My post-doc project in Dr. Jonathan Brennan's Computational Neurolinguistics Lab at the University of Michigan involved taking EEG recordings from interlocutors in unscripted conversation to build a neurolinguistic corpus of naturalistic native and second language speech. In a subsequent step, we tested competing models of L1 and L2 grammar processing to see which theories better align with the observed brainwaves. This was inspired by the excellent book chapter "Using free, unscripted conversation with synchronized neuroimaging data in linguistic inquiry" (Tremblay et al. , 2019. In The description, measurement, and pedagogy of words). For more details see NSF award number 2203723.
My doctoral dissertation at the University of Illinois - Chicago in Dr. Kara Morgan-Short's Cognition of Second Language Acquisition lab involved using machine learning analyses to pull apart neural indices of conscious vs. subconscious second language processing. In my experiment, participants were exposed to an artificial language with a hidden grammar pattern. By triangulating language task reaction times, subjective, reports, and EEG data, I investigated whether focusing on a grammar rule can help or hinder acquisition at a more implicit level. The long game is to find an optimal trade-off between form-based vs. meaning-based approaches to learning languages, for example, so that even individual L2 learners can deliberately change the way that they allocate attention during real-time second language processing to improve acquisition in the long run. The cherry on top would be to tailor these recommendations according to individuals' particular profiles (which may vary along dimensions of working memory, executive control, sheer tolerance for boredom during repetitive practice, etc., etc.).
I also performed research on how grammar learning can be conceived as the gradual acquisition of recurring statistical patterns in the input, such that grammatical constructions emerge gradually as the result of processes of abstraction over individual exemplars of a particular form. Language educators could facilitate this whole process by manipulating factors like token frequency ("How often do you encounter this particular form?"), type frequency ("How many different candidate forms do you usually see in this particular context?"), and input skewedness ("For this kind of context, to what extent does one or a few forms make up the lion's share of the input?"). Beyond informing psycholinguistic theory, this second research interest of mine could provide suggestions for language educators by addressing questions such as: does encountering a form more frequently necessarily make it easier to recognize and produce? Is it better to teach new grammar by using a wide variety of words in the examples, or by sticking to a few familiar vocabulary items? Should teachers approximate the input frequencies encountered “in the real world,” or is it possible to tailor classroom input to facilitate L2 acquisition?
I've also dipped my toes into corpus research with projects like:
Check out some recorded conference presentations below!
Drift diffusion modeling of reaction times in an artificial language experiment shows that apparent learning effects in participants without conscious awareness of an underlying rule were underlyingly driven by motor adaptation to predictable button-press sequences, whereas rule-aware learners showed both motor adaptation and sensitivity to noun semantics
Mass univariate analyses on EEG support dual-route models of morphosyntax processing (in other words, your brain composes "eat + s" instead of memorizing "eats" as one word)
Language learners with conscious awareness of a grammar rule perform similarly (with some caveats) no matter if they were told the rule or if they figured it out themselves:
Language learners with/without conscious awareness of a grammar rule that they've learned have differently-shaped reaction time distributions, hinting at different underlying cognitive processes
Assessing individual neurocognitive differences in native language morphosyntactic processing along an N400-P600 ERP continuum
Presented by first author and ex-labmate Dr. Irene Finestrat :)
Do first language neural processes for morphosyntax transfer to the second language?
Presented by first author Dr. Irene Finestrat
Neural oscillations measured via EEG as predictors of variability in second language learning
Yay resting state meditation!
Presented by first author, CogSLA undergraduate research assistant, and pre-medical student extraordinaire Victoria Ogunniyi
Examining individual variability in Event Related Potential responses and expanding the evidence to the second language
Presented by first author Dr. Irene Finestrat
New (as of 7 years ago) experimental methods in second language learning research
The methods have only gotten better since then, y'all! Machine learning-based brain decoding leaves univariate EEG and fMRI eating its dust.
From May 2022, titled "Disentangling neural indices of implicit vs. explicit morphosyntax processing in an artificial language."