# Week of February 27 February 27, 2023 ## Indo-European and Historical Linguistics Workshop Jorge Wong (Department of the Classics, Harvard University) Title : ‘the -εσσι datives in Homer and Lesbian Aeolic’ Friday March 3 | 5:00- 6:30 PM | Boylston Hall 335 ## Harvard LangCog Forrest Davis (MIT) Title: Principle B, Processing, and Neural Models of Language Abstract: Interpretation of neural language models has garnered, deservedly, much attention in the field of natural language processing. One approach in the literature that is particularly interesting to linguists and psycholinguists involves comparisons between the behavior of neural models and humans. Many aspects of human linguistic knowledge have been utilized in this approach, ranging from predictions about the form of a verb to the entailment relations between sentences. It remains unclear, however, whether we should reasonably expect a neural model to capture all of these human behaviors. In this talk, I will focus on behavior that is the most immediate, and, presumably, does not require auxiliary human capacities like reasoning. In particular, I will compare human and models on the interaction between Principle B (from Binding Theory; Chomsky, 1981) and incremental processing. Building on a variety of experiments, I will show that neural models exhibit both qualitative and quantitative gaps in capturing human behavior. Ultimately, I aim to connect these gaps to properties of training data to suggest that aspects of the human parser do not follow directly from linguistic data. I will conclude by commenting on the larger worldview that mediates comparisons between neural models and humans. Tuesday February 28 | 5:30-7:00 PM | William James Hall #1550 ## TU+ at Harvard The eighth Workshop on Turkic and Languages in Contact with Turkic will be hosted at Harvard on **March 4-5, 2023**. If you plan to attend the conference, please fill the form on the website: ## Harvard at ECO-5 [ECO-5](https://sites.google.com/view/eco-5/home?authuser=0) was hosted at UConn on February 23. The following members of Harvard Linguistics presented their work Aljoša Milenkovic *Symmetrical trade-offs are diachronically unstable: superlinear constraint cumulativity in Serbian* Annabelle Caso *Slay him naked: an interface approach to secondary predication* Daria Bikina *Why do some languages never relativize dative arguments? Evidence from Khanty* James Lee *Causativity in the Korean voice domain* Share on:- [ Facebook ](#) - [ Twitter ](#) - [ Linkedin ](#)