IBN 97-34350 MEL Neurons, the computing cells of the brain, typically possess extensive, highly branched tree-like structures called dendrites onto which tens of thousands of input connections from other neurons are formed. In contrast to the classical conception that the brain's computation power lies primarily in the "neural network", i.e. in the richly interconnected network of nerve cells, evolving understanding over the last 40 years and including modeling work in Dr. Mel's laboratory, among several others, has established that computations carried out within the dendritic trees of individual neurons can contribute in significant ways to the computing functions of neural tissue. One of the important principles to emerge from this modeling work is that different computations may be carried out simultaneously in different "subunits" of the dendritic tree, which could dramatically increase the computational capacities of individual nerve cells. A second principle is that the learning and memory may involve changes not only in the absolute potency of cell-to-cell connectivity but in the spatial arrangements of input connections onto the dendrites of individual neurons as well. Whether parallel subunit computations actually occur within the dendrites of real neurons remains to be determined. However, the design of experiments to answer these questions is complicated by the fact that the measured physiological responses of neurons can be due to either to internal operations of individual cells or the combined effects of networks of cells. In this work, Dr. Mel and his colleagues will model the roles and interaction of single- neuron computations and those carried out by the embedded network of interconnected neurons, in order to guide experimental approaches relating to critical questions of neural processing.