New paper with Olivier Michalon. We ask whether transformers that separate syntax and semantics into parallel streams classify sentences better than single-stream models. They do not. We think this matters for how the autonomy of meaning and grammar should be explained. 1/6 doi.org/10.1080/0952...

Amazing photo on @Science Advances cover today: 12,000 year old clay bead from Jordan showing the fingerprints of the kid who made it. www.science.org/toc/sciadv/1...

Computation is a particular kind of cognitive activity. It does not follow that cognition is entirely made of tiny computations (as cognitivism would make us believe). press.princeton.edu/books/paperb...

Bots have made their way to Prolific experiments. Our lab has stopped online testing of adults entirely now for this reason - we want to know if what we study is real. Probably data collected 2-3 years ago are ok, but moving forward we just can't know. www.pnas.org/doi/10.1073/...

Neurolinguistics in Sweden (NLS) 2026 — abstracts due 20 February! 📝 Conference at Stockholm University, June 11-12 @stockholm-uni.bsky.social Keynotes: Esti Blanco-Elorrieta, Leonardo Fernandino & me — Get your submission in! 🇸🇪🧠🗣️ www.su.se/english/divi...

The Marica De Vincenzi Foundation is inviting applications for its post-doctoral fellowship! The fellowship offers up to 2 years of post-doctoral support abroad for Italian psycholinguists. Spread the word! Deadline is 3/10 - see info here: nam10.safelinks.protection.outlook.com?url=https%3A...

«Nobody (...) has claimed that DeepMind’s AlphaFold is conscious, even though, under the hood, it is rather similar to an LLM. (...) AlphaFold, which predicts protein structure rather than words, just doesn’t pull our psychological strings in the same way.» @anilseth.bsky.social @noemamag.com

New paper out in @complingjournal.bsky.social 📄 With @elliot-murphy.bsky.social we respond to Mandelkern & Linzen on whether LMs’ words refer. Their paper prompted us to develop our own positive (internalist) account of machine reference. 🧵 1/20 doi.org/10.1162/COLI...

«Meaning is encoded in the brain in multiple formats: as stable, long-term representations of word meanings and as flexible, short-term structures that connect those meanings to sentence roles and to what is being referred to in the current context.»

«Into this context—a scientific system already optimized for measurable output, already decades into goal displacement, already reshaping research priorities around metrics rather than problems—arrive large language models. They did not arrive as disruptors. They arrived as intensifiers.»

¡Feliz de llegar a nuevos lectores con esta traducción al español de ‘Neurolinguística’! Gracias a los traductores, ilustradores y editores de Ediciones UC por este excelente trabajo 🧠🗣️📘🇨🇱 @edicionesuc #neurociencia #cerebro #lenguaje lea.uc.cl/neurolinguis...

A congenial essay by Michael Clune, modulo assumptions on AI’s persistence: “The skills that future graduates will most need in the AI era (…) are precisely those that are likely to be eroded by inserting AI into the educational process.” @theatlantic.com www.theatlantic.com/ideas/2025/1...

“Current LMs have limited linguistic common sense (…) the capacity to retrieve and exploit the kind of linguistic and world knowledge that would allow them to reliably make sense of complex, underspecified inputs.” Link to the published paper: aclanthology.org/2025.iwcs-ma...

“Our results demonstrate that poetic reformulation systematically bypasses safety mechanisms across all evaluated models. (...) This pattern suggests that existing alignment procedures are sensitive to surface-form variation and do not generalize effectively across stylistic shifts.” 1/2

If we don’t simply assume generative AI is here to stay, what follows for how we teach and assess students and how we develop our own skills? I jump into an ongoing debate on AI adoption at my university #NTNU #Universitetsavisa 🧐💭🗣️✍️ www.universitetsavisa.no/fremtidsscen...

‘Undead Signs: On the Possibility of a Computational Illusion of Meaning’ by Marvin Tritschler, Philosophy & Technology 🤖💭 #LLM #semantics link.springer.com/article/10.1...

I’m a big fan of Jamie Cohen-Cole’s work on the political history of cognitive science (check it out) and would attend this talk if I could teleport to Basel this afternoon: ‘How Children Became Model Organisms for Scientific Thought’ 👶🧠💭 @unibas.ch dg.philhist.unibas.ch/de/default-p...

