Phonetic Documentation in Three Collections: Topics and Evolution.

Phonetic aspects of many languages have been documented, though the breadth and focus of such documentation varies substantially. In this survey, phonetic aspects (here called "categories") that are typically reported were assessed in three English-language collections-the Illustrations of the IPA, articles from the Journal of Phonetics, and papers from the Ladefoged/Maddieson Sounds of the World's Languages (SOWL) documentation project. Categories were defined for consonants (e.g., Voice Onset Time (VOT) and frication spectrum; 10 in total), vowels (e.g., formants and duration; 7 total) and suprasegmentals (e.g., stress and distinctive vowel length, 6 total). The Illustrations, due to their brevity, had, on average, limited coverage of the selected categories (12% of the 23 categories). Journal of Phonetics articles were typically theoretically motivated, but 64 had sufficient measurements to count as phonetic documentation; these also covered 12% of the categories. The SOWL studies, designed to cover as much of the phonetic structure as feasible in an article-length treatment, achieved 41% coverage on average. Four book-length studies were also examined, with an average of 49% coverage. Phonetic properties of many language families have been studied, though Indo-European is still disproportionately represented. Physiological measures were excluded as being less common, and perceptual measures were excluded as being typically more theoretical. This preliminary study indicates that certain acoustic properties of languages are typically measured and may be considered as an impetus for later, fuller coverage, but broader consensus on the categories is needed. Current documentation efforts could be more useful if these considerations were addressed.

Extreme stop allophony in Mixtec spontaneous speech: data, word prosody, and modelling.

Word-level prosody plays an important role in processes of consonant lenition. Typically, consonants in word-initial position are strengthened while those in word-medial position are lenited (Keating et al., 2003). In this paper we examine the relationship between wordprosodic position and obstruent lenition in a spontaneous speech corpus of Yoloxóchitl Mixtec, an endangered Mixtecan language spoken in Mexico. The language exhibits a surprising amount of lenition in the realization of otherwise voiceless unaspirated stops and voiceless fricatives in careful speech. In Experiment 1, we examine the relationships between word position, consonant duration, and passive voicing and find that word-medial pre-tonic position is the locus of both consonant lengthening and less passive voicing. Non-pre-tonic consonants are produced with more voicing and shorter duration. We also find that the functional status of the morpheme plays a role in voicing lenition. In Experiment 2, we examine manner lenition and find a similar pattern - word-medial pre-tonic stops are more often realized with complete closure relative to non-pre-tonic stops, which are more often realized with incomplete closure. In Experiment 3, we model these lenition patterns using a series of deep neural networks and find that, even with limited training data, we can achieve reasonably high accuracy in the automatic categorization of lenition patterns. The results of this research both complement recent work on the phonetics of lenition in the world's languages (Katz and Fricke, 2018; White et al., 2020) and provide computational tools for modeling and predicting patterns of extreme lenition.

Disentangling the Effects of Position and Utterance-Level Declination on the Production of Complex Tones in Yoloxóchitl Mixtec.

Phrase-final position is cross-linguistically the locus of both processes of phonetic reduction and processes of phonetic enhancement. In tone languages, phrasal position is a conditioning environment for processes of tone sandhi/allotony, though such patterns emerge from local processes of tonal enhancement or reduction. The current article examines the production of tone in Yoloxóchitl Mixtec, an endangered language of Mexico with nine lexical tones and fixed, stem-final stress, across phrasal and utterance positions via three experiments. In the first two experiments, the findings show that speakers lengthen syllables and expand the tonal F0 range in utterance-final position. The effect of this range expansion is high tone raising, low tone lowering, and falling contour lowering. Rising contour tones undergo substantial leveling when produced in a non-utterance-final context, similar to Taiwanese Mandarin. These findings suggest that postural changes in F0 range are controlled, intonational effects in tonal languages and not paralinguistic. In the third experiment, we examine utterance-level declination and raising within sentences consisting entirely of level tones. We show that utterance-level F0 changes are independent from local tonal hyperarticulation effects in phrase-final position. Together, the experiments largely support prosodically-conditioned phonetic undershoot as a control mechanism in tone production and demonstrate how tonal complexity may constrain universal tendencies in speech production.

Cue weight in the perception of Trique glottal consonants.

This paper examines the perceptual weight of cues to the coda glottal consonant contrast in Trique (Oto-Manguean) with native listeners. The language contrasts words with no coda (/Vː/) from words with a coda glottal stop (/VɁ/) or breathy coda (/Vɦ/). The results from a speeded AX (same-different) lexical discrimination task show high accuracy in lexical identification for the /Vː/-/Vɦ/ contrast, but lower accuracy for the other contrasts. The second experiment consists of a labeling task where the three acoustic dimensions that distinguished the glottal consonant codas in production [duration, the amplitude difference between the first two harmonics (H1-H2), and F0] were modified orthogonally using step-wise resynthesis. This task determines the relative weight of each dimension in phonological categorization. The results show that duration was the strongest cue. Listeners were only sensitive to changes in H1-H2 for the /Vː/-/Vɦ/ and /Vː/-/VɁ/ contrasts when duration was ambiguous. Listeners were only sensitive to changes in F0 for the /Vː/-/Vɦ/ contrast when both duration and H1-H2 were ambiguous. The perceptual cue weighting for each contrast closely matches existing production data [DiCanio (2012 a). J. Phon. 40, 162-176] Cue weight differences in speech perception are explained by differences in step-interval size and the notion of adaptive plasticity [Francis et al. (2008). J. Acoust. Soc. Am. 124, 1234-1251; Holt and Lotto (2006). J. Acoust. Soc. Am. 119, 3059-3071].

Using automatic alignment to analyze endangered language data: testing the viability of untrained alignment.

While efforts to document endangered languages have steadily increased, the phonetic analysis of endangered language data remains a challenge. The transcription of large documentation corpora is, by itself, a tremendous feat. Yet, the process of segmentation remains a bottleneck for research with data of this kind. This paper examines whether a speech processing tool, forced alignment, can facilitate the segmentation task for small data sets, even when the target language differs from the training language. The authors also examined whether a phone set with contextualization outperforms a more general one. The accuracy of two forced aligners trained on English (hmalign and p2fa) was assessed using corpus data from Yoloxóchitl Mixtec. Overall, agreement performance was relatively good, with accuracy at 70.9% within 30 ms for hmalign and 65.7% within 30 ms for p2fa. Segmental and tonal categories influenced accuracy as well. For instance, additional stop allophones in hmalign's phone set aided alignment accuracy. Agreement differences between aligners also corresponded closely with the types of data on which the aligners were trained. Overall, using existing alignment systems was found to have potential for making phonetic analysis of small corpora more efficient, with more allophonic phone sets providing better agreement than general ones.