Eintrag

Titel: Geoffrey Schoenbaum, NIDA-IRP
Startdatum: Dez 11 2018
Startzeit: 11:15 am
Endzeit: 12:15 pm
Ort: Max Planck House Lecture Hall
Beschreibung: 


Associative learning is driven by prediction errors that occur in response to unexpected outcomes. Dopamine transients correlate with these errors, at least for rewards, however current interpretations limit these biological signals to transmitting errors in so-called cached or model-free value. This influential hypothesis is supported by much correlative data, but several key predictions have remained untested. One is that learning supported by such errors is content-free, consisting only of value and not including any specific information about the predicted future events. Another is that these errors are not elicited by surprising events unless the scalar value represented by those events is not as expected. In my talk, I will describe experiments we have done to directly test these predictions for VTA dopamine neurons recorded in rats. In each case, the results failed to match the straightforward predictions of this popular account and instead were consistent with a much broader view of the role of this biological system in supporting associative learning in the mammalian brain.

Host: Peter Dayan

More Information


Bild: