Focus prosody without accent in some Southern Bantu languages
Most work on the interaction of prosody and focus assumes that, cross-linguistically, there is a necessary correlation between main sentence stress (or accent) and focus. (See, for example, Gussenhoven 1984, 1996, 1999; Reinhart 1995, Samek-Lodovici 2005, Selkirk 1984, 1995, 2004, Root 1992, 1996, Szendröi 2003, Truckenbrodt 1995.) This correlation is mainly supported by European stress languages where cues for stress - like culminative pitch movement and duration - co-occur on the syllable with main sentence stress or focus, lending it unambiguous salience in the Intonational Phrase. Relatively little work has been done on the prosody of focus in languages where focused constituents condition non-culminative prosody, and where culminative prosody is independent of focus.
In this talk, I present the results of fieldwork on the prosody of various focus constructions in some Southern Bantu languages. This work has the following implications for the interaction of focus and accent. I show that phonological rephrasing, unaccompanied by culminative sentence prosody, is the most consistent correlate of focus in these languages. Although the syntactic positions favored for ex situ focus - sentence-initial and immediately after the verb - are ones that prosody highlights, focus does not necessarily correlate with culminative prominence in either position. I show further that focus-related morphemes, rather than the constituent they place in focus, are made salient by phonological rephrasing, creating a further mismatch between prosody and focus. The overall conclusion I will draw is that while culminative prosodies play some role in defining the syntactic positions favored for focus, neither pitch nor sentential stress nor rephrasing provide unambiguous cues to focus in these languages.
Multi-modal expression of Swedish prominence
Speech production has obviously both acoustic and visual effects. Traditionally the acoustic aspects have been the main concern of phonetician. During the last few years we have been interested also in the visual aspects of speech communication. We have studied several aspects, from detailed segmental audio-visual correlates to e.g. expressive prosody in human communication. In this contribution we will discuss different audio-visual aspects of prominence in Swedish, studied with different techniques, such as 3D point registration of facial movements and perceptual experiments with a multi-modal parametric speech synthesizer – a talking head. In one set of experiments the interactions between audio-visual prosodic expressions and different expressive modes have been studied. Prominence show up in virtually all visually registered facial points, but differently for the different expressive modes.
Yucatec Maya word tones in sentence perspective
Bruce (1977) demonstrated the usefulness of systematically varying a number of parameters in a corpus of read speech. In addition to the lexical tone contrast, position in the sentence and focus were varied, revealing effects on the prosodic structure for all three factors. Applying the same strategy to Yucatec Maya, a language with a three-way tone contrast in long syllables, shows effects of lexical tone and position in the sentence, though not of focus. The positional effects suggest that rearticulated syllables allow H-tone to associate with the first mora, while high long syllables associate H with the syllable. Information focus is left unexpressed in the prosodic structure.
Tone and intonation in Kammu
Kammu, a Mon-Khmer language spoken in Northern Laos, is a language that has developed lexical tones rather recently, from the point of view of language history. One of the main dialects of this language is a tone language of the “East Asian” type with (high or low) tone on each syllable, while the other main dialect lacks lexical tones. The dialects differ only marginally in other respects. Our investigation of tone and intonation in Kammu gives us an interesting opportunity to study the intonation contours of nearly identical sentences produced by speakers of dialects with and without tones with the aim of resolving uncertainty about what in the intonation contour is the contribution of lexical tones and what is phrase intonation. In our presentation we will report on a study of the production and perception of tones in different Kammu dialects and discuss possible interactions between tones and phrase intonation.
Alignment of tonal targets: 30 years on
One of the central theoretical innovations in Bruce's 1977 thesis was the idea that the F0 contour of individual syllables is largely irrelevant to intonational phonology. Instead, as he put it, "reaching a certain pitch level at a particular point in time is the important thing". This idea enabled Bruce to distinguish what is invariant about the Scandinavian word accents from what is lawful phonetically or sociolinguistically conditioned variation. In the following 30 years many empirical findings have been derived from the general insight that the key elements of intonational phonology are tonal targets that are aligned with the segmental string in specifiable ways. These include many findings of systematic response to time pressure (e.g. targets shifted earlier or later away from an adjacent tonal target) and many findings of robust segmental anchoring (i.e. consistent alignment of targets over a wide range of segmental contexts).
