[1986/2010] Some sandhi phenomena involving prosodic features (vowel length, stress, tone) in Proto-Slavic, Serbo-Croatian, and Slovenian [revised version]

Some sandhi phenomena involving prosodic features (vowel length, stress, tone)

in Proto-Slavic, Serbo-Croatian, and Slovenian [revised version, 2010]

Willem Vermeer

[Note on the revised 2010 version. This article appeared in Henning Andersen (ed.), Sandhi Phenomena in the Languages of Europe,

Berlin etc.: Mouton de Gruyter, 1986, pp. 577-603. It was written in 1983. At the time, the Department was under threat and my job was on the line, together with those of most of my colleagues. The unsettled conditions show embarrassingly in the textual quality of the article (language, signposting, …). Henning Andersen managed to avert some excesses and I am still grateful for what he accomplished, but he could do only so much. More thorough revision seemed called for.

In this revised version I have tried to get rid of what seemed to me the worst sources of embarrass-ment while sticking faithfully to the printed text where essentials are concerned. All examples have re-mained the same and so has the argument, wording apart. Material changes are few and insignificant, and explicitly marked as such. In the process of revision, most of the endnotes have ended up in the main text. In connection with that, section 3 has been split up (3, 3.1, 3.2). The remaining endnotes have been changed to footnotes. An appendix has been added in which the most important dialects mentioned in the text are listed in systematic order for ease of reference.

The page numbers of the original edition have been added as follows: “rising |588| character”.]

-----

1. Introduction

In descriptions of Serbo-Croatian (SCr) and Slovenian (Sln) dialects, sandhi alterna-tions are not usually treated exhaustively, the way alternations of other types are. Yet sandhi is frequently mentioned and it is obvious that there is quite a bit of it around in both languages. The present contribution is intended to draw attention to a number of sandhi phenomena that involve the prosodic features of vowel length, stress, and tone. Attention is focused on alternations which involve features that are contrastive in at least some positions, although, needless to say, it is the local non-contrastiveness of a given feature at or near word boundaries that tends to be responsible for the rise of san-dhi.

The article is primarily intended to inform the non-specialist in general terms about certain properties of SCr and Sln dialects. Exhaustiveness is not aimed at.

2. Accent symbols

In most SCr and Sln dialects three prosodic elements are contrastive:

(1) tone: rising vs. falling, (2) vowel length: short vs. long,

2

(3) the place of the stress.

The accent symbols traditionally used in the descriptive literature combine information about the three prosodic elements along the following lines:

‶ short falling tone. ‵ short rising tone. ⁀ long falling tone. |578| ′ long rising tone (in Sln and most SCr dialects). ˜ long rising tone (in certain SCr dialects; some systems have two differ-

ent long rising tones, one of which is transcribed as ′ and the other as ˜).

¯ unstressed long vowel. ˘ unstressed short vowel; most often, however, unstressed short vowels

are left unmarked, e.g. SCr. sèstra ‘sister’ = sèstră.

Indications like “short falling” reproduce the traditional way of referring to the con-trasts involved and serve as labels. There are different kinds of falling and rising. In certain cases (in particular in Sln), a contrast labeled in terms of falling vs. rising would perhaps better be described as one of high vs. low.

Some systems lack contrastive tone either completely or in part. In such cases both ‶ and ‵ are used to render tonally neutral short stressed vowels. The former is more com-mon in SCr, the latter in Sln. Similarly ⁀ and ′ are used to render tonally neutral long stressed vowels where that is appropriate.

In neither SCr nor Sln does contrastive tone ever occur in unstressed syllables, with the sole exception of the Poljanski dialect of Sln as described by Marija Stanonik (1977).

Nearly all systems operate without general rules determining the place of the stress on the basis of the phonetic environment, as in, say, Czech or Classical Latin, or other-wise restricting the place of the stress to certain syllables, as in Italian, Greek, Polish and many other languages. I am aware of only two well-authenticated exceptions to his, see Fancev (1907) and Lončarić (1977) for a system where the stress can only hit the two final syllables (Križevačko-Podravski Kajkavian), and Ivić (1961: 200-202) for a dialect where stress is restricted to the penultimate or antepenultimate syllable (Oštarije Central Čakavian). In both cases the choice of the stressed syllable is determined by vowel quantity in a way that is reminiscent of classical Latin, but not identical with it.

In quoting examples from sources, I have in a few cases simplified and/or normal-ized the transcription; clitics are always printed as separate words even if they are not in the source, which is rarely the case.

Technical terms referring to dialect groups (e.g. Štokavian) and indications about the location of dialects are given for the benefit of those who have some familiarity with the subject. Nothing depends on them.

[Addition 2010. See now the Appendix for a systematic inventory of dialects that re-ceive some attention in the article.]

3

3. Accent types and sandhi in Proto-Slavic

Most varieties of SCr and Sln have remained relatively close to the latest recon-structible Proto-Slavic prosodic system, especially as compared with other Slavic lan-guages, hence comparison with PSl is convenient to serve as point of reference.

3.1 Proto-Slavic (PSl)

PSl as traditionally reconstructed had contrastive tone (rising vs. falling) and vowel quantity (long vs. short). In addition any syllable could be stressed. As in nearly all of Sln and SCr, contrastive tone was limited to stressed syllables. Vowels were predicta-bly short, hence presumably neutral as to length, in pretonic positions further removed from the stress than the first pretonic syllable. It follows that long vowels could occur in:

(1) the stressed syllable, (2) posttonic syllables, (3) the first pretonic syllable.

Stress alternations were (virtually) limited to three quite distinct major patterns known in the literature as “accent types” or “accentual paradigms”. This classification was introduced by Stang (1957), who proposed to use the letters (a), (b) and (c) to refer to the three types.1

In each of Stang’s accent types a certain movement of the stress – or absence of such – was linked to quite definite tonal and quantitative phenomena, as follows: |579|

(a) Fixed stress on a given syllable of the stem.2 (b) Recessive stress, also referred to as recessive mobility, meaning stress alternating

between the final syllable of the stem and the first syllable of the ending. (c) Mobile stress, also known as lateral mobility, i.e. stress alternating between the ini-

tial syllable of the stem and one of the syllables of the ending (the last or the penul-timate, depending on the phonemic make-up of the ending).

Tone was generated by the following simple rule: any stressed vowel was rising, with two classes of exceptions in which the tone was predictably falling:

− Forms with initial stress belonging to type (c), e.g. gȍvoŕǫ ‘I am speaking’, klȃdǫ ‘I am loading’.

1 Stang’s reconstruction was elaborated by Illič-Svityč (1979, Russian original 1963) and by Dybo in a long series of publications (e.g. 1981). It constitutes the basis of modern Slavic accentology, on which see Kortlandt (1975; 1978). 2 [Note added in 2010. At a very early stage SCr, or perhaps rather the PSl dialect underlying it, or part of it, developed a situation by which certain di- or polysyllabic nouns otherwise belonging to type (a) stressed a different syllable in the genitive plural from the one stressed in all other case forms. This fun-damental departure from the PSl system plays no role in the article that follows.]

4

− Vocatives, apparently irrespective of accent type. This class of forms is mentioned here only for the sake of completeness, but see the example given below in the last paragraph of section 5.1.

Quantity was determined by more intricate rules, which likewise depended on accent type.

All reconstructible sandhi alternations involved combinations of stem-stressed forms with clitics and were closely connected with the properties of the accent types, as fol-lows:

(a) Fixed stress. No sandhi. (b) Recessive stress. Stem-stressed forms with monosyllabic endings shifted the stress

to the ending if a clitic followed. This is poorly attested. It is held to be continued by modern Russian examples like kúrit ‘smoke’ (3d. sg. pres.act.) vs. kurítsja ‘id.’ (3d. sg. pres. reflexive, formed by the addition of the particle ‑sja, originally a clitic pronoun; in quoting Russian, ′ denotes stress without tonal or quantitative implica-tions). See Ebeling (1967a: 591-592).

(c) Mobile stress. Stem-stressed forms became unstressed if they were followed and/or preceded by clitics, according to the following rules: (1) enclitics attracted the stress, e.g. gȍvoŕǫ ‘I speak’ vs. govoŕǫ bò (bo ‘for’). (2) In the absence of an en-clitic, proclitics received the stress, e.g. nȅ govoŕǫ ‘I do not speak’ (but cf. ne go-voŕǫ bò ‘for I do not speak’, with stress on bo because enclitics prevail over procli-tics). If two or more proclitics were present, it was the initial one that was stressed, e.g. ȉ ne govoŕǫ (i ‘and, also, even’).3

3.2 From Proto-Slavic to Serbo-Croatian and Slovenian (generalities)

The PSl accentual system as described in the preceding section does not seem to have been very stable. Modifications arose at an early stage even in SCr |580| and Sln. Exam-ples:

A. On short vowels, the difference between the tones was everywhere either elimi-nated (SCr) or continued in a way no longer involving a tone contrast on short vow-els (Sln). Tone contrasts on short vowels found in modern dialects are all of secon-dary origin.

