SSLS STORY-SCIENTIA LINGUISTICS SERIES 4 series editor Didier L. GOYVAERTS ‘also available: Vol. 1. D. L. Govvasnrs and G. K. Punta (eds.), Essays on the Sound Pattern of English, 1975, X +580 pp. Vol. 2. I. F. Hancock, E. PoLoMe, M. GoopMan and B. Hine (eds.), Readings in Creole Studies, 1979, XIV + 352 pp. Vol. 3. G. D. Pripeaux, B. L. Derwina and Wm. J. BAKER (eds.), Experi- mental Linguistics, 1980, V1+322 pp. in preparation: Vol. 5. K. Derez (ed.), Sociolinguisties in the Low.Countries, PHONOLOGY IN THE 1980’s Edited by D.L. GOYVAERTS 1981 E, STORY-SCIENTIA Scien Publishers GHENT/BELGIUM,© 1981 by B, STORY-SCIENTIA GENT [Nopartf hs work maybe reproduced nny for by print. phitoprint mieroim tr any other means without writen permission from the publisher [iets uit deve uitgave mag worden verveclvoudigdenlf openbaar gem door imidécl van dru ftokopie, microfilm of op welke andere wijze ovk. zonder yoorfgaande sehifelike Tosstemming van de wigever. i19a1i0008/33, ISBNS 906839 1505 (SSLS 4) ISBN 90 6439-2129 (te series) For Roger van de Velde, ‘with reepect and gratitudeSOMMER #.A. 1969, Knjen Photog: Synchrone end Diehront, Pate Lingus ‘BIL Caner (Orlinaly 1968 M.A. Thess, Univesity at Hawa, 1970, An Austin Languaze without CV Sylble’, UAL 36:57.8. 1972, Kunjen Syntae : A Generative View, Comber: ALAS. 19762,*A Problem of Metathesi in. Sutton (ed), 1976 19766, “Agent and Instrument in Contra Cape York Penns’ iP, Sutton 8), 1976, 1976*Umbuyzamu the Chasificstion of Cape York Peninsula Language’ Pace “inguin AA7=1331, Canter SUTTON P. (2), 1976, Languages of Cape York, Cantera: ALAS. RRSE. ON THE FUNCTION OF BOUNDARIES IN PHONOLOGICAL RULES Hang BASBOLL, (Nord Institut, Odense Universitet) INTRODUCTION ‘When the Editor of the present volume invited mé to contribute, he suggested ‘that I submit a paper on “The syllable in generative phonology” which had only been semi-published (in 1974). Tam glad to follow this advice, and 1 have taken the opportunity to add quite a few paragraphs concerning my current view on the function of boundaries (yllabic and grammatical alike) in phonology. ‘As I now see it, the problems discussed in the earlier version fit nicely into the ‘more general junctural framework put forward here (ee also Basbll 1978 b). ‘The present version of the paper is organized as follows. In section 1,1 first discuss some important views on boundaries in phonology put forward by CChornsky/Halle (1968), McCawley (1968) and Stanley (1973). Then (in section 1.2) present a general hypothesis on the function of boundaries in phonological rule, One consequence of this hypothesis is that boundaries should, generally, rot occur in the structural description of phonological rules. In section 2 (taken over from Basbgil 1974), I therefore discuss some rules which were taken by Hooper (1972) to prove that the syllable boundary ($) had to be allowed to occur in the SD of Perules, This section also contains a more ‘general discussion of the topic “syllables in generative phonology”. In section 3, [briefly consider the ranking function of (mostly grammatical) boundaries in French phonology. In the conclusion, finaly, some further perspectives and problems are mentioned. 1, GENERAL FRAMEWORK 1.1. Predecessors ‘According to Chomsky/Hall (1968), boundaries are units inthe phonological string which are cross-clasified by distinctive features, viz the features word boundary (WB) and formative boundary (EB). In addition to # (LHW, -H]) and + ([-WB, +FB]), they operate with a boundary =({-WB, -FB]) the postulation of which has been Gusty) criticized from many sides (apparentlym6 MANS BASHOLL CChoinsky/Hale do not consider the possibility of « boundary [+ WB, +B] If a specific boundary is mentioned in the SD of a phonological rule, it only spplivs to strings containing thé mentioned boundary atthe indicated place. ‘Apart from that, all oecurrences of + in the input sting toa rule are ielevant. for the application of the sul, wheres all occusrences of # (or =), on the ‘other hand, block its application (uses, of couse, a# (or=) is included in the SD at the appropriate place). ‘McCawley (1968), in contradistinction to Chomsky /Halle (1968), proposes that ‘boundaries be (linearly) ordered in a strngtv-hierarchy. The main function of boundaries is, eccording to him, that they serve to define the domain of rues. ‘Thus, exch phonological rule has a certain boundary as its “rank”, and each occurrence of a boundary of this rank as.well as of a stronger one serves to delimit (on one side) the extension of each chunk to be compared with the SD of the phonological rule. This has become known as the ranking function fof boundaries. I think itis more or less agreed today that McCawley’s Iierarchical model of boundaries is superior to,the SPE-model Stanley (1973) distinguishes between three functions of boundaries with respect to phonological rules: (1) rules ranked by a certain boundary (or stronger ‘one, ie, the function proposed by McCawley; (2) rules delineated by a certain ‘boundary (or stronger ones), e.. the rule which devoices final obstruents ia German; and (3) rules requiring a specific boundary in their SD. Stanley furthermore argues that if rules are formulated by means of variables, .g. if the mentioned final devoicing rule is formulated not as. @) fon} > [voi] | __# but instead as @) — X[ son} RANK: # 4 [ -voi] (where X indicates thet segments may occur at the lefthand side of the ‘environment and, consequently, the lack of a variable to the right of [son] indicates thatthe obstruent must be final in the relevant chain # __ #), then (1) and (2) above reduce to one type. (According to the conventions of ‘Chomsky/ Halle, on the other hand, rules of type (2) reduce to type (3), since stronger boundaries than # are symbolized by a sequence of #s; the SD [-son] #is thus aso satisfied by [son] ## etc) Vennemann (¢. 1972) has claimed that the correct environment for the Aevoieing rule in German mentioned above is /__$ where $ is the syllable boundary. Hooper (1972) argues extensively that this boundary should be allowed to occur in the structural description of phonological rues, se section 12.2. below. MeCawley (1968) p.36), on the other hand, recognizes the syllable 28.8 unit in the SD of P-rule in the following way (See section 2.1 below) “A rule which stresses the antepeaultimate syllable of words of three or more syllables, the penultimate syllable of two-yllable words, and the lone syllable of monosyllabic words could be written as @) 8 Estes) inom. __ SO) # ~ But MoCewley apparently has not used the syllable boundary as a jumcture in his phonological system, . 1.2. The hypothesis of the present paper On different points, the framework put forward here agrees with McCawley (1968), Hooper (1972) and Stanley (1973) = 1 agree with McCawley’ (1968) in his fundamental insight thatthe central function of boundaries is ranking. 