Pitch Peaks Research Articles

Perception of dynamic pitch and prominence in speech

The perception of dynamic – constantly changing – pitch in speech has been extensively studied in psychoacoustics and linguistics. In psychoacoustic studies, listeners are usually presented with short stimuli such as vowels or syllables, and their ability to discriminate a pair of stimuli is assessed. On the other hand, linguistic studies concern intonation over an utterance. Intonation entails not only acoustic prominence realised by pitch, duration, and loudness, but also listeners’ knowledge about the relative prominence between syllables or between words in their language. This paper discusses a series of speech perception experiments using both psychoacoustic and linguistic tasks. Participants judged either relative pitch height or prominence between two pitch peaks or valleys in an utterance. Native English speakers in different age and dialectal groups were tested. The results showed that, first, listeners’ pitch height discrimination in the utterance context seems to be more accurate than previously reported. Second, there is a robust perceptual asymmetry between pitch peaks and valleys, the valleys posing significant challenges in perception. Third, listeners’ perception of pitch height and prominence is disassociated. The findings taken together suggest an intricate interaction between the physical properties of the stimuli and listeners’ top-down knowledge in the perception of speech intonation.

Vocative Intonation in Language Contact: The Case of Bulgarian Judeo-Spanish

The present study investigates the prosodic realization of calling contours by bilingual speakers of Bulgarian and (Bulgarian) Judeo-Spanish and monolingual speakers of Bulgarian in a discourse completion task across three pragmatic contexts: (i) neutral (routine) context—calling a child from afar to come in for dinner; (ii) positive context—calling a child from afar to get a present; and (iii) negative (or urgent) context—calling a child from afar for a chastising. Through quantitative analyses of the F0 span between tonal landmarks, alignment of pitch peaks, intensity, and durational and prominence patterns, we systematically account for the phonetic characteristics of the contours and determine their tonal composition and meaning, thereby situating them within the intonation systems of Bulgarian Judeo-Spanish and Bulgarian. It is shown that both languages use the same inventory of contours: (1) L+H* !H-% (the so-called “vocative chant”), (2) L+H* H-L%, and (3) L+H* L-%. However, their distribution differs across contexts and varieties. Monolingual and bilingual speakers of Bulgarian, on the one hand, predominantly use (1) and (2) in neutral and positive contexts and clearly prefer (3) in negative contexts. In Bulgarian Judeo-Spanish, the bilinguals also more often recur to (3) in neutral and positive contexts and generally show more variation.

The influence of co-speech gesture presence on the timing of F0 peaks in a tonal language

The timing of co-speech gestures has been shown to correlate with the location of pitch peaks in speech in several non-tonal languages. While little research has examined this relationship in tonal languages, existing work suggests that pitch peak timing does not show the same synchronous alignment with manual gesture timing. In this paper, we examine whether the presence of a gesture can have an influence on timing of pitch peaks in Medʉmba, a tonal Grassfields Bantu language. Manual gesture data were analyzed from spontaneous interview speech of six Medʉmba speakers. In line with prior findings on tonal languages, manual gesture apexes and pitch peaks did not align, and gesture-accompanied vowels were not realized with higher or more dynamic pitch profiles (though they were realized with longer duration and greater intensity; ps < .01). The findings did indicate, however, that f0 peaks—particularly for high, falling, and low tone syllables—were realized earlier in gesture-accompanied vowels than non gesture-accompanied vowels (p < .05). We interpret these findings in the context of an integrated model of speech and manual gesture in which tonal gestures and manual gestures are both crucially timed to vowels, but timed to occur sequentially with one another.

Perceptual asymmetry between pitch peaks and valleys

Perceiving pitch in spoken utterances is an important part of the speech comprehension process. Previous studies on pitch perception concentrated on ‘peaks’, neglecting ‘valleys’, thus making it difficult to know how generalisable the findings are. In two experiments, we took a factorial approach and contrasted the pitch movement direction associated with a stressed syllable (‘peaks’ vs. ‘valleys’), the turning point shape (sharp turn, 25, 100 ms plateau), the accent height (varied in one-semitone steps), and pitch levels by shifting the whole contour up or down. We employed different sentences and also non-speech stimuli. Results showed that listeners were better at discriminating the pitch height between ‘peaks’ compared to ‘valleys’. In particular, ‘valleys’ in the low level posed challenges. A long pitch plateau forming a flat turn at a ‘peak’ or ‘valley’ made a ‘peak’ sound higher and a ‘valley’ lower compared to a sharp turn or short plateau of the same height. Similar results were observed across speech and non-speech stimuli. The perceptual asymmetry may lead listeners to allocate more attention to ‘peaks’ than ‘valleys’ in speech processing, while the effect of listeners’ native language deserves further investigation.