B. On the other hand the functional yield of the tone contrast on long vowels was in-creased by developments which caused long falling vowels to appear in non-initial syllables.

C. Phonetic stress shifts – most commonly leftward movements – took place in many areas, often giving rise to new tonal contrasts. Morphological stress shifts – for in-stance levelling of alternating paradigms – were even more widespread.

D. Redistribution of vowel length took place everywhere.

3 Examples freely adapted from Dybo (1977: 189-190).

5

As a result of all this, the boundaries between the three inherited accent types were usually blurred. It is only thanks to the comparative method on one hand, and to the availability of old accented texts and reliable grammars of a few archaic dialects on the other, that the PSl system can be reconstructed at all.

The most conservative SCr systems have retained the PSl inventory with three modi-fications:

(1) merger of the short rising with the short falling tone; (2) introduction of the long falling tone in non-initial syllables; (3) loss of contrastive quantity in vowels followed by syllable-final resonants in certain

positions.

Such maximally conservative systems are limited to a few tiny areas along the Adriatic coast, notably near Rijeka (e.g. Novi Vinodolski according to Belić 1909) and Zadar (e.g. the island of Vrgada according to Jurišić 1966 and 1973).

Most living systems have carried through additional innovations, of which the most common are:

A. Loss of length in posttonic position. B. Loss of the tone contrast also on long vowels. C. Leftward shifts of the stress – known as stress retractions –, usually accompanied

by the rise of novel tone contrasts or other complications.

These, and other, innovations have operated independently of each other. As a result, a bewildering variety of prosodic systems has arisen.

The three PSl sandhi phenomena described above have not fared equally well in the process:

− The type kúrit/kurítsja appears not to be extant anywhere in SCr or Sln. − The type gȍvoŕǫ/govoŕǫ bò does not seem to have survived either, unless it is con-

tinued by rare examples like spīš lȉ ‘do you sleep?’, znāte lȉ ‘do you (plur.) know?’, in which the stress has shifted from the verb onto the interrogative particle li (Rešetar 1900: 202 on Prčanj, Zeta-Lovćen Štokavian, no tones on short vowels).

− The type gȍvoŕǫ/nȅ govoŕǫ, on the other hand, is attested almost everywhere in SCr and Sln. It will be returned to below (section 5).

Most sandhi phenomena found in SCr and Sln are the consequence of post-PSl innova-tions. Like the inherited sandhi alternations, nearly all of them involve clitics.

In that connection it is vital to realize that the list of clitics is not everywhere the same. Indeed it may not even always be the same for all rules within a single system.

The list of proclitics usually includes most prepositions and verbal prefixes, and the negative particles ne and ni. In addition it often includes conjunctions (e.g. i ‘and’, da ‘that’), and occasionally numerals (for the latter see further in section 6.1). In what fol-lows, verbal prefixes will not be taken into account because they tend to obey the rules that obtain within the morphological word (although they do not do so invariably).

6

The list of enclities usually includes clitic pronouns, with considerable local differ-ences (e.g. me ‘I’, acc.; ni ‘we’, dat./acc.), particles (li interrogative particle; do particle modifying the meaning of the imperative; ka particle of unknown meaning), and the present tense of two or – less often – three verbs (sym ‘I am’, ćedu ‘they will’, reš ‘you go’) (the examples are SCr, though in most cases dialectal).

4. Vowel length

Vowel length alternations were common in the PSl dialect that underlies SCr and Sln. Most systems have subsequently undergone innovations which have had the effect of introducing new length alternations. It is therefore not surprising that there are several types of sandhi which involve vowel length.

4.1 SCr. (Srem) bȅrĕ vs. bȅrē se

Shortening of the posttonic long vowels of PSl is general in Sln and common in SCr, where it has given rise to sandhi in at least one area. |581|

In many Neo-Štokavian dialects spoken mainly north and west of Belgrade, long vowels have been shortened in open final syllables, while remaining long when not final. This has naturally resulted in an alternation between short and long vowels de-pending on the presence of a clitic, e.g. bȅrĕ vs. bȅrē se ‘pick up’ (3d. sg. pres. minus or plus reflexive pronoun respectively; Nikolić 1964: 296 on Batajnica in eastern Srem, see also 229-230). To the best of my knowledge, the phenomenon was first described by Milan Rešetar (1900: 30), whose attention had been drawn to it by the Hungarian linguist Oskar Asbóth.

It looks as if in cases like these quantity is neutralized in final position, so that what one could call the underlying quantity of a given morpheme shows up only if a clitic is present.

Length tends to spread to forms in which it has no etymological justification, e.g. Inđija (Eastern Srem) ònā je ‘she is’ < òna je, where ‑a is historically short (Nikolić 1964: 399). In many systems every word-final vowel is lengthened before a clitic (Pop-ović 1968: 83, Ivić 1979: 164). Elsewhere the process is going on as we speak. Unfor-tunately the details have not to my knowledge been adequately described. The process seems to get under way even before the contrast between long and short vowels has been completely eliminated in the relevant positions, i.e. when length is still optional (in the sense in which this term is used by Ebeling 1967b: 134-136 and Kortlandt 1972: 152-164; 1973), see Nikolić (1966: 195-201; 1969: 11-15) on dialects which are slower than eastern Srem in losing posttonic length.

Conversely, complete loss of the long alternant in dialects where the rules for the shortening of posttonic vowels would lead one to expect the alternation has also been recorded. For an examples see Ivić (1957: 40) on the Gallipoli dialect. It appears to be very uncommon.

In some systems, new long vowels have arisen which are not shortened, so that the alternation is no longer phonetically predictable, see Ivić (1958: 170) and Popović (1968: 31) on Begeč (central Bačka, northwest of Belgrade).

Factors complicating the study of the phenomenon are many and varied:

7

(1) The exact rules for shortening of posttonic vowels are complex and differ from sys-tem to system: they depend such factors as the distance of the final syllable from the stressed syllable, the presence of other posttonic long vowels, and/or the tone and quantity of the stressed vowels. (Note that the rule I have given in the text has been artificially simplified for the sake of clarity.)

(2) The task of determining the exact rules is made more difficult by the fact that there are usually positions in which length is optional and by the fact that morphological levellings – like the spread of length in cases like ònā je – tend to obscure the mat-ter.

(3) Owing to the composite nature of the population of the area the investigator cannot count on the speakers in a given village all being carriers of basically the same sys-tem.

(4) Even though long posttonic vowels are required by the norm, they sound rural and yokelish to educated city dwellers. As a consequence sociolinguistic pressures strongly favour loss of posttonic length, resulting in significant differences between speakers of different generations, at least in some areas (Ivić 1979: 162-163).

4.2 SCr (Omišalj) ȍko vs. v ȏko

In the SCr (Northwest Čakavian) dialect of Omišalj (spoken on the island of Krk in the Adriatic) any initial vowel is lengthened if it is preceded by a non-syllabic preposition like v ‘in(to)’, e.g. utȍrek ‘Tuesday’ vs. v ūtȍrek, ȍko ‘eye’ vs. v ȏko (no tone contrast; for some more examples see Vermeer 1980: 471). The alternation seems to occur whenever the conditions for it are present, which, to be sure, is not very often because the number of words beginning with a vowel is small, whereas non-syllabic preposi-tions are few and have more explicit syllabic variants which are preferred in many situations, e.g. va ‘in’. If one tries to elicit forms like v ȏko, for instance, informants almost invariably select the syllabic variant of the preposition, so the investigator just has to wait for examples to appear spontaneously. For references to related dialects where the phenomenon has been found see Vermeer (1982: 313). |582|

A somewhat similar alternation is common in dialects spoken around Zagreb and known as Kajkavian. It differs from its Northwest Čakavian counterpart in being lim-ited to stressed syllables on one hand and in not being limited to forms beginning in a vowel on the other, e.g. Prodindol (southwest of Zagreb) vèčer ‘evening’ vs. zvȇčera ‘in the evening’ (Rožić 1893: 107; 1894: 60), Ozalj (somewhat more to the west) zȅmlju ‘earth’ (acc. sg.) vs. v zȇmlju ‘into the earth’ (Težak 1981: 261, see also p. 241), Brezova near Začretje (northwest of Zagreb) vȍdu ‘water’ (acc. sg.) vs. h vȗodu ‘into the water’ (Junković 1972: 200n.; the long vowel has been diphthongized). These dia-lects lack tone on short vowels, with the possible exception of Prodindol. Lengthening produces a falling vowel. Judging by Ivšić (1936: 71) the alternation is more or less general in the Kajkavian dialect group of SCr.4

4 For a tentative historical explanation see Vermeer (1979: 372), where the term “compensatory length-ening” is due to an oversight, cf. Vermeer (1983: 468-469, notes 2 and 3).

8

4.3 SCr. (Gruža) kȏ vs. kȍ te

In Borač in Gruža (southeast of Belgrade, Neo-Štokavian of a non-standard type closely related to the Šumadija-Vojvodina variety) the long vowel of four frequent words is shortened whenever a clitic follows. Those are kȏ ‘who’, štȃ ‘what’, gdȇ ‘where’, and štȏ ‘that, which’ (relative), e.g. kȍ te pita ‘who is asking you?’ vs. kȏ tȅbe pita ‘who is asking yóu?’ (Stevović 1969: 470, 600; pita read pȋtā). The alternation is not phonetically conditioned.

4.4 Sln. (Prekmurski) zà ńäga vs. zá to

In most dialects the vowels of most prepositions are consistently short, except when preceding a non-syllabic or monosyllabic personal, reflexive or demonstrative pronoun like nj ‘him’, me ‘me’ (acc. sg.), se (acc.sg.) ‘myself, yourself etc.’, to ‘that’ (the pre-cise list is different in different dialects). This results in such alternations as zà ńäga ‘for him’ vs. zá to ‘for that’ (Pável 1909: 5, 10, 12, 14, 129 on Cankova in Prekmurje in eastern Slovenia; no tones).

The alternation appears to be very stable: it is found wherever the combinations of prepositions + short pronominal forms are in use and wherever the prosodic system has not been too drastically modified. Exceptions, though very rare, do occur, e.g. Zeta-Lovćen Štokavian nȁ nj (Ćupić 1977: 81 recorded in the Bjelopavlići village of Mio-kusovići), presumably connected with the fact that the relevant combinations happen to be very infrequent in the dialect.

The stability of the alternation is all the more remarkable because different dialects treat combinations like za me in different ways with respect to stress and tone. Speak-ing in PSl terms there are three different types, as follows:

− *zȃ me, with the reflex of the PSl falling tone on the preposition, as in zȃ me (Belić 1909: 235 on the Northwest Čakavian dialect of Novi Vinodolski, near Rijeka on the Adriatic coast);

− *zá me, with the reflex of the PSl rising tone on the preposition, as in noã me (na ‘on(to)’), as reported by Jurišić 1973: 78 for the Southeast Čakavian dialect of Vrgada near Zadar; |583|

− *zā mȅ, with the pronoun stressed, as in Neo-Štokavian zá me, reflecting earlier *zā mȅ, as shown by the rising tone, which owes its existence to a stress retraction for which see below, section 6.1. This accentuation is very widespread and also hap-pens to be the one that is recommended by normative grammars.

5. Stress I: the inherited alternation

We saw earlier (section 3.1) that in PSl, stem-stressed forms of words with mobile stress (type (c)) automatically lost the stress whenever a clitic was present. As far as proclitics are concerned, the alternation has been retained all over SCr and Sln. There is an important difference, however. Whereas in PSl the phenomenon was predictable on phonetic grounds – any form with a falling tone could be counted on to lose the stress to clitics –, the alternation is nowhere any more predictable in this sense, owing to in-novations that have hit the prosodic system.

9

5.1 SCr. plȍta vs. prȅko plota

In SCr, retention of the PSl sandhi alternation depends on a large number of factors, e.g. on the clitic involved and on the part of speech to which the form that is – or is not – to lose the stress belongs.

In combinations consisting of the 2nd/3rd pers. sg. of the aorist and the proclitic par-ticle ne ‘not’, retention of the alternation is universal or nearly so (Nikolić 1970: 53).

In the case of combinations consisting of nouns and proclitic prepositions, on the other hand, things are much more complicated.

Retention of the PSl system – with insignificant modifications – is required by the accentual norm of the standard language as codified by Daničić (1925), although it ap-pears not to be characteristic of the modern standard language as actually spoken by educated city dwellers.5 Complete or virtually complete retention of the alternation is also found in a limited number of mainly southern Štokavian dialects, e.g. Dubrovnik (Rešetar 1900: 207).

Quite a few systems have retained the alternation in a limited number of |584| cases which have to be listed as exceptions. Nedeljko Bogdanović, for instance, in his gram-mar of the Torlak dialects spoken in the area between Svrljig and Knjaževac in South-East Serbia, mentions a mere fifteen nouns that lose the stress to proclitics (1979: 49). Along similar lines Stevović repeatedly argues that in the Gruža dialect it is only in a limited number of fixed – adverbialized, as he puts it – expressions that retraction is found (Stevović 1969: 512, 562, 576-577, 582).

Although the laws that govern the development from complete retention to retention in a small number of exceptional cases are still in many essentials unclear, a few ten-dencies have been formulated by Ivić (1958: 167-168):

1. the longer a word, the less likely it is to lose the stress to a proclitic (see already Daničić 1925: 58, originally 1856);

2. frequent combinations and fixed expressions are more likely to have retraction than other combinations;

3. personal names are less likely and geographical names more likely than other nouns to lose the stress.

Several dialects have been reported to use the two types of accentuation side by side on more or less equal footing. Such systems appear to be particularly widespread in Bos-nia, see, e.g., Petrović (1973a: 19-20) on Zmijanje in western Bosnia, cf. also sugges-tive formulations of the kind “retraction is common, but not obligatory” in Ćerić (1981: 419) on Dobretići, and Brozović & Vuji čić (1981: 407) on Guber, located more to the east and to the south respectively. What one expects in cases like these is that the choice between the two possibilities involves some difference in meaning, in the wide sense in which that word is used by Ebeling (1978: 6-14). Since traditional dialect grammars concentrate on form, they cannot be expected to yield much information on

5 The date of publication of Daničić’s book is misleading; the studies it contains were originally pub-lished in the fifties and sixties of the nineteenth century.

10

this score. The way Josip Baotić describes his native Slavonian Štokavian dialect of Kostrč in the Bosanska Posavina (1979: 218-223) suggests that subtle semantic consid-erations play a role.

Some dialects manage to combine the three types of solutions that were enumerated just now, for instance some of the Montenegrin (Zeta-Lovćen Štokavian) dialects stud-ied by Pešikan (1965: 61-62). Slightly simplifying Pešikan’s rules, the following ap-pears to be the case:

− Retraction is obligatory with monosyllabic stems containing a long vowel, except – at least in part of the area – in the case of personal names, thereby exemplifying one of the tendencies noted by Ivić. Since stress retractions from final syllables have caused the PSl accent classes to merge in a large number of cases, retraction takes place irrespective of the original accent class of the noun involved.

− If the stem is monosyllabic and contains a short vowel in most or all case forms, retraction is also obligatory, but the link with the PSl accent types has to some ex-tent been maintained.

− If the stem alternates between a monosyllabic and a disyllabic form – which is often the case –, the shift is optional, and one would like to know more about the factors determining the choice between the two types of accentuation.

− With polysyllabic stems the shift is limited to a small number of items which have to be listed.

A related system is described by Petrović (1973b: 174 on speakers originally hailing from Vraka/Vrakë on Lake Scutari in Albania).

An interesting variant of the PSl retraction was found by Rešetar (1900: 206 on Prčanj) and later investigators in some Zeta-Lovćen Štokavian dialects of Montenegro: if the preposition contains an internal morpheme boundary, the second component re-ceives the stress – in other words: the stress does not retract beyond the final compo-nent of a preposition –, e.g. ispȍd zīda ‘under the/a wall’ (< iz+pod) vs. prȅko zīda ‘across the wall’ (no tones on short vowels).|585|

In word forms that lose the stress to proclitics, long vowels usually remain long, ex-cept, trivially, in dialects that shorten posttonic long vowels. Exceptions do occur, however. In Borač, for instance, the long vowel of štȃ ‘what’ is shortened if it loses the stress to a proclitic, e.g. zȁ šta ‘for what’ (Stevović 1969: 470).

The reverse phenomenon – lengthening of short vowels when the stress is shifted to a clitic –, though rare, has also been found, e.g. Prčanj mjȅsto ‘place’ vs. ȕ mjēsto (Rešetar 1900: 34).

For the sake of completeness it is worth pointing out that the rule is also attested in the vocative, to the very limited extent that it occurs with proclitics, e.g. nom. dum Péro ‘father Peter’ (referring to a priest) vs. voc. dȕm Pēro (Rešetar 1900: 212, report-ing on the Dubrovnik dialect).

5.2 Slovenian nočȋ vs. do nọ̑či

In Sln the PSl alternation has acquired a radically different look because at an early stage of the development of the language all PSl falling accents (on which see section

11

3.1) lost the stress to the following syllable, which received a long falling accent irre-spective of its original quantity, e.g. *lȍvili ‘having hunted’ (l‑participle, nom.pl. msc.) > lovȋli (Valjavec 1897: 204). As a consequence of the shift, a proclitic which is put before a word, rather than attracting the stress onto itself – as in PSl or SCr –, has the effect of causing the stress to shift from the ending to the stem (in the case of monosyl-labic stems) or from the second to the first syllable of the stem (in that of polysyllabic stems), e.g. okọ̑ ‘eye’ vs. na ọ̑ko (na ‘on(to)’), korẹ̑n ‘root’ vs. pod kọ̑ren (pod ‘under’) (Valjavec 1897: 198-200).

Since the progressive stress shift of examples like lovȋli , okọ̑ is not the only source of the Sln long falling tone, the alternation is not predictable on the basis of the pho-netic shape of a given word form. It is not strange that it has tended to be eliminated. For combinations of prepositions with nouns, for instance, older codifications of the norm require the alternation in many more types of cases than more modern treatmens, contrast Valjavec (1897: 195, 197, 198, 200) and Breznik (1916: 41, 69) with Toporišič (1976: 221, 227, 230, 234), who limits it to a small number of fixed expressions. It is noteworthy that in many cases the form that has been generalized is the one that is regular after a preposition.

6. Stress II: post-Proto-Slavic developments

It is only in marginally located SCr dialects that the PSl place of the stress has been consistently maintained, e.g.:

− much of the Adriatic coast and the islands (Novi Vinodolski, Omišalj, Vrgada); − a narrow strip west and northwest of Zagreb, near the border of Slovenia (Brezova

near Začretje); − part of the border area between Montenegro and Albania (Piperi); − a few points in the Boka Kotorska area in Montenegro (Prčanj); − one or two other marginal areas, e.g. a few points in the Burgenland (Austria) and

possibly points in Slavonia (Sekereš 1977: 187, contrasting strikingly, however, with other treatments of the same dialects, e.g. Ivšić 1913: 146).

Most of SCr and all of Sln have carried through leftward movements of the stress, known in the literature as stress retractions. Retractions have given rise to various san-dhi phenomena.

6.1 SCr. (Neo-Štokavian) prȁga vs. prekò praga

In the dialects of the central core area of SCr (known as Neo-Štokavian) the stress has consistently moved back one syllable towards the beginning of the word. The newly stressed syllables have a rising tone. Whenever the stress could not be retracted because it was on the initial syllable already, the tone is falling irrespective of the original tone. Examples: PSl *vodà ‘water’, *glāvà ‘head’, |586| *napíšē ‘he/she writes’ (perfective),

12

*ȍko ‘eye’, *mlȃdo ‘young’ (neuter), *pràga ‘threshold’ (gen. sg.), *píšē ‘he/she writes’ (imperfective) > Neo-Štok. vòda, gláva, nàpīšē, ȍko, mlȃdo6, prȁga, pȋšē.

This development is known as the Neo-Štokavian stress retraction. For a better un-derstanding of the synchronic workings of the systems it gave rise to some introduction is necessary.

Although the Neo-Štokavian stress retraction is patently a diachronic change, there is a tradition of interpreting it as tantamount to a synchronic rule forbidding the falling tone to occur in non-initial syllables. However, as Ivić (1979: 165) points out, it is a fact that, as he puts it, “more or less everywhere in Neo-Štokavian” non-initial falling accents are perfectly all right in words that have entered the language after the stress retraction operated, e.g. paradȁjz ‘tomato’. Such forms clearly prove that Neo-Štokavian accentual systems in fact admit non-initial falling accents (see Ebeling 1967b: 131 on the theoretical issues involved). Borrowings apart, morphological proc-esses have given rise to internal falling tones in widespread and frequent plural geni-tives like Dalmatȋnācā (nom. sg. Dalmatínac ‘man from Dalmatia’; the tonal alterna-tion is modelled on the example of shorter stems like vrábac ‘sparrow’, gen. pl. vrȃbācā), see Šimundić (1971: 18) on Central Dalmatia, Dešić (1976: 208-209) on Zmijanje in western Bosnia, and Milorad Simić (1978: 23) on Obadi in eastern Bosnia. All three dialects are in all other respects typically Neo-Štokavian and Dešić is quite right in interpreting such forms as the outcome of innovation on a strictly Neo-Štoka-vian basis, rather than sporadic retention of the pre-retraction state of affairs, as has been done so often that it is something of a tradition.

Examples like paradȁjz and innovating forms like Dalmatȋnācā show that it is sim-ply not true, as is often asserted, that an internal morpheme boundary of the kind that separates the two elements of a compound like očevȉdnī ‘augenscheinlich’ (Vuk 1852: 482) is a necessary condition for the appearance of non-initial falling accents.

Whatever the exact synchronic status of the systems produced by the Neo-Štokavian stress retraction, the shift has obviously given rise to a second layer of stressed procli-tics because now any word form that had initial stress in PSl casts back the stress if possible, not just the ones belonging to accent type (c). The difference between the two layers is always easy to spot since the Neo-Štokavian retraction yields a rising tone on the final syllable of the proclitic whereas the PSl retraction results in a falling tone on the first syllable, e.g. prekò praga (accent type (a)) ‘across the threshold’, as opposed to prȅko plota (accent type (c)) ‘across the fence’ (Ledinci in Srem, Nikolić 1964: 250-251).

The two types of retraction differ with respect to the treatment of numerals. In PSl, numerals did not count as clitics and they still do not do so in modern systems that con-tinue the PSl sandhi alternation. However, in certain areas numerals have subsequently become clitic. To give an example, in Montenegrin systems of the conservative type which has not carried through any stress retractions, virtually all numerals are un-

6 [Note added in 2010. The type of Neo-Štokavian on which the norm is based has mládo, which is the outcome of analogical levelling and does not directly continue the PSl accentuation.]

13

stressed when followed by nouns, unless a contrastive interpretation is intended (see Stevanović 1940: 148 on the Piperi dialect).

Against this background it is not surprising that in part of the Neo-Štokavian area some or all numerals count as proclitics, too, giving rise to such forms as Dubrovnik, Piva & Drobnjak, Gruža dvá brata ‘two brothers’ (Rešetar 1900: 211; Vuković 1940: 300; Stevović 1969: 606). In such systems, numerals followed by nouns are still un-stressed if they have not received the stress as a consequence of the retraction, e.g. Piva & Drobnjak pe sestárā ‘five sisters’ (Vuković 1940: 301). As Vuković points out, the consequence of all this is that synchronically mobile nouns like kamen ‘stone’, dinar ‘dinar (coin, unit of money)’ lose the stress in the PSl manner to prepositions (as in, e.g., zȁ kamenom ‘behind the stone’, zȁ dinār ‘for a dinar’), but in the Neo-Štokavian manner to numerals (dvá kamena ‘two stones’, stò dinārā ‘a hundred dinar’). This vio-lates the PSl principle that the clitic plays no role in the choice of the type of retraction.

As these examples show, a further complication arises from the fact that in some numerals the long vowel is shortened (e.g. in pȇt ‘five’, stȏ ‘hundred’), whereas in oth-ers the long vowel stays long (e.g. in dvȃ ‘two’, trȋ ‘three’).

Most Neo-Štokavian systems appear to have retained the original distribution – of PSl vs. Neo-Štokavian retraction – to a remarkable extent. Exceptions do occur, how-ever. To mention one example, in Piva & Drobnjak pronominal forms like tȅbe ‘you’ (gen./dat./acc./loc. sg.) lose the stress in the Neo-Štokavian way to monosyllabic prepo-sitions (e.g. zà tebe ‘for you’), but in the PSl way to disyllabic prepositions, e.g. prȅko tebe ‘by way of you, via you’, thereby again violating the PSl principle that the clitic plays no role in the choice of the type of retraction (Vuković 1940: 265).

A tendency towards elimination of the alternation is as strong in the case of the re-cent retraction as it is in the case of its PSl counterpart. The same types of systems can be distinguished: those in which the alternation is obligatory or virtually obligatory, those in which it is found only in a limited number of fixed combinations, and those in which it is optional, giving rise to the problem of determining what factors – semantic or otherwise – determine the choice between the two possibilities, cf. above, section 5.1.

