Marc L. Greenberg University of Kansas Multiple Causation in the Spread and Reversal of a Sound Change: Rhotacism in South Slavic Prispevek obravnava spremembo i > r (.rotacizem-) v juinoslovanskih jezikih. Zaradi zapletene razvrstitve spremembe tako po jezikovni geografiji kot po besednih vrstah se zdi, da se je ne da razloiiti z enim samim vzrokom. Ugotovlja se namret, da je sprememba nastala zaradi vrste gla- soslovnih in analognih vzrokov; umikanje inovacije proti zahodu pa so otitno povzrotili sociolin- gvistitni dejavniki. The paper treats the change i > r ("rhotacism") in the South Slavic languages. Its complex distri- bution both in its linguo-geographical dimension and parts of speech suggests that its develop- ment cannot be explained in terms of a single cause. It is found that a combination of phonolog- ical and analogical factors must account for the initial change and its spread; sociolinguistic fact- ors apparently account for its retreat towards the west. 0 Introduction 0.1 In Ivid's map phonological isoglosses in the W-SS1 dial area (= the dials corresponding to the Sn, Cr, Bs and Srb standard languages), the isogloss of the form moie > more 'can' (3rd pers sg) stands as a representative of the change i > r (IviC 1958: 31). The isogloss, which roughly bisects the area mentioned, seems clear cut, though, in fact, the change is much better represented in the W part of this territory than the E part. Nor is it clear whether the change is a phonetic or a morphological one. The complications with regard to this sound change are many and there is no consensus on the details of its development. To illustrate, the change occurs regularly in the form mentioned, but does not occur as might be expected in, for example, oieniti se 'to wed'. It occurs also in forms in which the phonetic motivation is either absent or isolated, for example, morati : mora 'must' (inf : 3rd pers sg). From the point of view of linguistic geography, the forms in which the change has occurred become sparser to the east as one approaches the moie > more isogloss, yet forms such as morati extend beyond the isogloss. Within a given dial there may be variation, for example, both moie and more are attested in the dials of NW-Bs (VujiCiC et al. 1979: 146) and in E-Slavonian localities that are transitional to the ~umadija-Vojvodina dialect (Ivid 1990: 87-89, 93). The change and the pecul- iarities of its distribution have been explained by some scholars as a result of phonet- ic change and by others through morphological analogy. In this paper it will be shown that neither of this forces can be discounted in accounting for the results of this change - not only are both explanations essentially correct (though with some amendment) - but they both must have worked together to give rise to the particul- ar pattern of change in part of the areal (Sn, Cr). It will also be suggested that the reversal of this change in part of the areal (Srb, Mac, Bg) can be accounted for by 64 Slovenski jezik - Slovene Linguistic Studies 2 (1999) considering sociolinguistic factors. In the continuation, the results of the change will be referred to as "r-forms," in order to avoid prejudging the motivations for the change! 0.2 The distribution of the change is restricted to a small number of lexical items and categories, as sketched here:' 0.2.1 The present tense of the verb moti 'to be able': morem, mores', more, as well as prefixed forms pomot/bi : pomore 'to help', premot/bi : premore 'to overcome'; !he form morda 'perhaps' (< *moie da). The area of the this change includes Sn, Ca, Kaj as well as Sto Ikavian, most of Ijekavian and W-Ekavian, including the environs of Novi Pazar. 0.2.1 In Sn and CrIBslSrb in the verb morati : moram, moras', mora 'to be obligated, must', apparently a derivative of mot/&. 0.2.3 Complementizers and adverbs built with the enclitic particle -ie, e.g., Sn kakor (< *kako-i(e)) 'as', kdor (< *kado-i(e)) 'who, whoever'; vendar (< *vem-da-i(e)) 'however'; it appears in Sto in the form jer 'because7. 0.2.4 Time expressions in some Sn dialects, e.g., nocar 'tonight' (Istria), nicor 'idem' (Styria). 0.2.5 The adv. bore 'pitifully, poorly', presumably built from the vocative of bog 'God', cf. boremi 'my God!'; the adj. derived from this, boren 'poor'. This is found in Sn and Cr dialects. 0.2.6 The present tense of the verb gnati 'to drive (cattle).' The present tense form renem is found in Inner Carniola, as well as a new infinitive built on this stem rdnit (attested in Inner Carniola, Kras and Upper Carniola); the prefixed forms odrene, prerene, prorene, proreneio are attested in the late 16th c. from Upper Carniola. A blend of gnati and renem is found in the forms yrdniti : yrdnem (Kras) (Ramovg 1924: 295, Bezlaj 1977: 175). NW-Bs dialects have izr2nuti, n2 more (VujiEiC et al. 1979: 52, 97). Skok finds that the Romanian form "porni nije od porinqti nego je pognati = pornati (14. v.), kako se vidi iz porni oile 'tjerati ovce', Mihai porni cete spre a lua in goani pe dugman 'Mihalj je poslao Eete da u potjeri zauzmu neprija- telja'. Odatle porneal2 (Skok I: 575). 