Linear and Logarithmic Effects Coexist in Language Predictability
coryshain · x · 2026-07-13
This thread summarizes the core findings of a new preprint:
- Researchers re-examined the impact of language predictability on reading/reaction times, finding evidence that linear effects and logarithmic effects coexist rather than being mutually exclusive.
- They argue that results from N400 research and a recent large-scale eye-tracking study are also consistent with the view that these effects can coexist.
- The authors further suggest that this research helps clarify current confusion in debates such as "cloze probability vs. LM probability," "controlled experiments vs. corpus studies," "prediction vs. integration," and "behavioral data vs. ERP."
- In a supplementary note to the thread, the authors add that the evidence for "predictive preactivation" is already quite strong and hard to explain by Bayesian belief updating alone. In their replication data, the original BK21 sample size was insufficient, but their own larger-scale replication successfully found the independent contribution of linear predictability.
Related event: Study Reveals Linear and Logarithmic Patterns in Word Predictability(3 posts)→
More from Research
- DeBias-CLIP tackles CLIP’s long-caption bias and hits state-of-the-art retrieval — Mila_Quebec · 2026-07-21
- Fable 5 is credited with a 3-variable counterexample to the Jacobian conjecture — Various-Affect4841 · 2026-07-21
- Anthropic says frontier models showed harmful behavior in tool-rich simulations — gerardsans · 2026-07-21
- Paper studies long-run behavior in linear-quadratic graphon mean field control — chaumian · 2026-07-21
- An interactive Zarr explainer shows how AI is changing technical education — MaxLenormand · 2026-07-21
- 3D-Fit finds LLMs can handle multiple molecular constraints, but still lag diffusion models — insilicomedicine · 2026-07-21