As far as can be determined at the present moment, the two alternations tend to be treated in the same way. If one is obligatory, then the other is too. If one is limited to a fixed number of combinations, then the other is too, etc. However, since investigators are probably inclined to assume beforehand that both alternations will be treated in the same way, it is conceivable that instances of differential treatment have remained unre-ported. |587|

6.2 SCr. (Gallipoli) gláva vs. glāvȁ je

In areas situated between the Neo-Štokavian core area – with its consistent stress re-traction – and the marginal dialects which have retained the PSl place of the stress, many types of partial stress retraction have been found.

Just as final syllables are more susceptible to loss of length than non-final ones (see section 4.1), they are more prone to lose the stress: in many dialects stress retractions have taken place from final syllables only. In such systems, words occur with retracted

14

stress or final stress depending on whether or not a clitic follows, e.g. dóđi ‘come’ (im-perative) vs. dōđȉ mi (mi ‘I’, dat. sg. clitic) (Ivić 1958: 227 on Kasidol, east of Bel-grade, Kosovo-Resava Štokavian; no tones on short vowels).

This alternation has been found in many mutually unrelated SCr systems situated around the Neo-Štokavian core area, e.g.:

− to the northwest in the Kajkavian dialects southwest and south of Zagreb (Rožić 1894: 103-110 on Prodindol; Šojat 1982: 383 on Turopoljski),

− to the north in the autochthonous dialects of the Posavina, midway between Zagreb and Belgrade (Ivšić 1913: 160; Baotić 1979: 163),

− to the northeast and east in the so-called Kosovo-Resava dialects (Ivić 1958: 226-232 with references; Radoje Simić 1972: 47-54 on Levač),

− to the southeast in Montenegro (Ivić 1958: 203-205 with references; Pešikan 1965: 26-27; Petrović 1966: 130, 1967: 231; Ćupić 1977: 64),

− to the west in the Central Čakavian dialects of Lika (see Finka – Pavešić 1968: 10 on Brinje and surroundings), and the dialects of western Istria (Bošković 1966-1967: 93).

Innumerable differences exist among individual dialects as regards the precise condi-tions under which retraction takes place, the prosodic characteristics of the newly stressed vowel, and the degree of retention.

The dialect of the Gallipoli Serbs described by Ivić (1957: 26-35) is typical. Final syllables containing a short vowel have lost the stress; the newly stressed vowel is long rising, irrespective of its original quantity, which is reminiscent of the outcome of the Sln progressive shift from PSl falling vowels, for which see above, section 5.2. Exam-ples: gláva ‘head’ < glāvà, óna ‘she’ < ŏnà. With the exception of a few fixed expres-sions (e.g. dubrȁ vi nȏć ‘a good night to you’) length has been carried over to end-stressed forms followed by clitics, e.g. ōnȁ se < onȁ se (se reflexive pronoun acc.). The retraction from final syllables is the only source of the long rising tone. By the way, its rising |588| character is not obligatory: it can always be replaced by a falling tone which does not differ from all other long stressed vowels, e.g. glȃva, ȏna. Long vowels have been shortened in stressed open final syllables, e.g. *vodé ‘water’ (gen. sg.)> vudȅ, a development which has given rise to a second type of sandhi because it, too, failed to take place before clitics, cf. the long stressed vowel in vudȇ mi se; other examples: tȉ ‘you’ vs. tȋ li (li question particle), disnȁ rúka ‘the/a right hand’ vs. disnȃ mi rúka ‘my right hand’. Analogical carry-over of the shortened vowel to the position before clitics is not unknown, e.g. vodȅ mu (mu ‘him’, dat.sg. clitic). Ivić does not specify the condi-tions under which this development takes place. Shortening of long vowels in stressed final syllables at a moment when short stressed vowels in final syllables have lost the stress is a natural development that is found in several other systems, e.g. in the dialects of southern Istria (Bošković 1966-1967: 93; Popović 1968: 83) and Šepurina on the island of Prvić (near Šibenik in Central Dalmatia), see the examples in Kursar (1979

15

passim). The change is taking place in the Kosovo-Resava Štokavian dialect of Landol (east of Belgrade), see Ivić (1958: 231).7

With respect to the retention of the alternation there are enormous local differences. In some areas it appears to be automatic, as it is in the dialect of the Gallipoli Serbs, e.g. in the Zeta-Lovćen Štokavian dialect described by Ćupić, or here and there in the Posavina. This has only been found in areas where retraction yields a rising tone that makes the newly stressed vowels different from all other stressed vowels. In some of these cases one may suspect the rising tone of examples like gláva to be in reality the realization of a pretonic long vowel in a form that carries the stress on the final syllable (glāvȁ), so that strictly speaking there is no alternation at all. This solution is favoured, for instance, by Ivić (1958: 205, 227, 291). The existence of such unique rising tones is not, however, a sufficient condition for the alternation to be automatic. In the Kosovo-Resava Štokavian dialect of Levač, to mention one example, retraction onto long pre-tonic vowels has yielded a unique long rising tone; nevertheless the alternation is op-tional, e.g. uzēšȅ mi ‘they took from me’, with the final stress that is to be expected before a clitic, vs. uzéše ni ‘they took from us’, with analogical carry-over of the accen-tuation that is regular in forms not followed by a clitic (Radoje Simić 1972: 53).

In those dialects where the alternation is not predictable on the basis of the tonal characteristics of forms not followed by a clitic, it is usually either dead or moribund. In the Montenegrin Zeta-Lovćen Štokavian dialects investigated by Pešikan (1965) the alternation has survived only in a tiny number of fixed combinations, e.g. fālȁ ti ‘thanks to you’ vs. fȃla ‘thanks’ (p. 35, see also pp. 26-27, 34; no tones, the example is from the village of Zagarač). In the Central Čakavian dialect of Lešće (in Lika) Ivić found the alternation in the single word kȁdi ‘where’ vs. kadȉ si ‘where are you’ (Ivić 1964: 127; no tones on short vowels). The alternation has been completely eliminated in a number of areas where the rules for stress retraction would lead one to expect it. |589|

6.3 SCr. (Podravina) mȕškārac/mȕškārȁc vs. muškārȁc je

In some of the autochthonous dialects of Slavonia – in particular in the area along the Drava river known as Podravina – we find retractions which differ from the ones treated in the preceding section in that it is the initial syllable – rather than the penulti-mate one – that receives the stress, e.g. muškārȁc ‘man’ > mȕškārac (or mȕškārȁc, with a secondary stress on the final syllable). This results in alternations like mȕškārac (or mȕškārȁc) vs. muškārȁc je (je ‘he/she/it is’, clitic) (Valpovo and/or surroundings, Hamm 1937: 74; Klaić 1936: 183). If possible the stress retracts to proclitics, e.g. Baćin (Hungary) ȉ za vreteno < i za vretenò ‘and for the spindle’ (Ivić 1961-1962: 123).

Unfortunately there are no satisfactory full-scale descriptions of dialects with this type of sandhi and some fundamental problems raised by them have not yet been eluci-dated. There is no consensus about the accentual properties of the forms in which re-traction has taken place. It is not even clear whether or not the falling accent that is the outcome of the retraction is or is not the same as the other falling tone of the dialects

7 The process has a parallel in Leskien’s law of Lithuanian, cf. Kortlandt (1977: 328) with references.

16

involved. On this non-trivial issue there is disagreement not only between Hamm and Klaić as regards the Valpovo area, but also between Ivić (1961-1962: 123 n.) and Bro-zović (1981: 604, 606-607) on Dušnok (Hungary). See further Ivić (1958: 286-287 with references), cf. also Sekereš (1974, 1977 and other publications by the same author) and Finka – Šojat (1981).

6.4 SCr. (Turopoljski Kajkavian) vȋdim vs. nȅ vidim

In the Kajkavian dialects of SCr – which are spoken in the area around Zagreb – sev-eral developments have given rise to new instances of the long falling tone.8

The new long falling accents have tended to cede the stress to the preceding syllable. If the latter contained a long vowel, retraction seems to be a general Kajkavian phe-nomenon (Ivšić 1937: 188). Since proclitics containing a long vowel in the required position happen not to exist, this development has not given rise to sandhi.

However, in some systems the stress has also been retracted to short vowels, pre-dictably giving rise to a new layer of stressed proclitics, e.g. vȋdim ‘I see’ vs. nȅ vidim ‘I do not see’ (Šojat 1982: 411 on Turopoljski; no tones on short vowels).

Not much is known about the pattern of retention of the alternation, notably in rela-tion to its PSl counterpart. The geographical distribution of the retraction was deter-mined by Ivšić (1936: 81-83 and map, with discussion of earlier literature). Junković (1956: 394) draws an isogloss between |590| areas with retention of the sandhi alterna-tion and dialects where it has been eliminated. However, in view of what we know about similar alternations in other dialects, it would be surprising if in all systems with a Turopoljski-like accentuation all types of cases would turn out to be eliminated at the same time, as Junković’s isogloss suggests, so obviously more research would be desir-able.

