Exploring Interspeaker Variation in Creaky Voice in Dutch.

OBJECTIVES: This study explored the extent and discriminatory potential of interspeaker variation in creaky voice in Dutch men. METHODS: Intervals of creaky voice for 30 speakers were manually segmented and annotated from a corpus of spontaneous speech data. For each speaker, at least 1500 syllables were analyzed. Total creakiness was calculated based on the proportion of creaky syllables. Creaky intervals were categorized into subtypes based on the degree of periodicity. Furthermore, acoustic measurements were taken from the intervals and tested for speaker-discriminating capacity by means of a linear discriminant analysis (LDA). RESULTS: Speakers differed in what percentage of syllables they realized with creaky voice, with a range of roughly 0-5% of all syllables. They likewise differed in the proportion with which they used different subtypes of creaky voice, such that some speakers have very distinctive profiles. The LDA resulted in correct classifications of creaky intervals to speakers at a rate above chance level. CONCLUSIONS: Interspeaker variation in creaky voice in Dutch male speech was confirmed and allowed for moderate speaker classification on the basis of speech acoustics.

Language Dependency of /s/ Production: Native Dutch Versus Non-Native English.

With forensic recordings being collected in multiple languages increasingly often, this study investigates the language dependency of the voiceless alveolar fricative /s/ in speakers of native (L1) Dutch and non-native (L2) English. Due to phonetic similarity between the languages, Dutch learners of English may exhibit language-independent /s/ acoustics, making it an interesting feature for multilingual forensic speaker comparisons (FSCs). However, the findings show that out of the four spectral moments, center of gravity, standard deviation (SD), skewness, and kurtosis, only SD remained stable across the languages; the other measurements were language-dependent. The results were largely independent of the /s/ tokens' contexts, although an interaction between language and context was found for skewness and kurtosis: With a labial right phonetic neighbor, language dependency was largely reduced. The findings have implications for FSCs: as /s/ is language-dependent in speakers of L1 Dutch and L2 English, it shows limited potential for cross-linguistic speaker comparisons in forensic casework.

The language dependency of /m/ in native Dutch and non-native English.

In forensic speaker comparisons, the current practice is to try to avoid comparisons between speech fragments in different languages. However, globalization requires an exploration of individual speech features that may show phonetic consistency across a speaker's languages. We predicted that the bilabial nasal /m/ may be minimally affected by the language spoken due to the involvement of the rigid nasal cavity in combination with a lack of fixed oral articulatory targets. The results show that indeed, L1 Dutch speakers (N = 53) had similar nasal formants and formant bandwidths when speaking in their L2 English as in their native language, suggesting language-independency of /m/ within speakers. In fact, acoustics seemed to rely more on the phonetic context than on the language spoken. Nevertheless, caution should still be exercised when sampling across languages when the languages' phoneme inventories and phonotactics show substantial differences.

Long-term within-speaker consistency of filled pauses in native and non-native speech.

Filled pauses are widely considered as a relatively consistent feature of an individual's speech. However, acoustic consistency has only been observed within single-session recordings. By comparing filled pauses in two recordings made >2.5 years apart, this study investigates within-speaker consistency of the vowels in the filled pauses uh and um, in both first language (L1) Dutch and second language (L2) English, produced by student speakers who are known to converge in other speech features. Results show that despite minor within-speaker differences between languages, the spectral characteristics of filled pauses in L1 and L2 remained stable over time.

Acoustic and speaker variation in Dutch /n/ and /m/ as a function of phonetic context and syllabic position.

In forensic speech science, nasals are often reported to be particularly useful in characterizing speakers because of their low within-speaker and high between-speaker variability. However, empirical acoustic data from nasal consonants indicate that there is a somewhat larger role for the oral cavity in nasal consonant acoustics than is generally predicted by acoustic models. For example, in read speech, nasal consonant acoustics show lingual coarticulation that differs by nasal consonant, and syllabic position also seems to affect realizations of nasal consonants within speakers. In the current exploratory study, the within- and between-speaker variation in the most frequent nasals in Standard Dutch, /n/ and /m/, was investigated. Using 3695 [n] and 3291 [m] tokens sampled from 54 speakers' spontaneous telephone utterances, linear mixed-effects modeling of acoustic-phonetic features showed effects of phonetic context that differed by nasal consonant and by syllabic position. A subsequent speaker-classification test using multinomial logistic regression on the acoustic-phonetic features seems to indicate that nasals displaying larger effects of phonetic context also perform slightly better in speaker classification, although differences were minor. This might be caused by between-speaker variation in the degree and timing of lingual coarticulatory gestures.

