Janez Orešnik Ljubljana CDU 803.959-4 MODERN ICELANDIC PREASPIRATION FROM THE PHONO- LOGICAL POINT OF VIEW Sumn:ary. The paper deals with the phonology of modern Icelandic preaspiration. The rule that produces the non-lexicalised cases of • preaspiration before what is written as .!?• !_, _! + 1, .!!: is different for haremad.i and linimel.i. The hare madi role generates preaspirati- on before an aspirated plosive + /1, n/. (The aspiration of the plo- sive is obliterated by a la ter deaspiration rule.) The linmael.i rule generates preaspiration before plosive + /l,n/ just in case there is at least one such form in the inflexional paradigm of the word-form to which the rule is to apply containing a long vowel immediately followed by a plosive (where the vowel and the plosive are those mentioned in the structural description of the preaspiration rule). - All instances of preaspiration in what is written as _EE, kk(j), and some instances of preaspiration in what is written as tt, are lexica- lised. The non-lexicalised instances of preaspiration in what is writ- ten as .!!._are produced by morphological rules. - Appendix A deals with Southern Icelandic quantity, Appendix B with such non-lexicali- sed / ~/ as occurs, say, in the nom. sg. stOll _of stol- "chair". 1 1 My thanks are due to Miss Margaret G.Davis, who has revised my English, and to my patient informants Aldfs SigurElardottir, Davfe Erlingsson, Jon Helgason, Jon Por, Magn.lls Petursson, Olga Sver- risdottir, Svavar Sigmundsson,. Vilhj:ilmur H. Gfslason, Porbjorg Helgadottir, and Porsteinn Mani .Arnason. 141 .!· Preaspiration is one of the characteristic features of modern Icelan- dic. It is pronounced as the initial part of what is written as .EE• tt, kk, before what is written as _E, __!, ~ + _!, m, .!!• and sometimes before or in what is written as tk. Examples: Kleppur name of a hospital in Reykjav{k, dottir "daughter", ekki "not", epli "apple",. lapm "~ld felow (used in addres- sing a person)", opna "open (vb)", , ~: _g, + l• m, _!!, or in bb, dd, ~· Kristjan Arnason and Hoskuldur Thrainsson were forced to posit underlying aspirated plosives in words such as jokulJ even ~ linm~­ li 2 and to assume that the preaspiration arises before / p , th, kj1, iJi; + /1, m,n/, for instance in the contracted jokl-, whereupon the aspiration of the plosives is obliterated in the position immediately after preaspirati- on: vowel syncope preaspiration deaspiration jok~l+ar jok~ar johk~ar johglar joklar3 2 The readers' attention is drawn to the fact that Hoskuldur Thrains- son uses the notation /p, t, k/ for the plosives with [ +spread glottis] (phonetically: aspirated), and (p,t,k] for the plosives with [-spre- ad glottis ] (i. e. unaspirated). 3 This derivation summarises Kristjan Arnason' § treatment. Hoskuldur Thrainsson' s derivation of words such as joklar is more complica- ted. Not only does he posit an underlying aspirated plosive in the root of jokull even in linm~i, but he also postulates, without any mo- tivation but the "free-ride" principle, a rule geminating aspirated plosives before /l,m,n/, so that the derivation begins as follows: vowel syncope gemination jok~l+ar .. h jok lar .. hh jok~-k lar The continuation of this derivation is somewhat uncertain. 1 assume that Hoskuldur Thrainsson' s role (28) first deletes the supralacyn- geal features of the left ;iJi; and thus generates /h/ (= the preaspi- ration) , hwhereupon Hoskuldur Thrainsson' s rule (34) deaspirates the right /k / by moving the feature [ +spread glottis ] (• the aspirati- on) of that segment to the newly. generated /h/. (But it can also be that (28) and (34)" apply in the opposite order; in that case the de- tail is a bit different, but the result is the same.) 143 Using the "free-ride" principle they have extended this treatment even to words in which there is no alternation between the presence and absence of preaspiration, such as epli "apple". Its putative derivation is, simplified, preaspiration deaspiration h ep 11 h ehp lI eh911 epli The postulation of intra-word aspirated plosives in the phonological representations seetns possible for haramaeli, where, say, the k of jokull is actually pronounced as aspirated in the non-contracted cases. In linmoieli such treatment, being a case of absolute neutralisation (Kiparsky 1968), is open to doubt. It is a purpose of the present paper to find a so- lution which would make it unnecessary to posit, in