iiinova — where your digital self keeps evolving

Adam Morgan @adumbmoron.bsky.social · 15/09/2026

Excited to take a look at this! Thanks for the ping

Adam Morgan @adumbmoron.bsky.social · 10/08/2026

Thanks, Fernanda! I do wonder if it might be related to broadly distributed representational systems generally. Lots of work to do to validate the finding and see how general it is!

010

Adam Morgan @adumbmoron.bsky.social · 10/08/2026

A huge thank you to the patients, my co-authors, PI Adeen Flinker, my labmates, and all of the generous reviewers who helped us tremendously throughout the publication process.

000

Adam Morgan @adumbmoron.bsky.social · 10/08/2026

Last, unsupervised clustering on patterns of neural activity & linguistic information revealed 5 distinct networks. Two were defined by activity, while three were characterized by linguistic information but low activity, recapitulating that dissociation.

Clusters of electrodes plotted on the brain. The first three appear to be defined by the type of linguistic information encoded in their electrodes; the last two are defined more by neural activity patterns and have spatial concentrations (prefrontal cortex for the "stimulus" cluster, where activity peaks early, and sensorimotor cortex for the "speech" cluster, where activity peaks at speech onset).

130

Adam Morgan @adumbmoron.bsky.social · 10/08/2026

Spatially, the information-sensitive electrodes showed a hybrid organization: broadly distributed across traditional language regions, but with focal concentrations: structure in IFG/MFG, event semantics in MTG/IPL, and (sub)lexical information in STG/SMC.

120

Adam Morgan @adumbmoron.bsky.social · 10/08/2026

If this generalizes, it has important implications: activity magnitude may be a poor guide to where higher-order language is encoded. But this is a first step. Replication and finer-grained manipulations are needed to isolate structure from correlated task differences.

220

Adam Morgan @adumbmoron.bsky.social · 10/08/2026

We found no evidence for a systematic relationship between activity levels and the presence of information about higher-order language -- neither semantic nor structural differences. By contrast, lower-level (sub)lexical information *did* correlate with neural activity.

Correlations between neural activity (vertical) and three types of linguistic information: structur (active/passive; left); event-semantics (middle), and information at the word level and below (right). Only the activity-word correlation showed a significant relationship.

230

Adam Morgan @adumbmoron.bsky.social · 10/08/2026

This led us to re-examine a widespread assumption: that greater neural activity indexes information processing. Using RSA multiple regression, we separately quantified each electrode’s sensitivity to 3 types of information: event-semantic, structural, and (sub)lexical.

RSA Regression analysis: We used a linear regression to define indices of three types of information in each electrode: structural (active/passive), event-semantic (GPT2 sentence embeddings), and information at the word level and below.

100

Adam Morgan @adumbmoron.bsky.social · 10/08/2026

These contrasts test different things, so some non-overlap is expected. But so little overlap was striking, especially because Sentence-vs-List comparisons are a standard approach to identifying higher-order language regions.

110

Adam Morgan @adumbmoron.bsky.social · 10/08/2026

First, we compared 2 approaches that we expected to identify brain areas involved in higher-order language processing: a Sentences-vs-Lists contrast and an Active-vs-Passive contrast. Surprisingly, among electrodes sensitive to either test, fewer than 5% overlapped.

Left: Electrodes that were sensitive to the sentence-list contrast (orange) and the active-passive contrast (purple). Only 6 electrodes (<5%) were sensitive to both contrasts (those apepar in black).

110

Adam Morgan @adumbmoron.bsky.social · 10/08/2026

New paper in #ScienceAdvances! 10 ECoG patients performed 3 language production tasks: word, word list, and sentence production. We found surprising dissociations between how active cortical sites were and whether they encoded linguistic information: 🧵 [1/9] doi.org/10.1126/scia... @science.org

doi.org

Hybrid spatial organization and evidence for magnitude-independent neural coding of linguistic information during sentence production

Brain recordings during speech reveal complex linguistic information throughout cortex, independent of neural activity levels.

1399

Adam Morgan @adumbmoron.bsky.social · 29/03/2026

Very grateful to have received the Gibson/Fedorenko Award at #HSP2026, and sorry I couldn’t attend this year. HSP feels like my academic home, so I’m especially touched. Thank you to the community, organizers, and donors who make it possible!

0100

Reposted by Adam Morgan

Liina Pylkkänen @liinapy.bsky.social · 12/09/2025

Spectacular talk by SNL Early Career Award winner Esti Blanco Elorrieta! Much NeLLab pride, congratulations Esti! 🎉🎉 #SNL2025 @snlmtg.bsky.social

1479

Adam Morgan @adumbmoron.bsky.social · 31/07/2025

There’s lots more work to be done here, including tinkering with prompts, model parameters, and extending to freely-available LLMs. In the meantime, we hope this is useful to folks and complements existing tools with something new: fast, scalable, and customizable VFF estimation.

