Morgan Sonderegger
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Morgan Sonderegger has a total of 61 co-authors. Here are the top co-authors by number of shared publications:
- Meghan Clayards
Department of Linguistics, McGill University 1 , Montréal, Québec H3A 1A7
4 shared publications - Michael Wagner
McGill University
4 shared publications - Jane Stuart-Smith
Department of English Language and Linguistics University of Glasgow United Kingdom
3 shared publications Max Bane
3 shared publications- James Kirby
Institute for Phonetics and Speech Processing, Spoken Language Processing Group, LMU Munich 1 , Munich, 80799
2 shared publications - James Tanner
Department of English Language and Linguistics University of Glasgow United Kingdom
2 shared publications
Latest Publications
- 17 October 2026
- Published by SAGE Publications in Language and Speech
The Role of Syllable Duration in the Perception and Imitation of Checked Versus Unchecked Tones in Taiwanese Southern Min
The imitation of phonetic features is often mediated by phonological contrasts, with a stronger mediation effect for critical cues. However, it remains unclear whether syllable duration, which has been only examined as a noncritical cue in previous research, is mediated in the same way as other features when it is critical to the imitated contrast. In this study, we conducted two experiments. In Experiment 1, we investigated the role of syllable duration in the perception of the checked versus unchecked tonal contrasts in Taiwanese Southern Min (TSM). Results showed that syllable duration is a critical cue for distinguishing the checked versus unchecked tones of the mid register but not for those of the high register. Therefore, in Experiment 2, we examined how syllable duration is imitated along the checked versus unchecked tonal contrast in the mid register. Results showed that, unlike other features such as voice onset time (VOT) and fundamental frequency (F0), the imitation of syllable duration is minimally mediated by the phonological contrast, even when it serves as a critical cue. This finding suggests that syllable duration has a unique property that allows it to bypass phonological mediation and adds to our knowledge of phonetic imitation being feature-specific rather than feature-general.
- 17 October 2026
- Published by SAGE Publications in Language and Speech
The Role of Syllable Duration in the Perception and Imitation of Checked Versus Unchecked Tones in Taiwanese Southern Min
The imitation of phonetic features is often mediated by phonological contrasts, with a stronger mediation effect for critical cues. However, it remains unclear whether syllable duration, which has been only examined as a noncritical cue in previous research, is mediated in the same way as other features when it is critical to the imitated contrast. In this study, we conducted two experiments. In Experiment 1, we investigated the role of syllable duration in the perception of the checked versus unchecked tonal contrasts in Taiwanese Southern Min (TSM). Results showed that syllable duration is a critical cue for distinguishing the checked versus unchecked tones of the mid register but not for those of the high register. Therefore, in Experiment 2, we examined how syllable duration is imitated along the checked versus unchecked tonal contrast in the mid register. Results showed that, unlike other features such as voice onset time (VOT) and fundamental frequency (F0), the imitation of syllable duration is minimally mediated by the phonological contrast, even when it serves as a critical cue. This finding suggests that syllable duration has a unique property that allows it to bypass phonological mediation and adds to our knowledge of phonetic imitation being feature-specific rather than feature-general.
- Open Access
- 26 August 2026
- Published by Elsevier in SSRN Electronic Journal
A Causal Mediation Analysis of Creaky Voice Acoustics: Evidence for F0 As a Mediator of Gender Effects
This study investigates how gender differences in creaky voice acoustics arise, focusing on the role of fundamental frequency (f0), using causal mediation analysis. Prior research largely shows greater acoustic creakiness in men's voices. F0 varies systematically with gender, due to both physiological differences in vocal fold anatomy and social factors; f0 is also plausibly, at least in part, causally upstream of other acoustic measures commonly used to quantify voice quality. This paper therefore asks how much of the gender effect on creaky voice acoustics is transmitted through f0 (an indirect effect) versus other pathways (a direct effect). The corpus contains spontaneous speech from 49 Canadian English-French bilingual speakers (92,679 vowels). Acoustic measures include the proportion of reliable f0 tracks, H1*-H2*, CPP, and HNR < 500 Hz. Implementing a causal inference framework using Bayesian regression models, direct and indirect effects are disentangled. Results show that the degree of f0 mediation varies across acoustic measures: the proportion of reliable f0 tracks is almost entirely f0-mediated, while H1*-H2*, HNR < 500 Hz, and CPP, show only partial mediation. This gradient is broadly consistent with physiological expectations, reflecting each acoustic measure's functional proximity to f0, and suggests that additional unmeasured factors-potentially social conditioning-also contribute to creaky voice variation. Beyond these empirical findings, the paper illustrates the value of causal mediation analysis for phonetics: by requiring assumptions such as the causal ordering of variables to be made explicit, it supports more principled reasoning about the mechanisms linking social factors to voice quality.
