es vowel in areas where AEs is systemically distinct from both Es and EE, they seem to fit regularly into the pattern illustrated in Table 1. That is to say, in my Burra dialect egg, beg and peg have the quality of vowel characteristic of laed (lead) [e] rather than that characteristic of led [EI] or need [i]. (This seems parallel to the merger of O [O] with U [ø] before /g/, giving dug, dog; bug, bog; lug, log). The mergers of EGG with Es and EE shown in Table 1 are thus regular for AE words. If the [i] pronunciation of EGG type words is indeed original it would seem that the Burra-type pronunciation of EGG with AEs must be accounted for by reverse analogy with HEAD type words, as it would otherwise seem reasonable to suppose that all words which have distinct AEs in Burra-type dialects, including the EGG, STONE, HEAL and TALE subclasses, merge systemically with EE in North Isles types and with Es in Mainland types, as suggested by Catford (cf. Mather, 1964, p. 44); and that the Burra-type dialects thus preserve a distinction which has been lost in the others.


(Click on numbers to return to appropriate place in text.)

(1) Speech Assessment Methods Phonetic Alphabet. See

For the modified form of SAMPA used in this article, see:

Modified SAMPA

(2) Consonant allophones - such as velarised and palatalised /l/ - are distinguished in neither the square bracket nor the forward slash notation.

(3) There is also a regional variant, luimie [lømi], which has a short [ø] vowel.

(4) Jakobsen (p. 810) considers that, in regions where the noun has a long vowel, it is owing to the influence of the verb

(5) Also before /r/, e.g. mour [mu:r] (snow heavily); mourie [mu:ri] and mouricaavie [mu:rikæ:vi], (blizzard); but moorit [mur@t], (shade of brown); mooratoug [mur@tu:g], (ant).

(6) Tack rhymes with neck in most parts of Shetland, this being a genuine case of different distribution from English and unrelated to soft raising of A.

(7) Compare e.g. felt [fElt], felled [fEIld]; skirt [sk3rt], bird [bIrd]; print [pr3nt], grinnd [grInd] (gate); Walter [walt@r], shalder [Sæld@r] (oyster-catcher).

(8) The fact that in my dialect their is pronounced with [I] whereas fir, our, etc. are pronounced with [3] could be interpreted to mean that the phoneme in their is AEh rather than Is. However, because the combination [Ir] is extremely rare in my dialect, and our is also pronounced as [wIr] in some other regions, it is perhaps better interpreted as an anomalous pronunciation of the sort which may occur in very frequent words.

(9) Set is used as an example rather than bet because bet - the word, not the class - has the AE rather than the E vowel in most dialects of Shetlandic.

(10) Less habitual speakers of English may, however, retain the Shetlandic vowel mutations even when they are speaking English. This variable bilingualism is not always fully recognised in investigations of Shetlandic phonology. I have heard of one (anecdotal) case where a tutor in English language refused to believe that a Shetland student naturally pronounced the word bad as [bæd] because he did not do so when he was speaking to the tutor in English. In A Corpus of Shetland English (Stockholm, 1985) Bengt Oreström gives a description of Shetland pronunciation with examples which I find incomprehensible unless I pronounce some as if by a Shetlander speaking Shetlandic, and others as if by a Shetlander speaking English.

(11) Not including the pronunciations of Ih as /ë-/ and Is as /7*/ which are recorded for Scousburgh - a curious reversal of the normal direction of soft mutation.

(12) Though in some cases the length of an original [o:] vowel seems to have been retained - in my pronunciation soak [sO:k] compared to sock [sOk].

(13) Shrunk and sunk are arguably oddities in an investigation of Shetlandic (as opposed to Shetland English), as the verbs shrink and sink typically have preterites shrank and sank and past participles shrunken [SrokN] and sunken [sokN] respectively.

TABLE 1 - The AEs merger.

English Spelling (1) Burra TypeMainland Type (2) North Isles Type
hailhe:l he:l he:l
heelhil hil hil
hellhEIl hEIl hEIl
wholehel hEIl hil
boneben bEIn bin
headhed hEId hid
eggeg EIg ig

(1) Hail, hell and heel, which do not have the AE vowel, are given for contrast and comparison with whole, which does.

(2) In other Mainland-type dialects both hell and whole have the vowel sound characteristic of whole [e] rather than that characteristic of hell [EI].

Notes on Table 2.

In the following table of Shetlandic vowels the first column is the vowel in my capital notation; the Pos column gives the positional ('hard' or 'soft') mutation; the Phon column my own generalised phonetic approximation; and the Sym column the notation that I use in word lists. The LAS column gives typical representations in LAS.