100

Adam Morgan @adumbmoron.bsky.social · 31/07/2025

📌 VFFs from Gahl et al. (2004)'s manually annotated (i.e. gold-standard) VFFs 📌 against preferences for competing frames (the dative alternation and NP/SC ambiguity) 🧵6/8

Evaluating the LLM's, benepar's, and the Stanford Parser's VFF estimates by comparison to Gahl et al.'s (2004) database. The LLM produced the best fit, across 7 different verb frames.

100

Adam Morgan @adumbmoron.bsky.social · 31/07/2025

We benchmarked it thoroughly. The LLM consistently outperformed benepar & the Stanford Parser: 📌 300 human-annotated sentences (LLM accuracy = 79%, vs. 69% for benepar and 59% for Stanford) 🧵5/8

Accuracy for the GPT-4o (LLM) parser, Berkeley Neural Parser (benepar), and Stanford Parser on three manually-parsed verbs. The LLM consistently showed higher agreement with manual parses.

100

Adam Morgan @adumbmoron.bsky.social · 31/07/2025

That’s particularly exciting because existing datasets don’t scale well. They’re hard to adapt to new verbs/contexts/languages according to experimental need. Our pipeline is simple, scalable, and adaptable. We release the full code + VFF norms for 476 English verbs. 🧵4/8

110

Adam Morgan @adumbmoron.bsky.social · 31/07/2025

So we got creative and tried asking an LLM to parse a bunch of sentences. As it turns out, not only did this work, but the LLM outperformed both the Stanford Parser and the Berkeley Neural Parser (benepar), a state-of-the-art deep-learning parser trained on treebanks. 🧵3/8

110

Adam Morgan @adumbmoron.bsky.social · 31/07/2025

We needed syntactic norms for an experiment -- specifically Verb Frame Frequencies (VFFs), or how often particular verbs appear in different syntactic frames (e.g., intransitive, prepositional object, etc.). Nothing in the literature quite fit. 🧵2/8

100

Adam Morgan @adumbmoron.bsky.social · 31/07/2025

📊 New Preprint! A Scalable Pipeline for Estimating Verb Frame Frequencies Using Large Language Models. We introduce another unexpected use for LLMs: custom treebanks via automated corpus annotation 🧵1/8 doi.org/10.48550/arX...

doi.org

A Scalable Pipeline for Estimating Verb Frame Frequencies Using Large Language Models

We present an automated pipeline for estimating Verb Frame Frequencies (VFFs), the frequency with which a verb appears in particular syntactic frames. VFFs provide a powerful window into syntax in bot...

141

Adam Morgan @adumbmoron.bsky.social · 05/06/2025

P.S. Yes, we know, Frankenstein wasn't the monster's name. 🤣

260

Adam Morgan @adumbmoron.bsky.social · 05/06/2025

More broadly, the field has largely assumed that the representations we study with single word production tasks are the same as those involved in sentences. By successfully using models trained on picture naming to decode words in sentences, we verify this 🔑 point. 🧵8/9

140

Adam Morgan @adumbmoron.bsky.social · 05/06/2025

These findings show that word processing doesn't always look like it does in picture naming: it depends on task demands. This complexity may even help explain why languages globally prefer placing subjects before objects! 🧵7/9

150

Adam Morgan @adumbmoron.bsky.social · 05/06/2025

We took a closer look at what was going on in prefrontal cortex. This revealed that these sustained representations traced back to different regions depending on a word's sentence position: when it was a subject, it was encoded in IFG, while MFG encoded objects. 🧵6/9

Density plots for the number of detections of subjects (left) and objects (right) during the production of subjects and objects in passive sentences, split by two prefrontal regions: IFG (top) and MFG (bottom). IFG sustained representations of subjects throughout both words while MFG sustained representations of objects.

150

Adam Morgan @adumbmoron.bsky.social · 05/06/2025

In passive sentences like "Frankenstein was hit by Dracula", we observed sustained neural activity encoding BOTH nouns simultaneously throughout the entire utterance. This was particularly true in prefrontal cortex. 🧵5/9

Decoding results from middle frontal gyrus during passive sentences showed sustained encoding of the object noun.

160

Adam Morgan @adumbmoron.bsky.social · 05/06/2025

For straightforward active sentences ("Dracula hit Frankenstein"), the brain activated words sequentially, matching their spoken order. But things changed dramatically for more complex sentences... 🧵4/9

Decoding results from sensorimotor cortex for active sentences: the subject noun is predicted above chance while it is being said, and the object noun while it is being said.