- 26 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA Causal Mediation Analysis of Creaky Voice Acoustics: Evidence for F0 As a Mediator of Gender Effects
This study investigates how gender differences in creaky voice acoustics arise, focusing on the role of fundamental frequency (f0), using causal mediation analysis. Prior research largely shows greater acoustic creakiness in men's voices. F0 varies systematically with gender, due to both physiological differences in vocal fold anatomy and social factors; f0 is also plausibly, at least in part, causally upstream of other acoustic measures commonly used to quantify voice quality. This paper therefore asks how much of the gender effect on creaky voice acoustics is transmitted through f0 (an indirect effect) versus other pathways (a direct effect). The corpus contains spontaneous speech from 49 Canadian English-French bilingual speakers (92,679 vowels). Acoustic measures include the proportion of reliable f0 tracks, H1*-H2*, CPP, and HNR < 500 Hz. Implementing a causal inference framework using Bayesian regression models, direct and indirect effects are disentangled. Results show that the degree of f0 mediation varies across acoustic measures: the proportion of reliable f0 tracks is almost entirely f0-mediated, while H1*-H2*, HNR < 500 Hz, and CPP, show only partial mediation. This gradient is broadly consistent with physiological expectations, reflecting each acoustic measure's functional proximity to f0, and suggests that additional unmeasured factors-potentially social conditioning-also contribute to creaky voice variation. Beyond these empirical findings, the paper illustrates the value of causal mediation analysis for phonetics: by requiring assumptions such as the causal ordering of variables to be made explicit, it supports more principled reasoning about the mechanisms linking social factors to voice quality.
- 1 January 2026
- Published by Open Library of the Humanities in Laboratory Phonology
Title Pending
- Open Access
- 1 January 2026
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 159 (1) , 647-663
Are individual differences in lexical and coarticulatory biases linked across speech tasks?
Individuals vary in their tendency to be biased by lexical and coarticulatory information during speech perception, though the sources of these differences remain debated. One suggestion is that lexical and coarticulatory bias trade-off with each other due to differences in processing style. This study explores the nature of this individual variation through three speech sound categorization tasks: one where both lexical and coarticulatory biases are present simultaneously for the /s-ʃ/ sibilants, a replication of Lai, Wade, and Tamminga [(2022) Linguist. Vanguard 8, 29–39]; the second where only lexical bias is present for /ɛ/-/ɪ/ vowels; and finally /d/-/g/ stops following /r/ or /l/ coarticulatory biasing contexts. Across 89 native Canadian English listeners, we replicated the within-task trade-off between lexical and coarticulatory effects (r = −0.19), weaker but consistent with Lai, Wade, and Tamminga [(2022) Linguist. Vanguard 8, 29–39]. There was no evidence for a trade-off across tasks; a joint Bayesian model that estimated individual-level correlations across tasks showed that lexical and coarticulatory effects were not credibly related. Participants' consistency in categorizing the acoustic continuum (effect of continuum step) was consistent across tasks. Overall, the findings suggest that while trade-off patterns can be replicated within specific tasks, they are not robust across tasks, indicating possible task-specific effects or contrast-specific perceptual strategies.
- 1 January 2026
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 159 (1) , 647-663
Open AccessAre individual differences in lexical and coarticulatory biases linked across speech tasks?