Cognitive constraints on advance planning of sentence intonation.

Pitch peaks tend to be higher at the beginning of longer than shorter sentences (e.g., ‘A farmer is pulling donkeys’ vs ‘A farmer is pulling a donkey and goat’), whereas pitch valleys at the ends of sentences are rather constant for a given speaker. These data seem to imply that speakers avoid dropping their voice pitch too low by planning the height of sentence-initial pitch peaks prior to speaking. However, the length effect on sentence-initial pitch peaks appears to vary across different types of sentences, speakers and languages. Therefore, the notion that speakers plan sentence intonation in advance due to the limitations in low voice pitch leaves part of the data unexplained. Consequently, this study suggests a complementary cognitive account of length-dependent pitch scaling. In particular, it proposes that the sentence-initial pitch raise in long sentences is related to high demands on mental resources during the early stages of sentence planning. To tap into the cognitive underpinnings of planning sentence intonation, this study adopts the methodology of recording eye movements during a picture description task, as the eye movements are the established approximation of the real-time planning processes. Measures of voice pitch (Fundamental Frequency) and incrementality (eye movements) are used to examine the relationship between (verbal) working memory (WM), incrementality of sentence planning and the height of sentence-initial pitch peaks.

POLITENESS/IMPOLITENESS IN CONFLICT DIALOGICAL DISCOURSE: PROSODY ASPECT

The given article presents an experimental phonetic investigation, which deals with a comprehensive analysis of English-language formal and informal conflict dialogues. It focuses on the study of various types of the investigated conflict dialogues intonational structure.The basic criteria of the practical material classification are: 1) the degree of relations formality between the collocutors and 2) politeness/impoliteness. According to these criteria the investigated conflict dialogues were classified as formal polite, formal impolite, informal polite and informal impolite. The recorded speech was investigated with the help of computer and statistical analyses which allowed to reveal those elements of prosody which actualize politeness/impoliteness in various types of the investigated conflict dialogues. The prosodic means that most clearly differentiate the investigated dialogical units are the pitch and intensity range, the pitch peaks, the average syllable duration, the average duration of the emphatic centre, the pause volume, types of pauses between the key remarks. The pitch and intensity peaks that coinside with the emphatic centre of a dialogical unity are greatly important for actualizing emphasis. The speech of collocutors who try to remain polite in a formal conflict situation is characterised by a normal tempo with the tendency to slowing down, a mid loudness, a mid pitch range that is due to the formal and official speech situation that forces the interlocutors to follow status subordination. An impolite formal conflict is characterized by a greater degree of emotionality than a polite formal one, as collocutors do not always follow the rules and standards of a formal situation behaviour being rude and harsh with their speech partners. At the prosodic level it is expressed by a faster tempo, a greater loudness and a wider pitch range than in the polite formal communication. The polite informal conflict communication is more emotional and natural than the polite formal one, but less emotional and natural than the impolite informal one. On the one hand the informal situation gives complete freedom to choose emotions and ways of their realization but on the other hand the desire to be polite in order not to offend the speech partner forces the collocutors to restrain their negative feelings in the polite formal conflict communication. The impolite informal conflict is the most emotional and uncontrolled. Expressiveness and uncontrollability in the impolite informal conflict are caused by the the speech situation informality, unwillingness and no need to restrain, complete freedom of choice of verbal and nonverbal means. At the prosodic level this type of conflict interaction is characterised by a high loudness, an accelerated tempo and a wide pitch range of the voice. The conducted research enabled us to state that conflict dialogical discourse represents a peculiar type of speech with a specific prosodic structure.

Automatic assessment of expressive oral reading

We investigated the automatic assessment of expressive children’s oral reading of grade level text passages using a standardized rubric. After a careful review of the reading literature and a close examination of the rubric, we designed a novel set of prosodic and lexical features to characterize fluent expressive reading.A number of complementary sources of information were used to design the features, each of them motivated by research on different components of reading fluency. Features are connected to the child’s reading rate, to the presence and number of pauses, filled-pauses and word-repetitions, the correlation between punctuation marks and pauses, the length of word groupings, syllable stress and duration and the location of pitch peaks and contours.The proposed features were evaluated on a corpus of 783 one-minute reading sessions from 313 students reading grade-leveled passages without assistance (cold unassisted reading). Experimental results show that the proposed lexical and prosodic features provide complementary information and are able to capture the characteristics of expressive reading. The results showed that on both the 2-point and the 4-point expressiveness scales, computer-generated ratings of expressiveness agreed with human raters better than the human raters agreed with each other. The results of the study suggest that automatic assessment of expressive oral reading can be combined with automatic measures of word accuracy and reading rate to produce an accurate multidimensional estimate of children’s oral reading ability.

