UDK 808.61/.63-4:807.1/.3-4 Willem R Vermeer Državna univerza v Leidnu
TRACES OF AN EARLY ROMANCE ISOGLOSS IN WESTERN BALKAN SLAVIC
Romunščina je od ostalih romanskih jezikov ločena z izoglosami, nastalimi, ko je bila odpravljena klasičnolatinska samoglasniška kolikost. Samoglasniški sestavi slovenskih in srbohrvaških narečij kažejo, da čakavska in štokavska narečna področja srbohrvaščine predpostavljajo substrat (jezikovno podlogo), ki je nadaljeval tip samoglasniškega sestava, ki je podlaga romunščine, medtem ko se samoglasniški sestavi, ki jih nadaljujeta kajkavščina in slovenščina, strukturno nanašajo na sestave, kakor jih med romanskimi jeziki najdemo drugje, npr. v furlanščini in veljotščini.
Rumanian is separated from the remaining Romance languages by an isogloss that arose at the stage when Classical Latin vowel quantity was eliminated. The vowel systems of Slovene and Serbo-Croat dialects show that the Cakavian and Štokavian dialect areas of Serbo-Croat presuppose a substratum that continued the type of vowel system that underlies Rumanian, whereas the vowel systems continued by Kajkavian Serbo-Croat and Slovene are structurally related to the systems found elsewhere, e.g. in Friulian and Vegliote.
1. Introduction.
Post-war research in the historical dialectology of Slovene and Serbo-Croat has considerably enhanced our insight into the disintegration of the Common Slavic vowel system in the area. The principal landmarks are Rigler's and Logar's work on Slovene (in particular Rigler 1963a, 1967, Logar 1963) and Ivié's publications on Serbo-Croat (in particular 1958a, 1959, 1968). Elsewhere I have proposed a number of modifications to Rigler's and Ivié's theories, intended to account for residual problems and to tie together Slovene and Serbo-Croat into a coherent whole (Vermeer 1979b, 1982a, 1983, 1987a).
In this contribution I would like to discuss the early Slovene and Serbo-Croat vowel systems in the light of what is known about the early Romance dialects which the speakers of Slavic encountered when, in the sixth and seventh century of our era, they arrived on the Balkans. There is a fairly general consensus that at that time, the present-day Slovene and Serbo-Croat linguistic area was to a considerable extent romanized.1 As I shall try to show, one of the few early isoglosses that are definitely known to have cut through early Balkan Romance has survived into present-day Serbo-Croat. Recognition of fact is interesting not only from the point of view of the early history of Romance, but it also points the way towards a solution of a few remaining problems connected with the history of the vowel systems of Serbo-Croat and Slovene.
1 Cf. Mihäescu 1978: 53-54, So 11 a 1980: 157-158, Desnickaja 1987: 30-31, all with copious references. Evidence of a period of Slavic-Romance bilingualism is present in the structure of Slovene and Serbo-Croat, roughly speaking in the same way that the structure of modern Russian betrays the former existence of Slavic-Finnish bilingualism. Although to my knowledge there is not a great deal of literature on the subject, the following statement by Ivič is typical of a feeling that is fairly general among scholars: »Mi danes možemo konstatovati da je glasovni sistem /.../ srpskohrvatskog jezika, iako on nije romanski, ipak bliži italijanskom glasovnom sistemu nego rumuński« (Ivič 1958c: 3).
2. Early Romance vowel systems.
As is well-known, Classical Latin possessed a contrastive distinction between long and short vowels (system Rl):
System Rl. Clasical Latin.
Hi	ulu
elë olö + au ae oe a/ci
Neither the distinction between long and short vowels, nor the diphthongs oe and ae survived as such into the Romance languages: everywhere these elements of the Latin vowel system coalesced at an early date into a single system of quantytati-vely neutral vowels. The pattern of mergers was different in different areas. Common to all systems are the loss of the distinction between long and short a and the loss of the diphthongs oe and ae. which merged with long ë and short e respectively. Three basic systems have been distinguished:
I.	Here and there in southern Italy, in the dialects of Sardinia, and presumably in some southern forms of Romance that are now extinct (Africa), all long vowels merged with their short counterparts, as shown in system R2 (Lausberg 1963: 147).
System R2. Proto-South-Romance.
i	u
e	o	+ au
a
II.	Here and there in southern Italy and in the dialects that were to develop into Rumanian, long vowels merged with their short counterparts only in the case of the pairs alä, olö and uh'i. In the case of the non-low front vowels Hi and elë, however, the pattern of mergers was different: long ë, rather than merging with short e, merged with short i, yielding a high-mid monophthong e which was distinct from the low-mid monophthong ę (< short e and the diphthong ae). The outcome was as shown in system R3. Since the product of the merger of o and ö developed in the same way as high-mid ę, it will in accordance with tradition be written as o for the sake of symmetry. For further details see Lausberg (1963:148).
System R3 Proto-Balkan-Romance.
/	u
e	o	+ au
?
a
III.	Elsewhere it was only in the case of the low vowel a/a that long and short counterparts merged. As in Proto-Balkan-Romance long ë and short /' merged into
a high-mid monophthong e whereas short e (and ae) yielded low-mid ę. In contradistinction to Proto-Balkan-Romance the opposite side of the system developed in parallel fashion: long ö and short u merged into a high-mid monophthong o, which remained distinct from the low-mid monophthong ç which was the reflex of Classical Latin short o. The outcome was a system of seven monophthongs (system R4). The system will referred to as the Proto-West-Romance vowel system.
System R4. Proto-West-Romance.
i	u
e	o + au
Ç Q a
Subsequently the low-mid vowels ç and ç of systems R3 and R4 were diphthongized in many Romance-speaking areas, including the dialects adjacent to Western Balkan Slavic (Rumanian, Friulian, Vegliote), yielding ie and ou, so that instead of R3 and R4 we find R5 and R6. For details and examples 1 refer to Lausberg (1963: 144-164), see also Hadlich (1965: 40) on Vegliote, and Francescato (1970: 16-18) on Friulian.2
System R5. Proto-Balkan-Romance after the diphthongization of ę (> ie).
i	u
ie	-
e o	+ au
a
System R6. Proto-West-Romance after the diphthongization of ę and ç (< ie, ou).
i	u
ie	ou
e o	+ au
a
Looking only at the surviving Romance languages it is of course impossible to tell where the isogloss between R5 and R6 ran just after it had arisen, because the arrival and the spead of Slavic on the Balkan peninsula caused a large section of Romance-speaking territory to be lost forever. Of the attested Romance dialects of the region Vegliote, Italian (apart from the south, which is not relevant in this connection) and Friulian continue R6, whereas Rumanian continues R5. The isogloss must have run somewhere on present-day Slovene or Serbo-Croat territory. As we shall see, it surfaces in Serbo-Croat.
; What is essential from the point of view of the present contribution is the rise of ie-like and но-like diphthongs out of earlier low-mid monophthongs. In many types of Romance, diphthongization was limited to certain positions. The literature on the subject is vast, see Wiiest (1979: 118-124) with references.
3. Early Western Balkan Slavic vowel systems.
The attested Slovene and Serbo-Croat vowel systems can be derived from system SI, which can be labelled "Common Western Balkan Slavic":
System SI. The Common Western Balkan Slavic vowel system.
i/i	и/п
ie/ië	-/-
e/ë з/з o/ö	+ ę/ę, ç/ç, r/r, III
a/ä
This system is the outcome of a series of innovations which affected an earlier Proto-Slavic system, of which the following have to be mentioned:
(1)	The number of nasal vowels, which at earlier stages of the development must have been greater than two (see in particular Kortlandt 1979 with references), has been reduced to two: ę/ę and ç>/ç>. In what was later to become Serbo-Croat (minus Istrian Čakavian and Lukovdol Kajkavian), ç/ç was raised and the nasal vowels were denasalized at a relatively early date. As Rigler used to point out (e.g. 1963a: 29, 1973: 120), the resemblance that has often been thought to exist between Kajkavian and Slovene is quite deceptive. In reality ç/ç was raised in all major dialect areas of Serbo-Croat (always excluding Istrian Čakavian and Lukovdol Kajkavian, which remained archaic).
(2)	Proto-Slavic ë, which originally must have been a low front vowel [ä], has been diphthongized and raised to a position between i/i and e/ë (>ie/Ië). On the Common Western Balkan Slavic character of this change, which was denied by Rigler (1963a: 27-29), but is implicit in Logar (1981: 29), see also Vermeer (1982a: 97-100).
(3)	The strong jers (primary as well as secondary) have merged in all positions, yielding з/з. Two early developments involving the jers differentiated Kajkavian from the remainder of Western Balkan Slavic:
-	Morpheme-initial *vt>- merged with *u- at a stage before the merger of b and Ђ and before rise of a contrastive distinction between strong and weak jers. This very early local difference within Western Balkan Slavic involves only a tiny number of lexical items and does not affect the properties of the vowel system as a whole (see further Vermeer 1979a: 363-365).
-	The reflex of the jers merged with that of *ë in most (if not all) positions. The evidence, which is not in all respects clear, seems to indicate that in Proto-Kajkavian long ä, which was not a very frequent sound to begin with, became even more rare or disappeared altogether (with local differences), whereas з became limited to unstressed position (see further Ivič 1968: 58 on long 5, Vermeer 1983: 444-448 on short a).
(4)	Loss of jers adjacent to tautosyllabic r and / has given rise to syllabic r/r and
Ш.
(5)	Proto-Slavic i and y have merged in all positions, yielding i/i. As is well-known, this must have happened relatively late, because the language attested in the Freising Fragments has not yet carried through this innovation in all positions.
Several of the early developments that took place in Western Balkan Slavic are shared with varieties of Slavic spoken to the north (Czecho-Slovak and beyond). This holds in particular for the diphthongization and raising of the reflex of *ë and for the raising of q/q (and subsequent denasalization), a change which failed to reach Slovene and a few adjacent areas of Serbo-Croat (Lukovdol Kajkavian, Istrian Čakavian). This shows that at the stage where these developments took place, it was possible not only for Western Balkan Slavic as a whole, but also for particular Western Balkan Slavic dialect areas to carry through innovations together with more northerly varieties of Slavic. The geographical continuity which made joint Western Balkan Slavic/Czecho-Slovak innovations possible was finally ruptured as a consequence of the arrival of the Hungarians around the year 900.
4. A Romance isogloss surfaces: raising of ö in the northwest.
Before going on I would like to draw attention to the resemblance between the diphthongization and raising of *ë in Slavic and the treatment of the low-mid vowels in the neighbouring Romance dialects (ę/ę > ie/uo)\ there may or may not have been a connection at this stage already.