Individuals vary in their tendency to be biased by lexical and coarticulatory information during speech perception, though the sources of these differences remain debated. One suggestion is that lexical and coarticulatory bias trade-off with each other due to differences in processing style. This study explores the nature of this individual variation through three speech sound categorization tasks: one where both lexical and coarticulatory biases are present simultaneously for the /s-ʃ/ sibilants, a replication of Lai, Wade, and Tamminga [(2022) Linguist. Vanguard 8, 29–39]; the second where only lexical bias is present for /ɛ/-/ɪ/ vowels; and finally /d/-/g/ stops following /r/ or /l/ coarticulatory biasing contexts. Across 89 native Canadian English listeners, we replicated the within-task trade-off between lexical and coarticulatory effects (r = −0.19), weaker but consistent with Lai, Wade, and Tamminga [(2022) Linguist. Vanguard 8, 29–39]. There was no evidence for a trade-off across tasks; a joint Bayesian model that estimated individual-level correlations across tasks showed that lexical and coarticulatory effects were not credibly related. Participants' consistency in categorizing the acoustic continuum (effect of continuum step) was consistent across tasks. Overall, the findings suggest that while trade-off patterns can be replicated within specific tasks, they are not robust across tasks, indicating possible task-specific effects or contrast-specific perceptual strategies.
- Open Access
- 28 July 2025
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 122 (31)
Semantic change in adults is not primarily a generational phenomenon
A central question in the study of language change is whether or not such change is generational. If a language changes over time generation-by-generation, the process looks as follows: New generations of speakers introduce innovations, while older speakers conserve their usage patterns, and the language changes as new generations replace older ones. At the opposite extreme, language change could be a zeitgeist phenomenon, in which changes are universally adopted by speakers simultaneously, regardless of age or generational cohort. This paper asks this question in the context of word meaning change. We analyze meaning change in over 100 words across more than 7.9 million U.S. congressional speeches, to observe whether, when a word sense rises or falls in prominence, adult speakers from different generations uniformly adopt it, or those from older generations conserve their prior usage. Using language model-based word sense induction methods, we identify different senses of each word, and then model the prevalence of each of these word senses as a function of time and speaker age. We find that most words show a small but statistically significant effect of speaker age; across almost 140 y of Congress, older speakers typically take longer than younger speakers to follow changes in word usage, but nevertheless do so within a few years. Our findings indicate that despite minor age-based differences, word meaning change among mature speakers is likely not a generational process, but rather a zeitgeist process, in which older adult speakers can readily adopt new word usage patterns.
- 28 July 2025
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 122 (31)
Open AccessSemantic change in adults is not primarily a generational phenomenon
A central question in the study of language change is whether or not such change is generational. If a language changes over time generation-by-generation, the process looks as follows: New generations of speakers introduce innovations, while older speakers conserve their usage patterns, and the language changes as new generations replace older ones. At the opposite extreme, language change could be a zeitgeist phenomenon, in which changes are universally adopted by speakers simultaneously, regardless of age or generational cohort. This paper asks this question in the context of word meaning change. We analyze meaning change in over 100 words across more than 7.9 million U.S. congressional speeches, to observe whether, when a word sense rises or falls in prominence, adult speakers from different generations uniformly adopt it, or those from older generations conserve their prior usage. Using language model-based word sense induction methods, we identify different senses of each word, and then model the prevalence of each of these word senses as a function of time and speaker age. We find that most words show a small but statistically significant effect of speaker age; across almost 140 y of Congress, older speakers typically take longer than younger speakers to follow changes in word usage, but nevertheless do so within a few years. Our findings indicate that despite minor age-based differences, word meaning change among mature speakers is likely not a generational process, but rather a zeitgeist process, in which older adult speakers can readily adopt new word usage patterns.
- Open Access
Automatic classification of stop realisation with wav2vec2.0
Modern phonetic research regularly makes use of automatic tools for the annotation of speech data, however few tools exist for the annotation of many variable phonetic phenomena. At the same time, pre-trained self-supervised models, such as wav2vec2.0, have been shown to perform well at speech classification tasks and latently encode fine-grained phonetic information. We demonstrate that wav2vec2.0 models can be trained to automatically classify stop burst presence with high accuracy in both English and Japanese, robust across both finely-curated and unprepared speech corpora. Patterns of variability in stop realisation are replicated with the automatic annotations, and closely follow those of manual annotations. These results demonstrate the potential of pre-trained speech models as tools for the automatic annotation and processing of speech corpus data, enabling researchers to 'scale-up' the scope of phonetic research with relative ease.
