Cognitive Symptoms

Several studies have shown that cognitive reinforcement learning processes are altered in Parkinson's disease, and are particularly sensitive to effects of dopaminergic medication. While off medication, patients are better at learning from the negative outcomes of their decisions than they are at learning from positive outcomes. Medication tends to improve learning from positive decision outcomes, but can actually impair learning from negative outcomes, potentially because this kind of learning depends on transient reductions in striatal dopamine, thought to act as a learning signal in D2-receptor expressing neurons.  Dopaminergic agonists and levodopa are likely to continually stimulates D2 receptors (regardless of whether dopaminergic neurons cease firing or not), and thereby prevent learning. It is critical to further identify the mechanisms underlying these effects, which may be implicated in the tendency for dopaminergic agents to induce pathological gambling and other impulsive behaviors in a subset of PD patients.

The new study by Palminteri et al replicates this basic cross-over interaction, in a computerized task that involved learning from financial gains and losses  There are two key novel aspects to the study. First, the tasks utilized required learning which button to press in response to visual stimuli presented on a monitor for as little as 50ms. At this stimulus presentation duration, participants were not able to visually discriminate the stimuli from each other, and reported not having seen the stimuli. Despite this ostensibly subconscious perception, some learning still occurred, and was sensitive to the valence of outcome and dopaminergic medication status. Indeed, medicated patients only showed any evidence for learning when the condition required learning whether or not to press a key to achieve a financial gain. After overnight withdrawal of their regular dose of levodopa, patients only showed learning in the loss condition.

The second, perhaps more novel, aspect of the study was that exactly the opposite pattern of results was found in a population of Tourette's syndrome (TS) patients. Unmedicated TS patients learned from gains but not losses, whereas those patients treated with D2-receptor antagonists (neuroleptics)  showed the opposite pattern, learning from losses but not gains (similar to unmedicated PD). These striking results are consistent  with the tendency for D2 antagonists or striatal DA depletion to induce progressively increased avoidance behavior in rats (Amtage & Schmidt, 2003, Wiecki et al 2009 for computational analysis), and with synaptic plasticity studies showing that a lack of D2 receptor stimulation is necessary for long term potentiation in striatopallidal cells (Shen et al, 2008, Science).

Moreover, these opposite crossover interaction effects observed in two patient groups as a function of dopamine agonism and antagonism provide stronger evidence that positive and negative outcome learning are both dependent on dopaminergic processes  -- even when such learning may not be consciously accessible.