Data recently collected by my students and collaborators in a number of languages indicate that alignment can be influenced by the SIZE of the word and phrase domains with which a tonal feature is associated. For at least some of these findings there is no obvious explanation in terms of time pressure from adjacent tonal events. For example, in short utterances in both English and Italian, the F0 peak of the nuclear accent is aligned slightly earlier or later depending only on the number of syllables in the sentence. Even after thirty years, in short, Bruce's target-alignment insight continues to generate new and unexpected findings and ideas.
Word accents in Mongolian
Mongolian (Halh dialect) is a language with word accents: words within an utterance have a rising gesture, a bitonal LH, aligned with the two first morae in the word (long vowels are counted as two morae, and short and non-phonemic epenthetic vowels ə as one mora):
LH L H || | | CVV.CV/ə CV.CV/ə || | | µµµµ
The Mongolian word accent has a lexical function in that it reflects word structure (phonological quantity relations), but the phonetic properties of its realisation within a word are decided at the postlexical level, and they cannot be described with reference only to the phonological representation of a word. This is due to the fact that the second mora position is not always lexically defined because of (rather irregular) assimilation processes in casual speech. The realisation domain of the word accent is rather the prosodic word.
Phonetic manifestation of word accents in sentence perspective: comparison of Tokyo and Kochi Japanese
Tokyo and Kochi Japanese were compared in their manifestation of word accent in the following contexts: N no N (with the genitive particle no), Ad V, sentence modifier, and NP ga V (with the nominative particle ga).
Two hypotheses tested were: (1) Syntax prosody mapping is less in the Kohi dialect, and (2) when the second accent appears at a reduced pitch register (downstep), the magnitude of reduction is less for the Kohi dialect. Total of 9 speakers (5 for Kochi and 4 for Tokyo) were analysed for their accent manifestation and accent peaks were measured in semitones to enable cross-speaker comparison more meaningful.
A clear difference in syntax prosody mapping was found for sentence modifier contexts. For a sentence modifier consisting of 3 to 4 accents, Tokyo speakers tended to read it as one unit with downstepped accents while the Kochi speakers did not. As for the magnitude of accent reduction, Kochi speakers consistently had lower magnitude than Tokyo speakers.
Lexical and post-lexical in Scandinavian tonogenesis
Theories about Scandinavian tonogenesis converge on the fact that it is the question of a post-lexical phenomenon that becomes lexical, rather than a phonemic lexical distinction (phonation) shifting expression in some context, as in South East Asian tonogenesis. Commonly, it is assumed that the number of syllables comes to occasion different melodies depending on whether or not there are syllables after the primary stressed syllable, before the end of the word (phrase). This phonetic variation is assumed to subsequently get phonologized (e.g. Öhman 1967, Elstad 1980, Lorentz 2002, Bye 2004, Kristoffersen 2004). In my contribution, I will focus on one general and problematic issue of these accounts, namely the fact that they fail to connect with broad generalizations of modern, synchronic post-lexical accent 2. Presentday accent 2 in compounds is deeply and demonstrably dependent on stress, rather than number of syllables, and in my view this ought to play a role in the analysis of tonogenesis in Scandinavian, or be fully accounted for otherwise. There are two issues:
A. Every dialect appears to have at least some, post-lexically, stress prosodically motivated contexts for accent 2 beside lexical accent 2. The core context appears to be stress clash where the first member of the compound is a monosyllable. This constitutes an argument in favour of the hypothesis that lexical accent 2 develops from a situation originally like that core context (Riad 1998).
B. The number-of-syllables hypothesis will not work for post-lexical accent 2 in all dialect types. If the number of syllables were the crucial generalization, then the accent melody of any dialect should be able to obey and signal it. However, the phonological structure of accent 2 (associations, spreading directions etc.) is relevant for whether or not accent 2 is general or generalizable in compounds. This constitutes an argument against the number-of-syllables rule.
Bye, Patrik. 2004. Evolutionary typology and Scandinavian pitch accent. ms. University of Tromsø.