Cross-linguistic filled pause realization: The acoustics of uh and um in native Dutch and non-native English.

It has been claimed that filled pauses are transferred from the first (L1) into the second language (L2), suggesting that they are not directly learned by L2 speakers. This would make them usable for cross-linguistic forensic speaker comparisons. However, under the alternative hypothesis that vowels in the L2 are learnable, L2 speakers adapt their pronunciation. This study investigated whether individuals remain consistent in their filled pause realization across languages, by comparing filled pauses (uh, um) in L1 Dutch and L2 English by 58 females. Next to the effect of language, effects of the filled pauses' position in the utterance were considered, as these are expected to affect acoustics and also relate to fluency. Mixed-effects models showed that, whereas duration and fundamental frequency remained similar across languages, vowel realization was language-dependent. Speakers used um relatively more often in English than Dutch, whereas previous research described speakers to be consistent in their um:uh ratio across languages. Results furthermore showed that filled-pause acoustics in the L1 and L2 depend on the position in the utterance. Because filled pause realization is partially adapted to the L2, their use as a feature for cross-linguistic forensic speaker comparisons may be restricted.

The effect of word class on speaker-dependent information in the Standard Dutch vowel /aː/.

Linguistic structure co-determines how a speech sound is produced. This study therefore investigated whether the speaker-dependent information in the vowel [aː] varies when uttered in different word classes. From two spontaneous speech corpora, [aː] tokens were sampled and annotated for word class (content, function word). This was done for 50 male adult speakers of Standard Dutch in face-to-face speech (N = 3128 tokens), and another 50 male adult speakers in telephone speech (N = 3136 tokens). First, the effect of word class on various acoustic variables in spontaneous speech was tested. Results showed that [aː]'s were shorter and more centralized in function than content words. Next, tokens were used to assess their speaker-dependent information as a function of word class, by using acoustic-phonetic variables to (a) build speaker classification models and (b) compute the strength-of-evidence, a technique from forensic phonetics. Speaker-classification performance was somewhat better for content than function words, whereas forensic strength-of-evidence was comparable between the word classes. This seems explained by how these methods weigh between- and within-speaker variation. Because these two sources of variation co-varied in size with word class, acoustic word-class variation is not expected to affect the sampling of tokens in forensic speaker comparisons.

The distribution of speaker information in Dutch fricatives /s/ and /x/ from telephone dialogues.

Although previous work has shown that some speech sounds are more speaker-specific than others, not much is known about the speaker information of the same segment in different linguistic contexts. The present study, therefore, investigated whether Dutch fricatives /s/ and /x/ from telephone dialogues contain differential speaker information as a function of syllabic position and labial co-articulation. These linguistic effects, established in earlier work on read broadband speech, were first investigated. Using a corpus of Dutch telephone speech, results showed that the telephone bandwidth captures the expected effects of perseverative and anticipatory labialization for dorsal fricative /x/, for which spectral peaks fall within the telephone band, but not for coronal fricative /s/, for which the spectral peak falls outside the telephone band. Multinomial logistic regression shows that /s/ contains slightly more speaker information than /x/ in telephone speech and that speaker information is distributed across the speech signal in a systematic way; even though differences in classification accuracy were small, codas and tokens with labial neighbors yielded higher scores than onsets and tokens with non-labial neighbors for both /s/ and /x/. These findings indicate that speaker information in the same speech sound is not the same across linguistic contexts.

Coding pitch differences in voiceless fricatives: Whispered relative to normal speech.

Intonation can be perceived in whispered speech despite the absence of the fundamental frequency. In the past, acoustic correlates of pitch in whisper have been sought in vowel content, but, recently, studies of normal speech demonstrated correlates of intonation in consonants as well. This study examined how consonants may contribute to the coding of intonation in whispered relative to normal speech. The acoustic characteristics of whispered, voiceless fricatives /s/ and /f/, produced at different pitch targets (low, mid, high), were investigated and compared to corresponding normal speech productions to assess if whisper contained secondary or compensatory pitch correlates. Furthermore, listener sensitivity to fricative cues to pitch in whisper was established, also relative to normal speech. Consistent with recent studies, acoustic correlates of whispered and normal speech fricatives systematically varied with pitch target. Comparable findings across speech modes showed that acoustic correlates were secondary. Discrimination of vowel-fricative-vowel stimuli was less accurate and slower in whispered than normal speech, which is attributed to differences in acoustic cues available. Perception of fricatives presented without their vowel contexts, however, revealed comparable processing speeds and response accuracies between speech modes, supporting the finding that within fricatives, acoustic correlates of pitch are similar across speech modes.

