Two fundamental issues - how does the brain work, and how can we build intelligent machines? - are the leitmotifs of this Handbook. There are many strategies for attacking these questions, depending on what particular aspects of these broad issues we are interested in. One approach lies in the behavioral-neuophysiological domain. The recording of the activity of single cells in the brain of behaving animals provides a tool for directly studying how a particular behavioral pattern is represented and generated. Studies along this line intend to answer the first question: How does the brain work?
In the framework of this approach, the firing of a single neuron or a population of neurons can be correlated with one or several behavioral variables changing in time. The main challenge is to solve a pair of complementary problems: the coding/specification problem and the decoding/implementation problem. The former addresses the question of how the information about a particular behavioral variable is encoded in the neuronal activity being produced (see POPULATION CODES). The latter concerns the neural mechanisms by which encoded variables generate a behavioral pattern unfolding in time.
In the behavioral-neurophysiological domain, the constructive framework for attacking the issue of building "intelligent" machines could be formulated in the context of the decoding problem, namely: How can we design adaptive systems that would transform neuronal signals recorded in the brain of behaving animals into the physiologically appropriate behavioral pattern generated by an artificial machine? (see BRAIN-COMPUTER INTERFACES).
The work reported here summarizes a series of studies based on experimental work and abstract modeling. It exemplifies the successful application of the above-mentioned paradigms to the study of the arm motor system of the monkey and to the design of adaptive systems that transform chronically recorded brain signals into the motor output of artificial actuators. For that purpose, relatively simple but behaviorally meaningful motor actions such as a reaching movement and an exertion of force were chosen. We address the question of how movement variables are encoded in the motor cortex and how this information could be used to drive a simulated actuator that mimics the primate arm