linmatli, underlying as- pirated plosives in words such as jokull. ~· My solution presupposes the division of all instances of preaspiration into lexicalised and non-lexicalised cases. In ·uninflected words, all cases of preaspiration are presupposed lexicalised, e.g. ekki [ -bgj- ] "not". In inflected words, the criterion is the distribution of the preaspiration in in- flectional paradigms: if the preaspiration occurs in all the forms of a para- digm, it is lexicalised; otherwise it is non-lexicalised. Examples. The word p{pa "pipe" is pronounced with a preaspirated .E in the genitive plural, p{pna [-h9n-] , and without any preaspiration in the remaining forms of its inflectional paradigm. In accordance with the crite- rion formulated above, the preaspiration of pfpna is not lexicalised. On the other hand, the preaspiration is pronounced in all the forms of brattur [-hg- ] "steep", and is therefore lexicalised. The criterion does not work in this simple fashion with the words whose inflectional paradigms display suppletion. The word-forms pertaining to such a paradigm must first be grouped into subsets whose members do not contract the relation of suppletion with one another, and then the criterion considers only one such subset at a time. Illustration: consider vatn "water", gen. sg. ~ [vas]. It can be assumed that the form [vas] and the re- maining forms of vacyi stand in a suppletive relation to each _other ,1 synchro- nically speaking, especially since there is also a regular gen. sg. vatns [ vahc;1:9.s J and the alternation between [ vah._!,~, + l,n, and with the preaspiration in..EE..i ~ kk.(All instances of preaspiration be- fore .E.• _!, ~ + !!!... and before tk known to me are lexicalised and thus out- side the domain of the preaspiration rule.) 4 .1. The preaspiration before „l» _!, ~ + !, n. 4.1.1. The situation in harElm)eli has already been discussed and the environ- ment of the rule can be informally formulated as follows: v __ [ +aspirated J~} I. e. preaspiration is generated between a vowel and an aspirated segment immediately followed by /1, n/. 4 (Only plosives can be aspirated in Icelan- dic.) 4 In the above formulation of the structural description of the harElm:tli preaspiration rule, a vowel is mentioned. As pointed out above, this is necessary because preaspiration is always immediately preceded by a vowel in pronunciation, and never by a consonant or anything at all. Let me mention,in passing, that these facts would follow automa- tically if the preaspiration were conceived (with Garnes 1976) as par- tial devoicing of the vowel that precedes the preaspiration. In that case the structural change of the rule would mention the vowel (the one to be partially devoiced), and there would be no need to include any vowel in the structural description of the ru!e. E.g. the haramieli version of the preaspiration rule would then b'e V - Vh / __ [ +aspirated J { ~} I. e. any vowel is partially devoiced before an aspirated plosive immediately followed by /1, n/. - Mutatis mutandis, this is also va- lid for linmatli. 145 Examples: jokull "glacier", nom.pl. joklar; depill "point", nom.pl. deplar~ fetill "s trap", nom. pl. fetlar; aukinn "augmented", nom. pl. m. auknir; ...,-;-- --,-. ,,,,. • -:::i----- etinn "eaten", nom. pl. m. etmr; batur "boat", definite dat sg. batnum; --.,- . ,...-- -r-r:- „ rjupa "grouse", gen. pl. rJupna; minuta "minute", gen. pl. minutna. - These examples are not all equally reliable. Those displaying preaspiration in the contracted case forms (i. e. the first five of the list) are best. In the examples such as batnum, rjupna, and mfuutna, the preaspiration can be a part of the ending, an infix that accompanies the desinence. Such two- -part endings are comµionplace in modern Icelandic; e.g. the gen. pl. s~la of s11ell "happy, blessed" contains the desinence -a and the infix /d/. (The rule for the formation of the forms batnum, -rjupna etc. would be: the desinences /nYm/ and /na/ are preceded by the infix /h/ if the root ends (a) in harem12li, in an underlying vowel + aspirated plosive, (b) in linmxli, in an underlying long or unstressed vowel + plosive. The unstressed vowel must be mentioned because of cases such as mfuutna, see section 4. 