Elstad, Kåre. 1980. Some remarks on Scandinavian tonogenesis. Nordlyd: Tromsø University Working Papers on Language & Linguistics 3, 62–77.
Kristoffersen, Gjert. 2004. The development of tonal dialects in the Scandinavian languages. Presentation, Typology of Tone and Intonation, Cascais, April 1–3 2004.
Lorentz, Ove. 2002. Delayed peak and tonal crowding in Scandinavian tonogenesis. Ms. University of Tromsø.
Riad, Tomas. 1998. The origin of Scandinavian tone accents. Diachronica XV:1, 63–98.
Öhman, Sven. 1967. Word and Sentence Intonation: A quantitative model. Speech Transmission Laboratory Quarterly Progress and Status Report (STL-QPSR) 2–3.20–54. Dept. of Speech Transmission, Royal Institute of Technology, Stockholm.
Why a Rising Tone is Falling in Mandarin Sentences
Mandarin Chinese is a contour tone language with four lexical tones. Each tone has its own distinctive tone shape and that serves to differentiate meaning: The syllable "ma" can mean "mother", "hemp","horse" or "to scold" depending on whether it has a high level, rising, falling-rising or falling tone.
Tonal production in sentences departs considerably from what is expected of the lexical tones. A rising tone may be produced with a falling shape and a falling tone with a rising shape. These variations arise because speakers shift strategies balancing two demands: to be efficient in the use of articulatory effort and to say things accurately so that the message is understood. Surface tonal variations result from different articulatory strength interacting with tonal context.
In this paper, I will present relevant data, argue that there are principled explanation to the observed phenomena and present modeling solutions.
Sentence-Level Effects on Alaskan Athabaskan Tone
The Athabaskan family is the largest indigenous language family in North America. It includes languages spoken in Alaska, western Canada, on the Oregon and California coasts, and in Arizona and New Mexico. While many of the languages have lexical tonal systems from the same proto-Athabaskan source (loss of final glottalic consonants; Leer 1979, Krauss 1978, 2005), these vary in reflex: some languages have high tone, some low, and some no tone at all. This is true within Alaska as within the whole language family.
Sentence-level intonation is also a factor in Athabaskan prosody. Navajo and Apache, with low tone from historic glottalization, have developed lexical high and mid tones on nearly all other prominent syllables (those with light vowels fall in with the low-toned), leaving few unmarked; these languages can be considered well-saturated. In Apachean, intonational boundaries are marked more by lengthening than by pitch, and lexical tones are clearly expressed (Tuttle and Sandoval 2002, Tuttle 2005).
In some of the eleven Alaskan Athabaskan languages, the majority of syllables are non-tonal, with only those reflecting proto-Athabaskan glottalization being “marked,” and the rest “unmarked,” to use terminology employed by Krauss and Leer. Unmarked syllables receive their pitch from intonation. The effect is one of competition between prosodic systems, with intonation frequently winning in the less-saturated languages. Tuttle (1998, 2003) compared the non-tonal Lower Tanana dialect of Salcha with the low-toned neighboring dialect, Minto, and found that both Salcha and sparse-toned Minto showed similar effects of intonation, including significant final lowering (neutralizing low and non-low final syllables) and pre-final high-pitched sentence accent (also neutralizing low and non-low syllables). Holton (2000) observed final lowering in Tanacross, which has high tone and strong saturation. John Ritter (p.c.) has reported similar facts for the tonal dialects of Upper Tanana.
Ritter (p.c.) and Holton (2000) also show evidence that different sentence types (interrogatives, imperatives) bear different intonational contours in Tanacross and Upper Tanana, contours that compete with lexical tones. In Lower Tanana (Tuttle 1998, 2003), interrogatives and negatives receive special pitches which are tied to verbal enclitics; an intonational source is likely. This represents a distinctly different source for new tone, independent of historic glottalization.
