Rethinking the Fundamentals of Sentence Comprehension

A nearly self-evident fact about human language is that knowing a language consists of two systems: a list of terms stored in long-term memory (the lexicon), and the procedural system for composing these terms into complex expressions (the grammar). This distinction is further reified in theories of language comprehension and typical psycholinguistic methodologies (Krauska & Lau 2023). By contrast, research in cognitive neuroscience of language paints a far murkier picture, with no obvious distinctions between brain areas that support sentence and word meanings (e.g., Shain et al. 2024) or syntactic vs. morphosyntactic structure (e.g., Stockall & Gwilliams 2023). How do we reconcile this?

Here, I present on two experiments that seek to bridge this gap. In the first experiment, I show that magnetoencephalography (MEG) responses to morphologically complex words in Bengali (Bangla) show sensitivity to both morphosyntactic and lexical-semantic properties in right anterior temporal lobe ~200ms. This suggests that retrieval of basic conceptual representations overlap with morphosyntactic structure building. In the second study, we use the 'rapid parallel visual presentation' (RPVP) paradigm, in which participants read short phrases in a single glance. This effectively prevents serial, word-by-word reading typically deployed in other experimental paradigms. We show that electroencephalography (EEG) activity elicited during the processing of noun-adjective phrases in Spanish shows simultaneous effects of lexical surprisal, neighborhood density effects, and syntactic processing ~250ms in posterior sensors. This further demonstrates that the brain simultaneously retrieves lexical items while processing morphosyntactic and syntactic structure. I then conclude by suggesting some boundary conditions for what a parsing architecture might look like that does not presuppose 'word-by-word' sentence processing nor a discrete boundary between the grammar and lexicon.