The Computational Neurobiology of Reaching and Pointing — A Foundation for Motor Learning by Reza Shadmehr and Steven P. Wise

Over the last three decades, computational neuroscience has become an increasingly important component in neurobiological research. Modern multi-electrode and multi-site recording techniques, brain imaging, brain-machine interfaces, data from chronically implanted electrodes, etc., together with the abundance recordings of behavioral data generate a wealth of information that can hardly be interpreted anymore without computational models of the underlying (usually nonlinear) processes. Computational neuroscience comes in various flavors, depending on which level of detail a nervous system is to be understood, and which particular information-processing problem is in focus. David Marr’s (Marr 1982) seminal work on computational neuroscience for vision made the useful distinction between the levels of theory, algorithms, and implementation in computational neuroscience. Theory addresses in a rather abstract manner how a particular problem can be solved at all, irrespective of whether it needs to be solved by a biological or an artificial system. Algorithms are concerned with how a particular theoretical approach can be put into working algorithmic procedures, and the implementation level addresses the realization of an algorithm in a particular medium, e.g., a computer or the neural architecture of a living being. Of course, these three levels are not mutually independent, e.g., the requirement of implementing a process with biological neurons may demand special algorithms, or even a special theory, but it is often useful to approach computational neuroscience with these three levels in mind in order to avoid getting lost in the overwhelming number of details of a nervous system. In terms of different specializations in computational neuroscience, one can at least distinguish between two areas, low-level computational neuroscience and systems-level computational neuroscience. Low-level computational neuroscience is primarily concerned with models of single neurons, channel dynamics, computational abilities of individual neurons and smaller neural networks, etc. Systems-level