Neuron
-
Learning models propose a role for both signed and unsigned prediction errors in updating associations between cues and aversive outcomes. In this issue of Neuron, Klavir et al. (2013) show how these errors arise from the interplay between the amygdala and anterior cingulate cortex.
-
The ability to switch flexibly between aversive and neutral behaviors based on predictive cues relies on learning driven by surprise or errors in outcome prediction. Surprise can occur as absolute value of the error (unsigned error) or its direction (signed errors; positive when something unexpected is delivered and negative when something expected is omitted). ⋯ We report that errors exist in different magnitudes and that they differentially develop at millisecond resolution. Our results support a model where unsigned errors first develop in the amygdala during successful learning and then propagate into the dACC, where signed errors develop and are distributed back to the amygdala.