Whatever may have been the case, structurally speaking the Western Balkan Slavic vowel system resembled in one important respect the Proto-Balkan-Romance system at the stage after the diphthongization of ę had taken place (system R5): both systems contained an isolated /e-like diphthong which was not counterbalanced by an MO-like diphthong on the opposite side of the system. In contradistinction to this, the Proto-West-Romance system, with its uo, was completely symmetrical.
In Western Balkan Slavic some systems turned out to tolerate the asymmetry, whereas in others long ö was diphthongized and raised to the same level as ië, turning into what will for the moment be written as do (cf. system S 2).
System S2. The Northwest.
Hi	ulu
ie/ië	-/üö
elë д/5 ol-	+ r/r, l/l (nasals)
a/ä
What is remarkable is not the fact that different kinds of Western Balkan Slavic reacted in different ways to the asymmetry, but the geographical distribution of both solutions: the asymmetry was retained in (roughly speaking) Čakavian and Štokavian Serbo-Croat, whereas it was eliminated by Kajkavian Serbo-Croat and Slovene. In other words: the Northwest adapted its systems to the Proto-West-Romance structure, whereas the Southeast retained the asymmetry in accordance with the Proto-Balkan-Romance system.
Although in the case of most modern dialects of Western Balkan Slavic it is a simple matter to determine from which of the two basic systems they have evolved, the relevant information has been lost here and there. Accordingly, three types of cases can be distinguished:
(1)	If the reflexes of long *ë and *<5 are treated in parallel fashion, we are dealing with a system that continues S2 and which is structurally identical to Proto-West-Romance.
(2)	if *ö goes together with *ë whereas *ë is treated in a different way, it is a continuation of the southeastern (Proto-Balkan-Romance) system we are dealing with.
(3)In	the case of systems with complete merger of long *ë and *e the difference between the two systems cannot be determinated.
The details require some discussion. We shall first take a brief look at the Northwest.
5. A glance at the Northwest.
In the Northwest, system S2 was modified in different ways in different places and soon considerable local differentiation developed. One innovation was common to the entire Northwest: everywhere the short diphthong ie was monophthongized: ie > ę. Since later developments show that the new monophthong remained the short counterpart of the reflex of ič, it is likely that the diphthongal character of ië, became subphonemic for a time: ië > /ë/, phonetically [ie] and the same thing can be expected to have happened to the reflex of *ö: üö > loi, phonetically [мој.
Local differences arose as a consequence of the different ways in which the remaining asymmetry of the system was eliminated.
In the western and central dialect areas of Slovene (with the exception of a small area in Istria), the following innovations took place:
(1)	Short vowels were lengthened in stressed non-final syllables: bràta > brâta. Short vowels affected by frra/a-lengthening merged with their long counterparts, with the single exception of short o, which had lacked a long counterpart since earlier long ö had been raised and diphthongized (> ö) and which accordingly turned into a new long ö.
(2)	Word-final short -o was raised (> o), providing long ö with a short counterpart (Rigler 1963b: 66f.).
(3)	Denasalization converted ç/ç and ç/ç into mid monophthongs elë and olö, which merged with the corresponding elements of the vowel system: elë (< PSI. el ë) and olö (< PS1. short o with or without brdfa-lengthening).
System S3. The Northwest I: the western and central dialects of Slovene.
Hi
u/u
ęlę [ie]
elë
olö [uo]
з/з alä
o/o
+ r/f, 111
For the elimination of S and the rise of local differences see further Rigler (1963a) and my discussion of Rigler's theory (Vermeer 1982a: 104-108).
There is tiny group of Slovene dialects spoken near Koper in Istria about which not much is known and which Rigler therefore refrains from incorporating into his
reconstruction (1963: 43, 47). The evidence, such as it is, shows that they cannot be derived from the system underlying the other western and central dialects (S3), because the reflex of long ç, which has yielded ö, has not merged with lengthened short o, which has yielded ü, merging instead with old long <5. This state of affairs can be derived from the common northwestern system (S2) on the assumption that short o turned into the short counterpart of uo at a stage before bra/a-lengthening took place and before the nasal feature was lost (for some further discussion see Vermeer 1982a: 108).3
Also in Istria, on the Croat side of the border, we find the highly specific dialects spoken around Buzet in northern Istria. Here the reflexes of o and ë have been treated in completely parallel fashion: when long, both o and ë have yielded high monophthongal reflexes ü and i respectively (in a large part of the area with secondary loss of cotrastive length: u/i); when short, the reflexes of o and ë are mid monophthongs o and e. The reflex of Proto-Slavic e and ę has been lowered to occupy a position opposite a. Hence, it is clear that these dialects, too, continue the northwestern vowel system. Somewhat more to the southeast, in the area near Boljun, we find the same system of short vowels, but in case of length ë has merged with e (instead of i, as in the dialects spoken around Buzet). For the facts I refer to Ivić (1961: 198, 207, 1963), for references to the literature and a brief discussion of the position of these dialects see also Vermeer (1982b: 318-322).
The same innovation which we find in a restricted area in Istria took place on a larger scale elsewhere: in the northern and eastern dialects areas of Slovene and in Kajkavian Serbo-Croat, short o was raised in all positions. As a consequence elë and a/a were reinterpreted as belonging on the same level. This change, which was first described by Ivic (1968: 58), can be referred to as »Ivic's vowel shift« (see further Vermeer 1982a: 109-110, 1983: 453^54, 1987: 249). The outcome was as shown in S4.