0.3.1 The majority opinion is that the source of the change is phonetic. Ramovg explains the change in articulatory terms: "prehod i > r (znan tudi shrv. in bolg.) doiivi le intervokalitni i, Ee mu sledi vokal e , ki je tvorjen s pridvignjenim delom jezikove ploskve proti sprednjemu trdemu nebu ..." (1924: 294; Vondrik 1924: 459-460). Following this formulation, Bezlaj assumes that the form rinem (Inner Carniola) arose under the conditions stated: "do rotacizma je priglo v sintagmah tipa krave iene" (1977: 175). Similarly, Cvetko Oregnik finds that "der Rhotacizmus ge- rade bei tere begann", the form attested in the FF and simultaneously the earliest attestation of the phenomenon (1987-88: 439). It is generally assumed that the change resulted in the alternation i lr, and that this r spread in word formation (see ' I am grateful to Andrew Corin (UCLA), Bill Darden (University of Chicago), Marko Snoj, Metka Furlan (InStitut za slovenski jezik, ZRC SAZU), and an anonymous reviewer for helpful comments on earlier drafts of this paper. A thorough review of the evidence is found in Cvetko OreSnik 1987-8. I therefore limit my statement of the facts to the minimum necessary for the purposes of the present paper. M. L. Greenberg. Rhotacism in South Slavic ... 65 Cvetko Oregnik 1987-88: passim). A further condition, that the segment appear in an unstressed syllable, has also been claimed (Cvetko Oregnik 1987-88: 434). 0.3.2 A fundamentally different scenario is presented by IvanEev, who claims that rhotacism was not a phonological change at all, but was caused by analogical spread from the relative adjective kateri. In the summary of his article he says: "ImajuCi u vidu da je domen pojave u leksitkom pogledu krajnje ograniten, moiemo smatrati da su reEi s r iz i u slovenatkom i srpskohrvatskom jeziku sluEajevi leksi- kalizacije pri kojoj zamena i sa r ne bi mogla da se obrazloii delovanjem nekog fo- netskog procesa kako se to obitno tumati. Pretpostavlja se da je u uslovima seman- titko-funkcionalne jednakosti leksiEke grade, -ieli u veCini odnosnih reEi, po utica- jem analogije zamenjeno sa -relr. To je prvenstveno bilo u reEima kateri, eventualno kater, -a, -0" (1981). This view assumes the metanalysis of kater- into pronominal (kate-) and relativizing (-r-) morphemes. The new relativizing suffix then spread and overtook relativizing constructions in -ie (e.g., iie). While such a development is a possible explanation for forms of the conjunctions and complementizers of the type kar, kdor, Ivantev's theory fails to provide a plausible mechanism for virtually all of the remaining types. Nor does it deal with the difficulties of motivating the spread even in the category he adduces. This theory is stretched the furthest in at- tempting to explain forms such as bori, boren, more, ien- as due to the same mech- anism (ibid.: 25); in these cases there is no model for analogy. The author concedes: "makar t e negtata tuk ne sa osobeno jasni" (loc. cit.). It is obvious that IvanEev's theory cannot be accepted as stated. However, with some qualifications, we shall demonstrate that there is a place for part of this analogical explanation in the fuller picture of the rhotacism phenomenon. 1 Phonological factors 1.1 As we have seen above, the question of whether the r-forms can be consid- ered a result phonological or morphological change in SS1 is fraught with ambiguity. Since the descriptive facts do not point to a definitive answer, it is reasonable to ask whether the possibility of phonetic change is systemically motivated. Andersen, not- ing the parallel between the change r' > i in WS1 and the W-SS1 i > r change, views each of these as natural changes in consonantal and vocalic languages, re- spectively. With respect to the i > r change, "since both stridency and voicing shift the acoustic structure of a consonant towards that of a vowel, it is easy to under- stand how learners of a language might evaluate voiced, strident, continuous ob- struents as realizations of liquids" (1978: 6). Thus the parallelism arises when the marked value of the feature + vocalic is replaced by the unmarked value, as in Table 1. 66 Slovenski iezik - Slovene Linguistic Studies 2 (1999) Table 1 l i l Ir 1 SS1 -VOC (M) > +VOC (U) +cons (U) +cons (M) lr'l If1 WS1 +cons (U) > +cons (U) +VOC (M) -VOC (U) This simplification is a specific example of the general tendency by which, accord- ing to IsaCenko, "[v]okalische Sprachen weisen die Tendenz auf, Konsonanten zu vokalisieren" and "die konsonantischen Sprachen entwickeln nicht nur keine sonan- tischen Konsonanten, sondern sie unterdrucken sogar die naturliche Sonoritat der Konsonanten" (1939-40: 72, 73). Thus the change parallels other changes that have increased the number of vocalic constrasts in Sn and WBsISrb (connection with in- tensive contact with Romance dialects) in contradistinction to the reduction of such contrasts in N-S1. This explains why the change is not found in WS1, although it probably occurred at a time when the connection between WS1 and SS1 had not yet been completely interrupted. For these reasons it seems plausible to assume that the change has a phonetic component. 