There is thus reason to imagine (as noted by Krauss 2005) that intonation at the sentence level could affect tonal systems diachronically, contributing to failure of transmission of tones with low functional load, and eventually to loss of low or high tone or development of new tonal patterns. However, the roles of social facts such as geography, exogamy and multi-linguality must also be taken into account. For example, Salcha texts show evidence that this dialect used to be tonal, but its position on the river between the now-extinct Chena dialect (low) and Tanacross (high) probably also pushed Salcha toward an intermediate, intonation-only state; Tanacross, which also has intonational effects, has not lost its tone.
There is thus considerable evidence for an ongoing and very lively interaction between tone and intonation in the highly varied tonal landscape of Athabaskan Alaska.
ERP effects of sentence processing and violations of information structure
The present auditory study aims at identifying ERP correlates for the processing of both sentence accents as such and their compatibility relative to a context question (information structure). The spoken German sentence material is based on that introduced by Steinhauer, Alter and Friederici (1999) but varies the positions of accents (as indicated by capitals in the examples below). Due to the wide focus in context-free utterances with 'neutral intonation', sentence accents in the Steinhauer et al. study were dependent on the syntactic structure only:
|A.||Peter verspricht Anna zu ARBEITEN und das Buero zu putzen.|
|Peter promises Anna to WORK and to clean the office.|
|B.||Peter verspricht ANNA zu entlasten und das Buero zu putzen.|
|Peter promises to support ANNA and to clean the office.|
In the present study, all sentences were preceded by context questions establishing a narrow focus on one sentence constituent such that only the corresponding accentuation pattern provided an appropriate answer. The following example illustrates both an appropriate and an inappropriate answer given the context question:
Question B1 (=sentence type B, focus on NP2):
WEM verspricht Peter zu entlasten und das Buero zu putzen?
(WHOM does Peter promise to support and to clean the office?)
compatible answer (B1):
Peter verspricht ANNA zu entlasten und das Buero zu putzen.
(Peter promises to support ANNA and to clean the office.)
incompatible answer (B2):
Peter verspricht Anna zu ENTLASTEN und das Buero zu putzen.
(Peter promises to SUPPORT Anna and to clean the office.)
Acoustic analyses of the 48 speech signals in each condition confirm that the respective prosodic patterns are reflections of the information structure rather than the syntactic structure. Note that incompatible answers violate the required information structure in two ways: (1) they do not provide the required accent on the element in focus and (2) they contain an inappropriate accent.
22 undergraduate students participated in this experiment and judged the prosodic compatibility of questions and answers.
ERP data suggest that in compatible answers, lexical elements carrying the respective accent elicit a centro-parietal positivity. The ERP patterns for incompatible answers appear to be more complex. Depending on both the type of violation and the position in the sentence we observed either a negative slow wave or a biphasic sequence of components. The nature of the ERPs will be discussed in the context of prosodic processing models taking into account the particular role of information structure in every-day conversation.
An 'early fall' in Swedish intonation: utterance or word prosody? - Evidence from a comparison with German.
German has a pitch accent with early peak timing, i.e. where pitch prototypically is high on the prestress, and low on the stressed syllable. This 'early peak' has originally been suggested by Kohler (1987), and has subsequently been incorporated in other models of German intonation (e.g. as H+L* or H+!H* in GToBI; Grice et al. 2005). The communicative function of this accent has been described as signalling 'given information'. 'New information', then, is signalled by a medial (H*) or a late peak (e.g. L*+H). According to Bruce's (1977; 2005) analysis of the Standard (Stockholm) Swedish word accents, non-focal Accent I and Accent II may be represented as H+L* and H*+L. That is, both German and Swedish have got a pitch accent timing distinction, where the early peak of German has received the same phonological description as Swedish non-focal Accent I. However, for German this timing distinction is located at the utterance level, while for Swedish it is located at the word level. The question then arises: What is the Swedish correlate for the German timing distinction? An obvious hypothesis is that an early peak in German functionally corresponds to a non-focal accent (either Accent I or II) in Swedish, while a non-early peak of German corresponds to a focal accent in Swedish (i.e. H+L* H- for Accent I and H*+L H- for Accent II).