In the Poljanski dialect of Sln (west of Ljubljana) a phonetic stress retraction has also had the effect of making words lose the stress to preceding words. The Poljanski retraction is intriguing – like much else about the dialect – because it is not only clitics that can receive the stress in this way. Unfortunately the amount of available infor-mation is limited. On this see Ramovš (1935: 97) and Stanonik (1977).

6.5 Sln. (Rožanski Carinthian) dorò vs. dòro je

In some of the Sln dialects spoken in Carinthia (Austria) end-stressed forms containing a short vowel in the stressed syllable retract the stress if an enclitic is added, e.g. dorò ‘good’ (neuter) vs. dòro je ‘it is good’ (Rigler 1977: 90 on Breznica pri Št. Jakobu v Rožu). A similar alternation is found whenever a disyllabic end-stressed word is lengthened through the addition of an ending or through replacement of one ending by a longer one, cf. zalàn vs. zȁlano ‘green’ (masc. vs. neuter), baràm vs. bàramo or bȁramo ‘I gather’ vs. ‘we gather’.

Evaluation of the matter is complicated by a number of factors; among other things the accentual characteristics of the forms with retracted stress are not quite clear, cf. the

8 For the diachronic background see Kortlandt (1976) with references.

17

inconsistent notation in the examples and Rigler’s admission (loc. cit.). Unfortunately very little credible information on the dialect has been published. According to Rigler (1977: 89n.) the accentuation in the data reported by Scheinigg (1881-1882) is in fun-damental respects unreliable. A one-page text published by Logar (1975: 63) is not very helpful.9

7. Tone

Most SCr and Sln dialects have tone contrasts. Unfortunately the difference between different tones is often slight, greatly complicating the job of discovering hitherto un-known tonal phenomena and giving rise to a lot of uncertainty and contradictory state-ments in the literature. It is not uncommon for dialects without contrastive tone to be described as if tone were contrastive, resulting in descriptions of daunting complexity, cf. on this Ivić (1959: 177), Steinhauer (1975: 24 and passim), Alexander 1975: 23). It also happens that investigators in approaching an unknown dialect with contrastive tone act as if they know beforehand how the tones sound, which can give rise to very con-fused accounts, see for some discussion Vermeer (1982: 304-309). On the other hand it is not rare for unusual tonal phenomena to be overlooked even by highly trained inves-tigators, cf. the example given by Stanonik (1977: |591| 296-297). Sometimes desperate steps have to be taken to determine whether or not a given dialect has contrastive tone (see Houtzagers 1982: 124-126). Occasionally even very experienced linguists have to acknowledge defeat, e.g. Rigler in the case of the retracted accents of Rožanski Slovene (see above, section 6.5). It is therefore likely that many sandhi phenomena involving tone have either not been reported at all or have been incorrectly described. I have se-lected three that seem to be reasonably well established.

7.1 SCr. (Turopoljski Kajkavian) pečȋ vs. pečĩ mi

Whereas numerous Sln and western SCr systems have retained the PSl tone contrast, quite a few of these dialects have lost it in word-final syllables, in which position all stressed long vowels are judged to be falling – in most cases – or rising, as in some Sln dialects).

In such systems the inherent tone of a given word form surfaces only if a clitic is present, e.g. pečȋ ‘stove’ (loc. sg.) vs. pečĩ mi (mi ‘I’, dat. sg. clitic) (Šojat 1982: 382-383 on Turopoljski Kajkavian). The mechanism is reminiscent of the types of sandhi described above, sections 4.1 and 6.2.

Unfortunately, though the phenomenon appears to be quite widespread, there are few if any satisfactory descriptions of it. The conditions for loss of tone are known to differ from dialect to dialect. On one hand there are systems where loss of tone takes place in sentence-final position only, so that a following word does not have to be a clitic in order to provoke the appearance of the underlying tone of a given form. On the

9 A tentative explanation by Ramovš starts from the observation that the dialects of the area tend to de-velop a secondary stress on the initial syllable; under certain conditions this secondary stress then turns into a new primary stress (Ramovš 1935: 15).

18

other hand there are systems where not even the presence of a clitic suffices for the original tone to surface, e.g. in Turopoljski Kajkavian as spoken in Trebarjevo Desno, according to Šojat (1981: 345). See further Rigler (1980), with references to some ear-lier literature on both SCr and Sln.

7.2 SCr. (Gruža) sȅstra vs. sèstra te

The Šumadija-Vojvodina dialect of Borač in Gruža (south of Belgrade) is Neo-Štokavian in most respects (see section 6.1). However, contrary to the normal Neo-Štokavian system, short rising vowels cannot occur in the penultimate syllable of words with a short vowel in the final syllable. Instead only falling vowels are found, so that it looks as if the tone contrast |592| is neutralized in the position involved. Example: sȅstra ‘sister’ instead of normal Neo-Štokavian sèstra (< sestrà by the Neo-Štokavian stress retraction), but gen. sg. sèstrē, where the short rising accent is maintained because of the long vowel in the final syllable, and sèstrama (dat./instr./loc. plur.) with a rising tone because it is not the penultimate syllable that is stressed (Stevović 1969: 420). The original town dialect of Belgrade has been reported to be similar (Belić 1929: 741).

One expects the original rising tone to reappear whenever a clitic is added. This does indeed happen in some cases. Stevović gives two types of examples:

(1) The nom. (sg. or plur.) of nouns in ‑a, e.g. sȅstra vs. sèstra te (te ‘you (sing.)’, acc. clitic) (Stevović 1969: 420).

(2) The present tense forms of the verbs meaning ‘be’ and ‘want/will/shall’, e.g. jȅsi ‘you are’ vs. jèsi li (li interrogative particle) (1969: 475).

This is all Stevović has to say about the matter, which is a pity, because there is evi-dently more to it than meets the eye:

A. In the first type of examples the alternation – in Stevović’s words – “can take place” (1969: 420), so presumably it is no longer automatic or obligatory.

B. In the second type of examples the falling tone has spread to forms with a disyllabic ending, where it would not strictly speaking be in place (cf. sèstrama), e.g. ȍćemo ‘we want/will/shall’ vs. òćemo li (Stevović 1969: 475, 622).10

7.3 SCr. (Novi) takȍ, pečȅ vs. takȏ j, pečẽ l

In some of SCr, vowels cannot be short if they are followed by a syllable-final reso-nant. The list of relevant resonants usually includes the nasal consonants (except some-times m), the laterals, r, j and often v. There are considerable local differences with respect to the positions in which neutralization is found and with respect to its present synchronic status.

The effect gives rise to sandhi lengthening whenever a word form ending in a short vowel is followed by a non-syllabic words consisting of a resonant that causes vowels to be lengthened in the system involved. Such words do not exist except where loss of

10 Alongside oćèmo li, with a different place of the stress (Stevović 1969: 475).

19

final vowels has given rise to them. That happens to be the case in Novi Vinodolski, where we find two, both very frequent:

− j (< je) ‘he/she/it is’, − l (< li ) interrogative particle.

What is intriguing here is not the fact of lengthening itself, which is predictable on the basis of the phonetic facts, but the choice of a tone: whereas short stressed vowels are neutral as to tone, long stressed vowels have to be either rising or falling. It so happens that j selects the falling tone and l the rising tone, e.g. njegȁ ‘him’ (gen./acc.) vs. njegȃ j, pečȅ ‘he/she/it is baking’ vs. pečẽ l (my material). Although the point is not discussed in Belić’s classical grammar of the dialect (1909), his texts contain a few examples:

− va pećȋ j ‘it is in the stove’ (p. 257); otherwise the loc. sg. of the group of nouns to which pȇć ‘stove’ belongs always ends in short ‑ĭ (e.g. p. 230);

− bi (auxiliary verb of the subjunctive mood, 3d.pers. sg. clitic) vs. bí l (Belić 1909: 252). |593|

8. Conclusions

Evaluation of the numerous sandhi alternations that have been found in SCr and Sln dialects is complicated by the absence of sufficient information on many points, e.g.:

(1) In several cases crucial facts of the tonal systems of the relevant dialects have not yet been elucidated (see 6.3, 6.5), which may seem strange, but which is only to be expected if one thinks of the difficulties an investigator faces when he wants to de-scribe a hitherto unknown accentual system (see the introduction to section 7).

(2) In those cases in which a given alternation has spread or dwindled, not enough in-formation is available for the process to be adequately reconstructed (e.g. 4.1, 5.1, 5.2, 6.1, 6.2).

(3) In those cases in which an alternation is optional, not enough attention has been paid to the factors that play a role in the choice the speakers make, including possi-ble semantic considerations.

It goes without saying that this is not intended as criticism of the SCr and Sln dialecto-logical traditions.