System S4. The Northwest II: the northern and eastern dialects of Slovene the Kajkavian dialects of Serbo-Croat.
Hi	ul ü
elë [ie]	з/о	olö [uo]	+ rlr, Ш, (nasals)
elë	al ä
As we have seen (section 3), Kajkavian and the relevant Slovene dialects started drifting apart at an early stage due to differences in the treatment of the nasals (where Kajkavian developed jointly with the remainder of Serbo-Croat) and the reflex of the jers (where Kajkavian went its own way). Hence, it is more realistic to treat both dialect complexes separately.
The northern and eastern dialects of Slovene can be derived from a system with nasal vowels which have retained their earlier mid values: system S5. for later
3 IfOrožen's description of the phonology of the Pomjan dialect is correct and representative of the Istrian Slovene dialect area as a whole, the crucial reflex (that of old short *o in stressed non-final syllables) is inconsistent: in some cases it has merged with *ö and in some with *ç (Orožen 1981: 105). The matter deserves further study.
developments see Rigler (1963) and my discussion of Rigler's theory (Vermeer 1982a: 108-118, 1987b: 249-255).
System S5. The northwest III: the system common to the northern and eastern
dialects of Slovene.
i/i	и/п
ç/ç [ie]	o/o	olö [uo]	+ rjr, IJJ, ç/ç, ç/ç.
e/ë	a/ä
As we have seen, Kajkavian innovated at an early stage by raising ç/ç and by eliminating 5 and з in a number of positions. Most of Kajkavian can be derived from a system that innovated beyond system S5 by denasalizing ç/ç and the reflexes ç/ ç (which had been raised ); ç/ç merged with e/ë, but a merger of the reflexes of ç/ç with those of и/п was prevented by the fact that и/п had previously been fronted. Fronting of и/п is a widespread phenomenon found in an area running from Čakavian Istria in the west to Štokavian Slavonia in the east, passing through the central and eastern dialect areas of Slovene.
When at a later stage vocalic Ifl was lost it was treated in the normal Serbo-Croat way: it turned into u, merging with the new и/п (< ç/ç). The outcome was system S6. For further details and discussion of Ivic's classical reconstruction (1968) I refer to Vermeer (1979, 1983: 451^68).
System S6. The Northwest IV: the Common Kajkavian vowel system.
i/i	ii/ii	и/п
e/ë [ie]	o/(o)	o/ô [uo]	+ r/r
e/ë	a/ä
6. The Southeast: principal solutions.
We now turn to those dialect areas that immediately continue system SI in the sense that they do not treat the reflexes of long ë and o in parallel fashion. By and large these areas correspond to the Čakavian and Štokavian dialect groups of Serbo-Croat. Below, the major solutions to be found here will be reviewed. Slavonia and some of the dialects now spoken in Rumania are problematical and will be discussed separately (section 8).
(1) In some ways one of the most archaic continuations of the southwestern system is exemplified by the familiar vowel system of the ijekavian norm of Serbo-Croat. Here short ie is reflected as a sequence je, whereas long îë has developed into the phonological unit that corresponds to ije in the standard orthography (on which see further Brozovič 1967). In some positions (notably before [j]) the reflex of short ë has merged with that of / (for detailed discussion of this important phenomenon see Ivič 1955-56: 106-116). On the opposite side of the system there is nothing that corresponds to these reflexes: long ö has persisted as a mid monophthong which is the counterpart of long ë from earlier ë and ç. As is well-known, systems fundamentally identical to the ijekavian standard are also widespread at the level of the dialects. Abstracting from areas which have received the
system as a consequence of recent migrations, such systems cover a large continuous area corresponding to Ivié's "East Hercegovina Al" and "Zeta Lovčen" units (1958b: 144-148, 202-226).
(2)	The westernmost fringe of the ijekavian area (again abstracting from the dislocating effects of recent migrations) differs in one important respect from the bulk of ijekavian: it shows evidence of the former existence of a system which had reintroduced symmetry by turning the reflex of vocalic / into the counterpart of the reflex of ë, or, speaking in terms of system SI, /// > uolüö. The language of old texts shows that the phenomenon has been on the retreat for quite some time now (on some phonological factors favouring loss of uo see Ivič 1959: 53). Evidence for the development / > uo has been found in three areas:
-	East Bosnian ijekavian, both as it was written in earlier centuries (e.g. by Divkovič), and, in remnants, in its present-day spoken form. This is Ivié's "East Herzegovina B" (1958b: 151-155, in particular 153), see also the discussion by Brozovič (1966: 135-137).
-	The old written language of Dubrovnik, i.e. the older stages of Ivié's "East Hercegovina AIII" (1958b: 148-150), see the summary of the facts in Rešetar (1952: 39-44).
-	The dialect of the island of Lastovo. Here not only short but also long ë has turned into a sequence consisting of j + (long or short) e. As a consequence the diphthong uo has lost its initial element, merging with the reflex of o (Oblak 1894: 431-435, Brozovič and Lisac 1981: 278-279).
