---
title: "LLM surprisal is necessary but not sufficient to capture English"
description: "The authors build a latent-process multinomial processing tree (MPT) model and fit it jointly to existing data from four… 102 references."
authors: ["Dario Paape", "Tal Linzen", "Shravan Vasishth"]
year: 2026
venue: "arXiv:2602.04489"
source: "https://arxiv.org/abs/2602.04489"
doi: "10.48550/arxiv.2602.04489"
arxiv: "2602.04489"
licence: "CC BY-NC-SA 4.0 (arXiv)"
licence_url: "http://creativecommons.org/licenses/by-nc-sa/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/paape-2026-llm-surprisal-necessary-but-sufficient/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/paape-2026-llm-surprisal-necessary-but-sufficient/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/35/) / LLM surprisal is necessary but not sufficient to capture English garden-path effects: Evidence from joint latent modeling of reading paradigms

# LLM surprisal is necessary but not sufficient to capture English garden-path effects: Evidence from joint latent modeling of reading paradigms

**Authors:** Dario Paape, [Tal Linzen](https://md.research.elamur.ai/free/en/authors/tal-linzen/) and Shravan Vasishth · **Year:** 2026 · **Venue:** arXiv:2602.04489 · **Licence:** [CC BY-NC-SA 4.0 (arXiv)](http://creativecommons.org/licenses/by-nc-sa/4.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2602.04489) · [PDF](https://arxiv.org/pdf/2602.04489) · [DOI 10.48550/arxiv.2602.04489](https://doi.org/10.48550/arxiv.2602.04489) · [arXiv 2602.04489](https://arxiv.org/abs/2602.04489) · [OpenAlex](https://openalex.org/W7127971132)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
Garden-path sentences, such as NP/Z structures where a comma would remove the temporary ambiguity, cause readers trouble. Each experimental method shows this trouble differently: as slowdowns, rereading, wrong answers or "ungrammatical" judgments. This makes results hard to compare across methods. It also leaves open whether LLM surprisal at the disambiguating word explains the difficulty on its own. The authors build a latent-process multinomial processing tree (MPT) model and fit it jointly to existing data from four paradigms: eye tracking, self-paced reading, bidirectional self-paced reading (BSPR) and Maze. The model separates three things: the chance of committing to the wrong initial parse, the cost of hitting an incompatible word, and the cost of reanalysis. It also allows for trials read inattentively. Under cross-validation, the MPT clearly predicts reading patterns and end-of-trial responses better than a surprisal-only model. Adding surprisal inside the MPT improves fit further. The conclusion is that surprisal is needed but cannot replace dedicated recovery processes, which matters for using LLMs as models of human parsing.

## Key points
- The model is fit jointly to four reading paradigms: eye tracking, unidirectional SPR, BSPR (which allows rereading) and Maze (forced word-by-word choice, no rereading).
- The latent cascade covers several processes:
  - attention versus inattention;
  - adopting the initial analysis;
  - covert reanalysis (slowing down in place) versus overt reanalysis (rereading).
- Misreadings and rejections are explained by guessing, failed reanalysis or pragmatic inference.
- Surprisal improved the MPT when it predicted reading time and initial garden-path cost. When it was linked to the probability of garden-pathing or to garden-path cost, fit did not improve and sometimes got worse.
- The surprisal baseline was given favourable conditions, yet the MPT still won:
  - only garden-path items were used, plus a precritical region as a baseline;
  - only first-pass times were used for eye tracking and BSPR.

## Interesting ideas
- Treat each trial as a mix of discrete latent paths (attentive or not, reanalysed or not) rather than one smooth cost, so that noisy trials stop distorting the cost estimates.
- Model the online process and the final judgment as outputs of one cascade, not as separate analyses, so that a wrong final answer can be traced to a specific failed step.
- Use an LLM-derived score as one additional input inside a structured model, and test where it actually helps rather than letting it replace the mechanism.
- Ask comprehension questions that probe lingering wrong readings after disambiguation, for example whether the team trained the striker.
- Make a comparison convincing by giving the rival baseline every advantage and still beating it.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (learn): silver data could carry a latent "careless witness" class, so that agreement statistics among witness parsers are not skewed by sloppy outputs.
- **I2** (tree): the parser could log an explicit initial attachment and a reanalysis event when a later word contradicts it, making repairs of NP/Z-type ambiguities visible in the explanation.
- **I3** (learn): LLM teacher signals could feed only into the decisions where ablation shows a gain, such as confidence, not into structural choices where they add nothing.
- **I4** (answer): add garden-path items with such probe questions as negative controls, checking that the world of the text kept the reanalysed reading and not the first one.
