This means that any effects found for surprisal always leave room for the possibility of latent factors driving both the probabilities and the human responses, and do not allow any conclusions about which factors are involved (and why).

So... Now what?

(image by Noémie te Rietmolen)

The power of surprisal stems from the fact that (lexical) surprisal can —and will— parametrically reflect variation stemming from any domain or representational level of language. Why? Because words form patterns for many reasons! Semantics, syntax, frequency... Surprisal does not distinguish.

Surprisal is the ‘everything bagel/nothing burger’ of predictors—it has everything baked in, which is the problem.

Hypothesis: the spell checker killed the variant "langauge"
en.m.wikipedia.org/wiki/Spell_c...

TIL the relative popularity of "langauge" peaks in the mid-80s

I had a fantastic time at #AMLaP2025 in gorgeous Prague! Great feedback on my poster, lovely talks, wonderful reunions, and a phenomenal keynote by @lindadrijvers.bsky.social (so relevant when looking at the photos of my own presentation... 👐 ).

work with Alexis Hervais-Adelman 🌐 link in comment

It's always a pleasure to read the research report from the @mpi-nl.bsky.social - but when your own work is mentioned, it does hit different. Thanks for the mention and, of course, the wonderful years of work! 🧠 💬 @andreaeyleen.bsky.social

Report: www.mpi.nl/sites/defaul...

6/ What did we see? The delta-band response to node count was slower OR started later for high surprisal words 🔴 than for low surprisal words 🔵. The difference was 150 to 190 milliseconds 🤯!

4/ Firstly, by comparing all combinations of our features, we investigated which features created the best model of our delta-band MEG signal. Those turned out to be bottom-up node counts 🔺 and surprisal (GPT2 -scalpmaps- and trigram) 🧮.

2/ We analyzed a naturalistic MEG-dataset: participants were listening to fairytales. Cas Coopmans, Michelle Suijkerbuijk and I manually parsed all sentences in the stories to get beautiful X-bar-theory-compliant tree structures. ✨ The node counts were our syntactic features 🔺