University of Leiden

Appendix

The systems discussed above are listed here more or less systematically, with refer-ences to the sections where they are mentioned. With minor changes dictated by the subject of the article, the terminology follows the one used by Ramovš (1935) for Sln, Ivić (1958) for SCr as a whole and Štokavian, Ivšić (1936) for Kajkavian, and Vermeer (1982) for Čakavian.

20

A. Slovenian

− Poljanski (2, 6.4). − Prekmurski (4.4 on Cankova) − Rožanski Carinthian (6.5 on Breznica pri Št. Jakobu). − Various normative systems (5.2).

B. Serbo-Croatian

B1 Marginal dialects without stress retraction

− Conservative Kajkavian (4.2 on Brezova near Začretje and Kajkavian as a whole). − Northwest Čakavian (3.2, 4.4, and 7.3 on Novi Vinodolski; 4.2 on Omišalj). − Conservative Southeast Čakavian (3.2 and 4.4 on Vrgada). − Conservative Zeta-Lovćen Štokavian (3.2 and 5.1 on Prčanj; 6.1 on Piperi).

B2 Intermediate systems with stress retractions only from final syllables

− Innovating Kajkavian (4.2 on Prodindol, Ozalj and Kajkavian in general; 6.2 on Prodindol and Turopoljski; 6.4 on Turopoljski; 7.1 on Turopoljski in general and as spoken in Trebarjevo Desno).

− Lika Central Čakavian (6.2 on Brinje and surroundings, and on Lešće). − Prvić South Čakavian (6.2) − Istrian Ikavian (6.2). − Posavina Štokavian (5.1 on Kostrč in the Bosanska Posavina; 6.2 on many locali-

ties). − Podravina Štokavian (6.3 on various locations, e.g. Valpovo, Baćin, Dušnok). − Innovating Zeta-Lovćen Štokavian (4.4 on Bjelopavlići; 5.1 on Vraka/Vrakë and

the dialects described by Pešikan (1965); 6.2 on various locations). − Kosovo-Resava Štokavian (6.2 on various locations, notably Kasidol, Levač, and

Landol). − Gallipoli (4.1, 6.2). − Torlak (5.1 on the area between Svrljig and Knjaževac).

B3 Mainstream Neo-Štokavian: general stress retraction from all syllables resulting in rising tones on preceding syllables

− Neo-Štokavian in general (4.4, 6.1). − Šumadija-Vojvodina (4.1 on various localities, notably Batajnica, Begeč, and In-đija; 6.1 on Ledinci in Srem as a representative of a Neo-Štokavian system).

− East Herzegovinian (5.1 and 6.1 on Zmijanje in western Bosnia; 6.2 on Piva & Drobnjak; 6.1 on Obadi in eastern Bosnia).

− Younger Ikavian (5.1 on Guber near Livno in southern Bosnia, shading into west-ern Herzegovina; 5.1 on Dobretići in central Bosnia; 6.1 on Central Dalmatia).

− Normative Neo-Štokavian (4.4, 5.1).

21

B4 Marginal Neo-Štokavian with stress retraction from all syllables not consistently resulting in rising tones on preceding syllables:

− Dubrovnik (5.1, 6.1). − Borač in Gruža (4.3, 5.1, 6.1, 7.2). − Belgrade (7.2).

B5 Systems with predictable stress

− Križevačko-Podravski Kajkavian (2). − Oštarije Central Čakavian (2).

References

Alexander, Ronelle 1975 Torlak accentuation. Munich: Sagner.

Baotić, Josip 1979 ‘Akcenatski sistem sela Kostrča u Bosanskoj Posavini’. Bosanskohercegovački dija-

lektološki zbornik 2, 161-267. Belić, Aleksandar

1909 ‘Zamětki po čakavskim govoram’. Izvěstija otdělenija russkago jazyka i slovesnosti 14 (2), 181-266.

1929 ‘Štokavski dijalekat’. In: St. Stanojević (ed.) Narodna enciklopedija srpsko-hrvatsko-slovenačka, vol. 4, Zagreb, 731-747.

Bogdanović, Nedeljko 1979 Govori Bučuma i Belog potoka. Beograd: Institut za srpsko-hrvatski jezik (= Srpski

dijalektološki zbornik 25). Bošković, Radosav

1966-1967 ‘Refleksi grupa tj, dj, tьj, dьj, stj, zdj, skj, zgj (ske, zge) u dijalektima južne i jugo-zapadne Istre’. Južnoslovenski filolog 27, 85-141.

Breznik, Anton 1916 Slovenska slovnica za srednje šole. Celovec/Klagenfurth: Tiskarna Družbe sv. Mo-

horja. Brozović, Dalibor

1981 ‘Dušnok (OLA 151)’. In: Ivić et al., eds., 603-610. Brozović, Dalibor – Dragomir Vujičić

1981 ‘Guber (OLA 46)’. In: Ivić et al., eds., 405-409. Ćerić, Snježana

1981 ‘Dobretići (OLA 48)’. In: Ivić et al., eds., 417-421. Ćupić, Drago

1977 Govor Bjelopavlića. Beograd: Institut za srpskohrvatski jezik (= Srpski dijalektološki zbornik 23).

Daničić, Đuro 1925 Srpski akcenti. Beograd: Srpska kraljevska akademija (= Posebna izdanja 58, Filo-

sofski i filološki spisi 16). Dešić, Milorad

1976 Zapadnobosanski ijekavski govori. Beograd: Institut za srpsko-hrvatski jezik (= Srpski dijalektološki zbornik 21).

Dybo, Vladimir Antonovič 1977 ‘Imennoe udarenie v srednebolgarskom i zakon Vasil’eva-Dolobko’. Slavjanskoe i bal-

kanskoe jazykoznanie, Moskva: Nauka, 189-272.

22

1981 Slavjanskaja akcentologija: opyt rekonstrukcii sistemy akcentnyx paradigm v prasla-vjanskom, Moskva: Nauka).

Ebeling, C. L. 1967a ‘Historical laws of Slavic accentuation’. In: To Honor Roman Jakobson, The Hague/

Paris: Mouton, 577-593. 1967b ‘Some premises of phonemic analysis’. Word 23, 122-137. 1978 Syntax and semantics, Leiden: Brill.

Fancev, Franjo 1907 ‘Beiträge zur serbokroatischen Dialektologie. Der kaj-Dialekt von Virje, mit Berück-

sichtigung der Dialekte Podravina’s (Koprivnica-Pitomača)’. Archiv für slavische Phi-lologie 29, 305-389.

Finka, Božidar – Slavko Pavešić 1968 ‘Rad na proučavanju čakavskoga govora u Brinju i okolici’. Rasprave Instituta za jezik

JAZU 1, 5-44. Finka, Božidar – Antun Šojat

1981 ‘Valpovo (OLA 41)’. In: Ivić et al., eds., 387-392. Hamm, Józef

1937 ‘Ze studjów nad akcentem serbskochorwackim. Kilka uwag metodologicznych’. Prace filologiczne 17, 66-75.

Hamm, Josip – Mate Hraste – Petar Guberina 1956 ‘Govor otoka Suska’. Hrvatski dijalektološki zbornik 1, 7-213.

Houtzagers, H. P. 1982 ‘Accentuation in a few dialects of the island of Cres’. Studies in Slavic and General

Linguistics 2, 117-130. Illich-Svitych – Vladislav Marković

1979 Nominal accentuatian in Baltic and Slavic. Cambridge, Mass./London: MIT Press. Ivić, Pavle

1957 O govoru galipoljskih Srba. Beograd: Institut za srpski jezik (= Srpski dijalektološki zbornik 12).

1958 Die serbokroatischen Dialekte. Ihre Struktur und Entwicklung 1: Allgemeines und die štokavische Dialektgruppe. The Hague: Mouton (= Slavistische drukken en herdrukken 18).

1959 Review of Hamm – Hraste – Guberina (1956). Zbornik za filologiju i lingvistiku 2, 171-183.

1961 ‘Prilozi proučavanju dijalekatske slike zapadne Hrvatske’. Godišnjak Filozofskog fa-kulteta u Novom Sadu 6, 191-211.

1961-1962 ‘Prilog rekonstrukciji dijalekatske slike srpskohrvatske jezičke oblasti’. Zbornik za filo-logiju i lingvistiku 4-5, 117-129.

1964 ‘O govoru ličkih čakavaca (okolina Otočca)’. Zbornik za filologiju i lingvistiku 7, 127-139.

1979 ‘Dve akcentološke novosti iz Srema’. Folia slavica 3 (1-2), 161-174. Ivić, Pavle, et al., eds.