Intonation of Mandarin speakers in their English as a Foreign language

Does the prosody of speakers' first language (L1) influence their prosody in their second language (L2)? The current work investigates this for tone languages (Beijing Mandarin and Taiwanese Mandarin) as L1, and an intonation/stress language, English, as L2. English uses F0 contours in the intonation system, to signal syllabic prominence in a word, word prominence in a phrase, and the difference between questions and statements. In English, there is a pitch peak delay, where the F0 peak occurs after the stressed syllable in two-syllable stress-initial words. Conversely, tone languages use F0 contours lexically, and in the case of Mandarin, the F0 peak is on the stressed syllable. Thus, F0 measurements were taken from three subject groups (Beijing Mandarin L1, Taiwanese Mandarin L1 and Native English) who produced narrow and broad focus statements and questions, to learn if the Mandarin speakers lack pitch peak delay in English like in their L1, or delay some of their pitch peaks beyond the corresponding syllable offset, like English speakers. Further, all groups produced contrastive focus statements and questions, to see to what degree differences in the L1 system impact L2 production, as Beijing and Taiwanese Mandarin differ in these structures. Data analysis is ongoing.

Realisation of two adjacent high tones: Acoustic evidence from Northern Sotho

We present findings of a multi-speaker production study that was undertaken to investigate the realisation of two adjacent high tones within the verb word in Northern Sotho, a Southern Bantu language. Experimental tokens are selected to ensure that the high tones originate from different combinations of morphosyntactic constituents. It is found that the morphosyntactic constituency determines how the adjacent high tones are realised. When both high tones originate within either the inflectional stem or the macrostem constituents, a single pitch peak is realised. Additionally, when the macrostem contains two high tones, the tone of the object concord is absorbed into the stem. Two adjacent high tones, of which one stems from the inflectional stem and the other from the macrostem, produce two pitch peaks, with the latter of the two delayed in order to satisfy the Obligatory Contour Principle. These generalisations are supported by acoustic data. A set of rules is formulated that describes the surface realisation of adjacent high tones in the verbal domain of Northern Sotho (with the exception of one unresolved issue).

Two speaker speech separation by LP residual weighting and harmonics enhancement

This paper presents a method for separating speech of individual speakers from the combined speech of two speakers. The main objective of this work is to demonstrate the significance of the combined excitation source based temporal processing and short-time spectrum based spectral processing method for the separation of speech produced by individual speakers. Speech in a two speaker environment is simultaneously collected over two spatially separated microphones. The speech signals are first subjected to excitation source information (linear prediction residual) based temporal processing. In temporal processing, speech of each speaker is enhanced with respect to the other by relatively emphasizing the speech around the instants of significant excitation of desired speaker by deriving speaker-specific weight function. To further improve the separation, the temporally processed speech is subjected to spectral processing. This involves enhancing the regions around the pitch and harmonic peaks of short time spectra computed from the temporally processed speech. To do so the pitch estimate is obtained from the temporally processed speech. The performance of the proposed method is evaluated using (i) objective quality measures: percentage of energy loss, percentage of noise residue, the signal-to-noise ratio (SNR) gain and perceptual evaluation of speech quality (PESQ), and (ii) subjective quality measure: mean opinion score (MOS). Experimental results are reported for both real and synthetic speech mixtures. The SNR gain and MOS values show that the proposed combined temporal and spectral processing method provides an average improvement in the performance of 5.83% and 8.06% respectively, compared to the best performing individual temporal or spectral processing methods.

Social action of syntactic reduplication

Reduplication has been shown to carry the semantic meaning of increased intensity, duration or emphasis. This study demonstrates that syntactic reduplication in Estonian is regularly used in responsive positions in action sequences. Instances of syntactic reduplication constitute specific social practices such as affiliative and disaffiliative urging, challenging the prior speaker, reinforcing answers to yes/no questions, and providing a non-elicited confirmation. Syntactic reduplication is a sedimented linguistic pattern grounded in the social actions it recurrently performs. Different reduplicative actions furthermore display characteristic prosodic features, involving initial prominence in affiliative actions and delayed pitch peak in disaffiliative ones. Mock repeats and disconfirming answers are produced with double pitch peaks. Grammar and prosody are complementary means of achieving social action in particular positions in interactive sequences. The paper shows that sequential and social contingencies may be essential in understanding a grammatical pattern.

