Contour Tones Research Articles

Tonal coarticulation in Thai

In tones, coarticulatory effects occur in much the same way as they do between consonants and vowels. Earlier studies on tonal coarticulation leave open the question as to whether the primary direction of coarticulatory effects is the same across tone languages, and whether the effect involves primarily a change in tone height or a change in slope as well. The present study investigated the magnitude and extent of tonal coarticulation in Thai. Ten speakers produced all 25 possible sequences of two tones from the five tones of Thai embedded in a carrier sentence. Tonal contours were analyzed in terms of height and slope of the fundamental frequency. Results from statistical analysis indicated that the coarticulatory effects were not symmetric. Thai tones were more influenced by carry-over than by anticipatory coarticulation. Carry-over coarticulation affected a greater number of Thai tones and extended farther into adjacent tones. Coarticulation affected primarily height; slope was relatively unaffected. Magnitude of the effects was fairly uniform regardless ofdirection of coarticulation.

Tone Sandhi and vowel deletion in Margi

Within the theoretical framework of nonlinear phonology, this paper proposes an account of tone sandhi and vowel deletion in Margi, a Chadic language spoken in Northern Nigeria. The database is Hoffman's Grammar of the Margi Language. Language-specific tonal processes in Margi are shown to originate in tone trapping, i.e., the impossibility for a tone to anchor to a skeletal slot by a universal mechanism. The paper identifies the circumstances leading to tone trapping (e.g., Vowel Elision) and formalizes the various tone-rescue processes available both word-internally and across words. Whereas trapped high tones are always saved (either taking over low-tone vowel positions or giving rise to contour tones), trapped low tones may remain trapped throughout a derivation and thus receive no phonetic realization (by universal convention).

Properties of the structure of multi-tone sequential patterns that determine the difficulty of perceptually isolating single target components

A method of adjustment was used to establish the importance of each of several structural properties of the context tones, in nine-tone sequences, in determining the perceptual isolability of target components. Successful ‘‘perceptual isolation’’ of a target tone was assumed to be achieved when frequency matches were as accurate as those achieved for tones presented in isolation, generally meaning less than 1%–2% for the 50-ms tones in these sequences. The context property that was found to primarily affect the frequency matches was the separation, in Hz, between the target tone and both the local and (to a lesser degree) the remote context tones. Other than its bandwidth, the form of the local pitch contour (the target tone plus the single tones immediately before and after it) had no clear effect on the ability to ‘‘hear out’’ the target tone, i.e., whether the local context was ascending, descending, concave up, or concave down. The contours of the remote context tones (first and last three in the patterns) likewise had no effect on performance. Performance ranged from 25% target tones isolated for the most difficult conditions to 90% for the easiest.

Rime length, stress, and association domains

Every regular Chinese syllable has a syllable tone (the tone we get when the syllable is read in isolation). In some Chinese languages, the tonal pattern of a multisyllabic expression is basically a concatenation of the syllable tones. In other Chinese languages, the tonal pattern of a multisyllabic expression is determined solely by the initial syllable. I call the formerM-languages (represented by Mandarin) and the latterS-languages (represented by Shanghai). I argue that there is an additional difference in rime structures between the two language groups. In S-languages, all rimes are simple, i.e., there are no underlying diphthongs or codas. In M-languages, all regular rimes are heavy. I further argue that a syllable keeps its underlying tones only if it has stress. Independent metrical evidence tells us that heavy rimes may carry inherent stress. Thus, in M-languages, all regular syllables are stressed and retain their underlying tones (which may or may not undergo further changes). In contrast, in S-languages, regular rimes do not carry inherent stress; instead, only those syllables that are assigned stress by rule can keep their underlying tones and hence head a multisyllabic tonal domain. The present analysis has theoretical implications for the analysis of contour tones, the tone bearing unit, and the interplay between stress and tone.

Tone Contours and Tone Clusters in Iau

Proceedings of the Eighteenth Annual Meeting of the Berkeley Linguistics Society: Special Session on The Typology of Tone Languages (1992)

Perceptual asymmetry in psychosis-prone college students: evidence for left-hemisphere overactivation.