1.2 To the extent that the change is phonetic, the chronology of -Vie- > -Vre- must be later than the first palatalization of velars (6th c) and perhaps earlier than the merger of the result of this change with the reflex of *z' from deiotation (8th c), since the latter remain unaffected. The change must have also preceded the writ- ing of the FF, where the change is first attested. The FF attest to a stage in which fluctuation between i - and r-forms could still occur, e.g., tere 'also' ( 2 ~ ) , ise 'which', rnosern 'I can' (2x), tornuge 'to this very' tige (= [tiie]) 'these same'. 1.2.1 It is not clear whether stress played a role in the change, as we find Sn dia- lect vre" 'already' (cf. StSn ik), which contradicts the notion that stresslessness in the syllable is a precondition of the change. It may be, however, that in this instance the r-form was an extension from other temporal expressions (see 2.3). 2 Morphological factors 2.1 Verbs 2.1.1 In the case of gnati : rene, the source of the change is ambiguous, since there is the possibility of contamination with the verb *r in~t i 'to push'. However, un- less the r-forms of -iene(ta) were already in place, there would not have been a formal match to pave the way for analogy. Alternatively, one might argue that the ilr alternation existed elsewhere in the lexicon and so the possibility of its extension to any morpheme containing i was opened. However, since the number of lexical items in which the alternation participated is so restricted, this scenario seems un- likely. Therefore it seems more likely that the identification with *r in~t i arose after the phonetic change had taken place in compounds of the type dorene. This led to an alternation between the simplex iene and the compound dorene. From this alter- nation, the distribution could have given rise to three logical outcomes: (1) the alt- M. L. Greenberg. Rhotacism in South Slavic ... 67 ernation could have remained; (2) iene could have been generalized, effectively re- versing the change; or (3) (do)rene could have been generalized. In the western areas (Kras, Inner Carniola, Ca, parts of Sto), outcome (3) occurred, where the sim- plex morpheme was replaced by analogy to rene. This regularization may be rein- forced by identification with *rinqti. Elsewhere outcome (2) occurred. 2.1.2 The verb moielmore lacks an alternation between i and r in the present tense and therefore the possibility of the change being reversed due to leveling can be ruled out. In contrast to -gnuti:-iene, the environment in question is non-alter- nating and this may in part account for the fact that the change is found over a much wider area for more than -rene. 2.1.3 If we contrast the outcome in each of these verbs we find that the greater areal of more in contrast to -rene is in accord with the principle that "change de- velops earlier in uniform environments because they represent the context for change more consistently than alternating evironments" (Timberlake 1978: 326; see also 1981). See Table 2. Table 2 uniform alternating morem, -eS, . . . ienem, -eS, ... I dorenem, -eS, . . . -r(e) ieniti se; *oreniti se 2.1.4 The verb morati occurs over an even greater area, including those areas of Sto that have moie. The reason for this appears to be that the verb was derived di- rectly from more and then imported from the W dialects E-wards. The formal mod- el for the derivation is unclear, though its semantic motivation is apparently due to language contact, cf. OHG muozan 'canlmust' > MoG miissen 'must' (see further Music5 1923; Ramovg 1924: 294; Grubor 1925-26; Snoj 1997: 354). This active con- struction replaced nominal constructions of the type R ja dolien 'I must' in a para- llel fashion to Cz musim, a direct borrowing and adaptation of G miissen. Lexical innovation seems to be an additional factor in the retention of the change (see 3.4.2, 4.1). 2.2 Pronouns and conjunctions 2.2.1 The possibility of analogy proceeding from the relative pronoun, proposed by IvanCev, cannot be ruled out. However, the analogy is hardly as straightforward as IvanCev suggests and a number of things need to be explained before such a pro- posal can be considered. The first major difficulty is that the forms as they appear today - Sn kdolkdd, kateri; Cr tko, khjt; Srb k6, khjt - do not match and as such cannot have given rise to analogy. This raises the question of the shape of the forms in question at a time early enough to have preceded the change i > r. The comparative evidence suggests fairly unambiguously that the CS1 input must be *k%to (< IE *kgo-s + tod): R ktb, Ukr, Br xto, Po chto, Cz kdo (OCz kto), Srb ko, 68 Slovenski jezik - Slovene Linguistic Studies 2 (1999) Bg, Mac ko, US Stb, LS chto. Although it is tempting to connect Cz kdo, Sn kdo and assume these to be a common innovation, the Sn dialect evidence suggests that the -d- arose in independently in both territories, e.g., Roi htu? Zilja tua, Prek- murje Stij." In C-Sn the -d- from the temporal and spatial interrogatives (*k(a)d- kdaj, kdelkje) is thought to have spread to *k(a)to (Bezlaj 1977: 27).5 This innova- tion cannot have preceded the change of i > r as we see in the form n i h 2 'no one' (< *ni-k5t.6-ie, cf. Po nikt 'idem'), a form common to both Sn and Kaj6 2.2.2 With regard to the relative adjective, the situation is complicated at the CS1 level as well as in Sn and Kaj (the form apparently