Elamur Research⌕
☰
home / authors / Ethan Gotlieb Wilcox

Ethan Gotlieb Wilcox

Ethan Gotlieb Wilcox studies what neural language models learn about the structure of human language and how closely their predictions match human sentence processing. The papers here borrow the methods of psycholinguistics. They build sentence sets that contrast minimal conditions and read a model's per-word surprisal as a stand-in for processing difficulty. Recurring questions include garden-path ambiguities, long-distance dependencies, grammatical knowledge that aggregate scores can hide, and whether surprisal predicts how long people take to read words.

The most cited work here is Neural language models as psycholinguistic subjects (2019). It asked whether models keep track of incremental syntactic state, for instance registering that a subordinate clause left open must later be followed by a main clause. It tested this with NP/Z and MV/RR garden-path items and analysed the results with mixed-effects regression. What do RNN Language Models Learn about Filler-Gap Dependencies? (2018) probed two pretrained LSTMs against a 5-gram baseline. It examined whether the models link wh-words to later gaps and respect island constraints. SyntaxGym (2020) turned targeted syntactic evaluation into a website with open-source tooling. With it, such tests no longer demand both experimental design skill and the engineering needed to run large models.

More recent work widens the scope beyond English syntax. Testing the Predictions of Surprisal Theory in 11 Languages (2023) used MECO eye-tracking data from eleven languages in five families. It drew surprisal and contextual entropy from mGPT and from monolingual models trained on corpora of different sizes. The 2026 paper on paired-focus constructions turns to meaning. It asks whether open-source models grasp both the form and the semantics of rare expressions such as "let alone" and "much less," where the two highlighted items are ranked along an unstated scale.

Wilcox is currently at Georgetown University, and a few recent papers also list the University of California San Diego. Earlier bylines name ETH Zurich, Harvard University and the Massachusetts Institute of Technology. The most frequent co-author on the papers here is Roger Levy, followed by Peng Qian, Richard Futrell and Takashi Morita. The work falls mainly under Language models & pretraining and Psycholinguistics & language acquisition.

Ethan Gotlieb Wilcox has 6 papers in our library, in 2018–2026, most of them on Language models & pretraining, Psycholinguistics & language acquisition, Syntax & grammar formalisms; often with Roger Levy, Richard Futrell, Takashi Morita.

6 papers in our library · 2018–2026

current affiliationGeorgetown University, Washington, United States; University of California San Diego, San Diego, United States

OpenAlex · ORCID · h-index 17 · 135 works · 1 263 citations OpenAlex, 2026-10-14

6 works in our library · 570 citations in our library · 4 more works in our library · ⭐ Paper Star rank #2073

The profile is the library's match of the name; two people with one name are not always told apart.

Elsewhere: Google Scholar · Semantic Scholar · dblp · ACL Anthology

Affiliations

From the bylines of their papers (OpenAlex, 2026-10-14), ORCID and Wikidata; institutions on papers are matched by machine. 2 more institutions on single bylines are not listed.

Topics

Language models & pretraining4Psycholinguistics & language acquisition4Syntax & grammar formalisms1

Works by year

2020s

20261 paper20231 paper20201 paper

2010s

20191 paper20182 papers

Papers

The most cited first (OpenAlex and our library).

  1. Neural language models as psycholinguistic subjects: Representations of syntactic state (2019 · 291 citations)
    Futrell and colleagues test whether neural language models track incremental syntactic state. This is the information a stack-based parser would hold, such as knowing that an open subordinate clause still needs a main clause. They borrow methods from psycholinguistics and treat per-word surprisal as a stand-in for…
  2. What do RNN Language Models Learn about Filler-Gap Dependencies? (2018 · 276 citations)
    The paper asks whether LSTM language models trained only on raw English text acquire filler–gap dependencies.
  3. Testing the Predictions of Surprisal Theory in 11 Languages (2023 · 158 citations)
    Evidence for surprisal theory, the idea that words harder to predict from prior context take longer to read, comes mostly from English readers.
  4. SyntaxGym: An Online Platform for Targeted Evaluation of Language Models (2020 · 145 citations)
    Targeted syntactic evaluation tests whether neural language models have learned specific grammatical generalizations.
  5. RNNs as psycholinguistic subjects: Syntactic state and grammatical dependency (2018 · 75 citations)
    The paper asks which aspects of grammar LSTM language models pick up when trained only to predict the next word.
  6. Language Models Learn Constructional Semantics, Not To Mention Syntax: Investigating LM Understanding of Paired-Focus Constructions (2026 · 1 citation)
    The paper asks whether open-source language models, and not only the largest prompted LLMs, understand both the form and the meaning of rare English PAIRED-FOCUS constructions: LET-ALONE, MUCH-LESS, NOT-TO-MENTION and NEVER-MIND.

Also in our library

4 more works of theirs are in our library without a page here yet; they will appear as we publish them.

Co-authors here

12 of 13 co-authors with a page, the most papers together first.

Cited by

Papers of our library cite their works 570 times (of the 140 884 papers whose bibliographies we read); 42 citing papers with a page here →.

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/authors/ethan-gotlieb-wilcox/