Rachel Ryskin

The first publication of the #ERC project ‘LaDy’ is a fact and it’s an important one I think:

We show that word processing and meaning prediction is fundamentally different during social interaction compared to using language individually!
👀 short 🧵/1

psycnet.apa.org/fulltext/202...
#OpenAccess

As our lab started to build encoding 🧠 models, we were trying to figure out best practices in the field. So @neurotaha.bsky.social
built a library to easily compare design choices & model features across datasets!

We hope it will be useful to the community & plan to keep expanding it!
1/

Planning to be incremental: Scene descriptions reveal meaningful clustering in language production How do speakers plan complex descriptions and then execute those plans? In this work, we attempt to answer this question by asking subjects to describ…

New paper: We argue that linearization in language production is a foraging process, with speakers navigating semantic and spatial clusters. Lead author: Karina Tachihara, former UC Davis postdoc, now faculty at UIUC!

www.sciencedirect.com/science/arti...

🚨 Postdoc Opportunity PSA! 🚨

🗓️ UC President’s Postdoctoral Fellowship Program applications are due Nov. 1 (ppfp.ucop.edu/info/)

Open to anyone interested in a postdoc & academic career at a UC campus.

I'm happy to sponsor an applicant if there’s a good fit— please reach out!

Predicting relative intelligibility from inter-talker distances in a perceptual similarity space for speech - Psychonomic Bulletin & Review Researchers have generally assumed that listeners perceive speech compositionally, based on the combined processing of local acoustic–phonetic cues associated with individual linguistic units. Yet, th...

Officially out! "Predicting relative intelligibility from inter-talker distances in a perceptual similarity space for speech" S.E Kim, B. R. Chernyak, @keshet.bsky.social, me, & A. Bradlow link.springer.com/article/10.3.... BONUS: free online similarity calculator so you can join in the fun! 1/

Humans largely learn language through speech. In contrast, most LLMs learn from pre-tokenized text.

In our #Interspeech2025 paper, we introduce AuriStream: a simple, causal model that learns phoneme, word & semantic information from speech.

Poster P6, tomorrow (Aug 19) at 1:30 pm, Foyer 2.2!

New paper with @rjantonello.bsky.social @csinva.bsky.social, Suna Guo, Gavin Mischler, Jianfeng Gao, & Nima Mesgarani: We use LLMs to generate VERY interpretable embeddings where each dimension corresponds to a scientific theory, & then use these embeddings to predict fMRI and ECoG. It WORKS!

LLM finds it FAR easier to distinguish b/w DO & PO constructions when the lexical & info structure of instances conform more closely w/ the respective constructions (left 👇). Where's pure syntax? LLM seems to say "🤷‍♀️" (right) @SRakshit
adele.scholar.princeton.edu/sites/g/file...

If you missed us at #cogsci2025, my lab presented 3 new studies showing how efficient (lossy) compression shapes individual learners, bilinguals, and action abstractions in language, further demonstrating the extraordinary applicability of this principle to human cognition! 🧵

1/n

(1)💡NEW PUBLICATION💡
Word and construction probabilities explain the acceptability of certain long-distance dependency structures

Work with Curtis Chen and Ted Gibson

Link to paper: tedlab.mit.edu/tedlab_websi...

In memory of Curtis Chen.

1/7 If you're at CogSci 2025, I'd love to see you at my talk on Friday 1pm PDT in Nob Hill A! I'll be talking about our work towards an implemented computational model of noisy-channel comprehension (with @postylem.bsky.social, Ted Gibson, and @rplevy.bsky.social).

Adaptation to noisy language input in real time: Evidence from ERPs Author(s): Li, Jiaxuan; Ortega, Alyssa Viviana; Futrell, Richard; Ryskin, Rachel | Abstract: Language comprehension often deviates from the literal meaning of the input, particularly when errors resem...

Adaptation to noisy language input in real time: Evidence from ERPs
escholarship.org/uc/item/8cm7...

The Role of Context Gating in Predictive Sentence Processing Author(s): Gokcen, Yasemin; Noelle, David C.; Ryskin, Rachel | Abstract: Prediction is a core computation in language, as humans use preceding context to implicitly make predictions about the upcoming...

The Role of Context Gating in Predictive Sentence Processing.
escholarship.org/uc/item/21w8...

Language experience and prediction across the lifespan: evidence from diachronic fine-tuning of language models Author(s): Chao, Alton; Cain, Ellis; Ryskin, Rachel | Abstract: Humans predict upcoming language input from context, which depends on prior language experience. This suggests that older adults' predic...

Language experience and prediction across the lifespan: evidence from diachronic fine-tuning of language models.
escholarship.org/uc/item/83b1...

Efficient Audience Design in LLMs Author(s): Ryskin, Rachel; Gawel, Olivia; Tanzer, Owen; Pailo, Viniccius; Kello, Christopher | Abstract: During human communication, speakers balance informativeness and effort by tailoring their lang...

Efficient Audience Design in LLMs
escholarship.org/uc/item/6zm2...

Looking forward to seeing everyone at #CogSci2025 this week! Come check out what we’ve been working on in the LInC Lab, along with our fantastic collaborators!

Paper 🔗 in 🧵👇

Skewed distributions facilitate infants' word segmentation Infants can use statistical patterns to segment continuous speech into words, a crucial task in language acquisition. Experimental studies typically i…

Some happy science news (a small light in times of darkness). New paper out with @luciewolters.bsky.social and Mits Ota: : Skewed distributions facilitate infants’ word segmentation. sciencedirect.com/science/arti...