1981 Fonološki opisi srpskohrvatskih/hrvatskosrpskih, slovenačkih i makedonskih govora obuhvaćenih Opšteslovenskim lingvističkim atlasom. Sarajevo: Akademija nauka i umjetnosti Bosne i Hercegovine (= Posebna izdanja 40).

Ivšić, Stjepan 1913 ‘Današńi posavski govor’. Rad JAZU 196, 124-254. 1936 ‘Jezik Hrvata kajkavaca’. Ljetopis JAZU 48, 47-88. 1937 ‘Osnovna hrvatska kajkavska akcentuacija u Pergošića (1574)’. Zbornik lingvističkih i

filoloških rasprava A. Beliću o 40-godišnjici njegova naučnog rada, Beograd, 183-195.

23

Junković, Zvonimir 1956 ‘Izvještaj o istraživanju kajkavskoga dijalekta u zagrebačkoj okolici’. Ljetopis JAZU

61, 391-395. 1972 Jezik Antuna Vramca i podrijetlo kajkavskoga dijalekta. Zagreb: Jugoslavenska akade-

mija znanosti i umjetnosti (= Rad JAZU 363). Jurišić, Blaž

1966 Rječnik govora otoka Vrgade 1: Uvod. Zagreb: Jugoslavenska akademija znanosti i umjetnosti.

1973 Rječnik govora otoka Vrgade 2: Rječnik. Zagreb: Jugoslavenska akademija znanosti i umjetnosti.

Klaić, A. 1936 ‘O podravskom akcentu i kvantitetu’. Južnoslovenski filolog 15, 181 -183.

Kortlandt, Frederik 1972 Modelling the phoneme: new trends in East European phonemic theory. The

Hague/Paris: Mouton (= Janua linguarum, Series Maior 68). 1973 ‘Optional features in contemporary Russian’. In: André van Holk, Dutch Contributions

to the Seventh International Congress of Slavists, The Hague/Paris: Mouton, 107-114. 1975 Slavic accentuation: a study in relative chronology. Lisse: Peter de Ridder. 1976 ‘The Slovene Neo-Circumflex’. The Slavonic and East European Review 54, 1-10. 1977 ‘Historical laws of Baltic accentuation’. Baltistica 13 (2), 319-329. 1978 ‘On the history of Slavic accentuation’. Zeitschrift für vergleichende Sprachforschung

92/1-2, 269-281. Kursar, Ante

1979 ‘O šepurinskom govoru (iz gramatike i čitanke)’. Čakavska rič 1979 (2), 3-49. Logar, Tine

1975 Slovenska narečja: besedila. Ljubljana: Mladinska knjiga (Knjižnica Kondor: izbrana dela iz domače in svetovne književnosti 154).

Lončarić, Mijo 1977 ‘Jagnjedovački govor (s osvrtom na pitanje kajkavskoga podravskog dijalekta)’. Hrvat-

ski dijalektološki zbornik 4, 179-262. Nikolić, Berislav M.

1964 Sremski govor. Beograd: Institut za srpskohrvatski jezik (= Srpski dijalektološki zbornik 14/2)

1966 ‘Mačvanski govor’. Srpski dijalektološki zbornik 16, 179-313. 1969 ‘Kolubarski govor’. Srpski dijalektološki zbornik 18, 1-71. 1970 Osnovi mlađe novoštokavske akcentuacije. Beograd: Institut za srpskohrvatski jezik (=

Biblioteka Južnoslovenskog filologa NS 1). Pável, Ágost

1909 A vashidegkúti szlovén nyelvjárás hangtana. Budapest: A magyar tudományos akadé-mia.

Pešikan, Mitar B. 1965 Starocrnogorski srednjokatunski i lješanski govori. Beograd: Institut za srpskohrvatski

jezik (= Srpski dijalektološki zbornik 15). Petrović, Dragoljub

1966 ‘Prilog poznavanju akcenatskog sistema u pješivačkom govoru’. Zbornik za filologiju i lingvistiku 9, 129-134.

1967 ‘O govoru Broćanca’. Godišnjak Filozofskog fakulteta u Novom Sadu 10, 231-238. 1973a O govoru Zmijanja. Novi Sad: Matica srpska. 1973b ‘Glavnije osobine akcenatskog sistema u govoru Vračana’. Zbornik za filologiju i ling-

vistiku 16 (2), 173-195.

24

Popović, Ivan 1968 Govor Gospođinaca u svetlosti bačkih govora kao celine, Beograd: Srpska akademija

nauka i umetnosti (= Posebna izdanja 175). Ramovš, Fran

1935 Historična gramatika slovenskega jezika 7: Dialekti. Ljubljana: Znanstveno društvo za humanistične vede v Ljubljani.

Rešetar, Milan 1900 Die serbokroatische Betonung südwestlicher Mundarten. Wien: Alfred Holder/Kaiser-

liche Akademie der Wissenschaften (= Schriften der Balkancommission, Linguistische Abteilung 1, Südslavische Dialektstudien 1.)

Rigler, Jakob 1977 ‘K problematiki daljšanja starega akuta’. Slavistična revija 25/KONG, 83-98. 1980 ‘Nekaj opažanj pri akutu na zadnjem zlogu v slovenščini’. Slavistična revija 28 (2),

219-222. Rožić, Vatroslav

1893 ‘Kajkavački dijalekat u Prigorju’. Rad JAZU 115, 68-136. 1894 ‘Kajkavački dijalekat u Prigorju (Svršetak)’. Rad JAZU 118, 55-115.

Scheinigg, J. 1881-1882 ‘Obraz rožanskega narečja na Koroškem’. Kres 1 and 2, in brief instalments, Celovec/

Klagenfurth. [Quoted on the basis of Rigler 1977: 89n.]. Sekereš, Stjepan

1974 ‘Govor Slavonske Podravine 1: Uvod’. Zbornik za filologiju i lingvistiku 17 (2), 125-166.

1977 ‘Akcenatske zone slavonskog dijalekta’. Zbornik za filologiju i lingvistiku 20 (1), 179-201.

Simić, Milorad 1978 ‘Govor sela Obadi u bosanskom Podrinju’. Srpski dijalektološki zbornik 24, 1-124.

Simić, Radoje 1972 ‘Levački govor’. Srpski dijalektološki zbornik 19, 1-586.

Šimundić, Mate 1971 Govor Imotske krajine i Bekije, Sarajevo: Akademija nauka i umjetnosti Bosne i Her-

cegovine (= Djela 49). Šojat, Antun

1981 ‘Trebarjevo Desno (OLA 35)’. In: Ivić et al., eds., 343-348. 1982 ‘Turopoljski govori’. Hrvatski dijalektološki zbornik 6, 317-493.

Stang, Christian S. 1957 Slavonic accentuation, Oslo: H.Aschehoug og Co. (W. Nygaard) (= Skrifter utgitt av

Det Norske Videnskaps-Akademi i Oslo, I. Hist.-Fil. Klasse 1957/3). Stanonik, Marija

1977 ‘Govor Žirovske kotline in njenega obrobja’. Slavistična revija 25, 293-308. Steinhauer, Hein

1975 ‘The phonemic system of the Susak dialect’. Linguistics 162, 17-39. Stevanović, Mihajlo S.

1940 ‘Sistem akcentuacije u piperskom govoru’. Srpski dijalektološki zbornik 10, 69-184. Stevović, Igrutin

1969 ‘Šumadijski govor u Gruži s osobitim osvrtom na akcente’. Srpski dijalektološki zbornik 18, 401-635.

Težak, Stjepko 1981 ‘Ozaljski govor’. Hrvatski dijalektološki zbornik 5, 203-428.

Toporišič, Jože 1976 Slovenska slovnica. Maribor: Založba Obzorja.

25

Valjavec, M. 1897 ‘Glavne tocke o naglasu kńiževne slovenštine’. Rad JAZU 132, 116-213.

Vermeer, Willem 1979 ‘Innovations in the kajkavian dialect of Bednja’. In Jan M. Meijer (ed.), Dutch Contri-

butions to the Eighth International Congress of Slavists, 347-381. 1980 ‘Die Konjugation in der nordwestčakavischen Mundart Omišaljs’. Studies in Slavic

and General Linguistics 1, 439-472. 1982 ‘On the principal sources for the study of čakavian dialects with neocircumflex in ad-

jectives and e‑presents’. Studies in Slavic and General Linguistics 2, 279-341. 1983 ‘The rise and fall of the kajkavian vowel system’. Studies in Slavic and General Lin-

guistics 3, 439-472. Vuk Stef. Karadžić

1852 Srpski rječnik, istumačen njemačkijem i latinskijem riječima. Vienna. Vuković, Jovan L.

1940 ‘Akcenat govora Pive i Drobnjaka’. Srpski dijalektološki zbornik 10, 185-417.