- **I5** (learn): when evaluating a new component against an LLM or bag-of-words baseline, tune the baseline on the target data so that any win is credible.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

📚 Shelves (6): [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works · [💭 Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/) · 1,552 works · [💭 Developmental and Educational Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/developmental-and-educational-psychology/) · 499 works · [🧠 Neuroscience](https://md.research.elamur.ai/free/en/shelves/neuroscience/) · 165 works · [🧠 Cognitive Neuroscience](https://md.research.elamur.ai/free/en/shelves/neuroscience/cognitive-neuroscience/) · 164 works

🧭 Topics (3): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,320 works · [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,263 works · [🧒 Psycholinguistics & language acquisition](https://md.research.elamur.ai/free/en/topics/psycholinguistics/) · 102 works

## Cite

Cite as: Dario Paape, Tal Linzen and Shravan Vasishth. 2026. LLM surprisal is necessary but not sufficient to capture English garden-path effects: Evidence from joint latent modeling of reading paradigms. arXiv:2602.04489.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{paape2026llm,
  title = {{LLM surprisal is necessary but not sufficient to capture English garden-path effects: Evidence from joint latent modeling of reading paradigms}},
  author = {Dario Paape, Tal Linzen and Shravan Vasishth},
  year = {2026},
  doi = {10.48550/arxiv.2602.04489},
  eprint = {2602.04489},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2602.04489}
}
```

- [← previous LLM Probe: Evaluating LLMs for Low-Resource Languages](https://md.research.elamur.ai/free/en/papers/teklehaymanot-2026-llm-probe-evaluating-llms-low/)
- [next → LLM-as-a-Jury for Machine Translation Publishability Assessment](https://md.research.elamur.ai/free/en/papers/yanishevsky-2026-llm-as-jury-machine-translation/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/35/)

## References

102 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Andrews, S., & Veldre, A. (2021). Wrapping up sentence comprehension: The role of task demands and individual differences. Scientific Studies of Reading, 25 (2), 123–140. [↗](https://doi.org/10.1080/10888438.2020.1817028)
2. Bader, M., Meng, M., et al. (2023). Processing noncanonical sentences: Effects of context on online processing and (mis) interpretation. Glossa Psycholinguistics, 2 (1). [↗](https://doi.org/10.5070/g6011117)
3. Bixler, R., & D’Mello, S. (2015). Automatic gaze-based detection of mind wandering with metacognitive awareness. International Conference on User Modeling, Adaptation, and Personalization, 31–43. [find](https://scholar.google.com/scholar?q=Bixler%2C+R%2E%2C+%26+D%E2%80%99Mello%2C+S%2E+%282015%29%2E+Automatic+gaze%2Dbased+detection+of+mind+wandering+with+metacognitive+awareness%2E+International+Conference+on+User+Modeling%2C+Adaptation%2C+and+Personalization%2C+31%E2%80%9343%2E)
4. Boston, M. F., Hale, J., Kliegl, R., & Vasishth, S. (2008). Surprising parser actions and reading difficulty. Proceedings of ACL-08: HLT, Short Papers, 5–8. [find](https://scholar.google.com/scholar?q=Boston%2C+M%2E+F%2E%2C+Hale%2C+J%2E%2C+Kliegl%2C+R%2E%2C+%26+Vasishth%2C+S%2E+%282008%29%2E+Surprising+parser+actions+and+reading+difficulty%2E+Proceedings+of+ACL%2D08%3A+HLT%2C+Short+Papers%2C+5%E2%80%938%2E)
5. Boyce, V., & Levy, R. (2023). A-maze of natural stories: Comprehension and surprisal in the maze task. Glossa Psycholinguistics, 2 (1). [↗](https://doi.org/10.5070/g6011190)
6. Christianson, K., Dempsey, J., Tsiola, A., & Goldshtein, M. (2022). What if they’re just not that into you (or your experiment)? On motivation and psycholinguistics. In Psychology of learning and motivation (pp. 51–88, Vol. 76). Elsevier. [find](https://scholar.google.com/scholar?q=Christianson%2C+K%2E%2C+Dempsey%2C+J%2E%2C+Tsiola%2C+A%2E%2C+%26+Goldshtein%2C+M%2E+%282022%29%2E+What+if+they%E2%80%99re+just+not+that+into+you+%28or+your+experiment%29%3F+On+motivation+and+psycholinguistics%2E+In+Psychology+of+learning+and+moti)