1.2.2 below. For underlying long vowels see Appendix A. The place of the in- fix is always immediately before the last consonant of the root.) However, this has no bearing upon the formulation of the structural description of the rule: it is the same whether the examples such as batnum, rjupna, mf- nutna are included or not. Neither the harematli nor the linmxli version of the structural description of the preaspiration rule can be V __ [-continuant ] [ ~ l (i. e. preaspiration is generated between a vowel and a plosive immediately followed by /1, n/) because then the rule has a great many systematic exce- ptions: the language is full of words containing /9 '} g/ + /1 n/ not preceded by preaspiration. Examples: afl [ -9!] "power", villa [ -91- J "mistake", ~ [-g! J "tail", hofn [ -9Jt J ''harbour", einn [ -9!1] "one", rigning [-in-] "rain". The harem~li preaspiration rule must be followed by a deaspiration rule in the ordering. The deaspiration rule deaspirates plosives preceded by preaspiration. Informally, [ +aspirated J-[ -aspirated ] / h__ in haremaeli I. e. any aspirated segment loses its aspiration if immediately preceded by preaspiration. 146 Bjorn Gue:finnsson 1946 mentions a sporadical substandard pronunciation (most likely to occur in East Iceland - Jon Helgason viva voce 1978) in which preaspirated plosives can have aspiration. Example: ekki [ ebgjhl ] "not". For the intersection of such speakers and of harEl.m2tli speakers the structural description of the deaspiration rule must be narrowed to h C, i. e. the deaspiration takes place between preaspiration and consonant only. This assures deaspiration in, say, joklar, but leaves the aspiration intact in, kk. 5 say,~ The harEl.mltli deaspiration rule must also apply immediately before /s/. Examples: gen. sg. bats [ -9s ] of natur ( -th_ J "boat", skips [-\>s ] of s kip [·Ph] "ship", leiks [-gs ] of leikur [-kh-J "play". 4.1.2.1. As to the preaspiration rule in linm~i, we will start with the fol- lowing temporary formulation of its structural description: v_x{~J where X is difficult to determine. As already mentioned, X cannot stand for the aspirated plosives, for linm:aeli lacks such segments outside word-i- nitial position in simplex words. For instance, both batur and bat are pro- nounced with the unaspirated [ g ] • On the other band, if X stands for /l> g g/ tout court, the rule has many systematic exceptions, as shown above in section 4.1.1. I propose the following solution: let the structural descrip- tion of the linmoieli preaspiration rule be V __ [ - continuant ] { ~ J i. e. (the rule applies) between a vowel and a plosive + /1 n/, where the plosive is not specified as aspirated. Let this formulation be accompanied by the following complicated condition on the application of the rule: in the inflectional paradigm of the word-form to which the rule is to apply (recall that the preaspiration role applies to forms of inflected words only) there must be at least one form containing a long vowel immediately followed by a plosive (where the vowel and the plosive are those mentioned in the struc- tural description of the preaspiration rule). 5 The existence of the preaspiration-cum-aspiration pr0nunciation corrobo- rates Hoskuldur Thrainsson' s implied hypothesis that the preaspiration and the deaspiration rules are two mutually independent processes. On the other hand, I do not know how Hoskuldur Thi'ainsson' s deaspiration rule would handle deaspiration of plosives before s, for the rule, as it is now formulated, would move the feature [ +spr;ad glottis ] from the /si to the immediately preceding plosive, thus making that segment aspi- rated. 147 Examples. (1) Nom.pl.m. etnir of etinn "eaten". The inflectional paradigm of the word contains forms such as etinn, where a long vowel @ is imme- diately followed by a plosive. Thus both the condition and the structural description of the preaspiration rule are fulfilled for etnir' to which the rule consequently applies. The pronunciation is [ jehQnI!' ]. (2) Nom. pl. m. h0ggrur of hogg(v)inn "hewn". In the inflectional paradigm of this word there are no forms containing a long vowel. Thus the condition of the preaspiration rule is not fulfilled, and the rule conse- quently does not apply to hOggnir. The pronunciation is [ hOgru_r]. (3) Nom.pl.m. lagnir of laginn "skHful". While forms containing a long vowel do occur in the inflectional paradigm of this word, e.g. in laginn, that long vowel is never followed by a plosive. Thus the condition of the preaspiration rule is not fulfilled, and the rule does consequently not apply to lagnir. The pronunciation is [ laignir ] • (4) Nom. sg. m. einn "one". The same situation as with (3) above. The preaspiration rule does not apply to einn. Its pronunciation is [ eiitt of l>inn "thy", sitt of sinn "his etc. ". 