Countering AI misuse in these crucial first-year courses requires breaking down larger classes into smaller groups, working in close contact with students, and monitoring their progress continuously, not through occasional tests — dear NTNU, there’s no way to do this while saving on resources.

Our special issue on ‘Parallelism in the Architecture of Language’ is now published! Huge thanks to all contributing authors, reviewers, Eva, Neil, and the topiCS journal team for their excellent work 📚🗣️🧠🤖🙏🏻 onlinelibrary.wiley.com/toc/17568765...

Antibiotic-resistant bacteria. This is now a complex problem, with many contributing factors and no easy solution. But it’s also simple at bottom: we’ve either ignored or failed to act on our understanding of evolution by natural selection #sciam 🦠🧬 www.scientificamerican.com/article/nigh...

“Without a base to anchor the composition of our representations of numbers, we would (...) never have managed to progress beyond the limits of working memory.” ‘A solid base for scaling up: the structure of numeration systems’, eds. A. Bender et al. 🧠🧮 royalsocietypublishing.org/toc/rstb/202...

I agree, physicists won’t like this: when was last time someone proposed a new ‘law of nature’? (Philosophers won’t like it either: we’re no longer supposed to think in terms of natural laws). This is for those puzzled by Darwinian evolution who want to try bring it under more general principles 🧬

It says a lot about the human mind that of all possible problems those mathematicians care about are precisely those that are hard to solve for *both* humans and computers, e.g., those requiring new schemata, not hammers. Fine paper by Dean & Naibo👇#AI doi.org/10.1093/phil...

Looking forward to reading Greg’s new book. Something tells me this is going to be a classic. 📖 (I’ll be wise to focus on the stuff I know less well and learn from it, but I’m curious to see how his views on IFG are evolving, and why IFG is still not part of the brain’s ‘combinatorial network’.) 🗣️👂🧠

🧠 Cognitive scientist Giosuè Baggio provides an entry on meaning for The Open Encyclopedia of Cognitive Science, a multidisciplinary, #openaccess resource to understanding the mind: oecs.mit.edu/pub/ivtprtc8... @oecs-bot.bsky.social #OECS @giosuebaggio.bsky.social

Important fMRI/RSA study by @marcociapparelli.bsky.social et al. Compositional (multiplicative) representations of compounds/phrases in left IFG (BA45), mSTS, ATL; left AG encodes constituents, not their composition, weighing the right element more, vice versa IFG 🧠🧩 academic.oup.com/cercor/artic...

My entry on Meaning in the Open Encyclopedia of Cognitive Science is live.🧠💭 OECS is a growing organism: it was satisfying to reference other entries and let my text be shaped by so much good content. Thanks to the editorial team for making this happen! oecs.mit.edu/pub/ivtprtc8...

Announcing the first (and perhaps only) Multilingual Minds and Machines Meeting! Come join us in Nijmegen, June 22-23, 2026, if you are interested in computational models of human multilingualism: mmmm2026.github.io

“(…) jargon infused with technology industry hype, such as shown in Table 1, does not meaningfully explain. (…) We strive to remain critical of the vocabulary the technology industry coopts and deploys, and to remain respectful of scientific terminology.” zenodo.org/records/1706... 8/🧵

Here we’re triggered by the words ‘Oxford’ and ‘ChatGPT’. But how many universities have already made Copilot available to staff and students through Office/365? Mine has, probably yours too. This double PR stunt signals big tech is eating up education upfront, not just via Microsoft’s trojan horse.

‘Institutional capture’ 🏫🎯

At IWCS soon, Matteo Radaelli shows 17 Norwegian LMs struggle to recover plausible covert events in complement coercion, in spite of positive (but not uniform across LMs) context effects 🤖💭 w/ Emmanuele Chersoni & Alessandro Lenci #compositionality #semantics iwcs2025.github.io/program

كتابي "علم اللغة العصبي" متاح الآن مجاناً باللغة العربية! شكراً للمترجمين الرائعين الذين جعلوا هذا ممكناً. أتمنى أن يصل إلى قراء جدد ويساهم في بحوث علم اللغة العصبي في العالم العربي. #علم_اللغة_العصبي #العربية #ترجمة #Neurolinguistics #Arabic #translation www.hindawi.org/books/905090...

