The only factor which might argue against this characterization of oe-ligature -Tensing as lexical in this set of dialects is its contravention of Structure Preservation ( SP ) , a constraint which states that " lexical rules may not mark features which are non-distinctive , nor create structures which do not conform to the basic prosodic template of the language " ( Borowsky , 1986 :sx 29) .sx SP is therefore intended to prohibit the introduction of non-contrastive features into the lexical phonology :sx as Kiparsky ( 1985 :sx 87 ) says , " if a certain feature is non-distinctive , we shall say that it may not be specified in the lexicon .sx This means that it may not figure in non-derived lexical items , nor be introduced by any lexical rule , and therefore may not play any role at all in the lexical phonology " .sx Since Harris ( 1989a ) assumes [ plus-minus tense] to be non - distinctive for English , and since oe-ligature -Tensing clearly refers to this feature , the rule contravenes SP .sx However , Harris tentatively suggests that newly - lexicalised rules may violate SP temporarily , with the reassertion of SP perhaps influencing the future direction of change , although he provides no clear evidence of this determinative role of SP .sx In other dialects , oe-ligature -Tensing has moved even further from its Neogrammarian source :sx in RP , for instance , oe-ligature -Tensing is no longer a synchronically productive rule , but has caused restructuring at the underlying level , such that the tense reflex of historical short / ae-ligature / has merged with /a :sx / ( > / unch / ) from other sources , including earlier /a :sx r/ , in path , laugh and so on .sx Harris' discussion of oe-ligature -Tensing suggests that sound changes may be phonologized as postlexical rules , but may subsequently acquire properties of the lexical syndrome in ( 2 ) and become ( initially non-Structure Preserving ) lexical rules , which may then also begin to diffuse .sx Ultimately , the number of lexical exceptions may increase , and the rule will be lost , its effects being incorporated into the underlying representations .sx If these suggestions are substantiable , LP gains considerably in a number of domains .sx Labov's two types of sound change can be matched with credible synchronic counterparts , and his notion of more and less abstract changes linked with the lexical-postlexical division , although the case of oe-ligature -Tensing shows that Labov's correlation of particular features with each type of change cannot be maintained .sx Furthermore , LP can be seen to be free of another shortcoming of standard generative phonology , in which the incorporation of change into the synchronic grammar was all too readily formalisable , with the result that the model cannot distinguish naturally - occurring from unattested types of change .sx At the same time , the resultant progressive differentiation of dialects and languages was entirely unformalisable , due to the static nature of the model and the resistance of early generative phonologists to change in the underlying representations .sx In LP , the lexicalization of rules and their eventual loss provides a mechanism for change at the underlying level and for the introduction of surface and underlying dialect variation .sx In the following sections , I shall show that the Scottish Vowel Length Rule ( SVLR ) provides further evidence for these proposals , and constitutes an arguably even clearer illustration of the 'life cycle' of changes and rules outlined above , albeit with some interesting differences from Harris' example of oe-ligature -Tensing .sx In what follows , I assume a two-level lexical phonology/morphology of the type proposed by Booij and Rubach ( 1987) .sx For English , irregular inflection and Class I derivation will operate on Level 1 , and Class II derivation , compounding and regular inflection on Level 2 .sx 2 .sx THE SCOTTISH VOWEL LENGTH RULE .sx 2.1. The synchronic SVLR :sx a preliminary discussion .sx A typically English dialect is one which preserves a reflex of the West Germanic system of phonemic vowel length , having one set of lexically short and one of lexically long stressed vowel phonemes .sx .. Scots dialects , on the other hand , are characterized by the disruption of this dichotomous pattern , resulting in the loss of phonemic length :sx vowel duration is to a large extent conditioned by the phonetic environment ( Harris , 1985 :sx 14) .sx The process generally said to be responsible for this historical loss of contrastive