1 also agree with his use ofthe syllable as a possible unit in the SD of phonological rules (eg. stress rules, see section 2.1 below). 1 agree with Hooper 1972 (and Vennemann 1972) that the syllable boundary plays « crucial role in phonology. I shall try to demonstrate in section 2.2 below, however, that her way of using this boundary is not always convincing. I shall claim that the syllable boundary also ranks phonological rules (in fact, it was shown in Basbgll (1972) that several mes accounting for the quality of certain short Danish vowols had the syllable as their domain, or, in other words, were tanked by the ayllable boundary) | agree with Stanley (1973) that of the tree functions of boundaries with respect to phonological rules which he distinguishes (ie. ranking, delineating and (cules requiring) a specifie boundary in the SD), the delineating function can be reduced to the ordinary ranking function by the use of variables, and I follow this proposal in my notations (see section 1.1 above). Furthermore, the ‘examples adduced as support for Stanley's type (3), Le. rules requ specific boundary in their SD, seem dubious to me : they are mostly taken ftom the analysis. of English stress and vowel shift by Chomsky/Halle (1968), and their account of these phenomena seems dubious by siny standard. I shall also claim (as suggested above) that the syllable boundary § - just like other ‘boundaries - does not occur in the structural description of phonological rules. 1 thus take ranking to be, ex hypothesi, the only function of boundaries (qllabic as well as grammatical) with respect to phonological rules But how is the use of the syllable as a unit. (0.2, in stress rules, seo section 2.1) related to boundaries?) The syllable is the unit of Rank (cf. Hooper 1972 p. 537);and similaly the grammatical boundaries define different units, like the morpheme, the minor phonological word, the major phonological word‘Phonological sentence (ee section 3); Just like syllables, che other units mentioned may occur as units in rules. Fg, rules for intonstion and stress pattems typically apply to strings articulated in such ‘higher units’ Gnformally : words, parts of a compound, word groups, ec.) Finally, let me state explicitly three strong postulates on boundaries in phonology : (1) The phonology of each language makes use of a small number of different ‘boundaries which are linearly ordered. (Comment: syllabic and grammatical ‘boundaries have a different origin, but the same function with respect to the phonotogical component), (2) Each boundary defines a domain, is, the extension of a phonological string, (Comment : the domain is said ts have the boundary as its rank). The domain is delineated on both sides by the boundary in question or a stronger one, whereas al its internal boundaries are weaker than its rank. (@) A domain can have two functions with respect.to a phonological rule (@)it can be the domain of the rule itself; (6) it ean occur as a unit in the SD of the rule. (Comment : in the latter case, the unit must be of a lower ‘rank than the rule), No boundary ean occur in the SD of a rule (till presupposing, of course, the use of variables as suggested by Stanley 1973), ‘From a traditional linguistic standpoint, there is certainly nothing revolutionary in these claims (in my view, this fact by no mes credibility). They embody the postulate that units lke the syllable, the morpheme, the word and the utterance can serve as the domain for phonolo ical processes, and that they themselves can be referred to by such processes ‘The boundaries are thus morely devices to delineate the units in question and hhave no independent function in phonology. In this context it should be remembered that units like the syllable, the word and the utterance can be defined independently of their boundaries. subtracts from theie 2. ON THE FUNCTION OF THE SYLLABLE IN PHONOLOGICAL RULES 2.1, The syllable asa uni in the structural description of phonological rules Consider how a very simple stress rule (like the one in Polish), placing stress on the penultimate syllable of polysyllabie words and stiessing monosplables too, would be formulated within an orthodox generative framework oo V> Erstress] /__Cy (V)Cy # or (i) V > [stress] /__C. (VC,) # orGi) = «V> [stress] /_ CYC, # ‘The notational istinotion between the three ways of parenthesizng the last syllable apparently does not corzetpond to anything empirically; this, of course, fs a shorteoming of the notation. Another drawback is the fact that one has to sate two instances of C, in the structural description of a rue. Now, what does C,, mean 7.C,, means that itis (relevant for the application of the rule whether there are any consonants after the vowels mentioned in the structural esoription or not. OF course, the over-all principle of what ought to be mentioned in the structural description of a rule is that only what is relevant for the ap- plication should be included in the notation. But in the case under discussion here it is the notationsl convention that stress cannot be ascribed to syllables which foreds one to state the irrelevant consonants. Compare this to the way in which the rule can be formulated when stress is ascribed to the syllable (he S-symbol) (remember thet we use variables, although this is irelevant for the present problem; notice that if stress were antepenultimate instead, there ‘would stil be a proudo-choice of parenthesization as there is in McCawley’s rule (3) here : (S(S)) or (S)S) (McCawley 1968 p. 36) © IF xs (S) + THEN: [stress] RANK: # Rule (5) avoids all the difficulties of rule (4). ‘The basic principle goveming the occurrence of the Danish stéd in foreign ‘words of a Latin-Greck type as well as the traditional school pronunciation of these languages (according to which the main stress is placed in agreement with ‘tho target languago, whereas the other feetures of pronunciation have in general ‘no connection with the classical languages) isthe following : if stress is on the ultimate or the antepenultimate syllable, the stressed syllable has fd i it has “sigd-basi” (Le. contains a long vowel and/or a postvocalic sonorant), whereas a stressed penultimate syllable doos not have stfd (the relation between non-main stresses and sida is more complicated and can be ignored for the present purpose) IF the notation does not allow one to asribe the stfd to a syllable, then the rule in question must be split up into two parts viz. in a traditional notation (here is an unacceptable alternative, vir. to postulate an ad hoe-transformation which in the notation moves the std around without its ever moving phonetical- ly, see Basbgll 1974 p. 481) © @o [ v Co (CVC) # ti | > Hstbal satroor250° HANS BASBOLL «@@ Pv “Hong tren] > Htéd] /__(C,Ve,VIC # o@) a) Tv long | > [+ stfa] |__C, (WCQYIC, # +ates, and a alee © ee [pen [sisal f: ong el seme a v 00 Fit] “tt sures] C(VEVIC, # Notice that rule part (a) as compared to rule part (bs notationally quite Afferent: in (a) the vowel i the affected segment (ie. the segment which i changed by the rule), in () iti the consonant; in (a) the affected segment is stressed, which i not the eas in (b); and in (a) the envionment tothe left is relevant which isnot the casein (b). I shall argue dhet all these differences have nothin to do withthe rule itself. Furthermot, thet is « notations] ambiguity in the pareathesizng of the enviroment to the sight which has no empirical content Quite independently of the phenomena under discussion here, there are gp0d arguments for a general “stqd-manifestation principle” according to which sta is manifested on the vowel if this is phonemally long, on the consonant ‘immediately fllowing the short vowel if this consonant isa sonoran, and not ‘manifested as a st¢d in the normal phonetic sense (but possibly by a “stéd-like’ Intonation, at least in certain Wester variates of Standard Dans) ifthe short, ‘vowel is immediately followed by an obstruent. Granted this stdd manifestation prinepl, the role in question canbe formulated quite simply ina syllabic framework 5 OD OF Fees} 65) 4 THEN: [+stid} RANK: # ‘The phonological behaviour of the Danish st#d thus strengthens the traditional concept of the syllable as bearer of prosodic features lke stress and tone. 22. Syllable boundaries and phonological rules with particular reference to Hooper (1972) “There is no doubt that phonological theory should allow phonological rules to be delineated by a syllable boundary (8), ic. when the rule exprestes a proces ‘aking place in absolute syllable initial or absolute syllable final position. (C use the quiliication “absolute” to zesolve the ambiguity of e.g, “syllable fins! postion” a meanig ether the final segment/consonant ofthe sylabe 2. lute syllable final postion”), of a sogment/consonant in the final part of the allble, ie. occuring after the sylsbic peak.) Examples of such proceses, which eccording to Hooper 1972 (and Vennemenn 1972) should be stated by means of the sylable boundary symbol $ in thetr structural description, are strengthening of glides (in Spenish) in absolute sable intial position (ct. Hooper 1972 p. 528, where the envisonment is given as /$ __) and tensing of nomlow stresed vowels (in Castilian) in open sylabes, ic. in absolute syllable final postion (ef. Hooper 1972 p. 530, where the environment is given ey___$). However, Hooper (1972) mentions the S-symbol in a much wider range of ‘Phonological rules than thove refered to above (je. cases where $ delineates the rule), and in most of these other cases I find that her use of $ in the structural escription of phonological rules is unjustified and in fact obscures the nature (of the procestes. This is the case with several ssximiation rules which, according to her statements, apply only across $. We shall now devote some attention to 4 typical example of this sort viz. the place-oF articulation assimilation of nasals ‘before consonants in Spanish (in “Allegretto” speech style, p. 525.9. ‘The problem to account for is that nasals assimilate to obstruents both within ‘words (eg. campo, ganga [-mp- -7-D, and across word boundaries (c ‘beso, un gato [ab ng-]). Before alides ((j, w]), on the other hand, the assimlation occuts only acest word boundaries (eg. un hielo, un huevo Fij-, nw-]), not within words @g- mil, lato, muevo, nvewo [my f rv. nv-]). Hooper waite : “With the sylable boundaries included in the strings, the solution tothe problem becomes obvious. Naslassintation occurs before a consonant ora glide only ia S-boundary intervenes. ‘Thus one rule generates the desired output for all esses (8: HB) [nasal] + [ocoronal ‘coronal B anterior Banterior fy back fs | rack 6 distributed 6 cistributedRule I accomplishes all that is required of & phonological rule : it generates all instances of nasal assimilation; it states the constraints on nasas followed by non-vowels; and it provides an explanation of the facts. Using $-boundares, ‘the difference between the glides of [un yelo} aid (nyeto] can be stated formally : the frst [y] begins @ syllable and thus is functioning as a non. syllabic segment, while the sccond is part of the syllable nucleus. Rule I states the appropriate generalization, that nasals asimilate only before segments that begin syllables, ie. segments in a non-syllabic function” (p. $26; my italics). ‘The frst thing to be noted in the passage quoted is that Hooper's “explanation” (hich “can be stated formally") does not follow from her rule formulation, but that it presupposes an additional axiom like the following : a slide which begins a syllable is « non-syllabie segment as opposed to a glide which does not bepin a ayllable, If this is not in plain contradiction to other passages of hers, 2, “two contiguous syllabic tegments form separate syllables. This rule applies to Spanish” (p. 534), at best ber terminology is confusing (the word, “syllabic” apparently meaning “part of the syllable nucleus” while the feature name “[sylabic]” seems to mean “constitutes the peak of a syllable"). Anyhow, ‘the “explanation” is based upon a dubious distinction between those prevocalic slides which do and those which do not form part of the syllable nucleus, and | find this far from convincing (it should be noted that in the speech style discussed here, the glides in absolute syliable-ntil position are not strengthened. to obstruents, p. 528). But if hor explanstion in the text is not connected with her rule formulation and {n itself unconvincing, let us then take a look at her notation. The cule in fact states that the assimilation takes place only ifthe segment to be assimilated is separated from the segment causing the assimilation by a $-boundary. But I find it hard to believe that itis the occurrence of a boundary which conditions ‘the assimilation. I suggest the universal principle that if certain assimilation ‘under certain non-junctural conditions takes place across a certain boundary, ‘then the Same assimilation under the same non junetural conditions will also take place in the absence of this boundary (and in the presence as well as absence of weaker boundary) This is, in fact, 2 corollary of the hypothesis ppt forward in section L2 above, ‘This i, of course, not to deny the observational adequacy of Hooper's assimi- lation rule, but I think the explanation should be sought in a different direction, although within the same basic framework, viz, that the syllable plays a crucial role for the assimilation. Even though the function of the glides in un hielo and nicto isthe sime, as I suggest is the cas the function of the nasals in the ‘ro examples is quite different. In the final part ofthe syllable there is no contrast of place-oftarticulation among nasals, and - with the Prague School ‘terminology - there is thus only one nasal archiphoneme syllable-finaly, ‘whereas syllable-iniially there is a three-way contrast between labial, dental, ‘and palatal nasels(simiazly the