Grant H. Lundberg Brigham Young University Typology of Tone Loss in Haloze, Slovenia: An Acoustic and Autosegmental Analysis The paper presents an acoustical analysis of pitch contour on accented syllables in Haloze micro- dialects and a comparative examination of other instrumental studies of pitch in selected South Slavic dialects. Based on these analyses, an attempt is made to place Haloze within a typology of tone loss which functions on autosegmental principles and is motivated by language contact. Prispevek predstavlja akustično analizo tonskega poteka v naglašenih zlogih ter primerjalno obravnavo drugih instrumentalnih raziskav v nekaterih južnoslovanskih narečjih. Na osnovi teh raziskav avtor skuša uvrstiti haloške govore v tipologijo izgube tonemskosti, ki temelji na avto- segmentnih načelih in je pogojena z jezikovnim stikom. Introduction This article is a theoretical reexamination of an instrumental study of the pitch contour on accented syllables in the dialects of Haloze, a small geographical region on the eastern border of Slovenia's historical province of Styria (see Figure 1 below). Based on the data from this instrumental study, the contemporary situation in Haloze will be explained in terms of a typology of tone loss, which is motivated largely by language-contact phenomena.1 Dialect information from this region is especially important because of the location of Slovenia in general and of Haloze in particular. The Slovene standard language and Slovene dialects are located at an important and central position in Europe. The Slovene speech area lies where Slavic, Romance, Germanic and Finno- Ugric, all of the major language families of Europe, meet. It is an ideal laboratory for observing the way that languages, some of similar and others of divergent genet- ic backgrounds, behave in contact situations. In addition to the interesting position of Slovenia in general, Haloze is located on the periphery of the Slovene-speaking world, a position where, typologically, one expects to find archaisms. Haloze is thus potentially a very rich area in terms of dialectology and historical linguistics. It is well known that at some point in its early development Common Slovene had word-level tonal distinctions. In fact, several important dialect groups stilI retain these oppositions. (For overviews of prosodic configurations from typological and di- achronic viewpoints, see Greenberg 1987 for Slovene; Lončaric 1979 for Kajkavian; Ivic 1966 for the Serbo-Croatian territory in the broad sense.) This is not true of the Styrian and Pannonian dialects of Slovene, and most discussions of the development of Slovene vocalic systems are clear on this point. For example, Logar states what 1 The fieldwork that gave rise to this study was supported by fellowships from IREX and Fulbright-Hays. 1 would also like to thank the anonymous reviewers of this paper for their helpful criticism and insights. The instrumental study referred to above first appeared as Lundberg 2001. 170 Slovenski jezik - Slovene Linguistic Studies 3 (2001) has be en said by others before him (Ramovš 1935; Rigler 1963; Leneek 1982): "Tako danes noben štajerski govor ne loči rastočih in padajočih naglasov niti v dolgih niti v kratkih zlogih. Tonski potek je tod en sam, namreč padajoč." Thus today no Sty- rian dialeet distinguishes between rising and falling aeeents, neither on long or short syllables. The tonal eontour is only one, namely falling' (1996: 395). This statement also expresses the eommonly held view that when tonal distinetions were lost in Styrian dialeets the falling tone was generalized to all positions. What exaetly is meant by "falling" is not elear in terms of phonetic realization. This term is of ten used interehangeably with "cireumflex" or "high" tone. It is not elear if this refers to piteh eontour , height or, more likely , default stress. These ambiguities will be eonsidered in more detail later. Figure 1. Dialect map of Slovenia ,-- .. ~I ... ~ "\ .L. ' '" , , , AUSTRIA (~~~ .. _/t·:.v"_' __ CROATIA A Carinthia BLittoral C Rovte D Lower Carniola E Upper Carniola F Styria G Pannonia Logar's eomments on prosody in Styrian dialeets are of partieular interest here beeause Haloze is generally eonsidered to be a part of this dialeet group? If these statements are extended to all dialeets in this group, then we must eonc1ude that there are no phonemie tonal distinetions in Haloze and that the unmarked word aeeent is "falling." The problem with extending the general Styrian dialeet eharae- teristies to Haloze is that these generalizations would have to be made without the benefit of reliable data from Haloze and in eontradietion to several statements in the dialeet literature. Ramovš, Rigler, Leneek and probably Logar did not have eonc1us- ive information about the prosodie systems in Haloze beeause they have not yet been thoroughly deseribed. Some new information based on the author's fieldwork in Slovenia will be offered below. (For a more detailed explanation of the voealie sys- tems of Haloze and the dialeet group to whieh it belongs see Lundberg 1999.) 2 There has been some debate regarding the dialect group to which Haloze belongs. See Ramovš 1935 and Rigler 1986. G. Lundberg, Typology of Tone Loss in Haloze, Slovenia: An Acoustic and Autosegmental Analysis 171 Prosody The vocalic system of eastern Haloze The vocalic systems of eastern Haloze make distinctions of word-Ievel prosody in the accented syllable only. This accent is free to fall on any syllable of the word, and its placement is phonemic, ki}sti 'bone' (loc. sg. fem.), kost? 'bone' (nom. pl. fem.) (Meje). These systems have not experienced acute relengthening, often called brata- lengthening, a process which is common to many Slovene dialects. In this sen se they are different form the Styrian dialects to the west and north of Haloze. Finally, the vocalic systems of eastern Haloze are strictly monophthongal, and distinctive quanti- ty has been retained, br{Jl 'brother', brat 'to pick' (supine) (Gorenjski Vrh).3 Long stressed Short stressed u u 9 9 e e a The vocalic system of central Haloze Like eastern Haloze, central Haloze has a vocalic system in which word-Ievel prosodic distinctions are realized only in the accented syllable. This system has dis- tinctive quantity, br{Jl 'brother', brat (sup.) 'to pick', although that quantity, stiH distinctive, carries less functional load in central Haloze because it is almost always accompanied by a quality distincti~n in the form of a diphthong, hltiiti 'to hurry', hitfin '1 hurry', #lati 'to work', diijlaš 'you work' (Belavšek). The place of the ictus is free, so this is also phonemic in central Haloze, k{}!fsti 'bone' (dat. sg.), kQstfi (gen. sg.). Long stressed Short stressed u u 9 a (~) 3 All examples, unless otherwise stated, are from the author's notes and recording s while in the field. 