7. Christianson, K., Hollingworth, A., Halliwell, J. F., & Ferreira, F. (2001). Thematic roles assigned along the garden path linger. Cognitive Psychology, 42 (4), 368–407. [↗](https://doi.org/10.1006/cogp.2001.0752)
8. Christianson, K., Luke, S. G., Hussey, E. K., & Wochna, K. L. (2017). Why reread? evidence from garden-path and local coherence structures. Quarterly Journal of Experimental Psychology, 70 (7), 1380–1405. [↗](https://doi.org/10.1080/17470218.2016.1186200)
9. Clifton Jr, C., & Ferreira, F. (1989). Ambiguity in context. Language and Cognitive Processes, 4 (3-4), SI77–SI103. [↗](https://doi.org/10.1080/01690968908406364)
10. Engbert, R., Nuthmann, A., Richter, E. M., & Kliegl, R. (2005). SWIFT: a dynamical model of saccade generation during reading. Psychological Review, 112 (4), 777. [↗](https://doi.org/10.1037/0033-295x.112.4.777)
11. Erdfelder, E., Auer, T.-S., Hilbig, B. E., Aßfalg, A., Moshagen, M., & Nadarevic, L. (2009). Multinomial processing tree models: A review of the literature. Zeitschrift für Psychologie/Journal of Psychology, 217 (3), 108–124. [find](https://scholar.google.com/scholar?q=Erdfelder%2C+E%2E%2C+Auer%2C+T%2E%2DS%2E%2C+Hilbig%2C+B%2E+E%2E%2C+A%C3%9Ffalg%2C+A%2E%2C+Moshagen%2C+M%2E%2C+%26+Nadarevic%2C+L%2E+%282009%29%2E+Multinomial+processing+tree+models%3A+A+review+of+the+literature%2E+Zeitschrift+fu%CC%88r+Psychologie%2FJournal+of+Psy)
12. Ferreira, F., Christianson, K., & Hollingworth, A. (2001). Misinterpretations of garden-path sentences: Implications for models of sentence processing and reanalysis. Journal of Psycholinguistic Research, 30 (1), 3–20. [↗](https://doi.org/10.1023/a:1005290706460)
13. Fodor, J. D., & Inoue, A. (2000). Garden path repair: Diagnosis and triage. Language and Speech, 43 (3), 261–271. [↗](https://doi.org/10.1177/00238309000430030201)
14. Forster, K. I., Guerrera, C., & Elliot, L. (2009). The maze task: Measuring forced incremental sentence processing time. Behavior Research Methods, 41 (1), 163–171. [↗](https://doi.org/10.3758/brm.41.1.163)
15. Frank, S. L., Fernandez Monsalve, I., Thompson, R. L., & Vigliocco, G. (2013). Reading time data for evaluating broad-coverage models of English sentence processing. Behavior Research Methods, 45 (4), 1182–1190. [↗](https://doi.org/10.3758/s13428-012-0313-y)
16. Frazier, L. (1979). On comprehending sentences: Syntactic parsing strategies [Doctoral dissertation]. University of Connecticut. [find](https://scholar.google.com/scholar?q=Frazier%2C+L%2E+%281979%29%2E+On+comprehending+sentences%3A+Syntactic+parsing+strategies+%5BDoctoral+dissertation%5D%2E+University+of+Connecticut%2E)
17. Frazier, L., & Rayner, K. (1982). Making and correcting errors during sentence comprehension: Eye movements in the analysis of structurally ambiguous sentences. Cognitive Psychology, 14 (2), 178–210. [↗](https://doi.org/10.1016/0010-0285(82)90008-1)
18. Fujita, H. (2025). The L-maze task and web-based data collection in second language sentence processing research. Research Methods in Applied Linguistics, 4 (3), 100251. [↗](https://doi.org/10.1016/j.rmal.2025.100251)
19. Gabry, J., Cešnovar, R., Johnson, A., & Bronder, S. (2024). cmdstanr: R Interface to ’CmdStan’ [[https://mc-stan.org/cmdstanr/](https://mc-stan.org/cmdstanr/), https://discourse.mc-stan.org]. [↗](https://mc-stan.org/cmdstanr/)
20. Gelman, A., Meng, X.-L., & Stern, H. (1996). Posterior predictive assessment of model fitness via realized discrepancies. Statistica Sinica, 733–760. [find](https://scholar.google.com/scholar?q=Gelman%2C+A%2E%2C+Meng%2C+X%2E%2DL%2E%2C+%26+Stern%2C+H%2E+%281996%29%2E+Posterior+predictive+assessment+of+model+fitness+via+realized+discrepancies%2E+Statistica+Sinica%2C+733%E2%80%93760%2E)
21. Gu, A., & Dao, T. (2023). Mamba: Linear-time sequence modeling with selective state spaces. arXiv preprint arXiv:2312.00752. [↗](https://arxiv.org/abs/2312.00752)
22. Hale, J. (2001). [A probabilistic Earley parser as a psycholinguistic model](https://md.research.elamur.ai/free/en/papers/hale-2001-probabilistic-earley-parser-as-psycholinguistic/). Second Meeting of the North American Chapter of the Association for Computational Linguistics, 1–8.
