2026 Spring

03/20 12:10-13:10  Prof. Andrei Munteanu (National Yang Ming Chiao Tung University, Taiwan)

Venue: HB 207 (人社二館 207), NYCU.

Registration link:  https://forms.gle/G55jXkfLnZ5ua9r18

"Supplementing traditional historical linguistics methodology with probabilistic modelling"

Abstract: I present Wordlist Distortion Theory, a probabilistic framework for evaluating comparative reconstruction attempts. The series of transformations – sound changes, borrowing, semantic change, etc. – acts as the input to the framework’s evaluation function. The output is the estimated probability that a randomly generated wordlist merits a reconstruction from the mother language using the same number of transformations (or fewer) than required by the daughter language. This estimate serves as a measure of the statistical reliability of the reconstruction: a reconstruction with a high probability of being substantiated by chance is defined to be less reliable while a reconstruction with a low probability of being substantiated by chance is defined to be more reliable. I briefly highlight the utility of this methodology by outlining some emergent theoretical insights and its uses as a cost function in machine learning approaches to historical linguistics.

05/06 (TIme: TBD)  Prof. Tsung-Lun Alan Wan (National Yang Ming Chiao Tung University, Taiwan)

Venue: TBD

Registration link:  TBD

"Variability in Sociolinguistic Cognition: Listeners’ Sensitivity to the Amount of Nonstandard Phonetic Features"

Abstract: In recent years, variationist sociolinguistics has undergone a cognitive turn, with growing attention to how mental representations process links between linguistic and social information. Within this line of work, Labov and colleagues (2011) proposed the notion of a sociolinguistic monitor, suggesting that the human mind includes a mechanism that tracks the social appropriateness of phonetic features in one’s own speech as well as in others’ speech. More recently, exemplar-theoretic scholars such as Kathryn Campbell-Kibler (2016) have challenged the idea of a dedicated monitoring mechanism for the social meaning of phonetic features, arguing instead that such processes are part of more general social cognition.

One empirical pattern identified in Labov’s sociolinguistic monitor account, however, has continued to attract scholarly interest. Labov argued that, after prolonged language socialization, the relationship between (i) the proportion of nonstandard phonetic features within a particular unit of speech (e.g., a passage)—such as [ɪn] instead of [ɪŋ] in American English—and (ii) the corresponding evaluative penalty on social judgments (e.g., professionalism ratings) is broadly logarithmic. Subsequent replication studies—including larger-scale work on American English as well as studies on Puerto Rican Spanish and British English—have instead tended to find linear relationships.

Using Taiwan Mandarin as a case study, this talk examines how different within-sentence proportions (0%, 20%, 40%, …, 100%) of socially salient phonetic variables—such as apical vowel rounding and retroflexion/detroflexion—map onto five evaluative dimensions: standardness, qizhi (refinement), maternal competence, approachability, and Taiwaneseness. I ask what mathematical form best characterizes these relationships and discuss which individual-level factors may modulate them. 

References:

Campbell-Kibler, Kathryn. 2016. Towards a Cognitively Realistic Model of Meaningful Sociolinguistic Variation. In Anna M. Babel (ed.), Awareness and Control in Sociolinguistic Research, 123–151. Cambridge: Cambridge University Press. 

Labov, William, Sharon Ash, Maya Ravindranath, Tracey Weldon, Maciej Baranowski & Naomi Nagy. 2011. Properties of the sociolinguistic monitor. Journal of Sociolinguistics 15(4). 431–463. https://doi.org/10.1111/j.1467-9841.2011.00504.x.