Thread by @LiAnnLeow

tweepaper time! in science.sciencemag.org/content/345/62… @herzfeldd & @reziliusReza showed that we learn a memory of errors that helps recall previously learnt behaviours 1/1

https://twitter.com/jjodx/status/1227267414406180869

We found that this memory of errors drives promotes faster learning in similar tasks.Without error, repeating the ideal action that “solves” the task didn’t help. journals.physiology.org/doi/full/10.11… 2/2

But what kind of errors promote faster future learning? Does the dog learn from from sensory prediction errors (errors predicting its movement outcomes)? Or task errors (i.e., #fail)? 3/3

Here, we manipulated #fail by enforcing #fail (target jumps away from cursor), or enforcing #win (target “jumps” to align with cursor). 4/4

Expts 1&2 show that correcting for #fails were important to encode memories that sped up subsequent learning. Sensory prediction errors alone were not enough for faster future learning. 5/5

What underpins correcting for #fails? We limited preparation time, finding that correcting for #fails didn’t only result from slow but smart strategies, but also fast and inflexible stimulus-response associations. 5/5

thank you reviewers and people who provided feedback on the preprint eg @Katinka3D raphael schween

@Katinka3D also to colleagues and mentors @TimCarroll_UQ aymar de rugy @w_marinovic @eva_reuter @BrendanKeane93 for making work fun!

unroll @threadreaderapp