In order to test this hypothesis, an investigation has been initiated where data from German and Swedish simulated dialogues are compared. Some preliminary results will be presented and discussed. As predicted from the current models of German and Swedish intonation, an early-timed pitch accent was found both for German and for Swedish Accent I in connection with 'given information'. However, for Swedish Accent II the results are not as expected, because basically the same intonation pattern was found as for Accent I. That is, the word accent contrast was hardly, and if so, only rather slightly maintained in the tested condition. In summary, the intonation pattern produced in the 'given information' condition looks and sounds very similar for German, Swedish Accent I and Swedish Accent II. That is, basically the same phonetic expression is used for the same communicative purpose in the two languages. However, the current phonological models of German and Swedish intonation would analyse this phonetic expression as belonging to different phonological levels (word vs. utterance). Thus, the results claim for a re-analysis of the early fall in Swedish intonation.
A stress and strength analysis of Slovenian clitics
Slovenian clitic pronouns show a set of crosslinguistically rare peculiarities concerning position (Franks 2000, Bošković 2001), stress and use (Dvořák 2003). The pronouns are not only functionally unusual but show a behaviour which goes against most definitions of what clitics are in general. Their ability to be used as short answers (1a) or stressed for contrastive purposes (1b) coincides with Höhle’s (1992) claims about the verum focus in that they can overtake strength functions when stressed. These functions can in some cases be interpreted by the syntax-phonology interface, since the 3rd person singular shows an exception to word order constraints in past tense constructions and there is a tendency towards an iambic rhythmic pattern in Slovenian (2a, 2b), but are explained by an independent phonological rule in most present tense situations, i.e clitic pronouns are independently stressed, as illustrated in (2c), and even in syntactic domains that generally do not allow strength, such as the imperative (3). Acoustic data will be used to illustrate the difference in prosodic form of the clitics in strong and weak positions.
|‘Do you see him?’||‘Do you really see him?’|
|‘I heard, but I didn’t see you.’|
|‘He heard, but he didn’t see you.’|
|Hear1. Sg||Cl.2.Sg.Acc||see1. Sg||Cl.2.Sg.Gen||Part.||Neg.|
|‘I (do) hear, but I don’t see you.’|
|‘I said that you SHOULD trust her.’|
Bošković, Ž. 2001. On the Nature of the Syntax-Phonology Interface.
Dvořák, B. 2003. Elliptische Prädikatisierung enklitischer Personalpronomina im Slowenischen. PhiN 26
Dvořák, B. & R. Gergel 2004. Slovenian clitics: VP ellipsis in yes/no questions and beyond. In: I. Comorovski & M. Krifka ESSLLI 16, Proceedings of the Workshop on the Syntax, Semantics and Pragmatics of Questions, 85-91
Dvořák, B. (2005): Slowenische Imperative und ihre Einbettung. PhiN 33
Franks, S. & T.H. King 2000. A Handbook of Slavic Clitics.
Höhle, T. 1992. Über Verum Fokus im Deutschen. Linguistische Berichte.
Ilc, G. and M. M. Sheppard. 2003. Verb movement in Slovene: a comparative perspective. Sprachtypologie und Universalienforschung (STUF) 56. 3: 266-286.
Repp, S. 2006. Gapping, Negation and Speech Act Operators. (Ph.D., to appear)
Zimmermann, I. To appear. Satzmodus. In: Tilman Berger, Karl Gutschmidt, Sebastian Kempgen and Peter Kosta (eds.).
FLEKKEFJORD REVISITED: Word accent realization and the occurrence of stød-like phenomena in the far South of Norway
The dialect of Flekkefjord has figured more or less prominently in a number of papers and books on Nordic prosodic phonology. The starting point is Larsen (1970), where the author claims that in low-sonority accentual domains, accent 1 is blocked in this dialect, and instead, accent 2 occurs. Larsen’s observations have been quite widely quoted and commented on, e.g. by Monsen (1971), Christoffersen (1981) Liberman (1982), Lahiri et.al. (1999) and Lorentz (2003).
In spite of this considerable amount of interest concerning Larsen’s quite controversial claims, instrumental evidence from Flekkefjord has so far not been available. In my presentation, I want to present data from a number of Flekkefjord speakers which could have a bearing on the questions under discussion by the authors mentioned above. I will also relate my data and analyses to Riad (2000), with particular reference to the phenomenon labelled ‘Eskilstuna curl’. The possible existence of a curl- or stød-like phenomenon in this area, situated on the Southern tip of Norway, is perhaps particularly interesting.