172 Slovenski jezik - Slovene Linguistie Studies 3 (2001) The vocalic system of western Haloze The vocalic systems of western Haloze are in most respects the same as those found in central Haloze. One important difference is that as a result of bra- ta-Iengthening and other processes all stressed vowels became long, and thus pho- nemic length distinctions were lost. This has also produced two new diphthongs, že and uo, which resulted from the lengthening of short *e and *<1 (> že) and *0 and *Q (> uo) respectively.4 Phonetically, it is not clear whether all stressed vowels are stiH long in these dialects. li u ie uo tt 9 e o ay r a Tone in Haloze To the extent that the dialect area of Haloze has be en described, the question of tone has not been dealt with adequately. Those who have conducted field work in Haloze have made comments on the topic of distinctive tonal contrasts, but these limited statements have not conclusively disposed of the issue. Kolarič, in his 1964 dialect description, states in a footnote that all word ac- cents in Haloze are falling (398), but there appear to be some gaps in his dialect data which bring the validity of this statement into question. Based on my own field work in central and eastern Haloze and judging from the examples of the possible reflexes of Common Slavic phonemes given by Kolarič, it does not appear that he had representative information on the vocalic systems of eastern Haloze. This state- ment is bas ed largelyon the fact that he does not mention a monophthong as a po s- sible reflex of Common Slavic *e and *Q, both of which are realized as mono- phthongs in eastern Haloze. He also fails to mention the merger of the *e and the *<1. This is a significant gap in his description because, as is evident from an exam- ination of the vocalic systems of eastern Haloze, eastern Haloze is separated from the rest of Haloze by several important isoglosses, therefore these dialects could also be different as regards tone (Lundberg 1999). A1though Kolarič states that there are no tonal distinctions in Haloze, he in- cludes a quote from J. Pajek printed in Zora in 1875 which brings up the problem of intonation and seems to contradict what was said earlier about word-Ievel pitch in Haloze. "Človek urnega jezika je Haložan; poskakuje namreč pri vsakej tret jej ali četrtej besedi z glasom na visoko in potem zopet nazaj; proti koncu stavka pa bese- de neprijetno zateza. To nenavadno, nenaravno naglašanje je krivo, da ga težko ra- zumeš, in te lahko smeh posili, dasi drugače dobro slovenščino govori, in le nekoliko sosednje hrvaščine vplete." 'A man from Haloze is of quick tongue; that is, every third or fourth word his voice jumps up high and then down again; at the end of a 4 The voealie information on western Haloze eomes from Zorko 1993. G. Lundberg, Typology of Tone Loss in Haloze, Slovenia: An Acoustic and Autosegmental Analysis 173 sentence the words are unpleasantly stretched. This unusual, unnatural accenting is responsible for it being hard for you to understand, and it can make you laugh, al- though otherwise he speaks good Slovene with onlyalittie influence from neighbor- ing Croatian' (Kolarič 1964: 396). Kolarič does not include this as a comment on word or sentence intonation in Haloze. He uses it only to emphasize how strange the speech of these border-region farmers can be. On the other hand, this statement does indicate that something in- teresting may be happening in the area of intonation, a1though it is not clear if Pa- jek is observing sentence or word -level prosody. Regardless , it is worth investigating further. In contradiction to Kolarič's comments on Haloze tonemes, Zorko (1991, 1993) states that there are distinctive tonal oppositions in eastern and central Haloze. "Na vzhodu in deloma v osredju je ohranjena še opozicija med visoko (cirkumfleks) in nizko intonacijo (akut). Nizka intonacija je možna le na kratko naglašenem zlogu, v katerem je samoglasnik refleks starega ali novega akuta ali pa akuta na sekundarno naglašenem zlogu." 'In the east and part of the center, the opposition between high (circumflex) and low (acute) intonation is stiH retained. Low intonation can occur only on short stressed syllables in which the vowel is the reflex of the old or new acute or from secondary retraction' (1993: 205). She does not follow this statement with any explanation of exactly what this opposition might entail. It is not clear what the contours of these tonemes might be or how one would distinguish them. We are left to assume that these tonemes, especially since she follows Toporišič in indicating that they should really be called "high" and "low" rather than "falling" and "rising" (1968: 315), are similar to the shape and contrasts of the tonemes in Do- lenjsko, Gorenjsko and Koroško. According to Zorko, this tonal distinction is found only on short syllables and only in eastern and part of central Haloze, but she does not specifically indicate which viHage dialects exhibit this phenomenon. This is significant because the dialects of eastern and central Haloze have retained the phonemic distinction between long and short syllables. This means that, according to Zorko, tonal opposi- tions have been lost on long syllables where it is easiest to distinguish tonemes and retained on short syllables where it is much more difficult to differentiate them. This claim by Zorko, a1though typologically unlikely, is intriguing, and, if true, it describes a situation, which to my knowledge is unique in world languages. The situation is complicated further by the fact that this feature would carry very low functional load on short syllables because all the reflexes of Common Slavic cir- cumflex on etymologically short syllables have lengthened in Haloze, bOuk 'god', mes8u 'meat'. That means that the short falling tonemes that would provide an opposition to the short rising would have to come from a different source, one that Zorko does not identify.5 It seems likely that Zorko has identified a phonetic feature that is not distinctive, i.e., it is not phonemic. These ambiguities require further examination. 5 It is true that in some areas of eastern Haloze the Common Slavic circumflex on short syl- lables has remained short, but this is a phenomenon with which Zorko appears to be un- acq uainted. 174 Slovenski jezik - Slovene Linguistic Studies 3 (2001) Spectrographic analysis In the following pages, a phonetic analysis of the question of tone in the eas- tern Haloze village dialect of Meje, which represents the tonemic situation in both eastern and central Haloze, will be discussed. This analysis was performed using software developed by Speech Technology Research, Ltd., Multi-Speech Signal Ana- lysis Workstation, Model 3700, Version 1.20, and was made available to me by the Phonetics Laboratory of the Linguistics Department at the University of Kansas. This software captures sound from audiotapes or live speech and converts it from analog to digital format. It then permits a spectrographic analysis on the data. The analysis described below samples both long and short stressed syllables which are the reflexes of original Common Slavic circumflex, acute and neo-acute as well as syllables which have received stress through later retractions. This section is not meant to be a complete or comprehensive phonetic analysis of the dialect of Meje, but as an examination of the questions raised by Zorko on this topic. (1) Are there distinctive tonal oppositions in Haloze? (2) Is this distinction only made on short syllables? This analysis will also provide data to help locate Haloze in a ty- pology of tone loss. Tonemic possibi1ities For comparative purposes, the speech of a native Slovene from the village of Jarše pri Domžalah, a tonemic region of Upper Carniola,6 was tested. A spectro- graphic analysis was performed on her speech in December of 1998. The contours of the two contrastive tones are represented in graphs 1 and 2. Graph 1 is of the "cir- cumflex" or "high" tone, and Graph 2 is the "acute" or "low" tone? The representa- tion is not of the entire word but the nucleus of the accented syllable only. The ver- tical line represents frequency in Hertz, and the horizontal line represents time in seconds. Graph 1 300Hz 200Hz 200Hz 190Hz 180Hz 100Hz 175Hz .05 sec. .10 sec. .15 sec. .20 sec. .25 sec. mesa 'meal' 6 The informant for this test was Marta Pirnat-Greenberg. She is a university educated Slovene from an Upper Carniola dialect base. 1 am very grateful to her for her help on this pro- ject. 7 These samples were recorded in asentence context. The form in Graph 1 comes from the sentence, Malo mesa sem pojedia '1 ate a little meat.' The form in Graph 2 comes from the sen- tence, Bela krava se pase 'The white cow is grazing.' For this short corroborative study, 1 sampled 30 words with acute tone and 9 with circumflex. The examples given here are representative of the other forms. G. Lundberg, Typology of Tone Loss in Haloze, Slovenia: An Acoustic and Autosegmental Analysis 175 Grapb 2 300Hz 200Hz 176Hz 186Hz 100Hz 143Hz .05 sec. .10 sec. .15 sec. .20 sec. .25 sec. krava 'cow' As is c1ear from the graphs, these two tonemes do have different contours, but the most important distinctive feature between the two tones is their relative height. This is illustrated by Toporišič's (1968) examination of the relationship between the arsis and thesis of acute and circumflex and the absolute height of the arsis of the acute compared to the arsis of the circumflex.8 Ris analysis shows that in most po- sitions and sentence constructions the acute has a low arsis and a high thesis. The circumflex has a high arsis, higher than the arsis of an acute in the same sentence structure, and a high thesis. The circumflex starts higher than the acute and has a peak that is higher than the acute peak. The following syllable is lower than the pe ak of the accented syllable. The acute starts lower relative to the cir- cumflex but rises more dramatically to a pe ak slightly lower than the pe ak of the circumflex. With the acute the pe ak of the post tonic syllable is as high or higher than the accented syllable (315). The important point is that the contour of the tone is not the most salient fea- ture. In fact, it is the least important. The highest phonemic load is given to the height of the tone. This is because the contour of the arsis and the relative differ- ence between the arsis and thesis of the acute as opposed to the circumflex can change in different types of sentences or in different positions within asentence. On the other hand, the relative height of the arsis of the acute compared to the height of the arsis of the circumflex is the most consistent difference between the two tonemes. Theoretically, this height difference could be enough to distinguish these tonemes even in word-final position, kralj 'king', meso 'meat'. It mayaiso be informative to review the study done by Lehiste and Ivic in their book Word and Sentence Prosody in Serbocroatian. I will focus on their ex- planation of the pitch contour distinctions in Neo-Štokavian Serbo-Croatian. Ac- cording to Lehiste and Ivic, knowledge of the tone on the accented syllable and on the post-tonic syllable is essential for distinguishing between the rising and the falling tone. The long falling tone is contained in one syllable. It generally starts low and rises to a high pe ak early in the syllable nuc1eus; the peak is in the first half of the duration of the syllable. Then it drops off sharply to a low at the end of the ac- cented syllable, and the post tonic syllable is lower stilI (44). 8 The arsis is the syllable under stress, and the thesis is the first post-tonic syllable. 176 Slovenski jezik - Slovene Linguistic Studies 3 (2001) The long rising tone has arising contour which usually reaches its pe ak near the end of the syllable nucleus, and the post ton ic syllable peak is as high or higher than the pe ak of the accented syllable. The change in the Fo is much less extensive than for the long falling tone. In fact, the contour is of ten almost a straight line (45). The most important distinctive feature is the height of the post-tonic syllable because in some environments the contours of the two tones are not distinctive. This is especially true for the short falling and short rising tones where post-tonic Fo height is the only reliable way to distinguish the tones (44). In the variety of Ser- bo-Croatian described by Lehiste and Ivic, tone is not distinctive in final syllables. Lehiste and Ivic also include a brief description of the pitch contour of a tonemic dialect in Croatian Kajkavian. The dialect described is Donja Pušca, which is reputed to be a very conservative Kajkavian dialect. This general area borders Ha- loze, so it is perhaps the most relevant area for comparison. In this area tone is di s- tinctive in final syllables as well as internally, and the tone contours are contained in one syllable. Graphs 3 and 4 are representations of the syllable nucleus under the Kajkavian rising and falling tones respectively. Graph 3 is of stniža 'sentry' , and Graph 4 is of škofa 'school'. Graph 3 300Hz 234Hz 200Hz 228Hz 204Hz 100Hz .05 sec. .10 sec. .15 sec. .20 sec. .25 sec. straža 'guard' Graph 4 300Hz 240Hz 200Hz 225Hz 213Hz 100Hz .05 sec. .10 sec. .15 sec. .20 sec. .25 sec. škola 'school' Note that the contour of the tone and the placement of the pe ak are much more important in this system than in the others that we have discussed thus far. Lehiste and Ivic mention that the pitch pe ak for Graph 3 is very near the end of the syllable nucleus. The duration of the syllable is 144 units and the peak comes at 129. G. Lundberg, Typology of Tone Loss in Haloze, Slovenia: An Acoustic and Autosegmental Analysis 177 ln contrast to this, the pe ak in Graph 4 is very early. The duration of the syllable is 117 units and the pe ak is at 41. This is well before the center of the syllable (81-82). Analysis of Meje in Haloze Informants and methodology All of the material for this spectrographic analysis comes from selections made from a corpus of approximately 20 hours of recordings made on electronic tapes of dialect speakers while the author was performing fieldwork in Haloze.9 From record- ings made of conversations with informants on various topics from daily life just over 100 words were selected. These forms represent the reflexes of Common Slavic circumflex on etymologically long and short syllables, Common Slavic acute and neo-acute on long and short syllables and words that have experienced an accent re- traction in more recent times such as zvezda 'star' or žena 'wife'. ln the process of choosing these words, the basic methodology of Lehiste and Ivic was followed (36). Words were chosen that were in neutral position in the sen- tence. This is a position in which the sentence intonation would have minimal in- fluence on the word intonation. These forms were taken from the central part of the sentence because in sentence initial position the word tone is higher than nor- mal, and in sentence final position word intonation is regularly neutralized. The words chosen also bore primary but not emphatic stres s in the sentence. Finally, al- though the software used for this phonetic analysis employs an algorithm that cal- culates the Fo at a number of predetermined intervals over the duration of the syl- lable, in order to simplify this explanation, the pitch height was recorded at the be- ginning, at the peak of the fundamental frequency and at the end of the syllable nuc1eus. The quantity of the syllables involved was also measured. Results This study was performed as ameans of confirming or refuting Zorko's c1aims about an unusually archaic prosodic system in Haloze. On the one hand, the phone- tic analysis did show that quantity is distinctive in eastern Haloze. Long syllables last between .2 and .24 of a second. Short syllables are between .12 and .15 of a se- cond, and unstressed syllables have a duration of between .05 and .08. It is reason- able to speak of two moras for a long syllable and one mora for a short syllable. On the other hand, a1though the situation is not completely as one might expect for a Styrian dialect, there is no distinctive tone in Haloze. There is only one pattern of pitch contour on short syllables. It is rising. On short stressed syllables the pitch level starts low at the beginning of the syllable and peaks 75% to 80% through the dura- tion of the syllable nuc1eus, then it falls off slightly but finish higher than it started. This is true of all short syllables under stres s in sentence neutral position, grant 'land' 234Hz-282Hz-245Hz (the duration was .14 sec.), člov~ka 'person' l67Hz- 306Hz-290Hz (the duration was .11 sec.), reku 'he said' 245Hz-324Hz-306Hz (du- ration was .12 sec.). Graph 5 below is a representation of the pitch contour of a short accented syllable. Even more interesting is that when lexpanded the study to inc1ude long syl- 9 The primary informant for the village dialect of Meje is a 73-year-old woman named Marija Arbajter. She was born in the same village in June of 1926. 178 Slovenski jezik - Slovene Linguistic Studies 3 (2001) lables as well as short syllables of all possible historical proveniences, 1 discovered that the exact same pattern of pitch contour occurred on long as the one found on short. This was the case for the reflexes of Common Slavic circumflex as well as neo-acute and secondary retractions onto etymologically long syllables, d~Zat 'to work' (supine) 239Hz-290Hz-256Hz (duration was .21 sec.), prepeZaZi 'they trans- ported' 200Hz-245Hz-212Hz (duration was .24 sec.), na br~gi 'on the hill' 190Hz-256Hz-225Hz (duration was .2 sec.). Note that for all of the long syllables the pitch peak comes near the end of the duration of the vowel. Graph 6 below is a representation of this pattern. Graph 5 300Hz 268Hz 200Hz ~45HZ 216Hz 100Hz . 05 sec. . 10 sec . . 15 sec . .20 sec . .25 sec . svetek 'holiday' Graph 6 282Hz 300Hz 245Hz 275Hz 200Hz 100Hz . 05 sec. . 10 sec . . 15 sec . .20 sec . .25 sec . obrnali 'turned' Because the contour is the same in all cases where the sentence intonation does not negate word prosody (this is largely a problem in sentence-final position because sentence intonation distorts word intonation here), it is possible to come up with an average contour for both long and short stressed syllables. According to the attested forms in this study the long syllable starts at 233Hz, peaks at 274Hz and then falls off to 249 Hz. Out of an average of 15 samples taken of the pitch contour of a long syllable the pe ak comes at the 11th and 12th samples. This puts the peak at about 80% of the way through the duration. The average short syllable starts at 239Hz, peaks at 274Hz and falls off to 257Hz. With an average of 12 pitch samples taken in the syllable, the pe ak is on the 8th or 9th. Again this puts the peak at about 75% of the way through the duration. Graphs 7 and 8 show the typical pitch contour for long and short syllables respectively. The most interesting point to be made about these averages is that the basic measurement s of the long and short syllables are almost exactly the same. The aver- G. Lundberg, Typology of Tone Loss in Haloze, Slovenia: An Acoustic and Autosegmental Analysis 179 age long syllable rises 41Hz to a pe ak at 274Hz and then falls 25Hz to the end of the syllable. The average short syllable rises 34Hz to a peak of 273Hz and then falls 16Hz to the end of the syllable. The striking difference is that the short syllable makes this contour fit within half the duration of the long syllable. The contour is therefore much sharper. This is significant and will be touched on more later.1O Grapb 7 274Hz 300Hz 249Hz 233Hz 200Hz 100Hz . 05 sec. .10 sec . .15 sec. .20 sec . .25 sec . Grapb 8 273Hz 300Hz ~ 257Hz 200Hz 239Hz 100Hz .05 sec. .10 sec. . 15 sec. .20 sec . .25 sec . Explanations Based on my own expectations and the descriptions by Kolarič and Zorko, the resu1ts of this study were not exact1y as expected. It is c1ear that phonemic tonal dis- tinctions have been lost in eastern and central Haloze. It is not strange that when the phonemic status of a feature is lost one of the two oppositions would be gen- eralized. This is what has happened in all of the Styrian dialects. As Logar's state- ment cited earlier in this paper indicates, Styrian dialects generalized the falling tone. This is also the case in Prekmurje where pitch has been lost. In these dialects there is a c1ear falling contour to the accented syllable of the word.l1 The interesting thing about Haloze is that when pitch distinctions were lost, the rising rather than the falling contour was generalized. This is not the rising contour of other Slovene dialects like the one represented in Graph 2 or the "high" tone 10 For the purposes of comparison, a limited examination was made of the pitch contour of the accented syllable in the western Haloze village dialect of Žetale. It is interesting that the ac- cented syllable is marked with high tone, but there is no distinct contour other than that the peak of the Fo is in the center of the syllable. The pattern is the same on the reflexes of original cir- cumflex and acute, doma 'at home' 120Hz-128Hz-117Hz, iz hiše 'from the house' 123Hz- BO Hz-126 Hz. This is an example of dialect in which all prosodic features other than stress have been lost. There are no phonemic tone or length distinctions, and a rise in pitch within the word serves only to mark the accented syllable. 11 Marc L. Greenberg, personal communication. 