contrast between dental and palatal laterals is neutralized in syllable-fnal position). These neutralizations in syllable-fnal position are only one manifestation of a more widespread phenomenon : that syllable-final consonants aze “weaker” than their initial counteryarts in # number of respects: they are shorter, have a laxer articulation, and are much mot susceptible to assimilatory” phenomena (of. Grammont 1933, p. 184f1), Whatever the correct formulation of the rule may be, I think this is the heart of the matter. Notice that Hooper must claim that her rule “states the constraints on nasals followed by non-vowels” in such a way that if « nasal should be followed by a word-final obstruent, the former would not necessarily be homorganic with the latter, Such a word-finl cluster does not, of course, occur in normal ‘Spanish words, but it seems improbable that a nasal preceding e.g. « word-final Jp] could be anything but (ra). ‘According to Hooper, “these assimilation rules are two examples of generalizations that cannot be stated without the $-boundary” (p. 526; my itlies). This assertion ‘ean only be due to lack of imagination, and I shall now sketch two alternative ‘ways of eccounting for the nasal assimilation in what seems to me a more insightful way (Chave at present no strong arguments in favour of one or the. ‘other of these two proposals). One possibilty isto reinelude the notion of the archiphoneme in phonology, in ‘the sense that a rule can refer to a segment which is not, fully specified, L2, which has zer0 a8 its coefficient for one or more features (but in Hooper 1975, ‘which argues for the reintroduction of archi-segments in phonology, the formy- lation of rule (8) is repeated). Furthermore, the place-ofarticulation assimilation expressed by the rule is obviously not a simultaneous but independent assimila- tion of four distin festures : [eoronal, anterior, back, distributed], as the formulation suggests. If we therefore use one multi-valued feature place-of- articulation (sbbreviated [art], ef. Ladefoged 1971, p. 4211), the rule can be formulated as follows = © © [ss] [.s] Sart art, + THEN: [art] “This rule expresses the generalization that the nasal archiphoneme is assimilated as regerds place-oF articulation to a following consonant, whereas a fully specified nasal is unchanged by the rule; this explanation also holds true for the lateral, assimilation. (This statoment in terms of archiphonemes may be interesting inview of evidence from: speech error analysis discovered by Niels Davidsen- Nislen (1977) which points to an encoding of sp, st sk- as /s/ followed by a stop srchiphoneme as regards the feature(6) distinguishing /ptk] from Podg/) ‘Another possible tolution would be that [-syllbic] segments are subeategorized into [+ final] and [nal]. Thus the underlying forms could be - but need not be - sequences of syllable, each syllable consisting of an unordered set of sepments which would then have to become a sequence by a process of Tinearization (cf. Perry ms.) in accordance with the syllabic hierarchy (Gee Basbjll 1974 and 1977). This idea of linearization has evident parallels in higher parts of the grammar (ef. Chafe 1970, p. 250 ff. Thus the assimilation rule could be stated lke this : (0) oF: “+ nasal c final art + THEN: — [acart] In favour of a solution along these (admittedly quite speculative) lines would be the fact that for a large number of assimilation processes the segmient affected could be defined as being [+ final], and this is true also in cases where there is zno noutralization in finel position, asin the voicing assimilation rule which Hooper states (p. $30) in th following way : (1) E> Fvoiee) /___ 8 [voice] Hooper (1972) apparently does not consider the possibility that syllable boundaries can block certain phonological rules, or that the syllable could be a domain of phonological rules : she seems implicitly to assign the same junctural status to the S-boundary as to the -+- boundary according to Chomsky/Halle (1968), i. that if a $ occurs in the structural description, then the rule applies only to input strings containing $ at the indicated place; where 8 rules not mentioning $ apply regardless of the distribution of $. Let us examine the following quotation from Hooper (1972) : "Many P-rules in| Akan ‘depend upon syllable shape and thus syllable boundaries (. ..]. For exemple, consider the nasalization of high vowels which occurs before a nassl consonant {in the'samo syllable, (Note that this environment is extiemely common for ‘vowel nasalization- it occurs in French for all vowels in just such an environment) It is possible to avoid the mention of ‘syllable’ in formally stating the rule, by recognizing two environments: before a nasal before another consonant, and before a morpheme-final nasal : eae [ M J? (+ nasal] /_ [+ nasal] \${ # +hieh Its simpler, however, to state the environment in terms of $-boundarios : (3;HB) [OV | [nasal + nal} Ss [asa] 7 est Crm Note tut Re (13 HB) NOT an brevinof(12 ) Rle 13; HB) the a statement ofthe evnent, wile Re (2; HB) merely a ae contrivance that produces the same results” (Hooper, p. $33; my italics) ‘The point to be emphasized is that the two formulations do notin all cases produce the same results, but thatthe “ad-hoc contrivance” in goneral gives the right output, whereas her rule (13; HB) may give a wrong output. For example, French words lke cent, sans [182] (1 use examples from French) in their ‘underiying forms end in obstruents (cf. centaie, eantenaire (stem, S6(ne:S] snd sans emour eSzam¥)-The latter example suggests that the phonological Soundary should oceur after the /2/ atthe point of the derivation where the vowel is natalized, since this isthe place of the word boundary (according to Hooper, one should not expect a readjustment of the $-boundary to the place before (x! until [2 i intervoclc, i. after the nasal has dropped, se p. $27, 1537). This conclusion seems strengthened by the observation that a word like sens [ss] has a nasalzed vowel before a word-fnal pronounced obstruent (Ga certain varotes of Fronch, at leat, theo is no underlying schwa in such words) A further argument for the postion that nasilzation of vowels in French can occur also before natals that are notin absolute sylsble-fina positon comes from words ike pente [pt] in which the vowel is naslized before a pronounced obsteuent followed by pause. Such words end in an tunderying schwa, and t might bo clsimed that the /t ssylable-intal atthe point ofthe derivation where the vowel is nsaized. However, this does not gree with the fact that in a word like étiquette [etket] (which is underlying /otikoto#/, of. iqueter [tite], the phonological syllable border must go ater the final // since the /2/ is converted to (se Dell 1973, p. 198ff and Bas, 1978a) Although each of these arguments taken alone is not inrefutable, they nevertheless give a cértain implausibility to Hooper’ forme lation. And at any rate iti certain thatthe two formulations are not equivalent in all eases; and I find it hard to accept a notation in which every form /VNS/ is naslized, but ao, form /VNCS/. 