(3)	To the northeast of the area where the archaic diphthongal value of *ë has persisted as an ijekavian reflex, we find systems that have innovated by carrying through monophthongization, but which are more archaic in having retained the reflex of ë as an independent phonological entity (a high-mid monophthong ę distinct from both i and e) in all positions, apart from those positions in which merger of short ë with i has taken place and which are by and large the same as the ones we find in the ijekavian systems. The most familiar example of such a "jatov-ski govor" (to borrow the convenient term coined by Petrovič 1982: 120) is the "Gallipoli" dialect (Ivič 1957: 42). Not so long ago it has turned out that a similar system is quite common in western Serbia in an area that was formerly thought to be ekavian in the Šumadija—Vojvodina and Smederevo—Vršac way (see further Remetić 1981, in particular 16, 105).
(4)	In Ivié's "Šumadija-Vojvodina" and "Smederevo-Vršac" dialects ë and e have merged in all positions apart from those cases that have merger of short ë and i in all systems treated so far ((1), (2) and (3)). For the latter phenomenon there is no counterpart on the opposite side of the system, which again betrays the previous status of ë as an isolated unit between i and e (see further Ivic 1958b: 167-188, 238-241). It is evident that Gallipoli-like systems have been on the retreat for a time, moving towards merger of ę and e, in other words, that "jatovski govori" have been changing into ekavian systems of type (4). See on this Ivič (1957: 435-436).
(5)	In moving further to the east we encounter systems with unconditional merger of e and ë in all positions, including those where we find i < ë elsewhere:
Kosovo-Resava proper and Torlak. Since these systems are neutral with respect to the difference between the two early Romance systems, it cannot be determined on the basis of the facts of the vowel system alone which system they continue. On the basis of the geographical facts, however, it is pretty obvious that they continue the southeastern system.
(6)	To the west and northwest of the ijekavian areas (again discounting the effects of the migrations) complete merger of ë with i is widespread. Today such ikavian systems occur in widely different areas: the neoštokavian Younger Ikavian dialect group (Ivic 1958b: 188-202), the ikavian dialects of Istria (including Klana near Kastav) and the related dialects of Molise in Italy (Ivic 1958b: 248-269, on Klana see Ivič 1961-2: 119n, Rigler 1963b: 13), the familiar ikavian Čakavian dialects of Dalmatia, the "Kajkavian" ikavian dialects spoken to the west and south of Zagreb (Ivšič 1936: 74-79), the dialects of the southern Burgenland (Neweklow-sky 1978: 126-186), Tribuče near Črnomelj in Slovenia (Logar 1958: 155). The present-day scattered distribution of ikavian systems is undoubtedly the consequence of recent migrations. Prior to the arrival of the Turks, the ikavian systems must have constituted a single large block, several sections of which now seem to be missing (for suggestions about the location and the principal characteristics of the missing parts see Brozovič 1963: 52-54).
(7)	To the west of the areas with an ikavian reflex we find systems that are ikavian as well, but have merger of ë and e in one important group of positions: before dental consonants (t, d, n, I, r) which are not followed in turn by a front vowel. The rule determining the distribution of i and e in these systems was discovered by Jakubinskij (1925). As in the case of ikavikan, the i/e-kavian systems have been scattered by migrations and are now found not only in the Croat interior, along the coastline and on the islands, where their presence is presumably old, but also in Istria, the Burgenland, the surroundings of Zagreb and other areas where they have arrived recently (for more precise locations and references see further Ivič 1961-2: 119f., Vermeer 1982b: 293-295, 297-302).
(8)	To the west and north of the i/e-kavian systems we find a small group of Čakavian dialects with a purely ekavian reflex, exactly as in Kosovo-Resava Štokavian (proper) and Torlak. This reflex is characteristic of a restricted area comprising the Čakavian dialects of the Hrvatsko Primorje from Bakar northward (possibly including Crikvenica on i/e-kavian territory), the northeastern coastline of Istria, the island of Cres and the northern part of Lošinj, areas in the interior of Istria around Labin, Pazin, Žminj and possibly Boljun (for references and further details see Vermeer 1982b: 309-318).4 Since a purely ekavian reflex is neutral as to the difference between the two early Romance vowel systems, it has to be determined on other grounds which system they continue. Travelling northward in Istria we encounter dialects which seem to continue the northwestern system, see above, section 5.
4 A responsible evaluation of the historical affinities of these dialects is still impossible because reliable full-scale grammars of the Istrian dialects are lacking. The type of ekavian Čakavian spoken on Cres/Lošinj has recently become much more accessible due to Houtzagers' work (in particular 1984-85, 1985).
7. Southeastern patterns.
In the different solutions found in the Southeast several patterns can be discerned.
To begin with, it is striking that the most archaic systems (the ones listed as (1), (2) and (3) occupy a relatively central position. This fact indicates that the most important innovations originated in marginal dialects, which suggests in turn that contact with neighbouring systems played a role.
Furthermore, it is significant that retention of the asymmetry of the original southeastern system seems to be limited to dialects spoken by the Orthodox population. All Catholic areas have eliminated the asymmetry either by mergers or, in the case of those that retained the ijekavian reflex of the ë, by the rise of uo from syllabic /, a development for which there is evidence in all three areas that are both ijekavian and Catholic: East Bosnia, Dubrovnik and Lastovo. The secondary loss of uo that is attested in both Bosnia and Dubrovnik is irrelevant in this context because it is known to be relatively recent.
