Simulation 3b: Conditioned inhibition — learning to predict the
omission of reward. (a) Empirical results from Tobler et al. (2003), adapted from Figure 3a,c, with
permission from Society for Neuroscience: Journal of Neuroscience, copyright
2003, showing the pattern of phasic dopamine signaling seen after conditioned
inhibition training, for the initially-conditioned CS (A+), the conditioned
inhibitor (X−), and their pairing (AX−) (top panels = single cell
histograms; bottom = population histograms). Note that the small early
activation phase seen for X− in the population histogram was attributed
to associative pairing with the A CS since it was eliminated by A- extinction
training (while the depression component persisted). (b) simulation
results showing qualitatively similar results produced by the PVLV model. For
AX− there are both positive (CeMPos; dashed red
line) and negative (LHbRMTg; speckled blue line)
components driving dopamine signaling (VTAp; solid black
line), but the model does not have the temporal resolution to see these
separately as in the empirical data. (c) empirical results from
Tobler et al. (2003), adapted from
Figure 6a,b, with permission from Society for Neuroscience: Journal of
Neuroscience, copyright 2003, showing the results of a summation test is which
the conditioned inhibitor (X−) is compounded with a different
separately-conditioned CS (C+) (top panel = CX− test, bottom = AX−
test.) (d) simulation results for the summation test showing
qualitatively similar results.