1 therefore claim that the nasal consonant in these examples doos not necessarily ‘occur immediately before the $ which Hooper’s formulation demands. In order to make hor rule (13; HB) work, it should be-changed to (for the French V > [+nasal]/__ [+aasal] Cy $But this ie only one more example where one is forced to state in the rule what is ixelevant for the application of it, viz. Cy. How should this rule, then, be formulated ? In the text Hooper in fact gives the correct condition, viz. “before a nasal consonant in the same syllable”. Thus it is irrelevant whether the S-boundary occurs immediately after the nasal or not, what counts is that ot occur between the vowel and the nasal consonant, If we recognize the possibility that a rule can have the sylisble as its domain, or can have the $-boundary as its rank, or can be blocked by the ‘occurrence of w $-boundary within its structural description (all these formu- lations here being equivalent), then the rule can be formulated like this, as) IF THEN: ( +nasel] Y Gnas RANK: $ ‘where it should be a property of the rule, but not of its structural description, that it has the syllable as its domain. Let us now turn to a Danish rule in which there is absolutely no way out of ‘mentioning C, in the SD if we are to accept Hooper’s framework. This rule is ‘productive and obligatory (“allophonic"), The short Ja/ phoneme in Ds pronounced as a back vowel in the environment of /s], 26 front vowel (Lal) before zero and dental and as a mid or back vowel ({a] or [a}) before a (ia most language usages) aia The fat thatthe fst vowel of pronounced [2] as opposed to the first ‘Absion, akkeloje suggests that this manifestation rule apples with the eyllable rather than the word es its domain. The sylabification presupposed for this purpose is identical to that presupposed for the prediction of the manifestation of /o/ and of consonant gradation (according to these principles a single intervocalic consonant goes to the preceding syllable ifthe following vowel is schwa, but to the following syllable if its vowel is e “full vowel", ie, non-schov). Particlarly suggestive of an explanation in terms of syllables are alternative ‘pronunciation patterns like the following. Amerikaner is ether pronounced [ameWikhe:7 na, amepikMe: ” na}, where the first /a/ is followed by /m/, ‘which again is fllowed by the full vowel /e/, ic. /a/ occurs in an open syllable and is thus pronounced [a]; or it s pronounced fomBikl's:%nalwhere the second vowel is dropped (possibly via a reduction to schva), and consequently the first /a/ ocours in a closed syllable ending ina labial, and itis thus pronounced [a] (or fa]. Forms of the stem abbed like sbbed, sbbeder, abbediooe can be pronounced as [ibeb, abe(‘)Ba, abedise], where the second vowel isthe full vowel fa (which can even bo long in abbeder), and (aj thus oecurs in an ‘open syllable and is pronounced [a; or it can be pronounced as [6b3B. obeba, q | i i | POUNDARIES IN PHONOLOGICAL RULES, 237 ‘abedise}, where the socond vowel reduces to schwa, and the /b/ therefore belongs to the first syllable with the effect of retracting the vowel /a ‘Since the colouring of /a/ by the following homosyllable consonant is quite independent of eny additional final consonants (eg. the [afin samt [son? A] is pronounced exactly like the /a/ i Sam [ta.m? ]), Hooper would have to Formulate the rule Uke tl (19) a > [back] |__ L-eor] 8 (Notice that in the SPE-framowork of distinctive features used by Hooper, there is no way to capture the assimilatory nature of the process in question,because ‘what is alike in the vowel and the consonant is not articulatorly defined, see below). However, | argue that the domain of the rule isthe syllable, and that its correct formulation is the following ao a Etarave] 4 THEN: [+ grave] RANK : $ 23, Syllabification By “syllabfication” [ understand the division of a phonological string (consisting of phonological sogmonts and grammatical boundaries) into consecutive syllables, ‘and it therefore conooms the border between contiguous sylabes, not their intemal structure in terms of peak, nucleus, margin, and the like. ‘As to the cases where the syllable functions as a unit in phonological rules (ie. typically in rules conceming prosodic features like stess, tone, and stdd, cf. section 2.1 above), syllabification is not requited for the correct application of ‘the rules : what is necessary is only that the number of syllables be known, and this information can possibly be given with an identification of the syllabic peaks 7 Obviously, if a syllable boundary delineates « phonological rule (section 2.2 above), the input sting to this rule must be syllbified. Similarly, regarding other rules having the syllable as their domain, their input strings «lso appear ‘to presuppose a previous syllabification. However, on this point I would like to claim that phonological syllable boundaries can be partly indeterminate, Le. they need not always be fully determined. In a great number of eaves where the syllable plays a role in Aetermining phonological processes, the only relevant distinction is one between ‘open and cloged syllables, whereas a distinction between syllables ending in ‘one, two, or three consonants does not matter. This is the case e.g. with the stress rule in clasical Latin and with the rule predicting the manifestation of Danish short /o/.HANS BASBOLE Let me briefly discus «thied example ofthis sort, vz. the rule E-Adjustment in Fronch which in certain contexts (which can tentatively be defined as “in closed sylables") converts an [efor [2 into Je/ (ee Basbgll, 1978. for a discussion of this rule in connection with syllabificational problems). Examples like genevosjetterons [ jnvwa, yet 31, derived from 1 ganov tuaadt, yack ort aN #2 #/, show that an interoclic consonant between two unstessed schwas belongs tothe following vowel if no morpheme boundary interenes between it and the consonant, otherwise to the preceding vowel. But where doss the syllable boundary occu in an example like stvrerons [sevi 9 88], derived from /ffsowrt at -++oN #2 #/ (CI. sevrer [rove] / # sove + a+r #/) 7 If we say that the syllable boundary coincides with the morpheme boundary, then the fis syllable ends in the otherwise unknown final consonant combination /vr/ ; but if we say that the syllable boundary occurs between /x/ and jx, then we must recognize that the factor Aetermining the syable boundary, viz the morpheme-boundary, occurs ata different place, ie, that it can only retract the sylable boundary one place, and this sounds incomprehensible to me. In fact I think the choiee between the two mentioned locations of the syllable boundary ta preudo-prober which only arises in a theory forcing one to state one unique