초점의 음성적 실현에 나타난 영어 화자의 한국어 운율 체계 습득 양상 연구

The study examined interlanguage of English learners of Korean in terms of prosodic features. Six Korean sentences which included focused words were recorded two times by four inexperienced Korean language learner(IEs, 17.8 average months of Korea residence), four experienced Korean language learners(EEs, 6 average years of Korea residence), and four Korean native speakers(Ks). AP-initiation, dephrasing after the focused word, frequency of the pitch peak, the location of the pitch peak, and pitch range were investigated. A perception test also was performed over the IEs, EEs, and Ks. The results revealed that even though length of L2 experience played a great role in acquiring L2 prosodic system, both the IEs and EEs were significantly different from the Ks in terms of the location of the pitch peaks, AP-initiation, the rate of dephrasing, and speech range. Some features were traceable to prosodic features of the L1, namely English. The rest, however, seemed universal features of interlanguage in L2 prosody. (University of Seoul)

An intonational grammar for Icelandic

The aim of this paper is to shed new light on the tonal grammar of Icelandic and to complement the tone inventory as previously described in the literature (Árnason 1998). Specifically, types of pitch accents and edge tones and their combinations in neutral declaratives and questions, and in utterances containing narrow focus are addressed. Two pitch accent types (H* and L*) and two edge tones (H- and L-) are identified, for which evidence has not been found in previous research. Moreover, the paper shows for declaratives, that along with downstep, Icelandic has upstep across Intonational Phrases. Upstep applies to a series of pitch peaks. It may occur in neutral declaratives and in utterances with final narrow focus. Overall, the results of this study provide a substantial addition to our knowledge of Icelandic intonational phonology.

How prosody correlates with syntax: An observation on Sakizaya, an endangered Formosan language

This pioneering study investigates the relationship between prosody and syntax of Sakizaya, an endangered Formosan language in eastern Taiwan. Both elicitation data and spontaneous data were collected from five informants (aged 48 to 74), in order to provide a more thorough sketch on this VSO language with clear case marking and voice systems. Preliminary results showed that in descriptive sentences, prosodic units and syntactic units do correlate with each other, depending on which of the four voices is used. (The four voices in Sakizaya: agent‐, patient‐, instrumental, and locative voices) However, once specific functions are imposed onto a sentence, they will override this "default" setting, and adopt corresponding prosodic patterns for the ease of communication. For example, in this language, the default position to receive a nuclear accent should be the ultimate syllable of a sentence; however, we observed an accent shift onto the penultimate syllable for yes/no questions, exclamations, as well as continuously progressive sentences. In negation sentences, pitch peaks are distinctively higher in the accented syllable of the negator (which posited sentence‐initially), instead of the ultimate syllable in sentence‐final position. Lastly, case markers and voice markers were found as preferred positions for speech planning and repair.

Pitch and voice quality in Burmese vowels

Burmese vowels are pronounced with a four-way contrast in laryngeal configuration, described as high, low, creaky, and killed registers [Bradley, Tonation, Pacific Linguistics Series A-62 117–132 (1982)]. However, acoustic descriptions differ as to whether the registers should be defined in terms of pitch or voice quality [Watkins SOAS Working Papers in Linguistics: 10, 139–149 (2000)]. This study investigates whether voice quality or pitch is the more consistent carrier of contrast across the four registers. Four speakers pronounced 20 words (5 vowels ×4 registers) inside a frame sentence surrounded by syllables of low register. Each sentence was repeated five times consecutively. For each syllable, F0 measures were taken at each quartile, as well as peaks and valleys. Voice quality was quantified in terms of spectral slope [Gordon and Ladefoged, J. Phonetics 29, 383–406 (2001)], specifically the difference in dB between the first and second harmonics in an FFT spectrum over a 50-ms window. Results indicate that pitch more reliably indicates register than voice quality. While there was no significant difference in spectral slope between registers, the four registers differed consistently in height and timing of pitch peaks.

Discourse-level contours in Nehiyawewin

This study describes declination and discourse-sized intonation contours in Nehiyawewin, an Algonquian language whose pitch and intonation systems have not been previously studied. The study draws on 270 min of recordings of two female Nehiyaw elders telling their life stories to another Nehiyawewin native speaker. Data is analyzed by using Praat’s default algorithm to generate f0 curves for each breath group. Preliminary results indicate: (1) When breath-group internal pitch peaks are considered, an obvious downward trend of f0 occurred in fewer than half the breath groups analyzed (about 40% or 37/90). This raises questions about the role of classical declination in natural discourse [Umeda, Journal of Phonetics 10 (1982)]. (2) When we abstract away from declination within a breath group by computing mean f0 and mean pitch peak for each breath group and tracking trends for these means, larger patterns seem to emerge; breath groups can be grouped into larger units based on raising and lowering trends. These units have a mean peak range of 150 Hz with a 30 Hz change from one breath group to the next and correspond to a domain of around five clauses (about 3–4 breath groups, about 45 syllables).