Open AccessAutomatic classification of stop realisation with wav2vec2.0
Modern phonetic research regularly makes use of automatic tools for the annotation of speech data, however few tools exist for the annotation of many variable phonetic phenomena. At the same time, pre-trained self-supervised models, such as wav2vec2.0, have been shown to perform well at speech classification tasks and latently encode fine-grained phonetic information. We demonstrate that wav2vec2.0 models can be trained to automatically classify stop burst presence with high accuracy in both English and Japanese, robust across both finely-curated and unprepared speech corpora. Patterns of variability in stop realisation are replicated with the automatic annotations, and closely follow those of manual annotations. These results demonstrate the potential of pre-trained speech models as tools for the automatic annotation and processing of speech corpus data, enabling researchers to 'scale-up' the scope of phonetic research with relative ease.
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 66-76
Unzipping the Causality of Zipf’s Law and Other Lexical Trade-offs
Amanda Doucette, Timothy J. O’Donnell, Morgan Sonderegger. Proceedings of the Workshop on Cognitive Modeling and Computational Linguistics. 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 66-76
Open AccessUnzipping the Causality of Zipf’s Law and Other Lexical Trade-offs
Amanda Doucette, Timothy J. O’Donnell, Morgan Sonderegger. Proceedings of the Workshop on Cognitive Modeling and Computational Linguistics. 2025.
- Open Access
- 1 January 2025
- Published by Open Library of the Humanities in Laboratory Phonology
- Vol. 16 (1)
A new perspective on the development of Quebec French rhotic vowels
Quebec French is reportedly developing rhoticity, with low F3 resulting from a bunched or retroflexed tongue (like English /ɚ/), in some or all of the front mid rounded vowels /ø, œ, œ̃ /. The source of this rare, understudied sound change is unclear from previous work: contact with English and contrast enhancement have been suggested, and phonologization of coarticulation is typologically plausible. We examine this issue, investigating the apparent time change in the F3 trajectories of the three vowels using generalized additive mixed models on a corpus of parliamentary speech (108 speakers from across Quebec). We observe rhoticity in /ø/ and /œ̃ /: men begin with low F3 in these vowels, and women show change in progress. Conversely, there is less clear evidence of change in /œ/. We suggest these findings are best explained by analyzing rhotacization as a two-phased change, originally due to borrowing and subsequently spreading through contrast enhancement. Rhoticity, we argue, is the combined product of intensive exposure to English (which led to frequent nonintegration of bunched/retroflex segments in loanwords) and an exceptionally large vowel inventory. It thus results from the unique interplay of social and phonological factors in Quebec French, which is consistent with such changes being cross-linguistically rare.
- 1 January 2025
- Published by Open Library of the Humanities in Laboratory Phonology
- Vol. 16 (1)
Open AccessA new perspective on the development of Quebec French rhotic vowels
Quebec French is reportedly developing rhoticity, with low F3 resulting from a bunched or retroflexed tongue (like English /ɚ/), in some or all of the front mid rounded vowels /ø, œ, œ̃ /. The source of this rare, understudied sound change is unclear from previous work: contact with English and contrast enhancement have been suggested, and phonologization of coarticulation is typologically plausible. We examine this issue, investigating the apparent time change in the F3 trajectories of the three vowels using generalized additive mixed models on a corpus of parliamentary speech (108 speakers from across Quebec). We observe rhoticity in /ø/ and /œ̃ /: men begin with low F3 in these vowels, and women show change in progress. Conversely, there is less clear evidence of change in /œ/. We suggest these findings are best explained by analyzing rhotacization as a two-phased change, originally due to borrowing and subsequently spreading through contrast enhancement. Rhoticity, we argue, is the combined product of intensive exposure to English (which led to frequent nonintegration of bunched/retroflex segments in loanwords) and an exceptionally large vowel inventory. It thus results from the unique interplay of social and phonological factors in Quebec French, which is consistent with such changes being cross-linguistically rare.
- Open Access
Exploring the anatomy of articulation rate in spontaneous English speech: relationships between utterance length effects and social factors
Speech rate has been shown to vary across social categories such as gender, age, and dialect, while also being conditioned by properties of speech planning. The effect of utterance length, where speech rate is faster and less variable for longer utterances, has also been shown to reduce the role of social factors once it has been accounted for, leaving unclear the relationship between social factors and speech production in conditioning speech rate. Through modelling of speech rate across 13 English speech corpora, it is found that utterance length has the largest effect on speech rate, though this effect itself varies little across corpora and speakers. While age and gender also modulate speech rate, their effects are much smaller in magnitude. These findings suggest utterance length effects may be conditioned by articulatory and perceptual constraints, and that social influences on speech rate should be interpreted in the broader context of how speech rate variation is structured.
