We studied reinforcement learning with the Go/NoGo RL task and used the wonderful #hBayesDM #Rstats package from @wooyoungahn (🙏) to estimate latent psychological variables underpinning observed behaviour 2/n
Patients with psychosis (on average) had impairments in overriding Pavlovian conflict, ⬇️learning rate and ⬇️sensitivity to both reward and punishment. Less widespread deficits were observed in ARMS. 3/n
In healthy adolescents and young adults (n=390 after various exclusions), there were no associations with polygenic risk for schizophrenia 4/n
Although it picks up group differences, in the general population (n=735) this task is unrelated to psychopathology. Here we look at a raft of traits to extend previous null general psychopathology findings by @M_Moutoussis + co 5/n https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1006679
In sum, RL deficits in first episode psychosis (less so in those at clinical risk for psychosis), but not predicted by polygenic risk for schizophrenia. The study does not support reinforcement learning (as measured by this task) as a psychosis intermediate phenotype 6/n
🙏 to a big team!, especially participants, + CAMEO colleagues @CPFT_Research, @m_montagnese for leading, Franziska Knolle, Juliet Griffin, @petra_vertes + @MRCcardiff for genomics, @M_Moutoussis for computational advice, @joosthaarsma, #NSPN Uchange led by @edbullmore 7/7
You can follow @gk_murray.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled: