iiinova — where your digital self keeps evolving

Shravan Vasishth @shravanvasishth.bsky.social · 01/10/2026

Applications are open for the 2027 summer school on statistical methods for linguistics and psychology: All details are here: smlp.science

smlp.science

The Eleventh Summer School on Statistical Methods for Linguistics and Psychology

01718

Reposted by Shravan Vasishth

Dario Paape @dariopaape.eurosky.social · 25/09/2026

Updated preprint with @tallinzen.bsky.social and @shravanvasishth.bsky.social : "LLM surprisal is necessary but not sufficient to capture English garden-path effects: Evidence from joint latent modeling of reading paradigms" arxiv.org/abs/2602.04489

032

Reposted by Shravan Vasishth

Charles Margossian @charlesm993.bsky.social · 26/06/2026

📘 With the release of our textbook "Bayesian Workflow" (avehtari.github.io/Bayesian-Wor...), I figured I'd also share the content of my graduate course on the topic at UBC. 🌎 charlesm93.github.io/stat547/ The course contains overlapping and complementary material, homeworks and reading.

avehtari.github.io

Bayesian Workflow book: Website – Bayesian Workflow book

Website for the Bayesian Workflow book by Gelman, Vehtari, McElreath, et al. — case studies, code, and exercises in R and Stan.

17419

Reposted by Shravan Vasishth

languagemit.bsky.social @languagemit.bsky.social · 20/06/2026

—— Open Mind is MIT Press’s Diamond Open Access Cognitive Science journal. We now have our first Impact Factor: 2.9 as of June 2026 Submit your papers to Open Mind direct.mit.edu/opmi

118372

Shravan Vasishth @shravanvasishth.bsky.social · 11/05/2026

I think you are right that this is weird. I have other websites running on github that work without any problems. I'm still trying to figure out what went wrong with this particular website. I'll contact you if I fail to work it out. Thanks for the offer!

010

Shravan Vasishth @shravanvasishth.bsky.social · 11/05/2026

It seems that github forces you to use something called a jekyll theme if you deploy from root, and if you don't use it and have your own css files or whatever, you can go and f**k yourself, slightly paraphrasing github.

100

Shravan Vasishth @shravanvasishth.bsky.social · 11/05/2026

I just learned that one easy fix to preventing github from allowing the user to deploy a web page from the root directory in the main branch. Just move your html page and any subdirectories into the docs directory and deploy the web page from there.

100

Reposted by Shravan Vasishth

Suyog Chandramouli @suyoghc.bsky.social · 06/04/2026

Excited to share that @shravanvasishth.bsky.social will be visiting @princeton.edu virtually tomorrow (Tuesday, April 7) 🗣️ Talk: "Can Context Eliminate Garden-Pathing? A Bayesian Approach" 12:00 PM ET 🛠️Workshop: "Theory-Informed Model Specification" 11:00 AM ET The zoom is open to all!

252

Shravan Vasishth @shravanvasishth.bsky.social · 02/04/2026

I have extended the deadline to 3 April 6PM Central European Time, as there was some confusion about when exactly applications close.

000

Shravan Vasishth @shravanvasishth.bsky.social · 31/03/2026

Tomorrow is the last day to sign up for the tenth summer school on statistical methods for linguistics and psychology (Aug 24-28, 2026, in Potsdam, Germany): vasishth.github.io/smlp2026/

vasishth.github.io

The Tenth Summer School on Statistical Methods for Linguistics and Psychology

073

Shravan Vasishth @shravanvasishth.bsky.social · 26/03/2026

Ever since Microsoft took over github they keep messing up the workflow. I have to go manually switch off their stupid Actions (which they enable by default), which breaks my websites for my data and other resources. WTF?

010

Shravan Vasishth @shravanvasishth.bsky.social · 22/03/2026

I'm betting the author(s) is/are famous. Something similar happened to me as a reviewer with a tabloid journal of great fame (no manipulation of data, just a trivial study that showed nothing that five mins of introspection would not yield). The lead author was too big to fail.

010

Shravan Vasishth @shravanvasishth.bsky.social · 10/03/2026

Our book also made a similar amount (divided over three authors).