inter-segmental location of every lable boundary, A more realistic solution isto define the notions open and closed syllable formally in sucha way-that the fist sha in stvrerons wil oceur in a closed syllable, but without commitment to the choles between several locations which will n any context unde tive the same result Now a word should be said about the level at which syllabifcation applies. Fisstof all, I do not think it makes much sense to speak of syllables and syllebification at very abstract levels of representation (cf. Fudge 1969). OF course, one can claim (Vlooper 1972, p. $38) that Spanish estable is derived from /Ssiaf ble) via (by an e-epenthesis rule followed by a resylabification) [SesStaSbieS), but I fail to see why there must be S-boundaries (before /s/) in the abstract string /SstaSbleS/. Similarly, she notes that Spanish pan is derived ‘rom /SpaSneS/ via (by an e-deletion rue followed by a resyllabifiation) Spans). tn fac, I think that the only arguments for abstract syllable boun- dares diferent from the phonetic ones are the impact of such boundaries on ‘the application of phonological rules, There is thus no reason to postulate an “abstract syllable Boundary” between /a/ and /n/ in pan if tis boundary can hhave no phonological effect whatever and, of course, no phonetic existence stall (hi ne with our proposal above that phonological syllable boundaries are in some cases paily indeterminate). Notice that the present Aiscussion concerns the level of description, i. Ido not intend to exclude an abstract sylaification /SpaSneS/ if this word-intemalsylable boundary coulé ‘be phonologically justified in other forms with e-dletion, circumstances OUNDARIES IN PHONOLOGICAL RULES 239 ‘The danger in using abstract morphemes a8 frame of reference for sylabifistional phenomena canbe sen in the following quotation from Hooper 1972; “In Spanish th division is [SY in ih, but in English i is /S. Again the constraints for syllable-initial position correspond to the constraint for word: intial postion, To account for auch exceptions (iz, to the gneral principle that te sylable boundary occurs before the obstrucat in an obsruentliquid sequence; HB] formally, I propor that, in addition to the genera rule for inserting $-boundatls, the metatheory provide alist of posible exceptions [che choice of exceptions applicable in a given language i determined by rmorpheme structure conditions of that language [...]. The fact hat such & relation obtains may bear onthe question ofthe existence ofS boundaries i the lexicon” (p. 538; my italics). Hooper jumps easily from the similarity between the slblesnital constraint and the word-intial constraint of /s/ in Spenish (which i rea), to an imaginary similarity between the syllblodniil constreint and the morpeme-stucture constraint which i in lagrant contradiction to her underlying forms on p. $38 like [SuabboS) which hab an initial sl. The correct concision i, of course, - that there is no relation between clusters in abstract morphemes and sylabif- cation, wnles these abstract elutes are also found at more concrete Ives of representation. ‘The postulation of phonetic syllable boundaries should be confirmed or #in a givon style under certain grammatical conditions) proposed by Selkirk (cf. section 4.2 below) have been applied. (Cente #4(Foman, (prefix #), ROOT (+sufix), ai] ene +) cca sonst Pet Boundary stitore of the French sentence, (Og ete hat the contents of (spree seo, oe of mare thats, and (Oy that prosent at lsat ones, The formu arrives no parectheesstructre, but ts purely in trms of morphomes and srammatical Sounders, ‘Tho domain ranked by # isa ‘minor phonological word’ and the domain ranked by # is a ‘major phonological word’. Compounds are not accounted for by the formula, ‘This model should be interpreted in relative (as opposed to absolvte) terms: it prodicts eg, that there isa stronger boundary between prefix and root than between root and sufix, and, in parallel fashion, that ther-is a stronger boundary between a verb and an encliticized object pronoun, compared to an enclticized subject pronoun. Eg. in most styles a root-final high vowel is desylabified| before a suff beginning with a vowel (4g. niant, maniaaue (nj, manjek] , ‘but nota profix-nal high vowel before a root beginning with a vowel (0 antintomsique, biannusl [4 itor, biangl], not *[je, bia] ); and there ae no styles which permit glide formation inthe latter but sot in the former case, Siilaly in many styles there is vowel harmony between & root and a suffix (under certain phonological conditions), eg. ofdant (ied, sad], but not betwoun a prefix and a root (under identical phonological conditions), «8 [bot], not *[pBeta] and there are no styles which permit vowel y ithe latter but sot in the former case, This offers evidence for the stronger boindary between prefix and root compared to root and suffix. But it does not, of course, exclude that in more reduced styles there can be pide formation and vowel harmony in all the situations mentioned (cf. setion 4.2 below), ‘Our parallel treatment of prefixes and procltie “words” is supported eg. by the identical treatment of en in both functions, compare enivrer, en avril [Znivke, Gnewil], emmener, en Mauritanie { Somne, SimoBitani] BOUNDARIES IN PHONOLOGICAL RULES ‘The independent status ofthe endings fz) and /t/ i shown by the fat that the part of the word before these endings i in all respects treated as if it ocourred independently, eg, with respect to schwatreatment and stress. One may also refer tothe promunciation [zaml] (in non-standard French) for amis! which suggests that /x is reanalysed as a plural prefix. Finally, a great simplification of French verb morphology is obtained by the proposed analysis ‘The particulary tight connexion between a verb and an enclticized subject ‘ronoun is indicated by the following fats: (1) /0/ is regularly deleted in esc, ete. Gus, este [yi es, ete), whereas it bears the word-stress and is never deleted in eg. prendslo, sur ce, parce que ! and others (2) Vowel hermony may oecur, even in relatively high siyles (aceptable to Grammont! } in esti, exttu [et ety] but only ifthe encltc is subject. (3) The distinction between /e, ¢ / is always neutralized in favour of in phonologially closed syllables (a notion which is defined in Basbgll 1978 2). Ifthe vowol is immediately followed by #, the syllable is never (phonologcally) closed. Thus ‘there is neutralization in favour of ¢ in parléje (prés, parla (ps. simple), paraisje (impft.), all pronounced [paB\e:3] (with vowel length conditioned by the following homosyllabic 4, which fils in with the hypothesis that je is treated asa suffix): on the oti? hand, [ais kept es /0/ in pares, parent S 532, Ranking of some French phonological rules The boundais dscused here, ie, the sentence boundary # #4 the strong word boundary ##, the weak word boundary (identical tothe strong word- jinternal boundary) #, and the morpheme