Our findings of uncorrelated activity of neurons in the output stage of the basal ganglia enable us to suggest a novel computational model for information processing in the basal ganglia. Briefly, the model claims that the main axis of the basal ganglia performs dimensionality (redundancy) reduction of the information space spanned by the cortical activity. This information compression is controlled by the modulatory neurotransmitters of the basal ganglia, e.g., the dopaminergic and cholinergic systems. This is why we are now trying to get an insight into the physiology of neurobiological reward systems. Recently, Schultz and co workers have suggested that dopaminergic neurons in the substantia nigra pars compacta (SNc) are feature detectors for the degree to which environmental events are different from predictions. The cholinergic interneurons of the striatum (the tonically active neurons, TANs) show similar responses. Both the dopaminergic and cholinergic systems are affected in Parkinson's disease (the dopamine-acetylcholine balance hypothesis), and prominent learning disabilities have been identified in this disorder.

We recorded from both neuronal populations in monkeys performing a probabilistic instrumental conditioning task. Our results suggest that despite the aforementioned similarities, the messages carried by the two systems are functionally distinct. While the response of the temporally uncorrelated dopamine neurons reflects mismatch between expectation and outcome, the highly correlated acetylcholine neurons respond identically to all behaviorally significant events, regardless of their predictability or rewarding value. We propose that while the acetylcholine reaction illuminates the timing of potentially significant events, the dopaminergic signal acts within this time frame to teach the system in view of those events. The combined effect of these two signals yields an appropriate alteration of the strength of cortico-striatal connections.

In parallel, we carried a detailed study of the basal ganglia activity following dopamine replacement therapy. We have shown that in the MPTP treated monkeys, dopamine replacement restore the abnormal uncorrelated activity of pallidal units. We are now studying the different neuronal behavior (firing rate, pattern and synchronization) following "optimal" dopamine therapy and during levo-dopa induced dyskinesia. In parallel, we are studying the neuronal activity in the substantia nigra pars reticulata (SNr) following involuntary oro-facial movements (tics). We believe that these studies will lead to better understanding of the pathophysiology of these common hyperkinetic disorders. In the near future we hope to test the hypothesis that in terms of the learning related function of the basal ganglia, the parkinsonian (dopamine depleted) state can be characterized as a state in which striatal neurons receive signals that erroneously indicate that predicted rewards failed to occur. Finally, the hypothesis will be studied that (tonic) dopamine replacement therapy improves Parkinsonism in part by normalizing the reward signals of cholinergic striatal cells.