23. Hammerly, C., Staub, A., & Dillon, B. (2019). The grammaticality asymmetry in agreement attraction reflects response bias: Experimental and modeling evidence. Cognitive Psychology, 110, 70–104. [↗](https://doi.org/10.1016/j.cogpsych.2019.01.001)
24. Heck, D. W., Erdfelder, E., & Kieslich, P. J. (2018). Generalized processing tree models: Jointly modeling discrete and continuous variables. Psychometrika, 83 (4), 893–918. [↗](https://doi.org/10.1007/s11336-018-9622-0)
25. Hofmann, M. J., Remus, S., Biemann, C., Radach, R., & Kuchinke, L. (2022). Language models explain word reading times better than empirical predictability. Frontiers in Artificial Intelligence, 4, 730570. [↗](https://doi.org/10.3389/frai.2021.730570)
26. Huang, K.-J., Arehalli, S., Kugemoto, M., Muxica, C., Prasad, G., Dillon, B., & Linzen, T. (2024). Large-scale benchmark yields no evidence that language model surprisal explains syntactic disambiguation difficulty. Journal of Memory and Language, 137, 104510. [↗](https://doi.org/10.1016/j.jml.2024.104510)
27. Huang, K.-J., & Dillon, B. (2023). An infrequent, large cost underlies garden path effects: an RT distribution approach. Talk presented at HSP 2023, University of Pittsburgh. [find](https://scholar.google.com/scholar?q=Huang%2C+K%2E%2DJ%2E%2C+%26+Dillon%2C+B%2E+%282023%29%2E+An+infrequent%2C+large+cost+underlies+garden+path+effects%3A+an+RT+distribution+approach%2E+Talk+presented+at+HSP+2023%2C+University+of+Pittsburgh%2E)
28. Huang, Y., & Ferreira, F. (2021). What causes lingering misinterpretations of garden-path sentences: Incorrect syntactic representations or fallible memory processes? Journal of Memory and Language, 121, 104288. [↗](https://doi.org/10.1016/j.jml.2021.104288)
29. Jurafsky, D. (1996). A probabilistic model of lexical and syntactic access and disambiguation. Cognitive Science, 20 (2), 137–194. [↗](https://doi.org/10.1016/s0364-0213(99)80005-6)
30. Just, M. A., Carpenter, P. A., & Woolley, J. D. (1982). Paradigms and processes in reading comprehension. Journal of Experimental Psychology: General, 111 (2), 228–238. [↗](https://doi.org/10.1037//0096-3445.111.2.228)
31. Keshev, M., & Meltzer-Asscher, A. (2020). The effects of syntactic pressures and pragmatic considerations on predictive dependency formation. Language, Cognition and Neuroscience, 35 (2), 256–272. [↗](https://doi.org/10.1080/23273798.2019.1650946)
32. Koornneef, A. W., & Van Berkum, J. J. (2006). On the use of verb-based implicit causality in sentence comprehension: Evidence from self-paced reading and eye tracking. Journal of Memory and Language, 54 (4), 445–465. [↗](https://doi.org/10.1016/j.jml.2005.12.003)
33. Levy, R. (2008). Expectation-based syntactic comprehension. Cognition, 106 (3), 1126–1177. [↗](https://doi.org/10.1016/j.cognition.2007.05.006)
34. Lewandowski, D., Kurowicka, D., & Joe, H. (2009). Generating random correlation matrices based on vines and extended onion method. Journal of Multivariate Analysis, 100 (9), 1989–2001. [↗](https://doi.org/10.1016/j.jmva.2009.04.008)
35. Lewis, R. L. (1998). Reanalysis and limited repair parsing: Leaping off the garden path. In J. D. Fodor & F. Ferreira (Eds.), Reanalysis in sentence processing (pp. 247–285). Kluwer. [↗](https://doi.org/10.1007/978-94-015-9070-9_8)
36. Li, A., Feng, X., Narang, S., Peng, A., Cai, T., Shah, R. S., & Varma, S. (2024). Incremental comprehension of garden-path sentences by large language models: Semantic interpretation, syntactic re-analysis, and attention. arXiv preprint arXiv:2405.16042. [↗](https://arxiv.org/abs/2405.16042)
37. Lin, Z. (2025). Six fallacies in substituting large language models for human participants. Advances in Methods and Practices in Psychological Science, 8 (3), 25152459251357566. [↗](https://doi.org/10.1177/25152459251357566)
38. Lopes Rego, A. T., Snell, J., & Meeter, M. (2024). Language models outperform cloze predictability in a cognitive model of reading. PLOS Computational Biology, 20 (9), e1012117. [↗](https://doi.org/10.1371/journal.pcbi.1012117)
39. Maina-Kilaas, A., & Levy, R. (2026a). Algorithmic consequences of particle filters for sentence processing: Amplified garden-paths and digging-in effects. https://arxiv.org/abs/2603.11412 [↗](https://arxiv.org/abs/2603.11412)