In other words: being Catholic seems to have predisposed carriers of the southeastern systems towards restoration of symmetry, a fact that seems rather extraordinary until one realizes that being Catholic implies a western orientation, involving contact with Romance dialects of the type that continue the Proto-West-Romance vowel system. If Vegliote is representative, the "Dalmatian" dialects of Romance spoken along the coastline and on the islands continued the Proto-West-Romance vowel system.
Finally, complete merger of ë and e is found at both ends of the southeastern continuum in areas where contact with carriers of basically different systems arose:
(1)	In the west, the merger eliminated the difference between on the one hand the southeastern systems underlying the i/e-kavian dialects of Krk and the Hrvatsko Primorje, and on the other the northwestern system as continued by the dialects spoken in Istria around Buzet and Boljun.
(2)	In the east (Kosovo-Resava in a narrow sense and Torlak) the merger eliminated an important structural difference between the Western Balkan Slavic vowel system (which had developed Proto-Slavic ë into an ze-like diphthong) and the Eastern Balkan Slavic (Bulgarian and Macedonian) system, which had retained ë as a low front vowel intermediate between e and a: *[ä]. The merger, which also took place in adjacent Macedonian and Bulgarian areas, eliminated a linguistic difference that must have been very striking to the speakers at the stage when contact between both branches of South Slavic was being reestablished in the course of the Middle Ages (on which see the extended discussion by Ivic 1986: 23-31).
Merger of ë and e is a simplification and as such it is likely to spread beyond the areas where it first arose. The different ways in which the merger spread are symptomatic of the differences between the east and the west: in the west, the Catholic population was rapidly eliminating the asymmetry of the original southeastern system. Hence the merger soon arrived in areas that had already eliminated the ë in a different way, so that it was prevented from going any further. In the east, on the other hand, where contact with speakers of Eastern Balkan Slavic was the
only factor endangering the independence of ë, the merger had no competitors and it could therefore proceed at a leisurely pace, without completely running out of fuel to the present day, as Ivič and Remetić have shown.
8. Residual problems: Slavonia, Rekaš, Banatska Črna gora, Karaševci.
The dialects of Slavonia require a separate treatment, which would have to take account of all local differences, not all of which have been adequately described. The most important points seem to be the following:
(1)	Most of the dialects of the Posavina can be most easily derived from a southeastern system. They have carried through a merger of short ë with i in the same environments we find in so many other Štokavian systems (the ones listed as (1) through (4) in section 6). In all other positions they show merger of long ë with i and merger of short ë with either i (as in normal ikavian dialects) or the sequence je (as in all ijekavian systems). Retention of ë as a separate unit with the position it has in the jatovski govori has however been reported in Gradište (Ivič 1959: 42, Finka and Sojat 1981: 443). Investigators disagree about the phonological status of the Gradište reflex of ë: according to Ivič it can be interpreted as a sequence /ej/; according to Finka and Šojat it is phonologically a monophthong which allows of ei-like and /e-like diphthongal realizations and which is only optionally distinct from the reflexes of *e and *i.5
(2)	On the other hand several properties of the Posavian vowel system that are found only in older written material or in the most archaic local dialects seem to presuppose the former existence of a Kajkavian-like system of the shape mentioned above (system S6), on which see Ivič (1958b: 304, cf. also Vermeer 1983: 463-468).
(3)	As for the Podravina dialects, they seem by and large to lack the otherwise widespread ikavian reflex of short *ë followed by [j]; until relatively recently they were generally assumed to be "purely" ekavian. However, among these dialects, too, systems with a distinct reflex of *ë have been found (see, e.g. Ivič 1961-62: 123 on Dušnok in Hungary). As in, say, Kajkavian, and differently from what we find in Gradište or in the jatovski govori of Serbia, the reflex of *ë is here a mid vowel directly opposite o, whereas the reflex of e and ę has turned into a low vowel on a level with a. In other words: the system has carried through what I have called "Ivié's vowel shift" and may have to be derived from a northwestern system.
At present the basis for a reliable evaluation of these facts is lacking. As it is, it seems likely that Posavian Štokavian basically continues a southeastern system which at a relatively early stage interacted with its original western neighbour: Kajkavian. At a later stage, contact with Kajkavian, in itself quite natural in view of the geographical facts, was ruptured for well-known historical reasons. Finally, after the departure of the Turks, a massive influx of carriers of Neoštokavian dialects caused the Posavian vowel systems to revert to their original southeastern shape.
5 In their earlier publication on the subject (1973: 7) Finka and Šojat do not mention ie-like realizations.
The striking retention of a distinct value of *ë in Gradište allows of several interpretations. Assuming for the moment that Posavian continues a basically southeastern system, it is not strange that elimination of the southeastern asymmetry, though otherwise general among Catholics, failed to reach an area that is so far removed from the coast as the eastern Posavina. Moreover, a distinct value of *é may have been protected for a time by the Kajkavian-like shape of the vowel system for which there is evidence among the most archaic Posavian dialects and which was later eliminated almost completely.
Rather similar problems are connected with a few important dialects spoken in what is now Rumania by the Catholics of Rekaš, the Orthodox of the Banatska Črna Gora and the Karaševci.
