Hotell Norge, Lillesand, Norway 7. – 9. June 2002
(Click on name to open abstract)
Jardar Eggesbø Abrahamsen, Norwegian University of Science and Technology
Jørn Almberg, Norwegian University of Science and Technology
Gösta Bruce, Lund University
Olle Engstrand, Stockholm University
Anders Eriksson, Stockholm University
Peter Gilles, University of Freiburg
Carlos Gussenhoven, University of Nijmegen
Jan K. Hognestad, Agder College
Gjert Kristoffersen, University of Bergen
Ove Lorentz, University of Tromsø
Randi Alice Nilsen, Norwegian University of Science and Technology
Tomas Riad, Stockholm University
In the West Norwegian dialect of Sunnmøre there are, as in most of Norwegian and Swedish dialects, two word-accents, accent 1 and 2, or toneme 1 and 2. These accents are realised within the Accentual Phrase (AP).
In every Sunnmøre AP there are two obligatory tonal entities: To the right there is the Low boundary tone of the AP. To the left there is a High prominence tone, defining the primary stress. The boundary tone is spread to the area in between, so that everything is either H or L, as opposed to East Norwegian dialects, where the area between the (Low) prominence tone and the (High) boundary tone is tonally unspecified, and displays a gradual phonetic transition between the two.
The prominence tone defines the presence of a primary stress. Thus the primary stress defines the left boundary of the AP.
The difference between the two word-accents is the presence (accent 2) or absence (accent 1) of a Low lexical tone to the left of the prominence tone. Accent 1 APs thus display the tonal sequens HL, while accent 2 APs display LHL.
Although the H prominence tone defines the presence of a primary stress, the tone itself may (in an accent 2 AP) be on another syllable than the stressed one. This is in fact the the common pattern in accent 2 APs with three or more syllables (four or more moras, as the first syllable of the AP is always stressed and bimoraic).
It seems to be the case that the lexical Low tone of accent 2 prefers to be associated to a syllable, while the prominence tone and boundary tone are always satisfied with a mora. Although trimoraic APs with accent 2 (LHL) frequently are realised with two tones on the first syllable (which has to be bimoraic, as it is stressed) and one tone on the last syllable, there are instances where even in such APs the lexical L tone insists on being associated to both moras of the first syllables. In those cases the L boundary tone will remain unassociated, and the sequence LHL is reduced to LH.
This is clearly demonstrated when the following AP is an accent 1 AP, with the tones HL. Normally there would be a dip between the H tones in two adjacent APs: (LHL) (HL). With the L boundary tone of the first AP unassociated this dip does not appear. Instead the H prominence tone of the second AP is downstepped: (LH) (!HL).
This presentation will discuss methodological aspects of an ongoing quantitative analysis of tonal contours in different Norwegian dialects.
Stylization. In order to eliminate perceptually less important tonal movements caused by e.g. micro-prosody, the tonal entities are being stylized. The conventions for the stylization process will be discussed in the presentation.
Normalization of time and pitch. The stylized entities constitute the basis for turning the tonal entities into figures. The values along the time- and frequency scales are normalized so that they can be compared, both within and between speakers.
Preliminary results. Some preliminary results, which are obtained by using the method described above, will be presented.
Further work: perceptual studies. The goal is to present the design of a future perception study based on the results from the production analyses.
Gösta Bruce, Lund University
One starting point of my talk will be the Bruce & Gårding prosodic typology and our Swedish intonation model from the late seventies. The main criteria used for the classification of prosodic dialect types were 1) pitch realization of focus and 2) timing of pitch gestures for the two word accents (accent I and accent II). According to the typology four basic prosodic dialect types were identified: SOUTH, CENTRAL, EAST and WEST. A fifth prosodic dialect type FAR EAST (Finland Swedish) was also recognized but not explored in the typology.
Another starting point is the SweDia 2000 project, The phonetics and phonology of Swedish dialects around the year 2000. This project is a cooperation between phonetics sections at the universities of Lund, Stockholm and Umeå, and is funded by the Bank of Sweden. A huge speech database has been collected from more than 100 dialects in Sweden and Finland, represented by a variety of speakers. For prosodic analysis, both elicited phrases with variation of focus and word accent as well as spontaneous speech will be exploited.
The general goal of this part of the project work is to devise a new, comprehensive prosodic typology. The more restricted purpose of the present study is to examine pitch contours from the elicited material, containing accent I and accent II words in focal position. A further restriction is that three speakers from each of approximately one third of the elicited material have been analyzed, representing the prosodic dialect types SOUTH, WEST and CENTRAL. The idea is to investigate internal variation within the dialect types.
According to the typology, pitch realization of focus is an either-or-feature, manifested in different dialect types either as a wider range of the word accent in focus or as the addition of a pitch gesture after the focal word accent. Using this criterion we would predict abrupt transitions between dialect types. Concerning the other criterion, timing of pitch gestures for word accent, several different timings appear to be available. This would predict some possible variation in terms of word accent timing within the same basic prosodic dialect type.
Preliminary findings related to the above criteria indicate only minor variation for different dialects within the same basic prosodic dialect type, e.g. SOUTH or WEST. In a transition area between two dialect types, variation could be manifested as the use of alternative pitch patterns, e.g. two distinct pitch realizations of focus by the same speaker or by different speakers of the same dialect.
Olle Engstrand, Stockholm University
(in collaboration with Gunnar Nyström, Uppsala University)
Do E.A. Meyer’s tonal word accents contours from the Swedish dialects provide a reliable basis for quantitative analysis? Measurements made on acute and grave tone-peaks in a number of dialects spoken in the province of Dalarna suggested that the timing of grave tonal peaks tended to vary systematically from south-east to north-west. The former dialects had relatively late and the latter relatively early tone-peaks. This finding suggests that Meyer’s accent data may be sufficiently accurate to reflect systematic variation within broad dialect areas. Implications for the historical development of the Dalarna dialects are discussed.
I will address some problems in connection with comparing accent contours between speakers and between dialects. Even though speakers of a given dialect may be said to produce the accent distinctions of their dialect in "the same" way, we all know that there is considerable individual variation and also considerable variation as a function of context. Typologies that have been suggested for Swedish (e.g. "The Swedish accent orbit" by Öhman, 1967 based on surveys of Swedish accent contrasts by Meyer, 1937; 1954) assume that the abovementioned sources of variation may be abstracted away from and that the given accent types may be summarised in one or two canonical forms. This may well be true, but there has, as yet, not been much discussion about how this should best be done. To illustrate the problem we may consider the effect on accent-contour timing as a function of speech rate. If one wants to normalise accent contours in the time domain, what is then the relevant time unit on which normalisation should be based - segments, syllables, words, something else? In my presentation, I will discuss the averaging/normalisation problem using illustrations based on analyses of two Swedish dialects, the Orsa dialect spoken in Dalarna, Sweden, and the Närpes dialect spoken in Österbotten in northern Finland.
In this talk I will present an overview of the historical development and the present day situation of a phonological tone contrast in the Central Franconian Areas of Germany and Luxemburg. This tone contrast is used to differentiate words (lexical items, morphological forms) by tone alone. In resemblance to the Scandinavian tonal contrast these pitch accents are called 'accent 1' and 'accent 2'; they serve e.g. to differentiate the singular and plural forms of [be:n] 'leg(s)'.
After an overview of phonological and phonetic aspects as well as the regional and lexical distribution of the tone accents I will discuss recent theories about origin and development as formulated by Gussenhoven (2000), Schmidt (in print) and Mihm (in print). In the second part of the talk phonetic and systematic aspects of the realization of the pitch accents are discussed for two different Franconian regions: While in the Cologne/Bonn area the tone contrast is still an integral part of the phonological system, it is gradually lost in the Luxemburg area. Here the former tone contrast has led to modifications of the prosodic system (mainly vowel shortenings in words that originally had accent 1).
Gussenhoven, Carlos (2000): On the origin and development of the Central Franconian tone contrast. In: Aditi Lahiri (ed.): Analogy, Levelling, Markedness. Principles of Change in Phonology and Morphology. Berlin/New York: de Gruyter, 215-260.
Mihm, Arend (in print): Graphematische Systemanalyse als Grundlage der historischen Prosodieforschung. To appear in: Peter Auer/Peter Gilles/Helmut Spiekermann (eds.) (2002): Silbenschnitt und Tonakzente. Tübingen: Niemeyer.
Schmidt, Jürgen Erich (in print): Die sprachhistorische Genese der mittelfränkischen Tonakzente. To appear in: Peter Auer/Peter Gilles/Helmut Spiekermann (eds.) (2002): Silbenschnitt und Tonakzente. Tübingen: Niemeyer.
The Franconian tone arose around 1250, probably in Cologne, as a H-tone on the second mora of stressed monosyllables containing short vowels, like /dax/ 'day'. This must have happened in a sociolinguistic context in which the ambient prestige pronunciation had a long vowel, /daax/ (cf. modern German /taak/). The lengthened pronunciation of the short vowel in the /dax/ dialect was however not intrepreted as a bimoraic (i.e., long) vowel because, in the dialect concerned, forms with long vowels, like /daax/, were in use as plurals or as singular datives. Morphological neutralisation was avoided in the tonal interpretation, leading to singular H*HL% vs plural H*L% contours on monosyllables. This contrast was subsequently extended to different phonological and morphological contexts. Understandably, the introduction of this lexical tone contrast strained the rich West Germanic intonational system considerably, and retention of this system must soon have become impracticable. The historical development to the present-day dialect of Roermond is sketched, and compared where relevant with developments in other dialects.
In my contribution I will draw on data from 3 dialects of South-Western Norway: Stavanger, Egersund and Mandal. I will try to outline the dialectal variation in the tonal accents of these three dialects. Although they all belong to the same main group of Norwegian dialects, traditionally labelled 'high tone dialects', tonal variation across dialect boundaries is considerable, perhaps more so in Southern and Western dialects than in most other parts of Norway. Speakers can quite easily identify a considerable number of dialects virtually on tonal criteria alone. However, this important phenomenon, in the judgement of linguists and non-linguists alike, has not been dealt with in the otherwise fairly comprehensive literature on dialects in these areas. It will be of interest to describe both the relationship between the two contrasting tonal melodies in each dialect, as well as variation in each of the two melodies across dialects.
This kind of synchronic description of tonal realisations in a number of Norwegian dialects is the main focus of the ongoing project 'Norwegian Tonal Typology'. I will, however, also deal with one other aspect in my presentation. One of the dialects in my data corpus, the city of Stavanger, has been subject to two earlier tonal investigations, the first of which took place as early as the 1920s (Selmer 1927) and the second in the late 1960s (Fintoft 1970). Together with my own 2001 investigation, this provides a unique opportunity to look at aspects of tonal dialectology in a diachronic perspective, and the question of a possible 20th century change in the tonal makeup of the accent 1 melody in Stavanger will be raised. The label 'high tone dialect' refers to the characteristic high tone signalling main stress in accent 1 words. I want to argue that this is perhaps no longer compatible with realisations found in younger speakers of the Stavanger dialect. My observations from Stavanger will be compared to my data from the Mandal and Egersund dialects respectively, and characteristic similarities and differences will be noted, particularly with respect to stability vs. change in tonal systems.
Fintoft, Knut 1970: Acoustical Analysis and Perception of Tonemes in some Norwegian Dialects. Oslo: Universitetsforlaget
Hognestad, Jan K. 1997: Tonemer i en høytonedialekt. Oslo: Det Norske Samlaget
Selmer, Ernst W. 1927: Den musikalske aksent i Stavangermålet. Opuscula Phonetica,
Fasc. VII. Oslo: A.W.Brøgger
In my talk I shall give an overview of the ideas underlying the project that provides the funding for this conference: Norsk tonelagstyoplogi (Typology of Norwegian Tonal Accents). The analysis of the data collected has unfortunately not yet reached the stage where major results can be presented, but I would like to present some of the methods of measurement that I have adopted in my own part of the project and try to show how these promise to throw light on the variation that is out there.
As an example I will offer a comparison between the dialects of Oppdal and Nord-Gudbrandsdal, spoken in neighboring communities in the northern part of Southern Norway. These dialects are normally considered to belong to the same type with respect to tonal accent. A quantitative analysis of data collected by means of structured recordings of speakers of these dialects, suggest that they are not, however, and that the Oppdal dialect represents a system that is difficult to account for within the analytical framework usually adopted in analyses of Scandinavian, where the difference between accent 1 and accent 2 consists of the absence vs. presence of a lexical tone.