vowel length in Scots , and for controlling the synchronic distribution of long and short vowel allophones , is the Scottish Vowel Length Rule .sx SVLR was first formulated in 1962 by A.J. Aitken ( after whom it is also called Aitken's Law ) , although its effects had been observed much earlier , in dialect studies like Patterson ( 1860 ) , Murray ( 1873 ) , Watson ( 1923 ) and Zai ( 1942) .sx A preliminary formulation of the synchronic SVLR , which applies in Scots dialects and in Standard Scottish English ( SSE ; the local sociolinguistic equivalent of RP , see Abercrombie , 1979 ) is given in ( 5 ) , and an illustration of its effects for two vowels appears in ( 6) .sx diagram .sx table .sx SVLR applies to the monophthongs /i u e o/ ( found in the standard lexical sets FLEECE/NEAR ; FOOT/GOOSE/CURE ; FACE/SQUARE and GOAT/FORCE ( Wells , 1982) ) ; to /a unch / ( lexical sets TRAP/PALM/START and LOT/THOUGHT/NORTH ( Wells , 1982) ) in some varieties , although these are consistently long in others , and to the first element of the diphthong / unch i/ ( lexical set PRICE ( Wells , 1982) ) , the only vowel to be affected both qualitatively and quantitatively .sx Its effects on the two other Scots diphthongs , /au/ ( MOUTH ( Wells , 1982) ) and / unch i/ ( CHOICE ( Wells , 1982) ) , is unclear ; / unch i/ is relatively rare in English in general , while /au/ is extremely peripheral in the Scots system , since Middle English /u :sx / did not diphthongize to /au/ by the Great Vowel Shift in Northern dialects , including Scots .sx The result is that lexical items like house , out , town have unch in modern Scots dialects ; where /au/ does appear , in formal registers and SSE , it may best be regarded as a borrowing from Southern English English , and might not then be expected to undergo a Scots-specific process like SVLR .sx The specification [+tense] in the structural description of SVLR in ( 5 ) is intended to include the input vowels listed above , while excluding the non - lengthening lax vowels / unches / .sx Lass ( 1974 ) , Aitken ( 1981 ) and Harris ( 1985 ) all accept that / unches / are exceptions to SVLR , but consider / epsilon / to be a lengthenable vowel , although they do not discuss it in any detail .sx However , there is little evidence for its inclusion in the lengthening set ; in fact , / epsilon / is consistently classified as non-lengthening in earlier dialect studies ( Dieth , 1932 ; Grant , 1912 ; Wettstein , 1942) .sx Recent experimental studies are inconclusive :sx Agutter ( 1988a , b ) , whose work will be discussed further below , did not test / epsilon / , and McClure ( 1977 ) , who did , was unable to examine it in as full a range of contexts as the other allegedly lengthenable vowels .sx For instance , / epsilon / does not appear in stressed open syllables , so that no examples of this vowel word-finally or before inflectional [d] or [z] are available .sx / epsilon is also rare before /r/ and voiced fricatives , and McClure was forced to resort to using the names Kerr [k epsilon r] and Des [d epsilon z] .sx McClure ( 1977 ) does claim to have found results broadly in line with the length modification expected if SVLR did affect / epsilon / .sx However , only one informant , McClure himself , was involved in this experiment , and his average vowel duration and range of durations were considerably higher than those of any speaker tested by Agutter ( 1988a , b) .sx This makes McClure's findings unreliable , since it is at least possible that they reflect " an exaggerated differentiation of vowel length in long and non-long contexts and extreme carefulness on the part of an informant who knew the purpose of the experiment " ( Agutter , 1988b :sx 15) .sx Since the early dialect evidence is the most conclusive currently available , I accept that / epsilon / , along with / unches / , is an exception to SVLR .sx There is distributional and behavioural evidence that these three vowels form a natural class of [-tense] vowels in Scots/SSE .sx All three fail to occur in morpheme-final stressed open syllables ; furthermore , in casual registers or under low stress , / unch / and / unch / tend to fall together for Scots speakers , and the entire set / unches epsilon / may also merge , as shown in ( 7) .sx diagram .sx The inclusion of [+tense] as an input condition for SVLR will , then , effect the appropriate exclusions , and is motivated insofar as the feature [ unch tense] itself is motivated .sx However , as Halle ( 1977 :sx 611 ) notes , " the feature of tenseness has had a long