In these dialects the reflex of PSI. *è has been retained as a monophthong which is distinct from both *e and *< and which in the vowel system occupies a position opposite o, as in Kajkavian, Dušnok and Istrian Čakavian, but unlike the Galiipoli dialect and the jatovski govori of Serbia. In other words: these dialects have carried through Ivic's vowel shift. For the time being it is unclear how we are to interpret this. On the one hand it is conceivable that these dialects carried through Ivié's vowel shift independently on the basis of southeastern vowel systems; on the other is it possible that they carried through the shift jointly with more westerly dialects (Kajkavian, Podravina Štokavian and the lost autochthonous dialects of the Vojvodina). The latter development would seem to be in accordance with Ivic's conception of the rise of these dialects (1958b: 280-283).
9. Conclusions.
In this contribution I have tried to provide an answer to a few questions that are bound to strike one the first time one looks at a dialect map of Serbo-Croat, such as the following. Why did the raising of ö > uo (i.e. the transition from SI to S2) fail to penetrate into Čakavian and Štokavian once it had taken place in Kajkavian? Why is it that the only hard evidence for the development of syllabic / into uo has been found in ijekavian dialects spoken by Catholics (despite the fact that the majority of ijekavian speakers are Orthodox)? Why is it that the area where syllabic / yielded uo has such a strange elongated shape? Why is it that purely ekavian systems are located on the one hand around Rijeka and on the other in southern and eastern Serbia? The answers to these questions turn out to involve the known properties of the vowel systems of the Romance-speaking population with which the Slavs that settled the western Balkans came into contact.
It is significant that Štokavian and Čakavian Serbo-Croat, though continuing the early Romance vowel system that underlies Rumanian, show no trace of the innovations that moved Rumanian beyond system R3, such as the rise of umlaut and the reduction of unstressed vowels, developments which have clear counterparts in Albanian and Bulgarian.
The results of research into historical dialectology easily lend themselves to misunderstanding by laymen who would like to project back present-day linguistic units into a past where they make no sense. Many of the time-honoured problems of Serbo-Croat historical dialectology are not primarily dialectological, but have to
do with the emotional value of words like Čakavian, Kajkavian, and Štokavian: laymen would like to believe that these words refer to clearly defined units with clearly defined geographical boundaries and clearly defined separate histories; they would also very much like to be told that in the past their preferred unit was bigger than it is now. Of course, at least since Jagic's times the most perceptive and responsible scholars have always realized that the truth is at the same time more prosaic and more complex. Jakob Rigler expressed it in the following words:
»Južnoslovanski jeziki se niso formirali na tak način, da bi se razvijali ločeno. Že v začetnem obdobju najbrž ni bilo trdnih samostojnih enot, ampak so se ob naselitvi najbrž pomešali razni elementi. Posamezne izoglose, ki se ne pokrivajo, niso bile tako izrazite, da bi že same izločevale jezike. Kake rasne čistosti v tem smislu ne moremo iskati ne pri slovenščini in ne pri srbohrvaščini. Niti kakih prajezikov najbrž ne moremo nastavljati. Sedanji južnoslovanski jeziki so produkt celotnega sklopa jezikovnih procesov, zlasti pa političnih in kulturnih dejavnikov, ki so delovali v njihovi zgodovini« (Rigler 1977: 36).
References
Brozovič, Dalibor:
1963 O rekonstrukciji predmigracionog mozaika hrvatskosrpskih dijalekata, Filologija 4, 45-55.
1966	O problemu ijekavskoščakavskog (istočnobosanskog) dijalekta. Hrvatski Dijalektološki Zbornik 2, 119-208.
1967	Some remarks on distinctive features, especially in Standard Serbo-Croatian, To Honor Roman Jakobson, The Hague/Paris: Mouton, 412-426.
Brozovič, Dalibor and Josip Lisac
1981 Lastovo (OLA 56), in: Ivič et al. (1981), 275-280. Desnickaja, A. V.
1987 K izučeniju latinskix elementov v leksike balkanskix jazykov, in: Romano-Balcanica (Voprosy adaptacii latinskogo jazykovogo èlementa v balkanskom areale), Leningrad: Nauka, 3-36. Finka, Božidar and Antun Šojat
1973 O slavonskom dijalektu ekavskog izgovora u okolici Vinkovaca, Rasprave Instituta za jezik 2, 7-19.
1981 Gradište (OLA 52), in: Ivič et al. (1981), 443-447. Francescato, Giuseppe
1970 Studi linguistici sul friulano, Firenze: Leo S. Olschki. Had lieh , Roger L.
1965 The Phonological History of Vegliote, Chapel Hill: The University of North Carolina Press ( = University of North Carolina Studies in the Romance Languages and Literatures 52). Houtzagers, H. P.
1984-85 Vowel systems of the ekavian dialects spoken on Cres and Lošinj, Zbornik za filologiju i lingvistiku 27-28, 885-893.
1985 The Čakavian dialect of Orlec on the island of Cres. Amsterdam: Rodopi (= Studies in Slavic and General Linguistics 5). Ivič, Pavle
1955-56 O nekim problemima naše istorijske dijalektologije, Južnoslovenski filolog 21, 97-129. 1957 O govoru Galipoljskih Srba, Beograd (= Srpski Dijalektološki Zbornik 12). 1958a Osnovnye puti razvitija serboxorvatskogo vokalizma, Voprosy jazykoznanija 7/1, 3-20. 1958b Die serbokroatischen Dialekte, 's Gravenhage: Mouton.
1958c Kakovy osnovnye zadači i problemy tipologii slavjanskix jazykov?, Južnoslovenski Filolog 22 dodatak, 3-5.