The notation in the Sym column only, demarcated by backslashes in the glossary, uses the 'Umlaut' diacritic to indicate the tendency to soft mutation. This convention is distinct from the normal use of this diacritic to indicate centralisation (as in the LAS column). In this notation the long vowel sign [:] is used only where the length is regarded as potentially contrastive. Incidental lengthening, such as before morpheme boundaries and the lengthening of A, E and O before voiceless fricatives, is indicated by the half-long mark [;], without any implication as regards the respective phonetic length.


Short Pos.Phon.Sym. Typical LAS.Examples
Aharda aa , ä bat, back, saps

softæ~E ä æ , a+ bad, plag, sab
AE hardI ee, ë paet, traep, laek

soft (1)e , i , EI ë e , i , e- haed, stael, staen
E hardE EE set, lep, seck

softEI Ë E;+^I , e , e;- led, neb, hen
I hard3 I3 , 7*- pit, trip, lick

softI Ï I , ë lid, lib, rig
O hardO OQ rot, rop, lock

softœ Ö Q , Q*+ , œ* rod, job, brod
Uhardo oO+ , V* cut, wup, luck

soft8 , ø ö O+ , V* , ø , ø*- cud, gub, lug
EE-i ii , i- leep, leet, need
OO-u uu room, smoot, rook
UI-ø ø ø , y cuit, guid, buil

LongPos. PhonSym.Typical LAS. Examples
AA harda: , A: a: a; , A; baak, faat, whaap

softæ: ä: æ; , O; , Q*; claag, faader, aaber
AI harde@ , E: e: E; , E;+^@ laik, waik, mair

softe: ë: e; maide, whaig, caib
EI-i: i: i;reib, peig, meid
OA -o: o: o; , o;+^@ voar, coag, affroad,
OU-u: u: -loumie, oub, roug
OE -ø: ø: ø; , y; roed, broel, oese

(1) The [i] and [EI] pronunciations indicate regional mergers with EE and E respectively.

Glossary to Examples in Table 2.

aaber \ä:b@r\ keen
\afro:d\ dissuade
\ba:k\ balk
\brÖd\ broad
\brø:l\ bellow
\bøl\ resting place
\ke:b\ thole pin
\klä:g\ cackle
\ko:g\ crane the neck
\køt\ ankle
\fä:d@r\ God
\fa:t\ fault
\göb\ lather
\gød\ good
\gøs\ goose
\hëd\ head
\lek\ like
\le:k\ leak
\lip\ parboil
\lit\ attend
\lIb\ castrate
\lu:mi\ oil slick
\lög\ ear
\më:d\ maggot
\me:r\ more
\mi:d\ landmark
\nËb\ beak
\ø:z\ use (v)
\u:b\ whimper
\pet\ peat
\pi:g\ small quantity; glimmer
\pläg\ garment
\ri:b\ stripe, corrugation
\rÏg\ spine
\rø:d\ speak nonsense
\ruk\ smoke (v)
\rOp\ rope
\ru:g\ heap
\sEk\ sack
\smut\ sneak
\stël\ steal
\stën\ stone
\trep\ argue
\vo:r\ spring
\we:k\ weak
\Wa:p\ curlew
\We:g\ heifer
\wop\ bind, tangle


Catford, J.C. Shetland Dialect, in Shetland Folk Book vol. 3.

Graham, J. J. 1993. The Shetland Dictionary, Lerwick (The Shetland Times Ltd).

Jakobsen, J. 1985. An Etymological Dictionary of the Norn Language in Shetland (2 vols), Lerwick (Shetland Folk Society).

Johnston, Paul. 1997. 'Regional Variation' in Jones, C., ed., Edinburgh History of the Scots Language. Edinburgh.

Mather, J.Y and Speitel, H.H. (eds), 1986. The Linguistic Atlas of Scotland, vol 3, Croom Helm.

Mather, J.Y. 1964. 'Dialect Research in Orkney and Shetland after Jakobsen'. in Froðskaparrit 13 Bók 1, Mentunargrunnur Føroya Løgtings, Tórshavn.

Oreström, Bengt, (ed). 1985. A Corpus of Shetland English. Stockholm (Almqvist & Wiksell International).

Tait, J.M. (trans.) 1999. Guid Unkens efter Mark, Mark's Gospel in Shetlandic. J.M. Tait.

(c) John M. Tait, 2000.

(This article was first published in Scottish Language Number 19, 2000)