The Scandinavian tonemic split occurred between originally monosyllabic words and originally polysyllabic ones. In this paper I propose that the default melody for an utterance final phrase was LHL, with a delayed peak. The reason for the tonal split was reduction of this default melody to HL in monosyllabic domains. When monosyllabic words later became polysyllabic by epenthesis and cliticization, the split occurred.
Reconstruction of the original tone pattern is possible by applying normal comparative methods to the tonal patterns of present Scandinavian dialects. The evidence from peripheral dialects is especially important in this regard. A distinction between early and late fall is the single characteristic that unifies most Scandinavian dialect (Gårding 1977:45).
Typological evidence from many languages will be brought to bear on our proposal. Neutralization in short domains is very common in tonemic languages. The initial rise in polysyllabic words also turns out to be a common phenomenon, which may be due to a universal constraint against having an initial H, sometimes ranked above a constraint that would otherwise assign H to an initial stressed syllable. The proposed constraint DelayPeak corresponds to the feature [+/-delayed tone] proposed in Lorentz 1981.
The discussion will also touch upon the question of markedness, where Accent 2 seems to be marked in Norwegian/Swedish, whereas Accent 1 is marked in the stød dialects of Denmark. I will also briefly discuss previous analyses of tonogenesis, such as Elstad’s 1980 proposal, which is similar to the present one, Riad’s 1998 proposal that the tonemic split was due to a contrast between words having tone marking of secondary stress and words without such stress, and my own 2001 proposal that a process of Tonal Doubling, known from e.g. Bantu languages and Bosno-Serbo-Croatian, was crucial to the tonemic split.
Elstad, Kåre. 1980. Some remarks on Scandinavian Tonogenesis. Nordlyd 3:6277.
Gårding. Eva. 1977. The Scandinavian Word Accents. Lund: CWK Gleerup.
Lorentz, Ove. 1981. Adding tone to tone in Scandinavian dialects. T.
Fretheim (ed.) Nordic Prosody II. Trondheim: Tapir.
Lorentz, Ove. 2000. Tonogenesis in pitch accent languages. Handout from the Workshop on Tone and Pitch Accent, University of Tromsø, June 2000.
Lorentz, Ove. 2001. Tone Doubling and Scandinavian Tonogenesis. Handout from the Stanford Phonology Workshop, Thursday, March 8, 2001.
Riad, Tomas. 1998. The origin of Scandinavian tone accents. Diachronica XV.1:63-98.
Assume that ’distinction’ refers to grammatically relevant non-identity, and that ’difference’ refers to grammatically redundant non-identity. In accounts for tonogenesis, it is common to assume that some distinction plays a role. Which particular distinction is involved varies between cases and proposals.
a) voiced vs unvoiced obstruents (South East Asian)
b) one vs two syllables (Scandinavian)
c) one vs two stresses (Scandinavian)
d) sg vs pl (Franconian)
Often, a difference is assumed to pattern with the distinction, where that difference is typically the raw material for the new distinction-to-be.
Sometimes the difference is directly caused by the distinction. A case like this is the difference in F0 pitch on the vowel following a voiced or voiceless obstruent in South East Asian tonogenesis (a). Many changes other than tonogenesis are like this, e.g. expansion of vowel systems under umlaut, which is later dropped.
Sometimes the difference just happens to largely co-pattern with the distinction. An example would be the assumed realization of an intonational tone on or after the only stressed syllable depending on word length in Proto-Nordic (b). Another would be the number of syllables which is typically higher with two stresses than with one, in Proto-Nordic (c).
Other times, the distinction is not phonological, but lexical or morphological. This distinction is initially accompanied by a phonological signal, e.g. a phonologically non-null morpheme. When this phonological signal is removed by other changes a situation occurs where the distinction loses its material support. The example here is Franconian (d), where it is assumed that tonogenesis can take place as a remedy the threat of this scenario, alternatively as a remedy to the situation itself.
In my talk I shall discuss the role of distinctiveness in Scandinavian tonogenesis and beyond.