020

Shravan Vasishth @shravanvasishth.bsky.social · 07/03/2026

LOL; we ourselves (me included) just forgot to anonymize a response letter. @mattgoldrick.bsky.social

020

Shravan Vasishth @shravanvasishth.bsky.social · 06/03/2026

Yes, I have also done that in the past and nothing happened. Of course, technically it breaks the anonymity, but I am actually unsure about the value of anonymity. Author names is valuable prior information for a reviewer (but it might penalize new people, I have to admit).

020

Shravan Vasishth @shravanvasishth.bsky.social · 05/03/2026

Authors sometimes forget to anonymize the code :). So double-blind doesn't really work.

120

Shravan Vasishth @shravanvasishth.bsky.social · 05/03/2026

I would ask the editor or look at their web page to see what their policy is. That said, usually even with double blind articles that I review, I can quite often easily guess who the authors are.

230

Reposted by Shravan Vasishth

Dario Paape @dariopaape.eurosky.social · 17/02/2026

New open-access paper with @shravanvasishth.bsky.social: "Context ameliorates but does not eliminate garden-pathing: Novel insights from latent-process modeling" doi.org/10.1016/j.jm...

doi.org

Redirecting

041

Shravan Vasishth @shravanvasishth.bsky.social · 25/02/2026

New preprint led by Pia Schoknecht: Local coherence effects are task sensitive: Evidence from event-related potentials in German doi.org/10.31234/osf...

doi.org

OSF

000

Shravan Vasishth @shravanvasishth.bsky.social · 14/02/2026

It is pure torture. That's probably why I felt compelled to become a linguist.

010

Shravan Vasishth @shravanvasishth.bsky.social · 14/02/2026

At best it has two interpretations, which is what is bugging me. The interpretation that jumps out at me and bugs me: Talks can not be short -> It is possible to set up a talk such that it is not short. Of course, the rest of the post makes it clear what the interpretation is, but still...

000

Shravan Vasishth @shravanvasishth.bsky.social · 14/02/2026

The distinction between "cannot" and "can not" completely changes the meaning of the sentence :) It always bugs me when people write can not when they really mean cannot. I want to shout: WHAT IS WRONG WITH YOU? :)

210

Shravan Vasishth @shravanvasishth.bsky.social · 11/02/2026

The content is less than 200 pages. My book, published by the same publisher in 2021, which is a bit longer, costs 28 Euros for the digital edition. Why the big discrepancy in price?

110

Shravan Vasishth @shravanvasishth.bsky.social · 11/02/2026

122 Euros for the digital edition? Really?

100

Reposted by Shravan Vasishth

Dario Paape @dariopaape.eurosky.social · 05/02/2026

New preprint with @tallinzen.bsky.social and @shravanvasishth.bsky.social: We jointly model reading data from four tasks (SPR, BSPR, eye tracking, Maze) with a latent-process mixture, and find that it outperforms LLM surprisal in terms of predictive fit: arxiv.org/abs/2602.04489

071

Reposted by Shravan Vasishth

Shravan Vasishth @shravanvasishth.bsky.social · 14/10/2025

Applications for the summer school on statistical methods for linguistics and psychology (Potsdam, Germany) are now open: vasishth.github.io/smlp2026/

vasishth.github.io

The Tenth Summer School on Statistical Methods for Linguistics and Psychology

141

Shravan Vasishth @shravanvasishth.bsky.social · 27/12/2025

Oral interactive discussion with instructor without any notes or electronic aids. Time-consuming but effective.

100

Shravan Vasishth @shravanvasishth.bsky.social · 04/12/2025

Here is your chance to weigh in on what the psycholinguistic and machine learning community needs regarding preprocessing pipelines, software tools, and data sharing standards for eye-tracking research: An anonymous survey that lasts just 10-15 minutes: www.soscisurvey.de/OpenEye/

soscisurvey.de

Questionnaire | page 1

011

Shravan Vasishth @shravanvasishth.bsky.social · 14/10/2025

Keynote speakers in 2026: Dale Barr and Lisa DeBruine.