disappeared in Medieval Srb). The S1 languages point to a number of CS1 variants: *koter% (OCS koteryi, Sk kotery], *kotor% (R kotbryj); *k%ter% (Cz ktery], *k%tor% (PO ktbry); *kater% (StSn kateri). Of these, only the first can be connected directly with the IE material, Gr pbteros, Li katrds, katards (Vilnius), Skt katardh. The StSn form is the most nar- rowly localized of these variants and may have its origins either in akanje or the influence of the pronoun kdk 'what kind', cf. kateri (Trubar, 16th c), kateri, ketiri (Dalmatin, Megiser 16th c), katir (Pohlin, 18th c), kateri (Gutsmann, 18th c) (Bezlaj 1977: 23). The form *koter% survived as a relic in the compound neikoteri 'certain ones' alongside nekateri in the language of Trubar, though as a pronominal adject- ive the form kateri is used exclusively (RamovS 1920: 274-275, 1952: 97; Rigler 1968: 185; Hamp 1980: 98) and is known also in Ca and Kaj koteri (Belostenec, HabdeliC, Voltiggi, Stulli, BudiniC, 16-18th cc) (Skok: 1977: 113). The forms UC ker, kerga (nom, gen sg), Pkm Stzri, Kaj Stari (Bednja) indicate a starting point in *katkra, which may have arisen under the influence of *k5t0.~ Of the three forms that are common to Sn and CrIBslSrb, the one that gained widespread productivity by the 9th c appears to be *k%ter%. If IvanCev's proposal is to be accepted, then the start- ing point for the spread of -r from the relative adjective originates in the analogy that must have developed from the juxtaposition of (the now archaic forms) *k%to/% Rigler cites the form with falling intonation (htu") for Breznica pri St. Jakobu v Roiu (Ri- gler 1981: 198). The innovation is known also in Kaj, e.g., Ozalj d8/gdB, in place of the usual form Sto. The form fto in US, Kaj and Pkm is n_ot the continuation of CS1 *&to, but represents the change ht > St, cf. Pkm Steti 'to want' < *xattti. Merger was avoided in US by means of a quantity contrast, viz. Sto 'what' : St6 'who'. In Kaj and Pkm it was avoided by the replacement of *&to with a new pronoun built from the complementizer kii < *ka (< IE *k*ehz, cf. Lat quii) and the pronominal formant *-jb (see Snoj 1997a). This is a question of relative chronology entailing the removal of one of the sources of ana- logy in the later history of a change. For a typological parallel, note the Sn and Cr dial change of *bez > brez under the influence of Erez. In the C-Sn dials the form Erez has become Eez by a regular phonological rule simplifying clusters of the type *Er, * i r > E, i, thus giving Eez, brez in StSn. The reconstruction of *ni-ksts in Sn, Kaj the proto-form is questionable, though can be based on the received forms i n Sn and Po. It may be that the root developed a variant parallel in shape to *Eb, as evidenced by Cak EL?, zaE 'what' (< *h, *za Eb), StSn nit 'nothing' (< *niEb), FF nizce. Alternatively, it could be assumed that the o was subject to an exceptional early reduction, i.e., from a form *ni-ksto-ie (Snoj 1997: 382). The Srb and Cr standard form kGi, -ii, -a are forms rebuilt on the oblique form koj- (OCS ky, kojego), a 13th-c development (Skok 1972: 112) which played no role in the formation in question. M. L. Greenberg. Rhotacism in South Slavic ... 69 : *k5ter5. Moreover, this can only be understood in a narrower sense than IvanEev intended, as this marker could have spread only with the meaning imparted by the function of -r as a relativizing particle. 2.3 Time expressions 2.3.1 One may be tempted to assign to the Sn dialect form noc& 'tonight' (De- kani pri Kopru nicdar, Skofije nacdar, Vodnik (Sty, 19th c) nizor (Sivic-~ular 1982: 415)), the same type of formation that arose in other temporal adverbs, such as Sn danas, CrIBslSrb ddnas 'today' (< *dbnksb 5 *dbnb + sb), Sn nocoj,8 in which the elements *-sb, *-jb added a temporal deictic function to the base form. On this view the r-element is regarded as an enrichment to the list of deictic particles used to build temporal adverbs. Such a proposal would not be without merit in light of the semantics of i e in N-S1, where the meaning can be deictic (cf. R on i e 'the very same one (masc.)'. However, this meaning is not characteristic in SS1. Instead, this -r may be an extension of the model built from *vZtera, in which the r-element is original (Gr hdsperos, Lat vesper, Lit vfikaras). This word developed two different meanings in SS1 dialects, one a noun meaning 'evening' and the other an adverb meaning 'in the evening', which after the fall of the jers became differentiated by accentuation: *vZter5 vs. *v% vetem, preserved in Pkm Sn vet& 'evening' vs. v&r 'in the evening' (StSn vec'gr with both meanings). That this -r- was interpreted both as a stem ending and as a deictic marker is evidenced by its removal in Sto ~ Z t e , as well as jute 'yesterday' (< *vbtera) (Skok 1973: 570) (see below on factors in the re- versal of the change). The temporal deictic meaning of -r- may have been enriched further by identification with both older *utro 'morning' (StSn jlitro, Pkm vutro, CrIBslSrb jutro) and newer Sn osorgj 'at this time, at this hour' (< *ob seji urZ < Lat hara, Sn dialect osorg, osovrg, osevrgj (Bezlaj 1982: 258)). Thus it is as likely that the -r in nocar is a result of the spread of -r marking temporal deixis from words like *v3 vetem as from -r(e) < *ie. 2.4 Other 2.4.1 The preservation of the form bore, and its derivatives, may have retained its shape due to taboo. Skok observed that "[klako je krgianstvo tabuiralo rijeE bog prema iidovskom uzoru da se Jahveh ne upotrebljava nego jehova, Eesto se ne upo- trebljava bog u potpunoj fonetici nego se g zamjenjuje sa r. Mjesto bogme govori se borme, tako ti bora mjesto boga; oj ti bora ti; za bora milogd' (Skok I: 191). This notion is confirmed by the Pkm usage, in which borme 'my God!' is a more polite form of bogme (Novak 1996: 22). 