Lexical Tones in Yucatec Maya: Downstep and Tonal Sandhi
Yucatec Maya spoken by a population of 700.000 speakers at the Yucatecan peninsula in Mexico is the only Mayan language that displays lexical tones (Fischer 1976). Despite the contradictory accounts on Yucatec Maya tone, there is some agreement that the tonal system comprises a high tone and a low tone. In addition to that, also syllables bearing no tone at all seem to exist. As far as information structure is concerned, no tonal interaction with intonational tones has been observed (Kügler & Skopeteas 2006, Gussenhoven 2006). Information structural categories such as focus and topic are syntactically expressed by means of clefting or left dislocation (Kügler, Skopeteas & Verhoeven 2006). The present study intends to provide more details on the tonal system of Yucatec Maya. In particular, purely tonal effects such as downstep, upstep or tonal sandhi are considered. Purely tonal effects such as H-raising have been reported for Yoruba (Laniran & Clements 2003), for Mandarin Chinese (Xu 1997) as well as for German (Féry & Kügler submitted). Downstep is well known for tone languages (e.g. Rialland & Somé 2000) as well as for intonation languages such as German (e.g. Truckenbrodt 2003, 2004; Féry & Kügler submitted). The main question is whether these tonal effects can be identified in Yucatec Maya as well.
The data of the present study has been recorded in August 2006 in Berlin where two native speakers have been invited by the universities of Erfurt and Potsdam for scientific purposes. The data have been elicited by means of question-answer pairs. Recordings were made on a portable DAT recorder. The speech materials consist of three different sets which are illustrated below. In all three contexts the target words are placed in the postverbal position, which is neutral with respect to information structure: In sentence (1) the target words are six coordinated nouns, in sentence (2) one single noun, and in sentence (3) a noun phrase consisting of a noun and adjective. In all sentences, target words were selected on the basis of their lexical tonal characteristics. The tonal combinations that have been inserted in the carrier sentences are given in (4). For each set, the nine different items of (4) have been recorded eight times resulting in 432 sentences (2 speakers x 3 sets x 9 items x 8 renditions).
|in-DEF||milpa-D2||LOC||EXIST||sorcerer||turtle||hawk||stinging nettle||and fawn|
|“In the corn field, there are a sorcerer, a turtle, a hawk, a stinging nettle, and a fawn.”|
|“There is ____ in the corn field. ”|
|“I have seen _____(N) _____(A) in the corn field. ”|
|(4)||Tonal combinations:||Similar tones||Alternations|
|H H||H L||H N|
|L L||L H||L N|
|N N||N H||N L|
A first inspection of the data reveals that tonal interactions rather concern sandhi phenomena than scaling phenomena such as downstep or upstep. The classical view of downstep is that a low tone in a tonal sequence H L H triggers downstep on the second high tone (Pierrehumbert 1980). This, however, does not appear to be the case in Yucatec Maya. Similarly, H-raising which is triggered by a following low tone such that the previous high tone is realised above a default reference line (Laniran & Clements 2003) could not be observed in the data. Finally, upstep has been occasionally observed for Yucatec Maya (Kügler & Skopeteas 2006). Yet, upstep seems to concern an interaction of a lexical tone with a phrase boundary tone since upstep between two or more lexical high tones as in ((1) – (3)) has not been observed. These findings will be discussed in the light of recent work concerning the interaction of tones and intonation.
Féry, C. & Kügler, F. submitted. German as a tone language. Submitted to Journal of Phonetics.
Fisher, W.M. 1976. On tonal features in the Yucatecan dialects. Mayan Linguistics 1, 29-43.
Gussenhoven, C. 2006. Yucatec Maya tone in sentence perspective. Poster at LabPhon 10, Paris.
Kügler, F. & Skopeteas, S. 2006. Interaction of Lexical Tone and Information Structure in Yucatec Maya. Proceedings of the second International Symposium on Tonal Aspects of Language (TAL-2), Université de la Rochelle, 83-88.