40. Maina-Kilaas, A., & Levy, R. (2026b). Revisiting real-time digging-in effects: No evidence from NP/Z garden-paths. https://arxiv.org/abs/2603.23624 [↗](https://arxiv.org/abs/2603.23624)
41. Marcus, M. P. (1980). A theory of syntactic recognition for natural language. MIT Press. [find](https://scholar.google.com/scholar?q=Marcus%2C+M%2E+P%2E+%281980%29%2E+A+theory+of+syntactic+recognition+for+natural+language%2E+MIT+Press%2E)
42. McLachlan, G. J., & Peel, D. (2000). Finite mixture models. Wiley Series in Probability and Statistics. [↗](https://doi.org/10.1002/0471721182)
43. McRae, K., Spivey-Knowlton, M. J., & Tanenhaus, M. K. (1998). Modeling the influence of thematic fit (and other constraints) in on-line sentence comprehension. Journal of Memory and Language, 38, 283–312. [↗](https://doi.org/10.1006/jmla.1997.2543)
44. Meng, M., & Bader, M. (2000). Ungrammaticality detection and garden path strength: Evidence for serial parsing. Language and Cognitive Processes, 15 (6), 615–666. [↗](https://doi.org/10.1080/016909600750040580)
45. Michaelov, J. A., Arnett, C., & Bergen, B. K. (2024). Revenge of the fallen? Recurrent models match transformers at predicting human language comprehension metrics. https://arxiv.org/abs/2404.19178 [↗](https://arxiv.org/abs/2404.19178)
46. Michaelov, J. A., Coulson, S., & Bergen, B. K. (2023). Can peanuts fall in love with distributional semantics? Proceedings of the 45th Annual Conference of the Cognitive Science Society. [↗](https://doi.org/10.48550/arxiv.2301.08731)
47. Mills, C., Graesser, A., Risko, E. F., & D’Mello, S. K. (2017). Cognitive coupling during reading. Journal of Experimental Psychology: General, 146 (6), 872–883. [↗](https://doi.org/10.1037/xge0000309)
48. Milne, R. W. (1982). Predicting garden path sentences. Cognitive Science, 6 (4), 349–373. [↗](https://doi.org/10.1016/s0364-0213(82)80009-8)
49. Mitchell, D. C., Shen, X., Green, M. J., & Hodgson, T. L. (2008). Accounting for regressive eye-movements in models of sentence processing: A reappraisal of the selective reanalysis hypothesis. Journal of Memory and Language, 59 (3), 266–293. [↗](https://doi.org/10.1016/j.jml.2008.06.002)
50. Nicenboim, B. (2023). The CoFI Reader: A Continuous Flow of Information approach to modeling reading. Paper presented at MathPsych/ICCM/EMPG 2023. [find](https://scholar.google.com/scholar?q=Nicenboim%2C+B%2E+%282023%29%2E+The+CoFI+Reader%3A+A+Continuous+Flow+of+Information+approach+to+modeling+reading%2E+Paper+presented+at+MathPsych%2FICCM%2FEMPG+2023%2E)
51. Nicenboim, B., Schad, D. J., & Vasishth, S. (2025). Introduction to Bayesian data analysis for cognitive science. CRC Press. [↗](https://doi.org/10.1201/9780429342646)
52. Nicenboim, B., & Vasishth, S. (2018). Models of retrieval in sentence comprehension: A computational evaluation using bayesian hierarchical modeling. Journal of Memory and Language, 99, 1–34. [↗](https://doi.org/10.1016/j.jml.2017.08.004)
53. Oh, B.-D., & Linzen, T. (2025). To model human linguistic prediction, make LLMs less superhuman. Trends in Cognitive Sciences. [↗](https://doi.org/10.1016/j.tics.2026.05.008)
54. Oh, B.-D., & Schuler, W. (2023). [Why does surprisal from larger transformer-based language models provide a poorer fit to human reading times?](https://md.research.elamur.ai/free/en/papers/oh-2023-why-does-surprisal-larger-transformer/) Transactions of the Association for Computational Linguistics, 11, 336–350. [↗](https://doi.org/10.1162/tacl_a_00548)
55. O’Leary, R. M., Omori-Hoffe, N., Dugan, G., & Wingfield, A. (2026). A drift-diffusion decomposition of conditions that influence shallow (“good enough”) processing of heard sentences. Memory & Cognition, 54 (1), 244–257. [find](https://scholar.google.com/scholar?q=O%E2%80%99Leary%2C+R%2E+M%2E%2C+Omori%2DHoffe%2C+N%2E%2C+Dugan%2C+G%2E%2C+%26+Wingfield%2C+A%2E+%282026%29%2E+A+drift%2Ddiffusion+decomposition+of+conditions+that+influence+shallow+%28%E2%80%9Cgood+enough%E2%80%9D%29+processing+of+heard+sentences%2E+Memory+%26+Cogniti)
56. Paape, D., Linzen, T., & Vasishth, S. (2026). Deconstructing sentence disambiguation by joint latent modeling of reading paradigms: LLM surprisal is not enough. In Proceedings of the 48th Annual Meeting of the Cognitive Science Society. [https://escholarship.org/uc/item/1q8735d3](https://escholarship.org/uc/item/1q8735d3) [↗](https://escholarship.org/uc/item/1q8735d3)