3 Reversal of the change 3.1 An intriguing factor in the development of r-forms is that its areal in the Medieval period was much more widespread than it is today. In StCrIStBslStSrb the The mysterious -c- in this form is a matter of some debate: k c - ~ u l a r finds this to be a relic of WSl infiltration (1982); Furlan assumes the form results from the simplification of a di- minutive *notJbcii from *natJb (1993: 220-2). She points out that an older explanation, deriving the base from an already deictic *not'bsb is untenable in light of Istrian najco, in which -jc- could not have arisen from *-t'bs- (ibid.). 70 Slovenski jezik - Slovene Linguistic Studies 2 (1999) relics of the change are few, which essentially reflects the fact that Ca and Kaj lit- erary traditions were abandoned in favor of Sto. The relics include the conjunction jer, which is connected with the older formation from complementizers built from *j- (related to OCS iie 'that, which' masc sg, jeie neut sg) that became replaced by those built on *&to and *kojb, a process that was underway in the 14th-15th cc (Grickat 1975: 287); the verb morati; and the word takoder 'also' (alongside the alt- ernant takode). The evidence in Bg and Mac is sparse, but the phenomenon is known there, too, e.g., Bg dori 'even' (< *do-ie), Mac duri 'up to, until', Bg (dia- lect) dordkldordkto 'until' (< *do-ie-d2), Bg tere 'and', Bg barlbarelbarkmlbarim 'at least' (< *ba-ie) (examples from MikloSiE, Vondrik, Lavrov, cited in Cvetko Oregnik 1987-1988: 435-436). As mentioned above, it was borrowed (from Bg?) into Romanian from the present tense of pogmati, attested in the 14th c (Skok 1971: 575). One can agree with Cvetko OreSnik and her predecessors that "man muss bei diesen Beispiel- en besonders vorsichtig sein" (loc. cit.), as they may well have been the result of in- filtration from Serbian usage. Nevertheless, it is clear that at a certain point in the high Medieval period, r-forms were on the wane in the E. Not only did they cease to be productive, but they also seem to have become systematically removed from the lexicon. 3.1.1 It is not surprising that the polarization of productivity is found in the con- junctions. As IviC has observed, "od gramatitkih reEi najvige razlika izmedu pra- slovenskog stanja i danagnjeg srpskohrvatskog pokazuju veznici. Objagnjenje moramo traiiti u Cinjenici da su od svih tih reEi veznici, naroEito oni u podredenim reEeni- cama, najzavisniji od stepena razvijenosti apstraktnog migljenja, Sto znaEi i od opgteg kulturnog nivoa sredine" (IviC 1991: 316). One does not have to look far to realize that in Slavic writings of the 10th-14th cc there is a striking variety in the forms that replace older subordinating structures (such as the dative absolute or "dative of subordination," on which see Andersen 1970). But why do the r-forms prevail in Sn and Cr and recede in Srb, Mac and Bg? As a working hypothesis, let us assume that in the formation of new complementizers, r-forms gained a high prestige value in a W milieu and a low prestige value in the E milieu, i.e., corresponding to S-Slavs under the influence of Rome vs. Constantinople? If this hypothesis proves correct, it helps explain why the development is more pervasive in Sn and Cr and undergoes at- trition in Srb (and, to the extent that they are relevant to the picture, Mac and Bg). 3.2 It is unfortunately very difficult to trace the details of the development in Sn, because there is a gap in the textual evidence in precisely the time frame under investigation. However, for the beginning and end points of the development, the picture is fairly clear, so that it is possible to surmise with a reasonable degree of assurance what occured in the dark period. In the 10th c the FF attest to the be- ginnings of a competition between r-forms and i - forms. The conjunction tere 'and, also' occurs twice; but other forms show unchanged i , e.g., ise 'which', rnosern 'I can' (occurs twice), tomuge 'to this very' tige, choiseih 'which' gen pl (< *ky-ie + ixa)." The number of examples here is too small to conclude anything except that r-forms The situation is reminiscent of Labov's famous study of r(1ess)-forms in New York City (that the segment in question is r is, of course, entirely fortuitous). Rather than class distinctions, however, the values would have been assigned to religious-ethnic communities (see Labov 1972). lo Forms from the Freising Folia are quoted from Bernik et al. 1993. M. L. Greenberg. Rhotacism in South Slavic ... 71 were possible variants at this stage. By the early 14th c., as evidenced by the Celo- vec1RateEe Ms.," the r-forms in subordinating conjunctions have been generalized: Otlcha rials kyr sy wnebessich . . . 