A mouse is released into a pool of water from any starting point. A platform is positioned in a specific location just below the water line. The platform is always at the same location. If platform removed normal animal will spend most of its time in the quadrant where it used to be. This behaviour ceases with damage to the hippocampus. Learning of this sort of spatial map depends on the hippocampus.
Packard and McGaugh (1992)
BG needed, not hippocampus when finding one of 2 cue marked platforms (only one big enough to hold the mouse) in a water maze when the platforms are constantly moving. Can't learn to associate a reward with a specific cue if BG is damaged and will split time equally between the 2 platforms.
Fiorillo, Tobler, and Schultz (2003)
When reward is totally unpredicted by the CS there will be no BG firing for CS and a lot for the reward. As you increase the probability that the CS will predict reward, the CS firing rate increases and the reward firing decreases. DA signal helps us learn about what predicts rewarding outcomes.
Lauwereyns et al. (2002)
Rewarding behaviours are performed faster, and produce greater activity in the caudate nucleus of the basal ganglia than non-rewarding.
Jackson et al. (1995)
SRTT: Only control Ss show the characteristic pattern of SRTT - not PD patients. DA neurons particularly important for learning motor habits - depleted in PD.
Botnivik, Niv, and Barto (2008)
Actor-Critic framework: Actor selects actions based upon a modifiable control policy. Critic determines the difference between what actually happened and what was expected (prediction error) which is used to update action selection policy and the value function. DA is the key neurotransmitter predictor of error.