000

Shravan Vasishth @shravanvasishth.bsky.social · 14/10/2025

Applications for the summer school on statistical methods for linguistics and psychology (Potsdam, Germany) are now open: vasishth.github.io/smlp2026/

vasishth.github.io

The Tenth Summer School on Statistical Methods for Linguistics and Psychology

141

Shravan Vasishth @shravanvasishth.bsky.social · 10/10/2025

a rare sighting of all three authors of The Book, all in one place:

080

Shravan Vasishth @shravanvasishth.bsky.social · 13/09/2025

New preprint investigating whether lossy-context surprisal can account for the locality and expectation effects found in Russian, Hindi, and Persian reading data: osf.io/preprints/ps...

osf.io

OSF

020

Shravan Vasishth @shravanvasishth.bsky.social · 31/08/2025

We are done with the ninth Statistical Methods for Linguistics and Psychology (SMLP) summer school, Potsdam, Germany. The tenth edition is planned for 24-28 August 2026.

0163

Shravan Vasishth @shravanvasishth.bsky.social · 17/08/2025

A common disease in academia is that for any position X there is an academic with an opposing position not-X. It's just how academics are. One can't just blindly follow one or another person's recommendation, but develop a good understanding of it oneself, and draw one's own conclusions.

110

Shravan Vasishth @shravanvasishth.bsky.social · 17/08/2025

I guess then we should also recommend against using p-values because they are sensitive to the likelihood function assumed? :) Who was it that gave this recommendation? I'm guessing SIngmann but maybe I'm wrong.

100

Shravan Vasishth @shravanvasishth.bsky.social · 17/08/2025

Because this is a frequently asked question in summer schools I teach, I am thinking about adding a video lecture on this in my online materials for the book. Do also read the online chapter from our book on this topic: bruno.nicenboim.me/bayescogsci/...

bruno.nicenboim.me

D Model comparison - Extended | Introduction to Bayesian Data Analysis for Cognitive Science

Introduction to Bayesian data analysis for Cognitive Science.

110

Shravan Vasishth @shravanvasishth.bsky.social · 17/08/2025

Also: Cumming, G. (2014). The new statistics: Why and how. Psychological science, 25(1), 7-29. Kruschke, J. K. (2011). Bayesian assessment of null values via parameter estimation and model comparison. Perspectives on Psychological Science, 6(3), 299-312. epub.ub.uni-muenchen.de/74222/

epub.ub.uni-muenchen.de

Bayesian Decisions using Regions of Practical Equivalence (ROPE): Foundations

110

Shravan Vasishth @shravanvasishth.bsky.social · 17/08/2025

Some other books to read: Royall, R. (2017). Statistical evidence: a likelihood paradigm. Routledge. Spiegelhalter, D. (2024). The art of uncertainty: how to navigate chance, ignorance, risk and luck. Random House.

020

Shravan Vasishth @shravanvasishth.bsky.social · 14/08/2025

Spiegelhalter, D. J., Abrams, K. R., & Myles, J. P. (2004). Bayesian approaches to clinical trials and health-care evaluation. John Wiley & Sons.

100

Shravan Vasishth @shravanvasishth.bsky.social · 14/08/2025

You can use the credible interval to think about whether the pattern is consistent with the predicted effect. Even better if you can compare that interval to a model's a priori predicted interval. Read Spiegelhalter's 2004 book:

110

Shravan Vasishth @shravanvasishth.bsky.social · 14/08/2025

That's what the likelihood ratio test (aka anova) does, you compare likelihoods under the null and non-null to decide whether to reject the null. Bayes factors compare marginal likelihoods but the idea is the same. See our BF chapter.

120

Shravan Vasishth @shravanvasishth.bsky.social · 14/08/2025

E.g., if my data point is 0, the likelihood of mu being 0 (assuming sd=1) is dnorm(0,0,1). The lik. of mu being 10 is dnorm(0,10,1), which is much smaller than that of dnorm(0,0,1). So I'd favor mu=0 over mu=10. If the data point were 10, the lik. of mu=10 is much higher, and I would reject mu=0.

100

Shravan Vasishth @shravanvasishth.bsky.social · 14/08/2025

So with Bayes factors you'd compute the marginal likelihoods under the two models and compare them.

100

Shravan Vasishth @shravanvasishth.bsky.social · 14/08/2025

To make a discovery claim, I would have to set up two models, m0 where the effect is 0, and mfull, where it is not. Then I can talk about the relative evidence assuming a null or a non-null effect. 2/2

100

Shravan Vasishth @shravanvasishth.bsky.social · 14/08/2025

I would say that both credible intervals in my example are *consistent* with the effect being positive, but I would not go so far as to make a discovery claim. 1/2

110