boundary (je. the weak (or irrelevant) word-interal boundaty) +, define four possible ranks of phonologied! rules (principles of structuring, inthe case of +) in French, When the syllable boundary § is included we have established flv posible ranks, each defining the extension of a phonological chain which ean sere at the domain for phonological proceses, Blow I shall enumerate these five phonological! chains and in each case mention one or more processes which can apply to the chains in question. As already mentioned, inreasng “word reduction” implies that sore and more processes apply to longer and longer cans (corresponding to a decreasing effect of the boundaries in question) This phenomenon is dsregacded in the following, where ve only consider a rather distinct level of style, with a high degeee of segmentalization (1) Phonological sentence. Rules of rank #4 #: phenomena atthe beginning and end of (phonological) sentences, concering schwa-dropping (ef. Dell 1973); in lower evels of style, certain “word-eduetions” (Such 88 assmuatons) apply with the phonological sentence as their domain.(02) ‘Major phonological word’, Roles of rank i ##:lsison and stress accent ddu syntagme’}; the ‘major phonological word’ includes pro- and encltics. @) ‘Minor phonological word, Rules of rank #f: word stress, vowel harmony and side formation, ‘Minor phonological words’ are the parts of the ‘major phonological words, eg. “bound pronouns’ (except encliticized subjects) and particles, but also, aecording to the present definition, prefixes and the endings Jo} and [t] (and the rest when all these morphemes are subtracted from the ‘major phonological word’, viz. a root or ¢ root plus suffix(s)). (4) The morpheme (or formative) There are no phonological rules proper of the rank +, but morpheme structure conditions (MSC) have + as their rank. It should be emphasized, however, that MSC are “abstract rules” with dubious (psychological) relevance, (The principles of phonological syllbification (a8 pro posed in Basbgil 1978.) have # as their rank, but under very restricted conditions they pay attention to a + in their structural description. (5) The syllable. Rule of rank $ : “closed syllable adjustment”, i. the neutra- lization of Je, €, @/ in phonologically closed syllable in favour of €, On more concrete leveis, the syllable seems to play an even more important role, cf. ¢.8. Schane 1973,-p. $2 ff, (It may be added that phonetic syllables represent a structuring of the linguistic expression (Helmsley's term) so that it becomes easier to encode and decode, whereas grammatical boundasies merely represent 1 projection of higher level information onto the sound chain). 4, CONCLUSION : FURTHER PERSPECTIVES By way of conclusion, I shall briefly enumerate some questions which are im- mediately raised when one works within the general framework of boundaries advocated here (this framework was summarized in section 1.2 above). Important ‘questions ike “where do boundaties come from 2” rust be left untouched. 4.1. Domains within the SD It isa consequence of the present hypothesis that all domains, ie. phonological strings ranked by a certain boundary, should be allowed to occur in the SD of rules. This applies, consequently, not only to syllables, but slso to what was called (in section 3) “minor phonological words” (ranked by #), “major pho- nological words” (ranked by ##), and to even longer chains. E.g. with respect to Danish stress rules (cf. Rischel 1972 and 1975), a eompound word fs treated “internally” ia one way, viz. by a rule whose rank is the whole major (eg, compound) word, the parts of which are treated as units in the SD; “externally”, however, it i treated in‘another way, viz. by rules of «higher ‘ROUNDARIES IN FHONOLOGICAL RULES. rank and which contain the major (eg. compound) word as units in the SD. 1 follows from the general conception of ranking used here that SD-intemal domains must have « lower ank than the rule itself, but it remains to be seen whether other restrictions can be found. 4.2. ‘The relation between a rule and its rank ‘There scems to exist at least a certain relation between the nature of a ‘phonological process and its rank (or domain). Eg. rules essigning consonant features to vowels and rules coalescing a vowel and the immediately following consonant often appear to have the syUable as their domain (ee section 2.2 above). However, this sltn i probably nt universal a ince certain ras seem tole to change thir dorcn Such conditions of rues changing thir domain aah be investigated; fr example, le (17) ove sees tobe inthe roca of enlarging its domain in younger seers, so tat words lik pai Skadeniker are now ofen pronounced [pho , oad” igo) (he pronunciation [pk suggests tat it not the ylable Boundary whish has, been mored). It isnot only by variation in time endjor space (social or geographical) that ‘what appears to be the same phonological process can have different domains. ‘Variations in speech tempo and (vaguely) “style” often implies that phonological rules change their domain, to that the domain gets larger (Le. the rank highs) as a function of faster and/or les formal speech (cf. Dressler 1976, § 15 : “In nnachlissgeren Stilen wird dieser Prozess auch [. ..] Gber Morphemrence hinweg ‘méglich [, .] In noch nachlissigeren Stlen kann dieser Prozess avch Wortgrenzen Aberscueiten [.. .}. Die Hierarche ist also: 1) innerhalb von Morphemen, 2) ‘bl lexikalisierten Prifigierungen, 3) bei produktiven Prifixen und semantischer ‘Transpatenz [ ..}, 4) in Kempositionsfugen und Phrasen”) . If there isa fixed relation (ie, not just & correlation) between the nature of a phonological process and its rank, it probably concerns its lowest posible rank, Le its minimal domain. ‘The concept of word reduction just referred to can be expressed by means of an ineressing amount of boundary deletion in faster and/or less formal speech (Ch Dresler i. : “L. Selkirk nimmt analog an, dass bei zunchmendem Tempo bow. Nachlissigheit immer mehr Wortgrenzen getilgt werden”), “This raises the further question whether all boundaries are erased at the end of ‘the phonological component (this ir one point on which Chomsky Hall ‘MoCawley and Stanley seem to agree) I happen to disagree on this, however, for the following reason : If the phonological component (in this context) isANS BASBOLE ‘meant to exclude ‘phonetic detailrule’, den all boundaries cannot be erased at the end of the phonological component since the phonetic detail rules un- doubtedly presuppose the boundaries for theic correct specification’of the phonetic output (¢, a8 to the quantity of intial vs. final allophones). IF phonetic detail rules are included in the phonological component, on the other hand, it i hard to see that the output of this component could be & mentally relevant phonetic structure, since we do not generally percsive the mentioned {quantity relations as such, but instead tse this information