'Our father who art in heaven . . .', odpulti nam dalge naue kakor yno my . . . 'forgive us our trespasses as we . . .'. Cf. also the 15th c StiEna Ms. zkemer 'with which' (2x), kakur 'as' (8x), kadar 'when' (3x), kygar 'whom' acclgen sg. 3.3 In contemporaneous Cr documents, r-forms are general on an equal footing with those found in Sn, albeit combined with partially different material. For exam- ple, in the Cr Glagolitic Paris Codex (Split, 14th c) we find more (4x), moreSb, premoreib 'you overcome', arelar' (5x) 'for, because', jureljur' 'already', tereltrb 'and', nii'tare 'nothing', nikogare 'nobody' gen sg, nigdarb 'never', nig'dtrb 'nowhere', nigdorb 'nobody' (Mali6 1972: 131). Examples of 2-forms in Cyrillic Cr works are numerous, though these are found in conjunction with other features of the OCS Cyrillic tradition, e.g., Povaljska listina (BraC, 1250) Tako jesm sliSal da sije zem- l(j)e jeie vy driite . . . 'Thus I have heard that these lands that you hold . . .' (MoguS 1993: 25; Mali6 1988: 38ff). In locally-colored texts r-forms abound, e.g., 3. MenCeti6 (Ca-Sto koin2, 15th c): ozriv se jak jelin ter ontas pode tja, ter gorti ner pelin i temer ostah ja (VonCina 1988: 82). 3.4.1 As was pointed out above in connection with the morphological motivation for r-forms in the relative adjective *katera, in $to there was a marked tendency to generalize other forms, in particular koi (< *koj- ,< *kajb 'which') (KopeEnjr, Saur and Polik 1980: 389). Subordinating pronominal forms built from *jb- (iie, etc.) be- came replaced by the general complementizer *&to (Sto), which began in the 14-15th cc to compete with koi (Grickat 1975: 286-91). Since *katera failed to be selected as a relative marker, it had no chance of becoming a model for analogical developments (in contrast to the developments in Sn, Ca, Kaj, discussed above). However, this did not eliminate the subordinating conjunction jer(e) (< *jeie), which is a relic of the older layer of *jb-forms (Grickat 1975: 72). The survival of this form is apparently connected with the widening of its semantics: OCS jeie functioned narrowly as a complementizer indicating causative subordination, e.g., tjuidaaxq se jeie kaSnjaaSe 'they were surprised that (s)he was late', whereas MoSrb jer ranges in function from strictly causatives meanings to (essentially para- tactic) general, summarizing meanings, e.g., tako je bolje, jer Sta bi ti radio ako bi slutajno saznao za tu nesrec'u! 'it's better this way for what would you do if you had found out about this misfortune!' (Grickat 1975: 126, 134-5 and passim). The new functions are far removed from the original meaning (anaphora), indicating that jer was felt to be operating natively and independently in the system. 3.4.2 In Srb Sto the process seems to have been reversed, leading to hypercor- rection~ of the type v2ter > vzte, juter > jute (Skok 111: 570), in which the -r element was identified with -r < -i(e). This gives some support to the notion that -r- became identified with a style that came to be felt as foreign to $to speakers. Note, however, that this applies only to productive types. This gives a plausible explanation for the reduction of the number of lexical items in which r-forms occurs in Stokav- ian: they persist most consistently in those lexical items in which the r-element is l1 Forms from the CeloveclRateEe and StiEna Mss. are cited from Mikhailov 1997: 101-102. 72 Slovenski iezik - Slovene Linguistic Studies 2 (1999) non-alternating and in which it has become lexicalized (morati, jer) and lost else- where. 4 Conclusion 4.1 To sum up, the phenomenon of rhotacism had a number of intersecting fact- ors determining its spread and reversal. The change occurred, first, only under nar- rowly defined phonemic conditions, thus limiting it to a small part of the lexicon. It persisted best in uniform environments and tended towards reversal in alternating environments. Its chances of persistence were increased if a semantically similar form with -r was available (*k@)ter- 2 kdo, kdor; rine 2 rene; veter 2 nocor). Forms that were high style (H) (conjunctions) were prestigious in the W and thus remained productive; the markedness was reversed in the E, and these cor- respondingly L forms tended to be removed, except for those that became integrated into the system in such a way that they were no longer identified as part of the same semantic field or level of prestige (jer, morati). The factors are summarized in Table 3. Table 3 Factor Favors retention? Alternating Uniform H style Analogical possibility Semantic innovation 4.2 With respect to chronology, the following rough scenario may be proposed: (1) 7-8th cc. - phonetic innovation i > r , (2) ca. 8-11th cc. enrichment of categories containing r-forms and E-ward expansion of lexical items containing r-forms; (3) 10-14th cc. attrition of productive r-forms spreading (E + W). 