Kügler, F., Skopeteas, S. & Verhoeven, E. 2006. Encoding Information Structure in Yucatec Maya: On the Interplay of Prosody and Syntax. To appear in ISIS – 8, (2007).
Laniran, Y. O. & Clements, G. N. 2003. Downstep and high raising: interacting factors in Yoruba tone production. Journal of Phonetics 31: 203-250.
Rialland, A., & Somé, P. A. 2000. Dagara downstep: How speakers get started. In V. Carstens, & F. Parkinson (Eds.), Advances in African linguistics. Trends in African Linguistics, Vol. 4. Trenton, NJ: Africa World Press, 251–263.
Pierrehumbert, J. B. 1980. The phonology and phonetics of English intonation. Doctoral dissertation, MIT.
Truckenbrodt, H. 2002. Upstep and embedded register levels. Phonology 19: 77–120.
Truckenbrodt, H. 2004. Final lowering in non-final position. Journal of Phonetics 32: 313–348.
Xu, Y. 1997. Contextual tonal variations in Mandarin. Journal of Phonetics 25: 61–83.
Japanese Downstep Revisited
In this study, I present the experimental results that show several new findings about the interaction of syntactic structure, downstep, and narrow focus.
Japanese intonation has been extensively studied since Poser (1984), Pierrehumbert & Beckman (1988), Selkirk & Tateishi (1988, 1991), Kubozono (1993), and many others. Downstep is considered to be triggered by lexical pitch accents (H*L) in Japanese. The so-called Major Phrase (MaP, aka intermediate phrase) is the domain of downstep. Selkirk & Tateishi (1991) proposed a syntax-prosody mapping principle, which basically predicts that downstep is blocked at the beginning of an XP. Focus has been sometimes analyzed in parallel to downstep and MaP formation (e.g., Nagahara 1994).
Stimuli of the conducted experiment were created to test the prosodic effects of the following factors: (i) lexical pitch accent; (ii) narrow focus; and (iii) syntactic (XP) boundary. Among the various findings from the results, noteworthy are the following two points:
- Downstep does not show a complete reset at the syntactic boundary.
- F0-related effects driven by focus show a different behavior from the pitch reset (after downstep) at the syntactic boundary.
These results raise several problems for the earlier analyses mentioned above in several respects. For example, the definition of MaP, as well as the proposed syntax-prosody mapping principle requires some revisions. Furthermore, a new analysis of focus intonation would be needed. We discuss the implications of the results, and potential solutions to the problems.
Intonation, word order, and the brain
The study aims at investigating how the brain processes the relationship between intonation and word order. In Swedish subordinate clauses, a sentence adverb such as inte 'not' canonically precedes the inflected verb, as in Han sa [att han inte sov] '(lit.) He said [that he not slept]'. Some matrix verbs also allow for their subordinate clause to have main clause word order, with the sentence adverb following the inflected verb, i.e. Han sa [att han sov inte ] '(lit.) He said [that he slept not]'. Roll (2004, 2006) found a high tone at the left boundary of subordinate clauses with main clause word order. Some verbs cannot take subordinate clauses with main clause word order, e.g. hoppas 'hope' in *Jag hoppas att han sov inte '(lit.) I hope that he slept not'.
We record EEG while subjects listen to sentences where the matrix verb may or may not take a subordinate clause with main clause word order. The matrix clauses are paired with subordinate clauses with main or subordinate clause word order. The subordinate clauses do or do not contain an initial high tone. The primary interest is to see whether the prosodic variation is enough for the brain to detect that main clause word order is going to be presented. In that case, there should be an effect for mismatch between matrix verb type and presence of an initial high tone in the subordinate clause. The prosodic variation should also influence a possible effect at the sentence adverb when there is a mismatch between word order in the subordinate clause and matrix verb type.
Roll, Mikael. 2004. Prosodic cues to the syntactic structure of subordinate clauses in Swedish. MA thesis. Lund University.
Roll, Mikael 2006. Prosodic cues to the syntactic structure of subordinate clauses in Swedish. In Gösta Bruce and Merle Horne (eds.), Nordic Prosody: Proceedings of the IX:th conference, Lund 2004. Peter Lang, 195-204.