to structure the sound chain, 4.3. Sound change as evidence It is of course an important question to find out whether boundaries have any psychological reality (it might be investigated in terms of “analyzability of. forms by the native speaker"). ‘The very notion of psychological reality can scarcely be considered well-defined for the time being (cf. Linll in this volume), Howover, one aspect of it lends itself readily to investigation by linguists, viz. the domain of sound change. Given the proposed sound change XA ~*XB, where X, A and B are sounds (or classes of sounds), we can investigate which sequences of X-houndary-A are turned into X-boundary-B, and which are not. From this investigation a esrtain rank or domain of the sound change can be established, and this may shed light on the (psychological) reality of boundaries in phonology. To take an example, recent sound changes in Danish most often do not apply across the boundaries between the parts of a compound and between words, For example (ef. Brink/Lund 1975), the diachronic rule a [- grave] unless before (+srave] Ge. roughly, [o] becomes [a] except before velas and labial, [slcontexts excepted, ef ts synchronic counterpart: rule (17) in section 2.2 above) accounts for the distinction tand ‘tooth (ten? J: tam “tame” [Lamm J, ete. but [a], does become [a] before a velar or a labial which is separated fiom [a] by a # within the present framework, .. sofapude ‘sofa eushion’ {co:fapacbo]. ‘When we find isolated examples of apparent compounds or phrases which do nonetheless undergo the change in question, this may be due to the fect that this compound or phrase has been lexicalized, and this can often be confirmed ‘on semantic grounds. I shall only mention one (probably uncontroversial) example: the phrase hvabehar meaning ‘(beg your) pardon’. Although it is etymological identical to hvad behager ‘what pleases (you) [va(’8)- behie:? Ya}, {tis pronounced [vowahé:? J The seduction fa: 7a) for ] hus w musnber of parallels (old doublet forms) which will not be discussed here. But the pronunciation [vo.be-] instead of [vabe-]indicetes that $ should occur after /b/, i.e. there should be no # between /a/ and fo/. This correlates withthe fact that ‘the phrase is completely frozen semantically and syntactically. BOUNDARIES IN PHONOLOGICAL RULES 261 NOTES 1 Ths contents of section 2 ofthe present paper Is taken fom Besbp 1974, particle fiom section 3 there (Barby 1974 content more dacusson on #yUsbiictonal probs in Danish and on the stvatore ofthe syle tn terms ofa hieachy of atnctive features lo ef, Dat 1977). The contents of section are exteses from Dusbfit 1978. Tam indebted to Laue Bove, Nise Dereon Nice, EH Fiche J4ecesen, Ifegon Rlachel and Olaf Thorwen for sugeetiong on ferent parts of the promnt paper 2 Chomsky Hae (196 271) consider the Roundary= to hav Prod + (Dreier 197 2k thus mistaken when he comers al since these raed forsatons are predictable, by fhe {ef a th tn [ome] ctgrain whch node show fo aot Tor fot of hared morgogy iy French tnd fo bntoauen the Boundary™ 1, ef tive words tke mmangebie{Bndebl]. {3 By the term “maltivaaed” Xmenn tat the fertre can take» finite number lazer {an two (but in racic «small nomber) of octets Sehane (1973, 9. 96) apparently onsiareSmultiaiued”eqavalent to “sel in the pnts that mulnod features (Ga. features ona ate, “phonetle” lve of representation wlthin hls fmework) an “opposed to “indspanden?”) ‘aying within certaln lint” (hei Infact «normal property of "seals" In agreement Irih Sehana’s ase of th word, whores "a multivalued linearly order {ie only late umber of values on a cetain "esi", Whatever terminalogy i chosen, the diatinctions mentioned above al soem nocsary, Le. both Schane's and Ladeford’s 4 pa mentone i ston 3 sor tn yt he oun edt i French fo {sae wh en tense co no ad ee atte ee tte enon lp so es EPETS CB earns tunation nls natch apne 6 PA ond woe #0, Sato to © BY, wht ney ete il naa ed te te Ws Tce felnioa ened motte Ba tara ate [puleni Stee net ler te opp wate SE A which is decisive for the occurrence of the *-bovndory e.g. «technical Stinteratare tk"HANS DASBOLL BIBLIOGRAPHY [BASULL Tins, 1972, ‘Some conditioning phonological factor for the pronunciation of ‘short vowels in Danish with sparn reference to sllabiieation”, ARIPUC 6, 185210. BASHPLL Hons, 1974, “The phonologiest sible with spect reference to Danish”, ‘ARIPUCE, 39.128. BASDPLL Hans, 1975, “Grammatical boundaies la phonology”, ARIPUC 9, 109-136 BASBOLL Hans, 1977, “The suctore ofthe ellsble and a proposed rary ‘of distinctive featoree” in W. Dresler (6) : Phonologtea 1976, Wen. BASBOLL Hane, 1978, "Schwa, joncturs et aylabfiation dan es veprésentations ‘honoloiquer di angle", ALI 16, 147-82. BRINK Lats and Jéen LUND, 1975, Dansk Rigs 1-2, Copennagen. CAIRNS Chases E1969, “Marksdaaes, neutazation, and unless redundancy rules", Le, 45, 863885. CCHAFE Wallace, 1970, Moonta and the Soucture of Language, Cheng. ‘CHOMSKY Noam and Mortis HALLE, The Sound Patter of lich (= SPE), New York DAVIDSENNIELSEN, Nis, 1977," phonological analyse of English 2 ‘vith special eference to epeach exor evidence", JTPA, 5,328, DELL, Prangol, 1973, Ler rélet et les sone Introduction & le phonologiegénéate, Pate DRESSLER W,, 1976, “Kinnen Marphemfusen die Domine phonoloicher Prozese ‘bearenzan 1", OpusculsSoice et Lingua (PstchitIsalschenko), Klgenfurt, FUDGE Erik, 1969, “Sylabes", JE 4, 47.68, GRAMMONT Maurice, 1933, Pté de phonique (Pcs GREENBERG Joseph, 1965, ‘Some generations concerning Initial and ial consonant equences" Ling. 18, 5:34, HOARD James, 1971,Aspzatio, tensones, and slabcation in English, La, 47, 133-140. [HOOPER Joan, 1972, "The sabe in phonological theory", Le, #8, 525540, HOOPER Joan, 1975,""The azehisegment is natural generative phonology", LE, So 536560, JAKOBSON Reman, 1941, Kinderspreche, Aphesleund allenic Lautgtetze, Uppsala; [English translation Old lenguag, aphasia end phonalogealunivesals, The Hse, 1968. LADEFOGED Peter, 1971, Petar to lnc phoneten, Cieag. MGCAWLEY James D., 1968, The phonological component of «grammar of Japenese, “The Hague, PERRY Thomas A, ms, "Motathes in SytsbiePonology” pape ead at he ‘Phonoloretegung, Vienna 1976 (pot published inthe proceedings) RISCHEL s9inen, 1972, “Compound erst in Danish without a cycle", ARIPUC6, 211-230, BOUNDARIES IN PHONOLOGICAL RULES 269 RISCHEL Hrgen, 1975, “Problemer ved en gonereti beset af dansk ey, Slab or ‘Nordisk lolol Araberetnng for 1971-73, Copenhagen. SCHANE S.A, 1973, Generative phonotogy, New Yersy. 'SELIIRK £.0,, 1972, The pase phonology of English and French, unpublised dstation, MOT. ‘STANLEY Richard, 1973, "Boundaries in phonology", in Anderson, SR and Kipasky, (ei) Festachrih for Morris Hale, New Yor, 185-206 \VENNEMANN Th 1972, “On the theory of syllabi phonology” Linguitche Berichte 18, Brauaachwelg, 1-8. (Addendum +) [BASBOLL Hans 1978, “On the Ure of “Dosis Ling, 183-166 In Phonology, Pro XIE Iarn Cong.