#EEGManyLabs

Null and Noteworthy—Learning theory validated 20 years later The first published paper from EEGManyLabs’ replication project nullifies a null result that had complicated a famous reinforcement learning theory.

The first published paper to emerge from @eegmanylabs.bsky.social settles a debate 20 years in the making. Read more in this month’s Null and Noteworthy. ‬‬‬‬

By @ldattaro.bsky.social

www.thetransmitter.org/null-and-not...

This is just the first in our #EEGManyLabs series—showing how collaborative EEG science can refine major theories. Watch this space for more. In the meantime, read the full paper for the deep dive: doi.org/10.1016/j.co...
Huge thanks to all labs involved!

One of the best parts? ✅ Minimal heterogeneity. ✅Across different EEG systems & participant samples, the pattern held strong, suggesting we have a robust and generalizable result.

The P300 also wasn’t as simple as “expectancy-only: we found both expectancy and valence effects. This implies that feedback evaluation is spread across multiple stages, rather than being sharply split into “FRN for valence” and “P300 for expectancy.”

The original study had only 17 participants—typical for its time but underpowered (~40% power). Our larger sample detected the small-to-moderate expectancy effect (ηp² = .08—identical to the original!).
🚫 Reminder: Absence of evidence ≠ Evidence of absence!

🚨 Results: The FRN isn’t just about valence! 🚨
It was significantly modulated by both:
✅ Valence (reward vs. no reward)
✅ Expectancy (expected vs. unexpected)
These results align more with Holroyd & Coles’ prediction error theory than Hajcak et al.’s original conclusion.

We put this to the test across 13 labs with 359 participants worldwide—a massive jump from the original n=17! Our goal? 🧐
🔍 Does the FRN really ignore expectancy?
🔍 Is the P300 only about surprises?

A new “two-stage” model proposed:
✅ FRN tracks valence (good vs. bad outcome)
✅ P300 tracks expectancy (surprise factor)
With 600+ citations, this study has shaped how researchers interpret feedback-locked ERPs.

But Hajcak et al. (2005) found something different: They found the FRN only distinguished reward vs. no reward, NOT whether an outcome was expected. 🤯 This challenged Holroyd & Coles’ reinforcement-learning theory and led to a new interpretation of feedback processing.

The original study (Hajcak, Holroyd, Moser, & Simons, 2005) tested a highly influential idea: Holroyd & Coles (2002) reinforcement learning model proposed that the FRN (feedback-related negativity) signals a better/worse-than-expected dopamine-driven prediction error.

🚨Exciting news! We now have the first-ever complete #EEGManyLabs replication. This large-scale multi-site study revisits a key debate in EEG & reinforcement learning. A thread! 🧵👇
📄 Full paper: doi.org/10.1016/j.co...