Thrilled to see this work published — and even more thrilled to have been part of such a great collaborative team!

One key takeaway for me: Webcam eye-tracking w/ jsPsych is awesome for 4-quadrant visual world paradigm studies -- less so for displays w/ smaller ROIs.

New paper w/ @ryskin.bsky.social and Chen Yu: We analyzed parent-child toy play and found that cross-situational learning statistics were present in naturalistic settings!

onlinelibrary.wiley.com/doi/epdf/10....

Post-Doctoral position - Department of Linguistics University of California, Davis is hiring. Apply now!

I'm hiring a postdoc to start this fall! Come work with me? recruit.ucdavis.edu/JPF07123

What are the organizing dimensions of language processing?

We show that voxel responses during comprehension are organized along 2 main axes: processing difficulty & meaning abstractness—revealing an interpretable, topographic representational basis for language processing shared across individuals

A schematic of our method. On the left are shown Bayesian inference (visualized using Bayes’ rule and a portrait of the Reverend Bayes) and neural networks (visualized as a weight matrix). Then, an arrow labeled “meta-learning” combines Bayesian inference and neural networks into a “prior-trained neural network”, described as a neural network that has the priors of a Bayesian model – visualized as the same portrait of Reverend Bayes but made out of numbers. Finally, an arrow labeled “learning” goes from the prior-trained neural network to two examples of what it can learn: formal languages (visualized with a finite-state automaton) and aspects of English syntax (visualized with a parse tree for the sentence “colorless green ideas sleep furiously”).

🤖🧠 Paper out in Nature Communications! 🧠🤖

Bayesian models can learn rapidly. Neural networks can handle messy, naturalistic data. How can we combine these strengths?

Our answer: Use meta-learning to distill Bayesian priors into a neural network!

www.nature.com/articles/s41...

1/n

Unfortunately, the NSF grant that supports our work has been terminated. This is a setback, but our mission has not changed. We will continue to work hard on making cognitive science a more inclusive field. Stay tuned for upcoming events.

Does the mind degrade or become enriched as we grow old? To explain healthy aging effects, the evidence supports enrichment. Indeed, the evidence suggests changes in crystallized (enrichment) and fluid intelligence (slowing) share a common cause. psycnet.apa.org/record/2026-...

title of paper (in text) plus author list

Time course of word recognition for kids at different ages.

Super excited to submit a big sabbatical project this year: "Continuous developmental changes in word
recognition support language learning across early
childhood": osf.io/preprints/ps...

from minicons import scorer from nltk.tokenize import TweetTokenizer lm = scorer.IncrementalLMScorer("gpt2") # your own tokenizer function that returns a list of words # given some sentence input word_tokenizer = TweetTokenizer().tokenize # word scoring lm.word_score_tokenized( ["I was a matron in France", "I was a mat in France"], bos_token=True, # needed for GPT-2/Pythia and NOT needed for others tokenize_function=word_tokenizer, bow_correction=True, # Oh and Schuler correction surprisal=True, base_two=True ) ''' First word = -log_2 P(word | <beginning of text>) [[('I', 6.1522440910339355), ('was', 4.033324718475342), ('a', 4.879510402679443), ('matron', 17.611848831176758), ('in', 2.5804288387298584), ('France', 9.036953926086426)], [('I', 6.1522440910339355), ('was', 4.033324718475342), ('a', 4.879510402679443), ('mat', 19.385351181030273), ('in', 6.76780366897583), ('France', 10.574726104736328)]] '''

another day another minicons update (potentially a significant one for psycholinguists?)

"Word" scoring is now a thing! You just have to supply your own splitting function!

pip install -U minicons for merriment

Figure 1. A schematic depiction of a model-mechanism mapping between a human learning system (left side) and a cognitive model (right side). Candidate model mechanism mappings are pictured as mapping between representations but also can be in terms of input data, architecture, or learning objective.

Figure 2. Data efficiency in human learning. (left) Order of magnitude of LLM vs. human training data, plotted by human age. Ranges are approximated from Frank (2023a). (right) A schematic depiction of evaluation scaling curves for human learners vs. models plotted by training data quantity.

Paper abstract

AI models are fascinating, impressive, and sometimes problematic. But what can they tell us about the human mind?

In a new review paper, @noahdgoodman.bsky.social and I discuss how modern AI can be used for cognitive modeling: osf.io/preprints/ps...

🚨 New Preprint!!

LLMs trained on next-word prediction (NWP) show high alignment with brain recordings. But what drives this alignment—linguistic structure or world knowledge? And how does this alignment evolve during training? Our new paper explores these questions. 👇🧵

Predicting relative intelligibility from inter-talker distances in a perceptual similarity space for speech - Psychonomic Bulletin & Review Researchers have generally assumed that listeners perceive speech compositionally, based on the combined processing of local acoustic–phonetic cues associated with individual linguistic units. Yet, th...

Out now! "Predicting relative intelligibility from inter-talker distances in a perceptual similarity space for speech" in Psychonomic Bulletin+Review. led by awesome postdoc Seung-Eun Kim, with B. R. Chernyak, @keshet.bsky.social and A. Bradlow. Supported by NSF. doi.org/10.3758/s134... 1/

Expectation violations signal goals in novel human communication - Nature Communications In the absence of language, there is a lack of common knowledge necessary for efficient communication. Here, the authors show that people solve this problem by reverting to commonly accepted physical ...

✨Excited to share that our paper on how expectation violations shape novel human communication has just been published in Nature Communications! ✨

📖 read the full paper: www.nature.com/articles/s41....

🧵 Detailed breakdown bellow! 👇