9 In hara:mDtli, the aspiration of the underlying stem-final dental of the types flatti, flattur, bfttu, hvftt is deaspirated by a deaspiration rule, on which see section 4 .1.1 above. - It is an open question wh- ether glatt "gladdened" is to be derived from the verbal stem : glaQ- . (this shape of the verbal stem not realised anywhere on the syste:ma- tic phonetic level ! ) in the same way as glatt "glad" is derived from the adjectival root . glaEl- , or from the past participle stem ·glad~ in the way indicated in the table above. In the latter case there is a complication in the area using the Northern Icelandic quantity sy- stem: the length of dd must be obliterated after the insertion of the preaspiration before dd, and for this a special rule is needed. 154 The grammaticalisation of the phonological rule which once upon a time ge- nerated preaspiration in .!!z into a set of morphological rules must be due to the inability of the learners of the language to construct, on the basis of the systematic phonetic forms, underlying representations sufficie- ntly alike in the relevant respect so that one . phonological rule could gene- rate the preaspiration in them. An example from linm•li: since the root of hvftur "white" ends in [ d J , and the strong nominative/ accusative singular ~~ . neuter ending of adjectives often is f g j , the learners of the language can only be led to construct the phonological representation hvf/ d+d/, which would most likely yield the unacceptable hv [ lwn, tbere are relatively large transitional zones bet- ween har&m~i and linmatli proper (Bjorn Gu&finnsson 1946). Tbe question arises as to wbicb of the two proposed preaspiration rules bas been incor- porated in to the grammar of tbe speakers of the transitional zones. This interesting question cannot be sol ved in the present paper. While it is not impossible tbat the speakers of the transitional zones master both versions of the preaspiration rule, it seems more likely that they make use of that of the two versions wbicb can generate botb the linm~li and tbe har&m~i farms. The linmaeli preaspiration rule bas in fact this property, wbereas the bar&maeli version does not have it (because it mentions word-internal aspirated plosives - a category of segmenta practically non-existent in non- initial position in linmaeli). It is therefore likely that the speakers of the transitional zones utilize the linmidi preaspiration rule. (This would also be in keeping with tbe fact that the linmieli pronunciation is tbe prevailing and tbe spreading pronunciation in Iceland.) If this is true, the issue revo- lves around the language universals that belp native learners of the langua- ge cboose between the bar&mzli and the linmMJ.i versions of the preaspirati- on rule. One possibility - easy to state, bard to prove - is that the lear- ners, ceteris paribus, cboose the least abstract of the available rules. This universal would dictate the cboice of tbe linmidi version of the preaspirati- on rule, not only in the transitional zones, but also in tbe bar&mXJ.i terri- tory proper. In that case tbe wbole of Iceland uses only one version of the preaspiration rule. i· Appendix A: Quantity in modern Icelandic. Tbe pbonetic and pbonolo- gical aspects of Icelandic quantity bave recently been treated in Orešnik & Petursson 1977. There are two quantity systems in modern Icelandic, a Nortbern Icelandic and a Soutbern Icelandic one. Tbe geographical boundary between the two has not yet been determined (Magnus Petursson 1978), but the working hypothesis is that most of the linlmdi territory uses Southern Icelandic quantity, and most of the har&maeli territory utilises Northern I- celandic quantity. Northern Icelandic quantity is described in the handbooks. Its characteristic trait is this: in a segment sequence consisting of a stressed vowel and an intervocalic consonant, either the vowel is long and the consonant short, or vice versa. 