The hypothesis that psychosis-prone students demonstrate a pattern of exaggerated perceptual asymmetry across both left- and right-hemisphere dichotic-listening tasks (consonant-vowel [CV] and tonal contour discrimination) was investigated. Subjects who scored high on the Perceptual Aberration or Magical Ideation scale or both (n = 20) demonstrated a significantly exaggerated right-ear advantage on a CV task in relation to normal control subjects (n = 27) but showed a reduced left-ear advantage on a tone task. The hypothesis of exaggerated functional lateralization across hemispheres in the psychosis-prone subjects was not supported, but the results are consistent with a hypothesis of left hemisphere overactivation in this sample.

Contour-tone spreading and tone sandhi in Danyang Chinese

An important contribution to our knowledge of tone sandhi among the Chinese dialects is Lü's (1980) article on the tones and tone sandhi behaviour of Danyang, a Wu dialect of Chinese. Lü's description of Danyang is, to date, our only source on the dialect. While it is a northern Wu dialect, the tone sandhi patterning in Danyang differs from Shanghai, Suzhou, Wuxi and other dialects in the vicinity. There are a number of interesting problems related to tone in the dialect. This paper restricts the topic to only one of these problems, namely the treatment of the six basic tone patterns in Danyang, focusing in particular on the pattern in which a contour tone is copied Onto adjacent syllables in the tone sandhi domain.

Generative and non-linear phonology By Jacques Durand (review)

REVIEWS363 Zubizarreta, Maria Luisa. 1982. The formal interaction of harmony and accent: The tone pattern of Japanese. The structure of phonological representations, ed. by Harry van der Hülst and Norval Smith, Part 11:159-212. Dordrecht: Foris. Department of Linguistics[Received 30 July 1990.] University of California Berkeley. CA 94720 Generative and non-linear phonology. By Jacques Durand. London & New York: Longman, 1990. Pp. xiii, 337. Reviewed by David Odden, Ohio State University Writing an introductory text in a field such as phonology, with its diverse approaches, must be a difficult and unenviable task. In GNLP Durand seeks to provide a broad perspective on issues in linear and nonlinear generative phonology, but two problems detract significantly from its usefulness as a textbook . First, the focus is on presenting theories, with little attention to developing the student's appreciation for good analysis and argumentation. There are no problem sets, and no extended analytic exemplars in the fashion of the Kenstowicz & Kisseberth text (1979). Arguments are often superficial and seriously flawed, and D does not provide the kind ofthorough and incisive argumentation that students need to model their own research on. For instance, the case against the linear treatment of contour tones in Ch. 7 is largely based on one weak argument that linear phonology does not explain the naturalness of certain assimilations. We are not given the stronger argument that the linear treatment is so powerful that it allows the formulation of legions of unattested rules, whose nonexistence is explained on principled grounds in the autosegmental theory. The second problem is the coverage of topics. Although D covers both linear and nonlinear phonology, only one third ofthe book is concerned with nonlinear theories, including Dependency Phonology (DP). Despite this, the coverage of nonlinear phonology almost turns out to be better than the coverage of linear phonology. The central technical issues of linear phonology are ignored, or treated only in passing. The linear theory of notation is never given a systematic treatment. Little is said of the rule ordering debate, iterative versus simultaneous rules, exception features, the cycle, or absolute neutralization. Ch. 1 provides an introduction and argues for using features in rules, with English aspiration given to show the superiority of features. The argument is bolstered with an analysis of onset clusters in English. Here it is noted that CCC clusters are of the form s - {p,t,k} - {l,r,y,w}. Again, {p,t,k} function as a natural class—this rule is stated with symbols, not features. The next step is to explain why y does not appear postconsonantally, contrary to the prediction of our rule; this motivates a condition against (C)CyV. But D does not point out that, if such conditions are stated in terms of symbols, we can eliminate the negative condition by removing y from the list of consonants in the third position: under closer scrutiny an argument for features becomes an argument against features. 364 LANGUAGE. VOLUME 67. NUMBER 2 (1991) The second chapter covers distinctive features. Having made a case for stating rules with features, we consider whether features might be abstract, lacking intrinsic phonetic content. It is stated that this is unsatisfactory, since the mapping between phonological and phonetic features becomes arbitrary. This is as far as the argument goes; we never consider why this is undesirable. The argument should be fleshed out: a theory in which phonological features are arbitrarily mapped onto phonetic features is more powerful than one in which phonological and phonetic features are the same, since the former theory properly includes the latter. D then provides a sketch of features. In the discussion of the features for glides there is a typo (45)—y and w are assigned the value [ + ant]. The assignment of [ + low] to the laryngeals h and ? is a repetition from SPE (Chomsky & Halle 1968:307), and certainly contradicts the definition of [low]. This anomaly goes unnoticed. Ch. 3 touches on the question of binary versus scalar or unary features, markedness, and the theory of 'gestures' in DP. D argues against a theory of privative features which assumes features like 'voice', 'voiceless', etc. The strongest argument for binarity is based on place assimilation of preconsonantal...