180 Slovenski jezik - Slovene Linguistic Studies 3 (2001) found on the circumflex of Slovene and Serbo-Croatian. It is cleady the rising con- tour represented in Graph 3, which is the Kajkavian pattern. This is not strange considering the strong historical and geographic connections of eastern Haloze to its Kajkavian neighbor. Asimilar situation exists in the variant of Standard Danish spoken on the island of Bornholm off the southern coast of Sweden. Bornholm is cleady a development of older East Danish, but, prosodically, it has much more in common with Stockholm Swedish than with Copenhagen Danish (Gronnum 1992: 70).12 The pitch contour of accented syllables in eastern Haloze is closer to that of its Kajkavian neighbor than to that of central Slovene. One other question remains. Why did Zorko maintain in two studies over the course of several years that eastern Haloze has a phonemic pitch distinction, which the spectrographic analysis cleady indicates it does not have? One possible ex- planation comes from general studies on the typology of tone and tonal languages. Grandour , as well as others before him, has pointed out that the human auditory sys- tem attaches different significance to changes in the contour of rising as opposed to falling tone. If the general direction of a tone is from high to low, the auditory sys- tem does not easily distinguish changes in the steepness of the contour . If, on the other hand, the general direction of the tone is from low to high, the auditory sys- tem is much more aware of differences in the contour of the rise (58). In Haloze the short syllable has the same basic change in pitch frequency from the beginning of the syllable to the peak of the Fo as that found in the long syl- lable. The difference is that this rise of 30 to 40 Hertz happens in half the time. That means that the pitch contour is much sharper and therefore more salient to the auditory system. This phenomenon makes the long syllable seem unmarked and the short marked for a tonal distinction. This rapid rise in pitch frequency is a salient feature to the auditory system, but it does not play the same role as tonal distinc- tions in other Slovene or Serbo-Croatian systems with tonemes. It may however play a role in helping speakers of the eastern Haloze dialects distinguish short and long syllables, a distinction which does carry phonemic significance. This analysis has succeeded in answering the questions asked above. (1) There is no distinction in eastern and central Haloze, at least in the dialects studied here, between circumflex and acute tonemes, and, (2) therefore, this distinction is not maintained on short syllables exclusively. Haloze does seem to be different from other Styrian and Pannonian dialects in that at the loss of tone it generalized a ris- ing contour rather than a falling contour . This distinction for now remains tentative because there are no pitch contour analyses of other Styrian or Pannonian dialects to confirm that the pattern is indeed falling in dialects outside of Haloze. It is sign- ificant that the pitch contour in Haloze is the same as the rising contour in Kajkav- ian dialects. The similarities could come from a genetic connection. There is good reason to at least consider this possibility based on the fact that eastern Haloze has a clear genetic connection to Kajkavian dialects in the vocalic system that it does not share with other Slovene dialects (Lundberg 1999: 94). On the other hand, inton- ation patterns are one of the easiest element s to borrow. They can often be bor- rowed with little other structural influence entering the language. It may be that as Haloze lost phonemic pitch, it generalized the pattern of the Kajkavian speakers 12 1 would like to thank Henning Andersen for pointing out this literature to me. G. Lundberg, Typology of Tone Loss in Haloze, Slovenia: An Acoustic and Autosegmental Analysis 181 with whom it had close contact. This question will be considered in the final section of this article. Typology of tone loss One possible explanation for the unusual tonemic situation in Haloze may be found in a typological approach to the data. Haloze may represent astage of tone loss that is different from other Styrian or Pannonian dialects. Salmons, in his book Accentual Change and Language Contact, sets up a typology of tone loss as a con- tinuum from a pure tonal language on one extreme to a system with only dynamic or fixed stress on the other, placing pitch accent languages like Slovene, Serbo-Cro- atian and Swedish in the middle (1992a: 185), which may be schematized in the fol- lowing manner: Tone> Pitch Accent > Stress Accent > Fixed Stress. Salmons defines this typological continuum in terms of high tone. A tonal lan- guage is one that specifies tone level s on virtually every syllable. This means that there can be more than one high tone per word. Pitch accent languages are those that allow only one high tone per word, and the placement of this high tone in re- lation to other prosodic features, such as quantity, determines the contour of that tone (70). Languages with simple stress do not specify tone at the word level at all. They indic ate intonation on the phrasal level only. This typology, then, moves from tonal languages, where pitch distinctions are bunched together in each word, to pitch accent languages, where tonal distinctions are spre ad out, so they only occur be- tween lexical items, and finally to simple stress systems which specify tone only on the phrasal level (71). Salmons further defines this typology in aseparate article on the typology of change from a tonal to simple stress system (1992b). He indicates that there are se- veral distinct sta ge s in this development. He suggests that the first stage in this pro- cess is for a language or dialect to generalize one high tone. Then it would lose dis- tinctions outside of the accented syllable and finally move to simple dynamic stress (274). As a mechanism for describing how this process might work, Salmons follows the non-linear formalism of Halle and Vergnaud, which allows for a tonal tier and a stress tier that function together (1987: 190-91). According to their model, I have represented the long tones of Serbo-Croatian as described by Ivic in Figure 2, where the top line is the tonal tier (high versus low) and the bottom is a stress line (* = stressed, o = unstressed). The first form, Radom 'with Rada' (instr. sg.), has rising tone, and the second form, radom 'with work' (instr. sg.), has falling tone. The double vowels in the accented syllable indic ate length.13 Figure 2. Autosegmental representation of Standard Serbo-Croatian word prosody (after Ivic 1987). LH H HL L s Raadoom s raadom * o * o The Kajkavian forms given earlier in this paper must be represented differently 13 Asimilar representation of pitch accent in Serbo-Croatian was proposed in Inkelas and Zec 1988. 