4.3 Further study, in the form of a survey of the appearance of r-forms in Medieval S-S1 texts, will be necessary to test the scenario put forth in this paper. One hopes that a more precise account of the paths of lexical enrichment and relat- ive chronology would emerge from such a study. This should, in turn, shed light on the social value of the variants, which will help to illuminate another aspect of the society of Medieval S-Europe. Abbreviations Bg = Bulgarian, Bs = Bosnian, C = central, Ca = Cakavian, Cr = Croatian, E = east(ern), FF = Freising Folia, G = German, Kaj = Kajkavian, LC = Lower Carni- olan dialect, Li = Lithuanian, LS = Lower Sorbian, Mac = Macedonian, Mo = Modern, N = north(ern), OCS = Old Church Slavic, OHG = Old High German, M. L. Greenberg. Rhotacism in South Slavic ... 73 Pkm = Prekmurje dialect of Slovene, Po = Polish, R = Russian, S = south(ern), S1 = Slavic, Sn = Slovene, Srb = Serbian, SS1 = South Slavic, St = Standard, Sto = Sto- kavian, Sty = Styrian dialect, UC = Upper Carniolan dialect, US = Upper Sorbian, W = west(ern), WS1 = West Slavic References AleksiC, R. 1937. Prilozi istoriji kajkavskog dijalekta. Juinoslovenski filolog 16: 1-98. Andersen, H. 1970. The Dative of Subordination in Baltic and Slavic. Baltic Lin- guistics (ed. by T. F. Magner and W. R. Schmalstieg): 1-9. University Park and London: The Pennsylvania State University Press. - -. 1978. Vocalic and Consonantal Languages. Studia Linguistica Alexandro Vasilii Filio Issatschenko a Collegis Amicisque Oblata: 1-12. Lisse: The Peter de Rid- der Press. Bajec, A. 1954. Prislovni paberki. Slavistitna revija 5-6: 195-226. Bernik, F., et al. 1993. Briiinski spomeniki. Znanstvenokrititna izdaja (= SAZU, De- la 39). Ljubljana: SAZU. Bezlaj, F. 1977 (I: A-J), 1982 (11: K-0), 1995 (111: P-S). EtirnoloSki slovar sloven- skega jezika. Ljubljana: Mladinska knjiga. Bidwell, Ch. E. 1961. The Chronology of Certain Sound Changes in Common Slavic as Evidenced by Loans from Vulgar Latin. Word 17: 105-127. Cvetko OreSnik, V. 1987-88. Zum Rhotazismus im Siidslawischen. Klagenfurter Bei- trage zur Sprachwissenschaft 13-14: 433-446. Furlan, M. 1993. 0 nekaterih slovenskih dvojnicah tipa rakitje : rokitje. Slavistitna revija 41: 219-229. Grickat, Irena. 1975. Studije iz istorije srpskohrvatskog jezika. Belgrade: Narodna bi- blioteka SR Srbije. Grubor, D. 1925-26. Morati. Juinoslovenski filolog 5: 150-161. Hamp, E. P. 1980. Slovenski koteri, katdri, in "saus-, Brii. slov. v uzrnazi in smug-. Slavistitna revija 2811: 97-101. IleSit, F. 1925-26. nMoCi<< i nmorati<< u slovenatkom jeziku. Juinoslovenski filolog 5: 162-170. Isatenko, A. 1939-40. Versuch einer Typologie der slavischen Sprachen. Linguistica Slovaca 112: 64-76. Ivantev, S. 1981. Za rotacizma v slovenski i sarboxarvatski ezik. Zbornik za filologi- ju i lingvistiku Matice srpske 2411: 23-25. IviC, P. 1958. Die serbokroatischen Dialekte: Zhre Struktur und Entwicklung, 1: All- gerneines und die Stokavische Dialektgruppe. The Hague: Mouton. - -. 1990. 0 starim izoglosama na tlu slavonskog dijalekta. Croatica-Slavica-Indo- Europza (ed. by G. Holzer) (= Wiener slavistisches Jahrbuch, Erganzungsband VIII): 83-93. 74 Slovenski jezik - Slovene Linguistic Studies 2 (1999) - - . 1991. Iz istorije srpskohrvatskog jezika. NiS: Prosveta. KopeEnf, F., V. Saur and V. Polhk. 1980. Etymologicky' slovnik slovansky'ch jazykj. Slova gramaticka a zajmena, sv. 2: spojky, tastice, zajmena a zajmenna ad- verbia. Prague: CSAV. Labov. W. 1972. Sociolinguistic Patterns (= Conduct and Communication, 4). Phila- delphia: University of Pennsylvania Press. MaliC, D. 1972. Jezik najstarije hrvatske pjesmarice (= Znanstvena biblioteka Hrvatskog filoloikog druitva 1). Zagreb: HFD. - - . 1988. Povaljska listina kao jezitni spomenik (= Znanstvena biblioteka Hrvat- skog filoloikog druitva 17). Zagreb: HFD. Mikhailov, N. 1997. I Monumenti Linguistici Sloveni dell' "Epoca dei Manoscritti" (= Studi Slavi, Dipartimento di Linguistics Universitd degli Studi di Pisa, No. 6). Pisa: ECIG. MusiC, A. 1923. MoEi i morati u slovenskom jeziku. Rad JAZU 227: 1-58. Novak, F., and V. Novak. 1996. Slovar beltinskega prekmurskega govora. Murska Sobota: Pomurska zaloiba. RamovS, F. 1920. Opazke k slovanskim tvorbam pronominalnega debla *qwo-. caso- pis za jezik, knjiievnost in zgodovino 2: 274-276. - -. 1924. Historitnu gramatika slovenskega jezika. Ljubljana: Znanstveno drugtvo. - -. 1936. Kratka zgodovina slovenskega jezika. Ljubljana: Akademska zaloiba. - -. 1952. Morfologija slovenskega jezika. Ljubljana: Univerzitetna gtudijska komisi- ja. Rigler, J. 1968. Zatetki slovenskega knjiinega jezika (= SAZU Razred za filoloike in literarne vede, Dela 22, Inititut za slovenski jezik 10). Ljubljana: SAZU. - -. 1981. Breznica pri St. Jakobu v Roiu (Friessnitz; OLA 147). Fonoloiki opisi sr- pskohrvatskih/hrvatskosrpskih, slovenatkih i makedonskih govora obuhvadenih Opiteslovenskim lingvistitkim atlasom (= Akademija nauka i umjetnosti Bosne i Hercegovine, Posebna izdanja 55, Odjeljenje druitvenih nauka 9) (ed. by P. IviC, et al.): 193-200. Sarajevo: ANUBiH. Shevelov, G. Y. 1965. A Prehistory of Slavic. The Historical Phonology of Common Slavic. New York: Columbia University Press. Sivic-~ular, A. 1982. Sln. nocbj in njegove vzporednice v slovanskih jezikih. Slavi- stitna revija 30: 415-418. Skok, P. 1971 (I: A-J), 1972 (11: K-ponil), 1973 (111: poni2-i), 1974 (IV: Kazala). Etimologijski rjetnik hrvatskoga ili srpskoga jezika. Zagreb: JAZU. Snoj, M. 1997. Slovenski etimoloiki slovar. Ljubljana: Mladinska knjiga. - -. 