Reflections on studying prosody in talk-in-interaction.

Rather than focusing on conversation as one context among many in which to study prosody, this paper approaches prosody as one set of resources and practices among many by which participants interactively produce conversation and other talk-in-interaction. Three episodes of conversation are examined, each exemplifying a different order of organization in which prosodic practices may be implicated. The first develops various lines of evidence to show that pitch peaks may be deployed and understood as projecting that a next syntactic possible completion is the designed end of the turn. In the second, the initial turns in the opening of a telephone conversation are examined as the site in which the participants work out the pitch level at which the conversation--or at least its first part--will be conducted, and thereby "negotiate" the tenor of the conversation's launching. The third episode focuses on the central part which prosody can play in the constitution of the action which an utterance is implementing. The paper closes with some reflections on what is needed for students of conversation in dealing with prosody--focusing especially on the need for a relevant way of describing the mediating operations which take the prosody as (partial) input and yield the action (or other conversational feature) being accomplished as outcome.

Differences between native and non-native speakers’ realization of stress-related durational patterns in American English

The purpose of this study was to investigate differences between native speakers’ and non-native speakers’ realization of stress-related duration and pitch patterns in English. Anderson-Hsieh et al. [Lang. Learn. 42, 529–555 (1992)] found that prosodic deviance showed a significant negative impact on native speakers’ judgments of non-native pronunciation. Japanese uses pitch to highlight accented syllables; English uses pitch, duration, and loudness. Therefore, it is expected that Japanese speakers use pitch, when speaking English, more than American speakers, but use duration less. Experimental results confirmed this. Three Americans and three Japanese were instructed to read English sentences which included words containing six target vowels, in different stress conditions (reduced, unstressed, and stressed), sentence-medially and -finally. It was found that in American speakers’ utterances, the difference in vowel duration showed constant increase from reduced to stressed. Japanese speakers’ utterances also showed an increase, but the differences among the three stress levels were much smaller than native speakers’. The frequency of the pitch peaks showed the opposite pattern: Japanese speakers showed greater differences than the Americans. These results suggest that Japanese speakers realize English stress using the parameters for realizing pitch accent, leading to less acceptable English speech.

Focus Conditions and the Prominence of Pitch-Accented Syllables

The purpose of the present study is to find out how the pitch peak heights on two pitch-accented syllables in one utterance relate to different focus conditions. The focus conditions are neutral focus, double contrastive focus, and single contrastive focus on either the first or the second pitch-accented syllable. In Experiment 1, subjects adjusted the height of one of two pitch peaks, so as to make the pitch contour express different focus structures. No systematic relationship was found between different fixed heights of one peak and the adjusted heights of the other, which suggests the existence of target values for focus-related pitch peaks. In Experiment 2, listeners judged which of the four focus structures was most likely represented by a given relation between peak heights. The results show that some pitch contours are ambiguous with respect to focus, but the majority of them is classified unanimously as signalling only one possible focus structure. The present results also shed new light on some unexplained findings of earlier prominence experiments.

Effects of stress, vowel duration, and vowel height on the timing of pitch peaks in Czech

Czech is described as having word stress fixed on the first syllable, independent of vowel quantity. The pitch peak of a word generally falls on that first, stressed syllable. However, in some environments, the pitch peak shifts to a later syllable, leading to the perception that this later syllable is more prominent. This study investigated the phonetic factors conditioning such divergent placement of stress and pitch peak in Czech. Native speakers produced contextually embedded trisyllabic nonsense words with high-pitch accents. Vowel height and quantity in the first and second syllable were independently varied to assess their effect on syllable pitch and the timing of the pitch peak. Whether a syllable was stressed constituted the third independent variable in the analysis. To determine whether stress remained on the first syllable, the timing of stress-related properties, i.e., amplitude and syllable lengthening, was examined. Results showed that a high vowel in either syllable tended to raise the pitch, but the second vowel could only attract the pitch peak if the first vowel was short and low. A long first vowel prevented accent shift to the second syllable. Stress-related properties remained largely confined to the first syllable, although the amplitude maximum occasionally shifted to the second syllable, if long.