Expectancies generated by tonal contour

Expectancies generated by tonal contour were examined in a two-part study. Listeners (N = 61) were presented with eight-note sequences followed by a probe tone. Subjects rated the fit of the probe to the sequence in a two-level (good or poor), forced-choice decision. For each probe, the proportion of good-fit replies provided a rating of its expectancy. In part 1, ascending and descending scales provided a key-defining context [Krumhansl and Kessler, Psychol. Rev. 89, 334–368 (1982)]. Cluster analyses on individual probe-tone profiles for all 13×2 contexts classified subjects into 12 groups: 5 were clearly identified as triadic (–n = 11), diatonic (n = 17, n = 3), or proximity (n = 3, n = 3). In part 2, 12 sequences representing degrees of higher-order sequential expectations provided eight-note contexts for the probe-tone task. For the triadic and the large diatonic groups, profiles from the 12 melodies when compared against the baseline profiles indicated effects of hierarchical contour. Thus sensitivity to higher-order contour structure influences expectations for particular tones. [Work supported by NSERC.]

Spectral slope as a cue for the perception of breathy and non-breathy stops in Shanghainese

In Shanghainese, breathy excitation following release distinguishes category 2 (so-called “voiced unaspirated”) stops from category 1 (“voiceless unaspirated”) stops. The most obvious acoustic correlate of bresthiness is relative spectral slope: During the first 50–100 ms after release, H1-H2 is greater for category 2 than for category 1 in both word-initial and word-medial (morpheme-initial) stops (Ren, unpublished). To investigate the perception of breathiness, a series of synthetic disyllables was prepared in which the spectral slope during the 100 ms following the release of the word-medial stop was increased step by step increasing the value of the “open quotient” in the computation of the glottal waveform—a technique previously used for the synthesis of breathy vowels [C. Bickley, Work. Papers MIT Speech Commun. Lab. 1, 71–81 (1982)]. Other potential cues (closure duration, pulsing during closure, tone contour) were neutralized. Ten Shanghainese speakers labeled the stimuli with greater spectral slope as category 2 and those with smaller spectral slope as category 1. [Work supported by NICHD Grant 01994.]

Evaluation of a physiologically based auditory model using psychophysical data

A physiologically based auditory model [O. Ghitza, J. Phon. 16(1), 109–123 (1988)] was tested against human psychophysical data. The model consists of a set of cochlear filters inferred from frequency selectivity of primary auditory fibers in cat, rate‐level processing, and synchrony extraction. Isoloudness contours for pure tones are calculated from the model and are compared to those measured in human by Fletcher and Munson (1933) and by Robinson and Dadson (1956). The model is also used to simulate Scharf's [J. Acoust. Am. 31, 365–370 (1959)] critical‐band measurements using complex sounds of four components. In addition, both calculated and measured critical‐band results are plotted as a function of cochlear place using formulas for human and cat cochlear maps in Greenwood's recent paper (submitted to J. Acoust. Soc. Am.). The comparison of calculated and measured data suggests possible improvements of the model. Laboratories, Acoustics Research Department.

