Abstract
A wealth of evidence supports the notions that (i) the phasic activity of dopamine cells reports a temporally sophisticated
prediction error for the delivery of rewards; that (ii)
targets of these cells, including the amygdala and ventral and
dorsal striatum, are involved in learning about and representing
both appetitive predictions of future rewards and
also the actions that will result in maximizing their delivery;
and that (iii) there is a complex interaction between
this scheme for the control of behaviour, which is associated
with outcome-insensitive habits, and a model-based scheme,
associated with outcome-sensitive or goal-directed actions,
which appears to be realized in prefrontal structures. Despite
an impressively extensive range of experimental findings, our
understanding of the nature and realization of aversive processing
appears to be rather less far advanced. One possibility
is that serotonin might play a role as the sort of aversive opponent
to dopamine that has been mooted and investigated
from a psychological perspective. However, appetitive and
aversive processing are far from being mirror images of each
other psychologically, and indeed dopamine and serotonin,
despite evidence for some mutually inhibitory interactions,
are far from being mirror images of each other neurally. We
will provide a computational examination of aversive processing in general, and opponency in particular.