‘Machine language as somniloquy: How the neuroscience of sleep illuminates generative AI’ — New quick paper to appear in THINK, exploring the idea that sleep talk and AI-generated language unfold in cognitive conditions with similar restrictions 🧠🤖😴 lingbuzz.net/lingbuzz/009...

RFK Jr.: “It’s one of the burdens of citizenship to do our own research”. Absolutely, please do your own research: set aside a few years of your life, join a lab, read the literature, run the studies, and become an expert, too. Everyone’s welcome.

“The goal of academic training is not to solve problems as efficiently and quickly as possible, but to develop skills for identifying and dealing with novel problems, which have never been solved before.”

Occhio all’inganno: alcuni al posto del ‘pensiero scientifico dominante’ vorrebbero far passare per buono il ‘pensiero antiscientifico dominante’, non certo il ‘pensiero scientifico non-dominante’. 👁️🤔👩‍🔬

Yes, reinforcement learning is powerful, but what put all that structure in the learning data in the first place? A system with what capacities? What architecture? What evolutionary and developmental histories? That’s where AI-as-engineering ends and Cognitive Science begins!🤖🧠

«In tempi in cui tutti dicono troppo, l’importante non è tanto il dire la cosa giusta, che comunque si perderebbe nell’inondazione di parole, quanto il dirla partendo da premesse e implicando conseguenze che diano alla cosa detta il massimo valore.» Italo Calvino, Palomar 📚💭

AI slop and the destruction of knowledge irisvanrooijcogsci.com/2025/08/12/a...

Language is not a monolith, but a set of connected cognitive capacities: what architecture explains best language structure, processing, evolution etc.? Our intro to ‘Parallelism in the architecture of language’, w/ @neilcohn.bsky.social & Eva Wittenberg is out! 🧠🗣️ doi.org/10.1111/tops...

Research systems can (and often do) survive funding cuts; it’s cuts combined with the belief that research can be done less by people (costly to train and employ) and more by ‘PhD-level’ AIs that could be catastrophic @theatlantic.com www.theatlantic.com/science/arch...

«En 2025, Ediciones UC (...) publicará (...) tres nuevas coediciones de la colección Conocimientos Esenciales del MIT: una sobre el ciclo de vida de los plásticos y su impacto en nuestras vidas y el planeta, otra sobre neurolingüística (...)» 🧠🗣️🇨🇱 www.uc.cl/noticias/los...

«Ma ciò che rende oggi drammaticamente attuale lo studio della scienza antica, e allo stesso tempo spiega la scarsa considerazione in cui è stata tenuta negli ultimi due secoli, è la sua tragica fine.» (Lucio Russo, 1944-2025) #rivoluzionedimenticata www.feltrinellieditore.it/opera/la-riv...

Do language models have semantics? Søgaard says non-trivial structural relations between LM vectors and independently obtained ‘referent spaces’ suffice for (weak) referential semantics: doi.org/10.18653/v1/... But problems (e.g., with indexicals) remain: doi.org/10.48550/arX...

«Vi kan ikke forklare hvorfor og hvordan Centaur gjør det den gjør, og den hjelper oss foreløpig ikke med å forklare hvorfor og hvordan mennesker tenker og handler som de gjør i disse oppgavene.» 🧠🤖 @forskning.bsky.social #KI www.forskning.no/kunstig-inte...

Imagine this at scale, unchecked: how is this not a fundamental threat to science? (Also, this is a perfect example of how irrational and ironic it is for large publishing companies to use “AI” to generate content that is of a much lower quality than the human-created content they already own.)

The paper 👉 ‘Compositionality and Event Retrieval in Complement Coercion: A Study of Language Models in a Low-resource Setting’ #CoNLL2025 #ACL2025NLP aclanthology.org/2025.conll-1...

«As well as to predict empirical data, the aim of cognitive science is to represent the functional architecture of a system and be able to explain its capacities for various cognitive functions. Cognitive theories strive not only to model observable data ... 1/2

In a new paper with @giosuebaggio.bsky.social and Eva Wittenberg, we introduce our special issue of Topics in Cognitive Science about "Parallelism in the Architecture of Language" onlinelibrary.wiley.com/doi/10.1111/...