Charles Yang: Predicting Language Change

Value of the Information & Strength of the Argument

The talk provides a valuable synthesis of linguistic theory and computational modeling, offering a concrete mechanism for language change. Yang’s argumentation is rigorous: he starts with a simple learning model, derives mathematical conditions for change, and then tests them against historical data. The French case is particularly compelling, as it explains a puzzling historical development (loss of V2) by incorporating pro-drop, and yields a specific quantitative prediction (18% dropped subjects) that aligns with historical texts. The discussion of the cot/caught merger illustrates the model’s applicability to ongoing changes, though the details are less developed. The talk is well-structured and the reasoning is transparent, though some steps are simplified for the seminar format.

Scientific Rigor, Source Quality, Title Accuracy

Yang demonstrates scientific rigor by grounding his model in established linguistic concepts (e.g., V2, pro-drop) and referencing prior work (e.g., Ian Roberts on French historical syntax). However, the talk does not provide explicit citations for all claims, and the sources are not listed in the video description beyond the seminar page. The title accurately reflects the content, as the talk is indeed about predicting language change. The presentation is technical but accessible to an informed audience, and the methodology is clearly explained. The lack of formal references in the video is a minor weakness, but the content is consistent with Yang’s published work.

The title accurately reflects the content: the talk focuses on predicting language change through computational models.

8/10

Presentation by a recognized expert (Charles Yang, UPenn) at a university seminar, grounded in established linguistic theory and quantitative modeling. The talk is technical and references specific historical data and prior work, but lacks peer-reviewed citations in the video itself.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 02:00Introduction to language change and parallels with biological evolution.
• 05:00Presentation of variational learning model and its analogy to natural selection.
• 10:00Derivation of the learning equation and its equivalence to selection equation.
• 15:00Discussion of two types of change: gradual syntactic/morphological vs. rapid phonological.
• 20:00Case study: loss of V2 in French, introduction of pro-drop factor.
• 25:00Derivation of numerical condition (18% dropped subjects) for V2 loss.
• 30:00Comparison with historical data from Ian Roberts' work.
• 35:00Case study: cot/caught merger in Massachusetts and Rhode Island.
• 40:00Discussion of rapid phonological change and winner-take-all learning.
• 45:00Conclusion: implications for predicting language change and future work.

• CLSP Seminar page — Seminar announcement and details for this talk.

• Ian Roberts on French historical syntax — Referenced in the talk for historical data on pro-drop percentages.