57. Paape, D., & Vasishth, S. (2022a). Conscious rereading is confirmatory: Evidence from bidirectional self-paced reading. Glossa Psycholinguistics, 1 (1). [↗](https://doi.org/10.5070/g6011182)
58. Paape, D., & Vasishth, S. (2022b). Estimating the true cost of garden pathing: A computational model of latent cognitive processes. Cognitive Science, 46 (8), e13186. [↗](https://doi.org/10.1111/cogs.13186)
59. Paape, D., & Vasishth, S. (2022c). Is reanalysis selective when regressions are consciously controlled? Glossa Psycholinguistics, 1 (1). [↗](https://doi.org/10.5070/g601139)
60. Paape, D., & Vasishth, S. (2026). Context ameliorates but does not eliminate garden-pathing: Novel insights from latent-process modeling. Journal of Memory and Language, 148, 104748. [↗](https://doi.org/10.1016/j.jml.2026.104748)
61. Peng, B., Alcaide, E., Anthony, Q., Albalak, A., Arcadinho, S., Biderman, S., Cao, H., Cheng, X., Chung, M., Derczynski, L., et al. (2023). RWKV: Reinventing RNNs for the transformer era. Findings of the Association for Computational Linguistics: EMNLP 2023, 14048–14077. [↗](https://doi.org/10.48550/arxiv.2305.13048)
62. Pitt, M. A., & Myung, I. J. (2002). When a good fit can be bad. Trends in Cognitive Sciences, 6 10, 421–425. [↗](https://doi.org/10.1016/s1364-6613(02)01964-2)
63. Pritchett, B. L. (1992). Grammatical competence and parsing performance. University of Chicago Press. [find](https://scholar.google.com/scholar?q=Pritchett%2C+B%2E+L%2E+%281992%29%2E+Grammatical+competence+and+parsing+performance%2E+University+of+Chicago+Press%2E)
64. Pynte, J. (1997). Syntactic ambiguity resolution while reading in second and native languages. The Quarterly Journal of Experimental Psychology A, 50 (1), 119–148. [find](https://scholar.google.com/scholar?q=Pynte%2C+J%2E+%281997%29%2E+Syntactic+ambiguity+resolution+while+reading+in+second+and+native+languages%2E+The+Quarterly+Journal+of+Experimental+Psychology+A%2C+50+%281%29%2C+119%E2%80%93148%2E)
65. R Core Team. (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria. [https://www.R-project.org/](https://www.R-project.org/) [↗](https://www.R-project.org/)
66. Radford, A., Wu, J., Child, R., Luan, D., Amodei, D., & Sutskever, I. (2019). Language models are unsupervised multitask learners. [find](https://scholar.google.com/scholar?q=Radford%2C+A%2E%2C+Wu%2C+J%2E%2C+Child%2C+R%2E%2C+Luan%2C+D%2E%2C+Amodei%2C+D%2E%2C+%26+Sutskever%2C+I%2E+%282019%29%2E+Language+models+are+unsupervised+multitask+learners%2E)
67. Ratcliff, R. (2006). Modeling response signal and response time data. Cognitive psychology, 53 (3), 195–237. [↗](https://doi.org/10.1016/j.cogpsych.2005.10.002)
68. Ratcliff, R., & Tuerlinckx, F. (2002). Estimating parameters of the diffusion model: Approaches to dealing with contaminant reaction times and parameter variability. Psychonomic bulletin & Review, 9 (3), 438–481. [↗](https://doi.org/10.3758/bf03196302)
69. Rayner, K., Chace, K. H., Slattery, T. J., & Ashby, J. (2006). Eye movements as reflections of comprehension processes in reading. Scientific Studies of Reading, 10 (3), 241–255. [↗](https://doi.org/10.1207/s1532799xssr1003_3)
70. Reichle, E. D., Rayner, K., & Pollatsek, A. (2003). The EZ Reader model of eye-movement control in reading: Comparisons to other models. Behavioral and Brain Sciences, 26 (4), 445–476. [find](https://scholar.google.com/scholar?q=Reichle%2C+E%2E+D%2E%2C+Rayner%2C+K%2E%2C+%26+Pollatsek%2C+A%2E+%282003%29%2E+The+EZ+Reader+model+of+eye%2Dmovement+control+in+reading%3A+Comparisons+to+other+models%2E+Behavioral+and+Brain+Sciences%2C+26+%284%29%2C+445%E2%80%93476%2E)
71. Reichle, E. D., Warren, T., & McConnell, K. (2009). Using EZ Reader to model the effects of higher level language processing on eye movements during reading. Psychonomic Bulletin & Review, 16 (1), 1–21. [find](https://scholar.google.com/scholar?q=Reichle%2C+E%2E+D%2E%2C+Warren%2C+T%2E%2C+%26+McConnell%2C+K%2E+%282009%29%2E+Using+EZ+Reader+to+model+the+effects+of+higher+level+language+processing+on+eye+movements+during+reading%2E+Psychonomic+Bulletin+%26+Review%2C+16+%281%29%2C+1%E2%80%932)
72. Riefer, D. M., & Batchelder, W. H. (1988). Multinomial modeling and the measurement of cognitive processes. Psychological Review, 95 (3), 318. [↗](https://doi.org/10.1037//0033-295x.95.3.318)