and complicated career in phonetics " , and objections have been raised against its integrity and validity .sx Certainly the claim that LP improves upon standard generative phonology , for instance in that it combats abstractness ( McMahon , 1989 ) , can gain nothing from avowed support for a " pseudo-feature " ( Lass , 1976) .sx One of the most vocal detractors of [ unch tense] is Lass ( see especially Lass , 1976 ) , who bases his case for the abandonment of the feature largely on the difficulty of locating distinct , measurable phonetic correlates for it .sx Lass holds that , when two vowels differ with respect to a cluster of factors such as relative height , backness and degree of rounding , these should be considered separately rather than ascribed as a set to " an explanatory abstraction " ( Lass , 1976 :sx 49 ) like tenseness .sx I believe , however , that these arguments can be countered , and outline five replies below .sx ( i ) It is true that tenseness is intimately connected with tongue height , frontness/backness and degree of lip-rounding .sx However , the tense-lax dichotomy does not rest on the individual contribution of these features , but on their variable conjunction ; and the weighting of contributory components is not constant in distinguishing different tense-lax pairs .sx So , although tense vowels tend to be more peripheral than their lax counterparts , the interpretation of peripherality is fluid .sx A high front tense vowel will thus be higher and fronter than its lax equivalent , while a low back rounded vowel expresses its peripherality vis- a-grave -vis its lax partner by being lower , more back and more rounded ( although some low tense vowels MAY be less peripheral than their lax counterparts) .sx It is this variable clustering of features , which would be difficult to relate using only the contributory elements , that [ unch tense] is intended to encapsulate ( for an example of a similar use of [ unch tense] , see Allen , 1965) .sx ( ii ) The use of [ unch tense] may make otherwise opaque processes characterizable ( see , for instance , Lieber's ( 1979 ) account of Middle English Open Syllable Lengthening) .sx This is surely one of the major tasks of linguistics , and a primary requirement of the formal and theoretical tools it employs .sx ( iii ) Lass asserts that tenseness is definable only according to its effects rather than " on the basis of a prior ( historically based ) partitioning of the lexicon " ( Lass , 1976 :sx 40) .sx However , a " historically based " characterization can be found for the three lax vowels / unches epsilon / in modern Scots/SSE , which form a historically motivated class as the only vowels in the system with no tense ( or long ) Middle English sources .sx /i e u o/ and the diphthongs have only tense sources , and /a unch / result from the merger of tense and lax vowels ( Lass , 1974) .sx However , / unches / have only short , lax sources , as does / epsilon / , given that all its possible tense source vowels collapsed with other long tense vowels during the Great Vowel Shift :sx ME / epsilon :sx / raised to /e :sx / and subsequently , in some cases , to /i :sx / , and although /a :sx / in turn raised to / epsilon :sx / , it afterwards continued to /e :sx / , leaving the long half-open front slot empty after the completion of the Vowel Shift .sx Synchronically , / unches epsilon / cannot be classified simply as short , since all Scots/SSE vowels are underlyingly short , but these alone fail to undergo SVLR .sx ( iv ) The property of tenseness may not always be reducible to a combination of length plus other phonetic features :sx for instance , it has been claimed that there are cases where both long and short vowels may be tense , as in Icelandic ( Anderson , 1984 :sx 95-96) .sx Furthermore , the association between length and tenseness may vary , not only cross-linguistically , but diachronically in one language .sx Thus , whereas in Middle English long vowels are consistently tense and vice versa , the advent of SVLR has altered this correlation for Scots/SSE , where tense vowels are now those that may become long under certain phonological circumstances .sx Similarly , Harris' ( 1989a ) work on oe-ligature -Tensing suggests that underlying restructuring in varieties like Philadelphia and New York City may be producing a distinction of short lax / ae-ligature / and short tense / AE-ligature / .sx ( v ) The above argument is ultimately circular , so long as 'tense' and 'lax' cannot be independently defined in phonetic terms .sx