140

Adam Morgan @adumbmoron.bsky.social · 05/06/2025

We trained machine learning classifiers to identify each word's specific neural pattern. 🔑We ONLY used data from picture naming (single word production) to train the models. We then used the models to predict what word patients were saying in real time as they said sentences.🧵3

Word-specific patterns of neural activity: electrodes that selectively responded to each of the six words.

140

Adam Morgan @adumbmoron.bsky.social · 05/06/2025

We recorded brain activity directly from cortex in neurosurgical patients (ECoG) while they used 6 words in two tasks: picture naming ("Dracula") and scene description ("Dracula hit Frankenstein"). 🧵2/9

Task screenshots (picture naming: a cartoon picture of Frankenstein; scene description: cartoon image of Dracula hitting Frankenstein) and mean neural activity per word for one electrode in middle temporal gyrus.

140

Adam Morgan @adumbmoron.bsky.social · 04/06/2025

Super proud of this! Thread to come soon…

1134

Adam Morgan @adumbmoron.bsky.social · 29/03/2025

Wow, thanks Laurel! Honestly one of the best compliments I’ve ever gotten given the quality of the other talks!

110

Adam Morgan @adumbmoron.bsky.social · 29/03/2025

Work with Jenny Yu, Lyn Ögate, Ismael Dono, and Hannah Sarvasy

000

Adam Morgan @adumbmoron.bsky.social · 29/03/2025

For folx at #HSP2025, tune in at 2:15 for our talk on the processing of Switch-Reference Marking in Nungon, a language spoken by ~1000 ppl that requires speakers to inflect the verb not just for features of its subject, but also for the UPCOMING subject! hsp2025.github.io/abstracts/15...

121

Adam Morgan @adumbmoron.bsky.social · 29/03/2025

Also just want to acknowledge how incredibly cool the other talks in this session were - Shota Momma showed evidence for null structure using really clever priming experiments & Ella Bohlman & Jessica Montag showed (that) (unnecessary) adjectives can make production easier by buying speakers time

230

Adam Morgan @adumbmoron.bsky.social · 29/03/2025

Just presented our work using #ECoG to decode words during sentence production at #HSP2025. Really grateful for all the great feedback. I got more clever ideas for future directions than I can possibly follow up on. Love this conference! doi.org/10.1101/2024...

Results of decoding words during the production of active and passive sentences. In actives, nouns were decoded in the order they were said, whereas in passives, prefrontal cortex sustained representations of both the subject and the object throughout the duration of the sentence while sensorimotor areas patterned with actives (showing “congruent” temporal representations).

1154

Adam Morgan @adumbmoron.bsky.social · 28/03/2025

Little late here but this talk at #hsp2025 yesterday was SO neat. Literacy effects disappear when you control for differences in SES. Work by Jessica Vélez Avilés and Paola (Giuli) Dussias hsp2025.github.io/abstracts/26...

080

Reposted by Adam Morgan

Laura Gwilliams @lauragwilliams.bsky.social · 13/12/2024

✨i'm hiring a lab manager, with a start date of ~September 2025! to express interest, please complete this google form: forms.gle/GLyAbuD779Rz... looking for someone to join our multi-disciplinary team, using OPM, EEG, iEEG and computational techniques to study speech and language processing! 🧠

forms.gle

Google Forms: Sign-in

Access Google Forms with a personal Google account or Google Workspace account (for business use).

210264

Adam Morgan @adumbmoron.bsky.social · 11/12/2024

Well thanks! Glad you thought so. Evenin’ from NYC 👋🍉🧠

010

Reposted by Adam Morgan

Cody Cao @neurocow.bsky.social · 26/11/2024

my first preprint with @dbrang.bsky.social is now live. we tested whether mouth movement improves auditory speech onset encoding and ongoing speech envelope tracking with distinct or overlapping mechanisms (1/7)

163

Adam Morgan @adumbmoron.bsky.social · 18/11/2024

It was seasonal when I was still trying to get this out by Halloween :)

010

Reposted by Adam Morgan

Desmond Warren, M.A. @desmondwarren.bsky.social · 17/11/2024

I've created a Black In Psychology Starter Pack to highlight the excellence of Black psychologists, #psychology trainees, psych organizations, & prospective students. Let me know, if you'd like to be added! 👨🏾‍🏫👩🏾‍🏫🧠 Help me spread the word by reposting 🙏🏽 go.bsky.app/KbrZvB8 #BlackSky #PsychSky #SciSky

41295162

Reposted by Adam Morgan

Nome @nome.bsky.social · 18/10/2023

So because I was reminded - the definitive answer to "Is a hot dog a sandwich?" is "by what taxonomy?" Categories matter, after all. Taxonomies exist for a reason. If you're not clear on defining your categories, then your mixed berry salad will consist of:

21482131