73. Rouder, J. N. (2005). Are unshifted distributional models appropriate for response time? Psychometrika, 70 (2), 377–381. [↗](https://doi.org/10.1007/s11336-005-1297-7)
74. Schröder, S., Morgenroth, T., Kuhl, U., Vaquet, V., & Paaßen, B. (2025). Large language models do not simulate human psychology. https://arxiv.org/abs/2508.06950 [↗](https://arxiv.org/abs/2508.06950)
75. Shain, C., Meister, C., Pimentel, T., Cotterell, R., & Levy, R. (2024). Large-scale evidence for logarithmic effects of word predictability on reading time. Proceedings of the National Academy of Sciences, 121 (10), e2307876121. [↗](https://doi.org/10.1073/pnas.2307876121)
76. Slaats, S., & Martin, A. E. (2025). What’s surprising about surprisal. Computational Brain & Behavior, 8 (2), 233–248. [↗](https://doi.org/10.1007/s42113-025-00237-9)
77. Smith, N. J., & Levy, R. (2013). The effect of word predictability on reading time is logarithmic. Cognition, 128 (3), 302–319. [↗](https://doi.org/10.1016/j.cognition.2013.02.013)
78. Snell, J., van Leipsig, S., Grainger, J., & Meeter, M. (2018). OB1-reader: A model of word recognition and eye movements in text reading. Psychological Review, 125 (6), 969–984. [↗](https://doi.org/10.1037/rev0000119)
79. Stan Development Team. (2024). Stan reference manual, version 2.36.0. [find](https://scholar.google.com/scholar?q=Stan+Development+Team%2E+%282024%29%2E+Stan+reference+manual%2C+version+2%2E36%2E0%2E)
80. Staub, A. (2025). Predictability in language comprehension: Prospects and problems for surprisal. Annual Review of Linguistics, 11 (1), 17–34. [↗](https://doi.org/10.1146/annurev-linguistics-011724-121517)
81. Sturt, P., Pickering, M. J., & Crocker, M. W. (1999). Structural change and reanalysis difficulty in language comprehension. Journal of Memory and Language, 40 (1), 136–150. [↗](https://doi.org/10.1006/jmla.1998.2606)
82. Timkey, W., Dillon, B., & Linzen, T. (2026). Why are language models less surprised than humans? testing the parse multiplicity mismatch hypothesis. arXiv preprint arXiv:2605.15440. [↗](https://arxiv.org/abs/2605.15440)
83. Timkey, W., Huang, K.-J., Oh, B.-D., Prasad, G., Arehalli, S., Linzen, T., & Dillon, B. (2026). Eye movements reveal a dissociation between prediction and structural processing difficulty in language comprehension. Proceedings of the National Academy of Sciences, 123 (32), e2532230123. [↗](https://doi.org/10.1073/pnas.2532230123)
84. Van Gompel, R. P., Pickering, M. J., & Traxler, M. J. (2000). Unrestricted race: A new model of syntactic ambiguity resolution. In A. Kennedy, R. Radach, D. Heller, & J. Pynte (Eds.), Reading as a perceptual process (pp. 621–648). Elsevier. [↗](https://doi.org/10.1016/b978-008043642-5/50029-2)
85. Van Schijndel, M., & Linzen, T. (2021). Single-stage prediction models do not explain the magnitude of syntactic disambiguation difficulty. Cognitive Science, 45 (6), e12988. [↗](https://doi.org/10.1111/cogs.12988)
86. Vasishth, S., Chopin, N., Ryder, R., & Nicenboim, B. (2017). Modelling dependency completion in sentence comprehension as a bayesian hierarchical mixture process: A case study involving chinese relative clauses. arXiv preprint arXiv:1702.00564. [↗](https://arxiv.org/abs/1702.00564)
87. Vasishth, S., Jäger, L. A., & Nicenboim, B. (2017). Feature overwriting as a finite mixture process: Evidence from comprehension data. arXiv preprint arXiv:1703.04081. [↗](https://arxiv.org/abs/1703.04081)
88. Vasishth, S., von der Malsburg, T., & Engelmann, F. (2013). What eye movements can tell us about sentence comprehension. Wiley Interdisciplinary Reviews: Cognitive Science, 4 (2), 125–134. [↗](https://doi.org/10.1002/wcs.1209)
89. Vehtari, A., Gabry, J., Magnusson, M., Yao, Y., Bürkner, P.-C., Paananen, T., & Gelman, A. (2022). loo: Efficient leave-one-out cross-validation and WAIC for Bayesian models. The R Foundation. [https://mc-stan.org/loo](https://mc-stan.org/loo) [↗](https://mc-stan.org/loo)
90. Vehtari, A., Gelman, A., & Gabry, J. (2017). Practical bayesian model evaluation using leave-one-out cross-validation and waic. Statistics and Computing, 27 (5), 1413–1432. [↗](https://doi.org/10.1007/s11222-016-9696-4)