182 Slovenski jezik - Slovene Linguistic Studies 3 (2001) because tonal opposl11ons are contained within one accented syllable in those di- alects, rather than spreading over two adjacent syllables as in Serbo-Croatian. Figure 3. Autosegmental representation of Kajkavian word prosody LH L straaža * o HLL ško ola * o Finally, forms from eastern and central Haloze could be represented as in Figure 4. Figure 4. Autosegmental representation of eastern and central Haloze word prosody LH L deelat * o LH L breegi * o This formalism used for Western South Slavic pitch accent systems brings up several interesting points. First, according to the typology formulated by Salmons, Neoštokavian Serbo-Croatian and probably most tonemic Slovene dialects falI someplace in-between a pure tonal language and a pitch accent language because they are not systems in which only one high tone per word is tolerated. The ton- emic distinctions spre ad beyond the accented syllable because the pitch peak of the posttonic syllable plays a role in distinguishing the tones. For Serbo-Croatian this is also true of other prosodic phenomena, such as length, which is distinctive outside of the accented syllable as well. Kajkavian is perhaps a more c1ear example of a pitch accent system, because all prosodic distinctions are made within the accented syllable.14 The second point illustrated by this formalism is that representing the dif- ferences between systems like those found in Serbo-Croatian, Kajkavian and finally Haloze is simply amatter of reducing the distinctions made in the tonal tier and finally eliminating them a1together. This leaves a system like that of central Haloze. In addition to this, western Haloze has lost quantity distinctions and eliminated the already non-phonemic contour found on accented syllables in eastern and central Haloze. Phonetic information about the contour of pitch on the accented syllable in Ha- loze contributes important data to what is known about the stages of tone loss. Ha- loze appears to represent the stage of tone loss just after an important distinction on the tonal tier has been lost. This rising pitch contour may represent the remnants of original phonological distinctions of pitch. This is very similar to the situation for Texas Swedish described by Ureland (cited by Salmons 1992a). Swedish Accent 1 14 For an alternative view see Inkelas and Zec (1988), in which they argue that the underlying representation of the phonological word in Serbo-Croatian contains only one high tone, which spreads from right to left according to a Spreading Rule. This allows the language to observe the Obligatory Contour Principle. The problem with this is that the Spreading Rule does not function in Slovene because falling tone can be found on final as well as on initial syllables. G. Lundberg, Typology of Tone Loss in Haloze, Slovenia: An Acoustic and Autosegmental Analysis 183 (acute accent) has taken over completely for some of Ureland's informants, and Ac- cent 2 has disappeared (1971: 72, 106). This kind of tonal restructuring is not uncommon in language contact situ- ations. Salmons believes that accentual phenomena are extremely vulnerable (1992a: 1) and that tone loss can very of ten be explained sociohistorically in terms of lan- guage contact (35). There are many examples of the vulnerability of tone. One of these is Hausa, an Afro-Asiatic language that adopted tone from its neighbors in the Niger-Kordofanian family (26). Another example can be found in bilingual Finnish and Swedish villages in Finland where speakers have generalized one prosodic sys- tem for both languages (Gčhding 1977: 98). There are Slavic examples of this kind of development as well. Thomason and Kaufman, based on the work of Veenker (1967), indic ate that dose contact between northern Russian dialects and the Ura lic group, specifically Finnic, is the reason that some northern Russian dialects have fixed initial stress (1988: 241). They also describe asimilar situation in the north of the Serbo-Croatian speech territory near the Hungarian border where a particular dialect has fixed stress on the penultimate mora of the word (62).15 Finally, Thomason and Kaufman propose that strong Hun- garian influence may be the reason that Czech and Slovak have fixed initial stress (41). These contact explanations for prosodic changes are supportive of such an ex- planation in Slovenia and in Haloze in particular. As mentioned before, the dialects of the Slovene speech territory have experienced strong contact with all the major language families of Europe as well as divergence and convergence relationships with neighboring Slavic dialects. Greenberg indicates that, a1though tone loss in Slo- vene dialects is not well understood, at the very least, contact, especially with Italian and Hungarian, has accelerated the processes involved in the loss of tone (2000: 160-61). Haloze is in an area with strong German and Hungarian sociolinguistic in- fluence. It is known that in the very early feudal period the lands to Haloze were under the control of Ptuj. This situation, especially for western and central Haloze continued largely unchanged until the mid nineteenth century. But, the situation is different for eastern Haloze, where the church and state boundaries were less stable. This is due partly to the fact that the Hungarians took eastern Haloze in 907 and did not lose control of it until Frederick of Ptuj redaimed the Ormož region on East- er 1200. At the same time the castle Tranbek, located near the present day village of Dravinjski Vrh, took back the lands of eastern Haloze and probably the castle Borl (Bračič 1967: 57). From the mid thirteenth century on most of eastern and part of central Haloze was controlled from Borl, and the Slovene state and ethnic borders developed much as they are today (62). German contact has been an important factor in all Slovene dialects, in fact it continued as a contact language in Slovenia well into the first half of the 20th cen- tury (Greenberg 2000: 55). A1though German linguistic influence is strong, Hun- 15 They are likely referring to the Kajkavian dialect of Virje, which was described in 1907 by Fancev and later treated in March 1981. This connection is interesting, but not obvious because Hungarian has fixed initial stress, and it is not clear how that might influence the development of a system with fixed stress on the penultimate mora of the word. 184 Slovenski jezik - Slovene Linguistic Studies 3 (2001) garian and Italian contact se em to be more important for tone loss (56). Haloze is on one of the borders of tone loss in Western South Slavic. To the west and north of Haloze are Styrian and Pannonian dialects which have lost tone as well as all quantity distinctions. They have systems of simple stress accent. To the south and east of Haloze are Kajkavian dialects which have phonemic tone in a pitch accent system. Haloze is on the periphery of tone loss, a development, which has its center in Pannonian dialects with strong Hungarian contact. Conclusion The spectrographic analysis of the Fo contour of accented syllables in eastern and central Haloze indicates, in contradiction to several claims in the dialect liter- ature, that Haloze does not have phonemic pitch distinctions. However, unlike other Pannonian and Styrian dialects which have lost pitch distinctions, Haloze has gener- alized arising contour to all accented syllables. According to the typology of tone loss as formulated by Salmons, Haloze ap- pears to be near the end of this typological continuum. This continuum begins with Indo-European, which Kortlandt believes may have gone through a tonal stage (1986: 156).16 Several of the Indo-European daughter languages, Balto-Slavic being one of them, developed pitch accent systems. The tonemic dialects of the Serbo-Croatian and Slovene speech territories represent variations of this stage. Eastern and central Haloze represent the middle ground between pitch accent and stress accent. Western Haloze and the Styrian dialects that surround it are systems with no prosodic dis- tinctions other than stress. This typology of tone loss is a useful tool for understand- ing the otherwise strange tonemic situation in Haloze. Additional spectrographic ana- lyses of Western South Slavic dialects on the border between tonemic and nontonem- ic systems would contribute significantly to what we know about tone loss. References Bračič, Vladimir. 