1959 Der Vokal č als lebendiges Phonem in den serbokroatischen Mundarten, International Journal of Slavic Linguistics and Poetics 1/2, 38-53.
1961 Priloži poznavanju dijalekatske slike zapadnę Hrvatske, Godišnjak Filozofskog fakulteta u Novom Sadu 6, 191-211.
1961-62 Prilog rekonstrukciji predmigracione dijalekatske slike srpskohrvatske jezičke oblasti. Zbornik za filologiju i lingvistiku 4-5, 117-130.
1963 Parelele poljskome "pochylenie" na srpskohrvatskom terenu, Studia linguistica in honorem Thaddaei Lehr-Sptawiński, Kraków.
1968 Procesi rasterečenja vokalskog sistema u kajkavskim govorima. Zbornik za filologiju i lingvistiku 11, 57-68.
1986 Srpski narod i njegov jezik, drugo izdanje, Beograd: Srpska književna zadruga. Ivič, P. et al.
1981 Fonološki opisi srpskohrvatskih/hrvatskosrpskih, slovenačkih i makedonskih govora obuhva-čenih Opšteslovenskim lingvističkim atlasom, Sarajevo (= Posebna izdanja ANUBiH 55), 29-33. Ivšič, Stjepan
1936 Jezik Hrvata kajkavaca, Ljetopis JAZU 48 (za godinu 1934/35), 47-88. Jakubinskij, L.
1925 Die Vertretung des urslav. č im Čakavischen, Zeitschrift für slavische Philologie 1, 381-396. Kortlandt, F. H. H.
1979	On the history of the Slavic nasal vowels. Indogermanische Forschungen 84, 259-272. Lausberg, Heinrich
1963 Romanische Sprachwissenschaft I: Einleitung und Vokalismus, zweite, durchgesehene Auflage, Berlin: Walter de Gruyter & Co. (= Sammlung Göschen 128/128a). Logar, Tine
1958 Dialektološke študije X: belokranjski govori. Slavistična Revija 1 1, 145-155. 1963 Sistemi dolgih vokalnih fonemov v slovenskih narečjih. Slavistična Revija 14, 111-132. 1981 Izhodiščni splošnoslovenski fonološki sistem, in: Ivič et al. (1981), 29-33. Neweklowsky, Gerhard
1978 Die kroatischen Dialekte des Burgenlandes und der angrenzenden Gebiete. Wien: Verlag der Österreichischen Akademie der Wissenschaften (= Schriften der Balkankommission 25). Oblak, Vatroslav
1894 Der Dialekt von Lastovo, Archiv für slavische Philologie 16, 426-450. Orožen, Martina
1981	Pomjan (OLA 11), in: Ivič et al. (1981), 103-109. Petrovič, Dragoljub
1982	O problemu jatovskih govora u severnoj Srbiji, Zbornik za filologiju i lingvistiku, 25/2, 113-128.
Remet ič, Slobodan
1981 O nezamenjenom jatu i ikavizmima u govorima severozapadne Srbije, Srpski Dijalektološki Zbornik 27, 7-105. Rešetar, Milan
1952 Najstariji dubrovački govor, Beograd (= Posebna izdanja SAN 192). Rigler, Jakob
1963a Pregled osnovnih razvojnih etap v slovenskem vokalizmu. Slavistična Revija 14, 25-78. 1963b Južnonotranjski govori, Ljubljana.
1967 Pripombe k Pregledu osnovnih razvojnih etap v slovenskem vokalizmu, Slavistična Revija 15, 129-151.
1973 Smeri glasovnega razvoja v panonskih govorih. Študije o jeziku in slovstvu, Murska Sobota: Pomurska Založba, 113-128.
1977 O slovensko-kajkavskih jezikovnih razmerjih, XIII. seminar slovenskega jezika, literature in kulture. Zbornik predavanj, Ljubljana 29-38. Sol t a , Georg Renatus
1980	Einführung in die Balkanlinguistik mit besonderer Berücksichtigung des Substrats und des Balkanlateinischen, Darmstadt: Wissenschaftliche Buchgesellschaft.
Vermeer, Willem
1979a Innovations in the kajkavian dialect of Bcdnja, in: Jan M. Meijer (ed), Dutch Contributions to the Eighth International Congress of Slavists, Lisse: Peter de Ridder, 1979, 347-381.
1979b Proto-Slavonic *u in kajkavian, Zbornik za filologiju i lingvistiku 22/1, 1979, 171-177.
1982a Raising of *ë and loss of the nasal feature in Slovene, Zbornik za filologiju i lingvistiku 25/1,
1982,	97-120.
1982b On the principal sources for the study of Čakavian dialects with neocircumflex in adjectives and e-presents, Studies in Slavic and General Linguistics 2, 279-340.
1983 The rise and fall of the kajkavian vowel system, Studies in Slavic and General Linguistics 3,
1983,	439-477.
1987a Rekonstruiranje razvoja samoglasnikih sestavov v rezijanskih govorih, Slavistična-Revija 35/ 3-, 1987, 237-257.
1987b Further evidence of fronted reflexes of PSI. *u in kajkavian. Zbornik za filologiju i lingvistiku 30/2, 1987, 99-100. Wiiest, Jakob
1979 La dialectalisation de la Gallo-Romania: Problèmes phonologiques, Berne: Francke.