91. Vehtari, A., Gelman, A., Simpson, D., Carpenter, B., & Bürkner, P.-C. (2021). Rank-Normalization, b for Assessing Convergence of MCMC (with Folding, and Localization: An Improved R Discussion). Bayesian Analysis, 16 (2), 667–718. [find](https://scholar.google.com/scholar?q=Vehtari%2C+A%2E%2C+Gelman%2C+A%2E%2C+Simpson%2C+D%2E%2C+Carpenter%2C+B%2E%2C+%26+Bu%CC%88rkner%2C+P%2E%2DC%2E+%282021%29%2E+Rank%2DNormalization%2C+b+for+Assessing+Convergence+of+MCMC+%28with+Folding%2C+and+Localization%3A+An+Improved+R+Discussion%29%2E+Bayes)
92. Verdonck, S., & Tuerlinckx, F. (2016). Factoring out nondecision time in choice reaction time data: Theory and implications. Psychological Review, 123 (2), 208–218. [↗](https://doi.org/10.1037/rev0000019)
93. von der Malsburg, T., & Vasishth, S. (2011). What is the scanpath signature of syntactic reanalysis? Journal of Memory and Language, 65 (2), 109–127. [↗](https://doi.org/10.1016/j.jml.2011.02.004)
94. Waltz, D. L., & Pollack, J. B. (1985). Massively parallel parsing: A strongly interactive model of natural language interpretation. Cognitive Science, 9 (1), 51–74. [↗](https://doi.org/10.1016/s0364-0213(85)80009-4)
95. Wang, D., Sadrzadeh, M., Stanojević, M., Chow, W.-Y., & Breheny, R. (2025, January). Extracting structure from an LLM - how to improve on surprisal-based models of human language processing. In O. Rambow, L. Wanner, M. Apidianaki, H. Al-Khalifa, B. D. Eugenio, & S. Schockaert (Eds.), Proceedings of the 31st international conference on computational linguistics (pp. 4938–4944). Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Wang%2C+D%2E%2C+Sadrzadeh%2C+M%2E%2C+Stanojevic%CC%81%2C+M%2E%2C+Chow%2C+W%2E%2DY%2E%2C+%26+Breheny%2C+R%2E+%282025%2C+January%29%2E+Extracting+structure+from+an+LLM+%2D+how+to+improve+on+surprisal%2Dbased+models+of+human+language+processing%2E+In+O%2E+Ra)
96. Warner, J., & Glass, A. L. (1987). Context and distance-to-disambiguation effects in ambiguity resolution: Evidence from grammaticality judgments of garden path sentences. Journal of Memory and Language, 26 (6), 714–738. [↗](https://doi.org/10.1016/0749-596x(87)90111-2)
97. Wilcox, E. G., Pimentel, T., Meister, C., Cotterell, R., & Levy, R. P. (2023). [Testing the predictions of surprisal theory in 11 languages](https://md.research.elamur.ai/free/en/papers/wilcox-2023-testing-predictions-surprisal-theory-11/). Transactions of the Association for Computational Linguistics, 11, 1451–1470. [↗](https://doi.org/10.1162/tacl_a_00612)
98. Witzel, N., Witzel, J., & Forster, K. (2012). Comparisons of online reading paradigms: Eye tracking, moving-window, and maze. Journal of Psycholinguistic Research, 41 (2), 105–128. [↗](https://doi.org/10.1007/s10936-011-9179-x)
99. Wolf, T., Debut, L., Sanh, V., Chaumond, J., Delangue, C., Moi, A., Cistac, P., Rault, T., Louf, R., Funtowicz, M., et al. (2019). Huggingface’s [transformers: State-of-the-art natural language processing](https://md.research.elamur.ai/free/en/papers/wolf-2020-transformers-state-art-natural-language/). arXiv preprint arXiv:1910.03771. [↗](https://arxiv.org/abs/1910.03771)
100. Yarkoni, T. (2022). The generalizability crisis. Behavioral and Brain Sciences, 45, e1. [find](https://scholar.google.com/scholar?q=Yarkoni%2C+T%2E+%282022%29%2E+The+generalizability+crisis%2E+Behavioral+and+Brain+Sciences%2C+45%2C+e1%2E)
101. Yoshida, R., Isono, S., Someya, T., Oseki, Y., & Kuribayashi, T. (2026). [An existence proof for neural language models that can explain garden-path effects via surprisal](https://md.research.elamur.ai/free/en/papers/yoshida-2026-existence-proof-neural-language-models/). In M. Liakata, V. P. Moreira, J. Zhang, & D. Jurgens (Eds.), Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) (pp. 36556–36569). Association for Computational Linguistics. [↗](https://doi.org/10.48550/arxiv.2604.18293)
102. Zhang, L., Carpenter, B., Gelman, A., & Vehtari, A. (2022). Pathfinder: Parallel quasi-Newton variational inference. Journal of Machine Learning Research, 23 (306), 1–49. [find](https://scholar.google.com/scholar?q=Zhang%2C+L%2E%2C+Carpenter%2C+B%2E%2C+Gelman%2C+A%2E%2C+%26+Vehtari%2C+A%2E+%282022%29%2E+Pathfinder%3A+Parallel+quasi%2DNewton+variational+inference%2E+Journal+of+Machine+Learning+Research%2C+23+%28306%29%2C+1%E2%80%9349%2E)

*Updated 1 November 2026*