Larynx frequency information for lipreading speech in Mandarin Chinese

Prosodic contrasts are important sources of information in a tone language like Chinese. Each syllable is characterized by a tonal contour that determines the lexical meaning of the word in isolation. In continuous speech, the overall stress and intonation patterns contribute significantly to understanding speech. These contrastive speech patterns are directly related to the fundamental frequency of the speaker's voice, which is relatively invisible. When speech communication is dependent on lipreading, auditory voice pitch information can thus be the most effective aid. That this is so in an English‐speaking context has been shown in numerous studies. The present work evaluates the contribution of larynx frequency information to lipreading connected discourse, and specifically to resolving ambiguities that are primarily signaled by contrastive stress patterns and tone patterns in Mandarin Chinese. The results have important implications for the design of speech processing aids and the development of speech perceptual assessments in a tone language community.

Contour tones

Recent work (Clements 1985; Sagey 1986) on the structure of distinctive features has analysed affricates and prenasalised stops as involving branching for the features [continuant] and [nasal] respectively. This analysis explains the edge effects associated with such segments, while simultaneously identifying them as single melodic elements that associate as units to prosodic templates. This paper will argue that some languages have contour tones that show parallel properties to those of affricates: they associate as units, but also exhibit edge effects. Such behaviour can be simply understood if tonal features hang off a tonal root node (Archangeli & Pulleyblank forthcoming), and this tonal root node is allowed to branch. A high rising contour tone will have the structure shown in (i), where [upper] is the tonal root node, and [raised] is free to branch.

Sensitivity of Single Neurons in Auditory Cortex to Contour: Toward a Neurophysiology of Music Perception

The responses of single neurons in the primary (AI) and secondary (AII) auditory cortical fields of waking cats were recorded during presentation of sequences of five iso-intensity tones. Tonal contour was varied by employing sequences that either increased or decreased monotonically in frequency, or in which contour was non-monotonic. The vast majority of neurons in both auditory fields (84%) exhibited sensitivity to tonal contour, that is, the response to at least one frequency (usually more) in a sequence differed significantly between two or more types of contour. Although all of the mechanisms involved have not been identified, the serial position of a tone within a sequence could explain some of the findings. Responsivity may also depend upon the magnitude of the interval difference between two consecutive tones. Additionally, contour per se could be effective even for neurons that were not sensitive to differences in type of contour; for example, responses to the same tones could be facilitated when presented in sequences of different frequencies compared to sequences of the same frequency. Further, the range of frequency response and the best frequency of a neuron may depend upon whether these parameters are determined using isolated tones or contoured sequences. Finally, omission of one tone within a sequence produced significant modification of response to other tones in most (87%) cells, with effects extending up to 3 sec following the the gap. Neurons in both primary and secondary auditory cortical fields displayed all effects, the only difference was that neurons in All were more likely to exhibit sensitivity to tone omission than those in AI. The findings are discussed with reference to neurophysiological and psychophysical data. We conclude that the effects of contour and other aspects of tonal sequences are detectable at the level of single neurons in auditory cortex of waking animals and that the use of dynamic stimuli is likely to be critical for a neurophysiological account of music perception and the perception of other complex acoustic stimuli.

An acoustic study of tone sandhi in Taiwanese

In Taiwanese, there are five long tones: high level, high falling, mid falling, low rising, and mid level. These tones undergo sandhi changes when they do not precede a major syntatic boundary: mid level→mid falling→high falling→high level→mid level←low rising. The aim of this study was to validate these phonological observations by investigating the acoustic properties of tones in sequence. Another issue of interest was whether tonal coarticulation occurs in Taiwanese. Six native speakers produced several repetitions of a sequence of two syllables /si/ and /do/ in sentence‐final position with all combinations of tones. These utterances were analyzed to determine the average fundamental frequency (F0) contours of the five tones on each syllable. The tonal contours on /do/ syllables were highly similar to those observed previously in isolated syllables. Some slight perserveratory effects of the preceding /si/ tones on the F0 onset of /do/ tones were found. The tones on the penultimate syllable /si/ underwent sandhi changes, as expected. No coarticulatory effects were found on /si/ tones. However, the sandhi tones differed in their F0contours from those on the final /do/ syllable. [Work supported by NICHD.]

Floating tones and contour tones in Kenyang

Tonal alternations in the Bantu language Kenyang appear on first consideration to be rather complicated but yield to analysis into a small number of rules, which reveal interesting properties of floating tones, contour tones, and the tone-bearing unit in the language. This study focuses on the following problems. First, there is a phonetic contrast, found only at the end of the utterance, between the downgliding L of eket and the unreleased L of basemo. Unreleased L will be shown to derive from rising tone. Second, I argue that syllable final consonants may be tone-bearing, a claim supported by analysis of tone alternations resulting from postlexical resyllabification. Third, Kenyang uses floating L prefixes to form morphological verb tense distinctions. There is a behavioral contrast between the free L tone marking the progressive, which triggers downstep and blocks a spreading rule, versus the free L used in the recent past, which docks to the first root vowel, thereby causing the root tone to shift rightward. The analytic problem is to find a way to represent these two types of floating L. The distinction can be handled by assigning them to different levels of the lexical phonology, so that the shift-inducing L is added when verb roots are inserted, but the float-only L is added at a later stratum-. Finally, I show that the interaction between the two rules H Spreading and Fall Simplification provides evidence for the cyclic application of postlexical rules.

Tone production deficits in nonfluent aphasic Chinese speech

Previous studies demonstrating a right hemisphere lateralization of musical, tonal, and intonational stimuli had suggested that in aphasic tone language speakers, the tonal phonemes might be subject to a lesser degree of deficit than consonantal phonemes. Using a word repetition task, this research demonstrates that left-damaged nonfluent aphasic speakers of Chinese experience a tonal production deficit which is both quantitatively and qualitatively equivalent to the deficit experienced by these speakers in the production of consonants. It is suggested that in tone languages, lexical specification of tone contour information results in left hemisphere lateralization of that information, thus making the tonal phonemes vulnerable to left hemisphere damage.

Musical Imagery

An objective indicator of musical imagery is developed that involves tracking the up and down movements of the tonal contour of an imagined musical phrase or tune. In two experiments, college students' imagery of music was examined. In both experiments, subjects learned musical phrases with words (songs) and without words (melodies). They then indicated as rapidly as possibly the tonal contour. In Experiment 1, the primary issue was whether musical imagery (as distinct from kinesthetic or visual imagery) drew on the same representation as overt song. Subjects processed the phrases by using either an imaginal or overtly sung representation. No difference in processing time was found between the imaginal and overt modes of representation, consistent with a common representation. A second issue was "tonal primacy," the priority of tonal coding over verbal or word coding in musical phrases; in fact, songs (with words) were processed as well or better than melodies (without words). No evidence favoring tonal primacy was found. In Experiment 2, the issues examined were possible kinesthetic or visual image coding of pitch representation and possible sharing of tonal and verbal generation processes for musical imagery and auditory imagery. Spoken responses for classifying tonal relations took longer than written responses, indicating that kinesthetic and visual image coding was unlikely and that the pitch generation of musical imagery shared resources with a more general auditory imagery.

The Comparative Study of a New Pattern Recognition Based Exact Coding Method with Conventional Coding Methods for Two Tone Graphics

A new Pattern Recognition (P R) based coding method for two tone contour type of graphics is introduced in this paper. The basic P R concept used here is the detection of a digital line segment in accordance with the Freeman's definition of digital line. The relative data compressibility of the new method is compared with that of different widely used representative coding methods. The relative performances of all methods are studied both on normal and noisily degraded data. The effect of gradual increase of noise level on data compressibility is studied for all the methods. The extension of the new methods is also proposed.

On the perception of intonation from sinusoidal signals: Tone height and contour

Our recent investigations of intonation have employed signals that consist of three or four time‐varying sinusoids. Each tone reproduces the changes in formant center frequencies from a natural utterance. Such sentences produce a clear impression of intonation though there is no fundamental frequency common to the sinusoids composing the pattern. We find that the tone corresponding to the first formant is the basis for sinusoidal intonation, much as signal periodicity in the region of the first formant (not coincidentally, the dominance region) provides the information for pitch in natural speech. A series of differential similarity tests was used to determine whether listeners represent sinusoidal intonation as patterns of specific tone heights or as patterns of relative pitch changes. The results are discussed in terms of auditory frequency coding of the speech signal, intonation, whispered speech, and musical pitch. [Supported by the National Institute of Child Health and Human Development.]