158 Southern Icelandic quantity is not mentioned in the handbooks. It distinguis- bes sbort a.nd long vowels, wbile the consonants are sbort, with the excep- tion of _!:, which can be sbort or long. (Tbe consonants are somewhat longer after sbort vowels tba.n after long vowels, but the difference is perceptually negligible, except in the case of the r. 10) An example: noro. sg. {s "ice" and its gen. sg. {ss both contain a il"iort consonant in Southern Icelandic, whereas tbe vowel is long in the nominative and sbort in the genitive: [ Ts, fs]. The hypothesis that there are two quantity systems in modern Icelandic has only been lmown after Magnus Petursson bad publisbed his pbonetic investi- gations of Icelandic quantity. These investigations have been corroborated by Garnes' (1976), altbougb the latter' s research bas been limited to the speecb of subjects from Reykjavik. The remainder of this Appendix will discuss Southern Icelandic quantity only, partly because acquaintance with this system will facilitate the under- standing of tbe main body of tbe present paper, and partly because I can now offer a somewbat improved version of tbe description of Southern Ice- landic quantity system as presented in Orešnik & Petursson 1977. As is well lmown, in Icelandic the opposition between short and long quanti- ty is only possible in syllables that bear (at least some) stress. The dis- cussion is bere limited to stressed vowels in initial syllables of simplex words. It is necessary to distinguisb between lexicalised and non-lexicalised quan- tity along tbe same lines as between lexicalised and non-lexicalised preaspi- ration (see section 3 above). Examples of lexicalised quantity: mjOg "very" (long ..§>, villa "mistake" (sbort .!} • Examples of non-lexicalised quantity: dalur 11valley" vs. dals (long vs. sbort _!!), gefinn "given" vs. gefnum (long vs. sbort !Y. The quantlty rule bas nothing to do with the lexicalised quantity, except that the rule must be formulated so as not to affect sucb quantity. On the other band, the quantity ru.le regulates the non-lexicalised quantity. Tbe ba- sic hypothesis upon wbicb the formulation of the rule is based is that all underlying non-lexicalised stressed vowels are long, and that the task of the quantity rule is to sborten those long vowels in certain contexts. 10 Another language with long and sbort r, and otherwise with sbort consonants only, is Spanisb (Magnus Petursson 1978). 159 An indication that the underlying quantity of stressed vowels is long is the statistical situation in the inflectional paradigms in which long and short vowels alternate on the systematic phonetic level. In the nominal. paradigms the stressed vowels are prevailingly long. In the verbal paradigms, the long and the short vowels are approximately equally divided. So on the who- le, the long stressed vowels predominate. That the underlying quantity of stressed vowels cannot be short, follows from examples such as hygginn "clever" and auki.nn "augmented". If the un- · derlying vowel quantity were short, the underlying representations of these words would be /hlgln/ (lexicalised quantity) and /Otgln/ (non-lexicalised quantity), respectively. The two representations contain no clue that would help the quantity rule determine which of the two stressed vowels is to be lengthened. There are of course many such examples. The underlying quantity cannot be left unmarked, either. This can again be illustrated with the example pair hygginn and aukinn. If the underlying quan- tity of the stressed vowels of these words were ;not marked, the underlying representations would be /hlgln/ and/oigln/, with the stressed /1/ and /oi/ unmarked for quantity. Consequently the representations would contain no clue that would tell the quantity rule which stressed vowel to lengthen and which to shorten. If the underlying quantity of stressed vowels is neither short nor unmarked in the non-lexicalised cases, it must be either long, or long· in some cases and short in others. The latter alternative has never been seriously inves- tigated, although it is theoretically possible. Such an investigation will not be undertaken in the present paper. Orešnik & Petursson 1977 and the pre- sent paper are based on the simpler hypothesis that the underlying quantity of stressed vowels is long. The quantity rule shortens underlying long vowels in four types of environ- ments: (1) before continuant + consonant. Example: hrust of laus "loose". Exception: there is no shortening before / sj/, / sr/, · maybe also / sv /; e. g. lausra of laus. (2) before plosive + /l,n/. Example: gefnir [ -t>n-] of gefinn "given", einn [ ei~ ] •ione". (3) before long .!:· Example: st8rra of stor "big, great". (4) in the following morphologically defined contexts: 160 (a) in the genitive singular of monosyllabic nominal s-final roots if the word contains no genitive case ending (the postpositive - article disre- garded). Example: fs "ice", gen. sg. fss [is ]. (b) in those forms of ~-final ve~bal roots in which /e/ is replaced by /'}/. Examples: bfea "wait", impl. biddu; lofla "cleave", pret. lOddi. (c) in the dative singular feminine and the genitive plural of mono- syllabic vowel-final noun roots whose only ending is the postpositive arti- cle. Example: a "river", definite dat. sg •. futni, gen. pl. fuina. - ~- -~ Apart from the limitation concerning the clusters / sj, sr, ? sv /, mentio- ned sub (1) above, there are no exceptions to the quantity rule as formu- lated bere. Nevertheless the present formulation is not satisfying, for the following reasons, among others: (i) The rule is not formalised, and it seems that any attempt at its formalisation would encounter major obstacles. Consider, e.g., point (4b) above. (ii) Also needed is a phonological rule regulating vowel quantity in non-initial syllables bearing the so-called rhythmical stresses. The formu- lation of such a rule, when it is attempted, will undoubtedly have repercus- sions on the part of the rule regulating the quantity of the vowels in initial syllables. The Southern Icelandic quantity rule can also account for Northern Icelan- dic quantity, whereas the Northern Icelandic quantity rule (for which see Orešnik & Petursson 1977) cannot account for Southern Icelandic quantity (because it presupposes the existence of long consonants in the language, and these,lbarring the !• are absent from Southern Icelandic). Hence it is conceivable that the Southern Icelandic quantity rule is also used in Nort- hern Icelandic and in the transitional zones between linmaeli and haremxli, of course in conjunction with another rule lengthening eonsonants immedia- tely following short vowels. 7. Appendix B: On /dl, dni .1 This Appendix deals with the descrip- tive origih of /<}./ in the clusters /<}.l, 9nJ not preceded by preaspiration. (A less complete treatment of the same subject is Orešnik 1973.) It is necessary to distinguish between lexicalised and non-lexicalised /r)/' s, along the same lines as between lexicalised and non-lexicalised preaspira- tion (see section 3 above). Examples: the /'}./ of villa [vI9J.a J "mistake" is lexicalised, because it occurs in all the forms of the word. The /d/ of . einn [ ei~) "one" is not lexicalised, because it occurs only in some forms of the word ( e. g. not in its dat. sg. m. einum). 161 There is some doubt concerning the treatment of the cases such as fall "fall". The word contains [ d lin ali its forms, except in the gen. sg. falls [fa.Is ]. I will here assume, ~ithout argumentation, that the /!}/ of fall is-- lexicalised and deleted in the gen. falls by a phonological rule which dele- tes / i}/ in the environment -- V CD where C stands for any consonant, and D for dental plosi- ves or /s/ (This rule is responsible, e.g., for the nom./acc. sg. n. allt "all" wit- hout [ ~ ].) 11 (Another special case are those non-lexicalised /O/' s standing between / r/ and /1, n/ (where / r/ sometimes disappears), ~.g. ferill "track, trace", nom. pl. ferlar [ -rraefi i/W is not pronounced in words such as beinn, ~ (Bjorn Gu&finnsson 1946:129). · 163 This exposition has shown that the non-lexicalised /iJ/ is not likely to be due to the operation of a phonological rule: there is no suitable source of / (}./, and the vowels that can immediately precede / ()/ - or the complement of these vowels - do not form a natura! class. Moreover, the language a- bounds in minimal pairs such as noro. sg. aamfrall - acc. sg. aamfr3.l, which seem to postulate that the rule generating the non-lexicalised /9/ also mention morphological environments. Valfells 1967 and Anderson 1969 have suggested - in passing, without argu- mentation, and without discussing the complete material - desinence-ini- tial /r/ as the source of the non-lexicalised /g/. Thus the gen. pl. einna of einn "one" would be derived from / ein+ra/,' and be parallel to the gen. pl. van-ra of vanur "accustomed". This abstract solution cannot be correct. Parallel to all the adjectival forms containing non-lexicalised /iJ/ except the nominative singular forms there are colloquial forms containing both the /C)/ and an /r/-initial case marker: gen. pl. einna and coll. dat. sg. f. einni gen. sg.f. einnar einnra [ -<}nr-] einnri [ -