Alexandra Krauska, Ellen Lau (2023), Frontiers in Language Sciences 2, 1125127 (Hypothesis and Theory article) — DOI-10.3389/flang.2023.1125127
Key takeaways
- Standard production and comprehension models built on “lemmas” (stored triads of meaning, syntax and form, with syntax stopping at the word) inherit a lexicalist assumption that typology has refuted. Replacing that assumption needs new algorithms, not just new representations: separate meaning↔syntax and syntax↔form mappings with one integrated stage for retrieving and assembling syntactic structure. This predicts different aphasia profiles.
Seeded from the abstract and a skim, before the reading in NOTE-029. What follows is what the work says about itself.
Mainstream psycho- and neurolinguistic models retrieve “lemmas” or “lexical items” and combine them. This assumes lexical items store meaning, syntax and form together, that lexical and syntactic processes are separate, and that syntax does not go below the word. Decades of cross-linguistic work contradict this: syntax operates both above and below the word, and meaning and form depend partly on syntactic context. The authors review the arguments against lexicalism, show how lexicalist assumptions shape production models, and list phenomena those models struggle with. They sketch a non-lexicalist production model without lemmas, using separate meaning–syntax and syntax–form mappings.
Standing in the record
Filed from the survey of 2026-09-25 of work the anthology set aside as out of scope (tier B): 160 seconds of active reading over 5 sessions in the papers-feed tracker. It was filed Deferred, unread. NOTE-029 is the close reading of 2026-09-25, and it placed the work: Proposed — This is a clear diagnosis of where lemma models build in lexicalism, but its positive model is a verbal sketch (the full model is “in prep.”) and its aphasia predictions are untested; an implemented model, or lesion or aphasia data that separate its predictions from lemma-model ones, would settle it.
When seeded, before the reading in NOTE-029, the entry gave a priority for a deeper reading: medium — 160 s over 5 sessions. The skim captures the argument and model outline, and a deeper read matters only if psycholinguistic processing models become a thread.
What the seeding asked a deeper reading to check (NOTE-029 is that reading):
- It carries the typological argument (c61, c63) into cognitive and neural models. The claim that wordhood intuitions come from orthography matters for how word-level units in language models are interpreted.
- The proposed model is a sketch without computational implementation or new data. Check which predictions are concrete enough to test.
- It depends on Bruening (c63) and Haspelmath (c61) for the linguistic case, so filing these together makes sense.
Access when seeded: Crossref (published 2023-02-13) and the full PDF from frontiersin.org (17 pp., CC BY), which I read: front matter (received 15 Dec 2022, accepted 26 Jan 2023), abstract, §1 introduction, §2.1 on lemmas, the §3.1 model outline (8 stages), §3.2 on aphasia and §4 conclusion. The article keywords were not captured, so the keywords above are my own.