Open AccessExploring the anatomy of articulation rate in spontaneous English speech: relationships between utterance length effects and social factors
Speech rate has been shown to vary across social categories such as gender, age, and dialect, while also being conditioned by properties of speech planning. The effect of utterance length, where speech rate is faster and less variable for longer utterances, has also been shown to reduce the role of social factors once it has been accounted for, leaving unclear the relationship between social factors and speech production in conditioning speech rate. Through modelling of speech rate across 13 English speech corpora, it is found that utterance length has the largest effect on speech rate, though this effect itself varies little across corpora and speakers. While age and gender also modulate speech rate, their effects are much smaller in magnitude. These findings suggest utterance length effects may be conditioned by articulatory and perceptual constraints, and that social influences on speech rate should be interpreted in the broader context of how speech rate variation is structured.
- Open Access
- 7 January 2024
- Published by Open Library of the Humanities in Glossa: A Journal of General Linguistics
- Vol. 9 (1)
Investigating the universality of consonant and vowel co-occurrence restrictions
Certain phonotactic constraints on the co-occurrence of segments appear to be much more common across the world’s languages than others. In many languages, similar consonant co-occurrence is restricted through Obligatory Contour Principle (OCP) effects, while there are some exceptions for identical consonants. In vowels, the opposite pattern appears to hold: many languages have vowel harmony processes, where vowels within a domain are required to share some feature. Languages that encourage similar consonant co-occurrence or restrict similar vowel co-occurrence appear to be exceedingly uncommon. However, evidence of this pattern so far only comes from studies of individual languages or families, or of only consonants or vowels. We investigate patterns of co-occurrence in vowels and consonants in 107 Northern Eurasian languages across 21 families using Bayesian negative binomial regression to explicitly model the effects of aggregate similarity and segment identity on co-occurrence counts (the results of which can be interpreted similarly to observed/expected ratios). We find that the effect of similarity is remarkably consistent across languages: Similar consonant co-occurrence is disfavored, while aggregate similarity has no effect on vowel co-occurrence. Identical segment co-occurrence effects are much more variable across languages, with a tendency towards disfavoring identical consonants, and favoring identical vowels. We also find larger effects in consonants than in vowels, suggesting that consonant co-occurrence is more strongly constrained than vowel co-occurrence. We also find that there is no evidence for or against any correlations between vowel and consonant co-occurrence, suggesting that more data is needed to evaluate this possibility.
- 7 January 2024
- Published by Open Library of the Humanities in Glossa: A Journal of General Linguistics
- Vol. 9 (1)
Open AccessInvestigating the universality of consonant and vowel co-occurrence restrictions
Certain phonotactic constraints on the co-occurrence of segments appear to be much more common across the world’s languages than others. In many languages, similar consonant co-occurrence is restricted through Obligatory Contour Principle (OCP) effects, while there are some exceptions for identical consonants. In vowels, the opposite pattern appears to hold: many languages have vowel harmony processes, where vowels within a domain are required to share some feature. Languages that encourage similar consonant co-occurrence or restrict similar vowel co-occurrence appear to be exceedingly uncommon. However, evidence of this pattern so far only comes from studies of individual languages or families, or of only consonants or vowels. We investigate patterns of co-occurrence in vowels and consonants in 107 Northern Eurasian languages across 21 families using Bayesian negative binomial regression to explicitly model the effects of aggregate similarity and segment identity on co-occurrence counts (the results of which can be interpreted similarly to observed/expected ratios). We find that the effect of similarity is remarkably consistent across languages: Similar consonant co-occurrence is disfavored, while aggregate similarity has no effect on vowel co-occurrence. Identical segment co-occurrence effects are much more variable across languages, with a tendency towards disfavoring identical consonants, and favoring identical vowels. We also find larger effects in consonants than in vowels, suggesting that consonant co-occurrence is more strongly constrained than vowel co-occurrence. We also find that there is no evidence for or against any correlations between vowel and consonant co-occurrence, suggesting that more data is needed to evaluate this possibility.