1967. Vinorodne Haloze. Maribor: Založba Obzorja. Fancev, Franjo. 1907. "Beitrage zur serbokroatischen Dialektologie." Archiv fur Sla- vische Philologie 29: 305-89. Garding, Eva. 1977. The Scandinavian Word Accents. Malmo, Sweden: CWK Glee- rup. Grandour, Jackson T. 1978. "The Perception of Tone." in Tone (ed. Victoria A. Fromkin). 141-76. New York: Academic Press. Greenberg, Marc L. 1987. "Prozodične možnosti v slovenskem knjižnem jeziku in slovenskih narečjih." Slavistična revija 35: 171-86. - -. 2000. AHistorical Phonology of the Slovene Language (= Historical Phonolo- gy of the Slavic Languages, vol. 13). Heidelberg: Universitatsverlag C. Winter. 16 The idea that Indo-European may have gone through a tonal stage is speculative. Many scholars believe that the pitch distinctions found in Balto-Slavic are not necessarily the direct con- tinuation of a posited PIE system but may be explained as separate developments in adialectal stage of Indo-European. G. Lundberg, Typology of Tone Loss in Haloze, Slovenia: An Acoustic and Autosegmental Analysis 185 Gmnnum, Nina. 1992. The Groundworks of Danish Intonation. Copenhagen: Muse- um Tuscu1anum Press. Halle, Morris, and Jean-Roger Vergnaud. 1987. An Essay on Stress. Current Studies in Linguistics. Cambridge, MA: MI. Inke1as, Sharon, and Draga Zec. 1988. "Serbo-Croatian Pitch Accent: The Interaction of Tone, Stress, and Intonation." Language 64/2: 227-48. Ivic, Pavle. 1966. "Fonološki aspekt genetičkog odnosa izmedju štokavske, čakavske i kajkavske dijalekatske grupe." Orbis Scriptus. Dmitrij Tschižewski zum 70. Ge- burtstag: 375-83. Munich: Wilhelm Fink. Johnson, Keith. 1997. Acoustic and Auditory Phonetics. Cambridge, Massachusetts: Blackwell. Kolarič, Rudolf. 1964. "Haloški govor." Prace filologiczne 18/2: 395-401. Kort1andt, Frederick H. H. 1986. "Proto-Indo-European Tones?" Journal of In- do-European Studies. 14: 153-60. Lehiste, lIse. 1961. "The Phonemes of Slovene." International Journal of Slavic Lin- guistics and Poetics 4: 48-66. Lehiste, I1se and Pavle Ivic. 1986. Word and Sentence Prosody in Serbocroatian. Cambridge, Massachusetts: MIT Press. Lencek, Rado L. 1982. The Structure and History of the Slovene Language. Ohio: Slavica. Logar, Tine. 1975. Slovenska narečja. Besedila. Ljubljana: Mladinska knjiga. - -. 1996. Dialektološke in jezikovnozgodovinske razprave. Ljubljana: Znanstveno- raziskovalni center SAZU, Inštitut za slovenski jezik Frana Ramovša. Lončaric, Mijo. 1979. "Naglasni tipovi u kajkavskom narječju." Rasprave Zavoda za jezik, knj. 4-5: 109-17. Zagreb: Zavod za jezik IFF. Lundberg, Grant. 1999. "Preliminary Report on Dialectological Fieldwork in Haloze, Slovenia." Slovenski jezik / Slovene Linguistic Studies 2: 91-108. - -. 2001. "Tone Loss in Haloze, Slovenia: An Instrumental Study." Balkanistica 14: 83-99. March, William. 1981. "Kajkavian Inflectional Morphophonemics: An Analysis of the Dialects of Velika Rakovica, Virje and Bednja." Rad Jugoslavenske akademije znanosti i umjetnosti, Razred za filologiju, knj. 388: 237-312. Zagreb: JAZU. Ramovš, Fran. 1935. Historična gramatika slovenskega jezika, VII: Dialekti. Ljubljana. - -. 1936. Kratka zgodovina slovenskega jezika. LjUbljana. - -. 1950. "Relativna kronologija slovenskih akcentskih pojavov." Slavistična revija 3/1-2: 16-23. Rigler, Jakob. 1963. "Pregled osnovih razvojnih etap v slovenskem vokalizrnu." Slavi- stična revija 14: 25-78. - -. 1986. "Smeri glasovnega razvoja v panonskih govorih," in Jakob Rigler. Raz- prave o slovenskem jeziku (ed. by Franc Jakopin): 116-28. Ljubljana: Slovenska matica. Salmons, Joe. 1992a. Accentual Change and Language Contact. Stanford, California: Stanford University Press 186 Slovenski jezik - Slovene Linguistic Studies 3 (2001) - -. 1992b. "Diachronic Typology and Tone-to-Stress Shift." Journal of Indo-Euro- pean Studies 20/3-4: 269-81. Thomason, Sarah and Terence Kaufman. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley: University of California Press. Toporišič, Jože. 1967. "Pojmovanje tonemičnosti slovenskega jezika." Slavistična revi- ja 15/1-2: 64-108. - -. 1968. "Liki slovenskih tonemov." Slavistična revija 16: 315-94. - -. 1987. "Slovensko narečjeslovje." 1. Toporišič, Portreti, razgledi, presoje: 217-56. Maribor: Obzorja. Ureland, Sture. 1971. "Observations on Texas-Swedish Phonology." Studia Linguistica 25: 69-110. Veenker, Wolfgang. 1967. Die Frage des finnougrischen Substrats in der russischen Sprache (= Uralic and Altaic Series, vol. 82). Bloomington: Indiana University Press. Zorko, Zinka. 1991. "Vzhodni haloški govori." Nemzetkazi szlavisztikai napok 4: 51-69. - -. 1993. "Daljšanje akuta v severovzhodnih slovenskih narečjih." Slavistična revija 4111: 193-207. Prispelo januarja 2001, sprejeto marca 2001 Received January 2001, accepted March 2001 Tipologija izgube tonemskosti v Halozah: akustična in avtosegmentna razčlemba Prispevek predstavlja teoretični pregled instrumentalne analize tonskega poteka v naglašenih zlogih v haloških govorih. Na osnovi podatkov instrumentalne analize se obravnavajo sodobne razmere v haloških govorih v okviru tipologije izgube to- nemskosti, pogojene z jezikovnim stikom. V nasprotju z nekaterimi izjavami v literaturi spektrografska analiza, predstav- ljena v razpravi, dokazuje za haloške govore odsotnost razločevalne tonemskosti tako v dolgih kot v kratkih naglašenih zlogih. V nagla šen ih zlogih obstaja le ena vrsta tonskega poteka, in sicer rastoča. V naglašenih zlogih se ton začne nizko in doseže vrhunec po 75 do 80 % trajanja zlogovnega jedra, nato rahlo pade, vendar se konča više od začetne točke. Povprečni dolgi zlog se dvigne za 41 Hz do vrhunca pri 274 Hz, nato pa proti koncu zloga pade za 25 Hz. Povprečni kratki zlog se dvigne za 34 Hz do vrhunca 273 Hz, nato pa pade za 16 Hz proti koncu zloga. Razlika je v tem, da je v kratkem zlogu trajanje sicer enakega poteka kot v dolgem za polovico krajše, tj. potek je v kratkem zlogu znatno bolj navpičen. Pomembno je, da se je po izgubi tonemskega nasprotja v haloških govorih po- splošil rastoči in ne padajoči tonski potek. Ta pa ni ne rastoči potek drugih sloven- skih narečij ne »visoki" ton cirkumfleksa v slovenskem ali srbohrvaškem knjižnem G. Lundberg, Typology of Tone Loss in Haloze, Slovenia: An Acoustic and Autosegmental Analysis 187 jeziku, temveč rastoči potek, ki se najde v kajkavskih narečjih. To ni nepričakovano glede na močne zgodovinske zveze med vzhodnimi Haložani in njihovimi kajkavsko govorečimi sosedi. Ena od možnih razlag za nenavadno tonemskost v haloških narečjih izhaja iz tipološke razčlembe podatkov. Haloški