Vocalic correlates of pitch in whispered versus normal speech.

In whispered speech, the fundamental frequency is absent as a main cue to pitch. This study investigated how different pitch targets can acoustically be coded in whispered relative to normal speech. Secondary acoustic correlates that are found in normal speech may be preserved in whisper. Alternatively, whispering speakers may provide compensatory information. Compared to earlier studies, a more comprehensive set of acoustic correlates (duration, intensity, formants, center-of-gravity, spectral balance) and a larger set of materials were included. To elicit maximal acoustic differences among the low, mid, and high pitch targets, linguistic and semantic load were minimized: 12 native Dutch speakers produced the point vowels (/a, i, u/) in nonsense vowel-consonant-vowel targets (with C = {/s/, /f/}). Acoustic analyses showed that in addition to systematic changes in formants, which have been reported before, also center of gravity, spectral balance, and intensity varied with pitch target, both in whispered and normal speech. Some acoustic correlates differed more in whispered than in normal speech, suggesting that speakers can adopt a compensatory strategy when coding pitch in the speech mode lacking the main cue. Speakers furthermore varied in the extent to which particular correlates were used, and in the combination of correlates they altered systematically.

Asking or Telling--Real-time Processing of Prosodically Distinguished Questions and Statements.

We introduce a targeted language game approach using the visual world, eye-movement paradigm to assess when and how certain intonational contours affect the interpretation of utterances. We created a computer-based card game in which elliptical utterances such as "Got a candy" occurred with a nuclear contour most consistent with a yes-no question (H* H-H%) or a statement (L* L-L%). In Experiment I we explored how such contours are integrated online. In Experiment 2 we studied the expectations listeners have for how intonational contours signal intentions: do these reflect linguistic categories or rapid adaptation to the paradigm? Prosody had an immediate effect on interpretation, as indexed by the pattern and timing of fixations. Moreover, the association between different contours and intentions was quite robust in the absence of clear syntactic cues to sentence type, and was not due to rapid adaptation. Prosody had immediate effects on interpretation even though there was a construction-based bias to interpret "got a" as a question. Taken together, we believe this paradigm will provide further insights into how intonational contours and their phonetic realization interact with other cues to sentence type in online comprehension.

Perception of prosody in normal and whispered French.

The current study explored perception of prosody in normal and whispered speech using a two-interval, two-alternative forced-choice psychophysical task where listeners discriminated between French noun phrases pronounced as declaratives or interrogatives. Stimuli were either presented between 50 and 8000 Hz or filtered into one of three broad frequency regions, corresponding to harmonic-resolvability regions for normal speech (resolved, partially resolved, unresolved harmonics). Normal speech was presented against a speech-shaped noise masker, whereas whispered speech was presented in quiet. The results showed that discrimination performance was differentially affected by filtering for normal and whispered speech, suggesting that cues to prosody differ between speech modes. For whispered speech, evidence was mainly derived from the high-frequency region, whereas for normal speech, evidence was mainly derived from the low-frequency (resolved harmonics) region. Modeling of the early stages of auditory processing confirmed that for whispered speech, perception of prosody was not based on temporal auditory cues and suggests that listeners may rely on place of excitation (spectral) cues that are, in contrast with suggestions made by earlier work, distributed across the spectrum.

Development of the A§E test battery for assessment of pitch perception in speech.

OBJECTIVES: The auditory speech sounds evaluation 2009 test battery for assessment of speech pitch perception is presented. It was designed to (a) assess perception of pitch in linguistic contexts without the confounds of secondary acoustic cues, (b) be usable with listeners from different language backgrounds, and (c) be suitable for use in a clinical setting. The need for this test battery arises from increased awareness of the importance of prosody in clinical practice, and the development of methods for improving pitch perception in listeners with profound hearing losses. METHODS: Identification and discrimination tasks based on linguistic contexts were developed to establish listeners' just noticeable differences (JNDs) for pitch changes. Stimuli were pseudosentences and pseudowords based on speech from a female speaker, overlain with stylized pitch contours. Target pitch excursions were varied from the 200 Hz baseline to a maximum of 349 Hz. Ninety normal-hearing listeners participated in test validation that assessed goals (a)-(c), established test-retest reliability, and gathered normative data. RESULTS: The JNDs on non-linguistic, control tasks were lower than on linguistic ones, showing that non-linguistic tasks may overestimate pitch perception in speech. Listeners from different language backgrounds scored comparably on most linguistic tasks, and test-retest differences were non-significant. Test usability as evidenced by task duration and subject experience seemed satisfactory for clinical use.