1997a. Kaj je kaj? Skrabteva misel 11. Zbornik s Simpozija '96: 187-192. Nova Gorica: FranEigkanski samostan Kostanjevica. Timberlake, A. 1978. Uniform and Alternating Environments in Phonological Change. Folia Slavica 2: 312-328. - -. 1981. Dual Reflexes of *dj in Slavic and a Morphological Constraint on Sound Change. International Journal of Slavic Linguistics and Poetics 23: 25-54. M. L. Greenberg. Rhotacism in South Slavic ... 75 Vondrik, W. 1924. Vergleichende slavische Grammatik, I. Band. Lautlehre und Stammbildungslehre. Gottingen: Vandenhoeck & Ruprecht. VujiEiC, D., D. BrozoviC, M. DesiC, and A. Peco. 1979. Govori sjeverozapadne Bosne. Bosanskohercegovatki dijalektoloiki zbornik 11: 7-157. Sarajevo: Institut za je- zik i knjiievnost. Prispelo junija 1998, sprejeto novembra 1998 Received June 1998, accepted November 1998 RazliEni vzroki za Siritev in umik fonetiEne spremembe: rotacizem v juinoslovanskih jezikih Prispevek obravnava spremembo i > r (nrotacizem<<) v juinoslovanskih jezikih, kot se najde npr. v sedanjiku glagola "moie(t5) > sln. more, hrv./bos. nar. miire, v oziralnih veznikih (*kako i(e) > sln. kakor), v razliEnih prislovnih tvorbah, npr. ("boie > bore), v nekaterih sln., hrv./bos. nareCjih tudi v sedanjiku glagola ("gma- ti :) *iene(ta) > rene. Izvor te spremembe je razliEno obravnavan kot fonetiEna ali oblikoslovna (analogna) sprememba, vendar pa nobena od razlag ne pojasnjuje ome- jene uresnititve spremembe (tj., sprememba je omejena na majhno gtevilo leksemov in slovniCnih kategorij) in zelo razliCnih arealov spremembe v vsakem posameznem primeru, v katerem je do nje priglo. V priCujoEem prispevku se dokazuje, da je iz- vor spremembe fonetiCen in da izhaja iz teinje, ki podpira razloCevalne lastnosti pri samoglasnikih (za razliko od severnoslovanskih razvojev), kar je teza, ki se navezuje na predvojno IsaCenkovo razpravo. Vseeno pa se da Siritev spremembe in njen umik delno razloiiti z oblikoslovnimi dejavniki. Eden od dejavnikov je ta, da stalno okolje (npr. sedanjik morem, morei ...) vspodbuja njeno giritev, v nasprotju s spremenljivimi okolji (-ienel-rene), ki so povzroEila njen umik ali zaustavitev. Drugi dejavnik je Si- ritev -r-ja kot produktivnega oziralnega veznika in Casovnega znaka po modelu ob- lik *kate/o-r- oz. vete-r. Zdi se, da je pri medmetu borme dodatno vlogo pri ohran- janju oblike z -r- igrala tabuiranost. Umik spremembe v gtokavSCini je imel lahko stilistitne vzroke, saj so bile oblike z -r v nasprotju z oblikami z i(e) obEutene kot izrazito zahodne, katoligke. Ta opaianja izhajajo iz dejstva, da oblike z -r izginevajo prav v tistih primerih, v katerih so bile produktivne v EakavSEini, kajkavgEini in slovengCini, tj. v oziralnih veznikih in Casovnih prislovih. In ne nazadnje, najbolje so ohranjene oblike, ki predstavljajo semantitne inovacije, npr. morati, jer, in to celo v StokavSCini, kjer so oblike z -r- leksikalizirane. Multiple Causation in the Spread and Reversal of a Sound Change: Rhotacism in South Slavic The paper treats the change i > r ("rhotacism") in the South Slavic languages, such as that, e.g., found in the present tense of the verb "mdie(t.6) > Sn more, WBs dialect more; complementizers, e.g., "kako i(e) > Sn kakor; in various adverb- 76 Slovenski iezik - Slovene Linguistic Studies 2 (1999) ial formations, e.g., *boie > bore; and, in some Sn and CrIBs dialects, the present tense of the verb (*gmati:) *iene(ta) > rene. The origin of the change has been treated variously as a phonetic or a morphological (analogical) change, though neith- er explanation can account for the limited realization of the change (i.e., it is re- stricted to a small number of lexical items and grammatical categories) and the vast- ly differing areals of the change for each item in which it occurs. The present paper argues that the origin of the change is phonetic and follows from a tendency to fav- or vocalic distinctive features (in contrast to N-S1 developments), a notion that goes back to a pre-War paper by IsaEenko. Nevertheless, the spread of the change, and its reversal, can be partially explained by morphological factors. One factor is that uni- form environments (e.g., the present tense morem, morei ...) favor its spread vs. alt- ernating environments (-ienel-rene) which have reversed or inhibited it. Another factor is support for the spread of -r as a productive complementizer and temporal marker on the models of metanalyzed forms *k5te/o-r- 'which' and *vete-r 'evening', respectively. Taboo seems to play an additional role in pres~rving r-forms in the in- terjection borme 'my God!'. The reversal of the change in Stokavian may have been due to stylistic considerations, where r-forms were felt to be distinctly western, Catholic, in contrast to i(e) forms. This observations follows from the fact that r-fo~ms disappear in Stokavian in precisely the forms in which they were productive in Cakavian, Kajkavian and Slovene, namely, the complementizers and temporal ad- verbs. Finally, those forms which represent ~emantic~innovations, uch as morati 'must', jer 'for, because' are best preserved, even in Stokavian, where the r-forms are lexicalized as such.