govori morda predstavljajo drugo razvojno fa- zo v izgubi tonemskih nasprotij kot drugi štajerski oz. panonski govori. V prispevku so haloški podatki predstavljeni v okviru tipologije izgube tonemskosti na kontinu- umu od jezikov s popolno tonemskostjo na eni strani do jezikov, ki imajo le naglas oz. ustaljeno mesto naglasa na drugi. Ta tipološki kontinuum je mogoče definirati v smislu visokega tona. Jeziki s popolno tonemsko st jo imajo skoraj v vsakem zlogu tonski potek, kar pomeni, da ima beseda lahko več kot dva visoka tona. Jeziki s to- nemsko naglašenost jo imajo le po en visoki ton v besedi in mesto tega visokega tona v odnosu do drugih prozodičnih prvin, npr. količine, pogojuje tonski potek. Jeziki z dinamičnim naglasom sploh ne opredeljujejo visokega tona na ravni besede, temveč na ravni stavka. Tipologija se torej premika od jezikov, v katerih so tonemska nas- protja nakopičena v besedah, preko jezikov, v katerih so visoki toni razporejeni ta- ko, da se nahajajo le med leksikalnimi enotami, do jezikov s preprostim naglasom, v katerih je ton opredeljen le na stavčni ravni. Za opis, kako v okviru te tipologije poteka jezikovni razvoj, se v razpravi upo- rablja nelinearna formalizacija, ki predstavlja ton in naglas ločeno na tonski in na- glasni ravni, ki delujeta vzajemno. Jezik se lahko premakne od tonemsko naglašene- ga do naglasnega sistema tako, da postopoma zmanjšuje število nasprotij na tonski ravni. Navajajo se primeri iz drugih slovanskih in svetovnih jezikov, ki ponazarjajo občutljivost tonske ravni v pogojih jezikovnega stika. Po tej tipologiji se haloški govori približujejo koncu kontinuuma. Lahko bi rek- li, da se je kontinuum začel z indoevropščino, ki je morda, kot domnevajo nekateri znanstveniki, prešla obdobje s tonemskimi nasprot ji. V nekaterih hčerinskih jezikih, med njimi tudi v baltoslovanščini, so se razvili tonemsko-naglasni sistemi. Slovenska in srbohrvaška narečja s tonemskim naglaševanjem predstavljajo različice te razvojne stopnje. Vzhodni in osrednji haloški govori predstavljajo vmesno razvojno stopnjo med tonemskim in dinamičnim naglaševanjem. Zahodni haloški in štajerski govori, ki jih obkrožajo, pa so sistemi brez prozodičnih nasprotij, z izjemo mesto naglasa. Ta tipologija izgube tonemskosti je koristna za razumevanje sicer zelo nenava- dnih tonskih razmer v haloških govorih. Dodatne spektografske analize zahodnih južnoslovanskih narečij na meji med tonemskimi in netonemskimi sistemi bi lahko veliko prispevale k našemu dosedanjemu vedenju o izgubi tonemov. Typology of Tone Loss in Haloze, Slovenia: An Acoustic and Autosegmental Analysis The article presents a theoretical reexamination of an instrumental study of the pitch contour on accented syllables in the dialects of Haloze, a geographical region on the eastern border of Slovenia's historical province of Styria. Based on the data from this instrumental study, the contemporary situation in Haloze is discussed in terms of a typology of tone loss that is motivated by language contact phenomena. 188 Slovenski jezik - Slovene Linguistic Studies 3 (2001) ln contrast to some statements in the literature, the spectrographic analysis pre- sented here indicates that there is no distinctive tone in either long or short syllables Haloze. Rather, there is only one pattern-a rising pitch contour-on accented syl- labies. On stressed syllables the pitch level starts low and peaks 75 to 80% through the duration of the syllable nucleus, then it falls off slightly but finishes higher than it started. The average long syllable rises 41Hz to a peak at 274Hz and then falls 25Hz to the end of the syllable. The average short syllable rises 34Hz to a peak of 273Hz and then falls 16Hz to the end of the syllable. The difference is that the short syllable makes this contour fit within half the duration of the long syllable. The contour is, therefore, much sharper. It is significant that when pitch distinctions were lost in Haloze, the rising rather than the falling contour was generalized. This is not the rising contour of other Slovene dialects nor the "high" tone found on the circumflex of Slovene and Serbo-Croatian. It appears to be the rising contour found in Kajkavian dialects. This is not unexpected in light of the strong historical and geographic connections of eas- tern Haloze to its Kajkavian-speaking neighboring population. One possible explanation for the unusual tonemic situation in Haloze may be found in a typological approach to the data. Haloze may represent astage of tone loss that is different from other Styrian or Pannonian dialects. In this paper Haloze is placed within a typology of tone loss that is set up as a continuum from a pure tonal language on one extreme to a language with only dynamic or fixed stress on the other. This typological continuum can be defined in terms of high tone. A tonal language is one that specifies tone levels on virtually every syllable. This means that there can be more than one high tone per word. Pitch-accent languages are those that allow only one high tone per word, and the placement of this high tone in re- lation to other prosodic features, such as quantity, determines the contour of that tone. Languages with simple stress do not specify tone at the word level at all. They indic ate intonation on the phrasal level only. This typology, then, moves from tonal languages, where pitch distinctions are bunched together in each word, to pitch-ac- cent languages, where tonal distinctions are spre ad out, so that they occur only be- tween lexical items, and, finally, to simple stress systems, which specify tone only on the phrasal level. As a mechanism for describing how linguistic development might occur within this typology, this study follows a non -linear formalism, which represents pitch and accent using a tonal tier and a stres s tier that function together. A language might move from a pitch -accent system to a system of dynamic stress simply by reducing distinctions made on the tonal tier. Examples are given from Slavic and other world languages that show the vulnerability of the tonal tier in language contact situa- tions. According to this typology of tone loss, Haloze appears to be near the end of the typological continuum. This continuum might be said to have begun with In- do-European, which, some scholars believe, may have gone through a tonal stage. Several of the Indo-European daughter languages, Balto-Slavic being one of them, developed pitch-accent systems. The tonemic dialects of the Serbo-Croatian and Slo- vene speech territories represent variations of this stage. Eastern and central Haloze represent the middle ground between pitch accent and stress accent. Western Haloze G. Lundberg, Typology of Tone Loss in Haloze, Slovenia: An Acoustic and Autosegmental Analysis 189 and the Styrian dialects that surround it are systems with no prosodic distinctions other than stress. This typology of tone loss is a useful tool for understanding the otherwise un- usu al tonemic situation in Haloze. Additional spectrographic analyses of Western South Slavic dialects on the border between tonemic and non-tonemic systems would contribute significantly to what we know about tone loss.