2025 Spring

6/10 10:00am-12:00pm  Dr. Andrei Munteanu (McGill University, Canada)

"When Experimental Results go beyond Corpus Trend Extrapolation: Russian Homophony Avodiace"

Registration link: https://forms.gle/SapUbUS5kH75T4e5A

Abstract: I argue for the importance of experimental evidence in phonology through a case study of homophony avoidance in Russian nominals. Corpus data shows that, while the genitive singular morpheme /-a/ is potentially homophonous with one of the nominative plural allomorphs /-a/, little to no homophony is found in the language. Potentially homophonous pairs of genitive singular and nominative plural forms differ in either stress assignment or in the stem (suppletion). A subsequent nonce word production task shows that speakers strongly prefer non-homophonous combinations of genitive singular and nominative plural, even going beyond the patterns observed in the corpus. As such, while a homophony avoidance effect is implied in the corpus data, it is the experimental results that directly evidence homophony avoidance in the synchronic grammar.

5/29 10am-12pm Dr. Kellen Parker van Dam (Universität Passau, DE)

"Tonal Frontiers: Reconciling descriptive linguistic fieldwork with modern computational analyses"

Registration link: https://forms.gle/53TwduM5DbuEfdmYA

Abstract: A considerable amount of work has been conducted in recent years relying on large data corpora for the purposes of answering some of the unknowns in linguistics. This includes significant phylogenetic work using Bayesian inference (Sagart et al 2019, Zhang et al 2019), employment of decision trees for assessing tonal inventories, as well as the use of computational tools for determining cognacy across varieties (List & Forkel 2024). However, despite major advancements in computing power and the continued development of tools and resources, much of the data on which such analyses rely may have unseen complications. Methodological approaches may be opaque, and often data are not provided in open formats when presenting linguistic conclusions. For many computational linguists, data sets are taken at face value for the sake of analytical work without a clear picture of the practical issues that may be behind the data collection, while many fieldwork-focused descriptive linguist may be wholly unaware of the ways in which their data may eventually be employed for the sake of large scale data analysis. In this talk, the issues around data collection and presentation are discussed with a specific focus on linguistic tone in undescribed Tibeto-Burman varieties of the Indo-Burmese border. How are data best normalised for proper analysis without losing replicability? How can this work be done in a post-COVID world where "patchwork anthropology" (Günel & Watanabe 2024) has gained wider acceptance? What issues arise in describing tone systems which have never been described before, and what new avenues of research are opened up by adequately addressing these issues? The talk will focus on a sub-branch of Tibeto-Burman known as Patkaian, with examples from the past decade of new investigations into the languages.

4/11 12:10pm-1:00pm Dr. Alvin Cheng-Hsien Chen (Department of English, National Taiwan Normal University)

"Exploring Pitch Variability in Lexical Production: Implications for Usage-based Grammar"

 Registration link: https://forms.gle/5LmTdb6Buk1wRfGj8 

Abstract: In this talk, I’d like to explore pitch variability in language production to understand processing advantages of holistic units. Focusing on disyllabic words in spontaneous Taiwan Mandarin speech, we investigate how pitch variability is shaped by lexical frequency, predictive contingencies, and retrodictive contingencies. Using a 185-million-word native corpus as a proxy for the statistical properties of native usage, we examine how pitch variability is influenced by lexical frequency, predictive contingencies, and retrodictive contingencies. We introduce a pitch variability metric (f0PVI) based on the duration-based pairwise variability index (PVI). Our analyses reveal three key findings. First, disyllabic words show significantly lower f0PVI values compared to non-holistic partword counterparts, highlighting the metric’s ability to identify holistic units. Second, we find an inverse relationship between pitch variability and lexical frequency, suggesting reduced prosodic prominence in frequently used words. Third, retrodictive contingency effects on pitch variability are moderated by prosodic juncture alignment, emphasizing the role of contextual predictability in speech planning. Our findings establish f0PVI as a reliable prosodic measure for assessing the automatized processing and entrenchment of linguistic units through repeated usage.