Any differences ~ depression/anxiety?

- We'd thought WM preoccupation with punishment might explain putative RL punishment differences
- Here, otoh, model parameters robustly relate to task performance in expected directions
- Yet, we find strikingly little in differences ~ depr/anx (or trait rum)

But little retention of this into test, captured by low RL learning from from negative PEs

What about learning to avoid punishment?

- In learning phases, pts indeed prefer the neutral vs. punishing option

Building on that work, we modeled RL contributions blunted under high WM

- But big ind diffs here — and a qualitatively different pattern among pt subsets:
-- Higher blunters: worst in low set sizes even at final test
-- Low blunters: at final test same pattern as learning of best at low set sizes

Also a twist on paradoxical pattern found in past studies by Collins and Rac-Lubashevsky — where better learning performance in lower set sizes *reverses* at test

- Midway/when learning resumes, we find an inverted U
- Then learning resumes, more PEs — and no overall set-size pattern at final test

We find the standard pattern of parametrically increasing choice of the correct option ~ set size, but with a twist because performance as expected plummets after the break/test phase

- So we do indeed get more error trials than would otherwise

W/ Joonhwa Kim and @lnccbrown.bsky.social

We designed a variant of the RL-WM task with:
- punishment not just reward as a a possible outcome each trial
- a break and test phase not just end of task, but also midway — hence more errors and better ability to assess punishment-avoidance learning

Now out in JEP: General, "How working memory and reinforcement learning interact when avoiding punishment and pursuing reward concurrently"

psycnet.apa.org/record/2026-...

Preprint with final version: osf.io/preprints/ps...

1/n

I'll be talking about two projects coming soon out of the lab at a CCNP talk next Wednesday, Sept. 10 at 10am EDT (Eastern) — giving a sense of our work and future directions

Join the list here (ccnp.princeton.edu/upcoming-mee...) or shoot me a message for the link.

Looking forward to #RLDM2025!

🌀 I'll be talking about why we get caught in repetitive negative thinking patterns like rumination and worry at the Thursday poster session (poster #21) — come say hi if you're around!

Looking forward to being at APS for the first time in years.

Come say hi on Friday at 4 if you're around — at this symposium chaired with @shirleybwang.bsky.social

Presentations by Ann Haynos, @julianburger.bsky.social, and myself and discussion by @aidangcw.bsky.social

1/2

Looks like a phenomenal issue!

Looking forward to reading through, and thrilled to have been part of it w/ this commentary — on why a key focus in my lab is on deriving powerful treatment principles.

psycnet.apa.org/record/2025-... (for access: translational-lab.com/publications)

We develop a four-stage account of how meta-control — selecting, executing, and learning from mental actions — can go awry and lead to repetitive negative thinking.
(3/3)

What are the shared computational bases of rumination and worry? (2/3)