Foundations of Neuronal Representations - Brain-Machine Interface Engineering - page 22

of objects [ 2 ], spatial coordinate transformations [ 3 ], and motor hemispheric specializations [ 4-6 ].
Historically though, the theoretical support for these functional divisions in the context of neuronal
organization for computational motor control is largely incomplete. Namely, the challenge is to
determine what is the neuronal substrate of mental functions [ 7 ] and how the system contributes to
the principal organization of elements that are supporting mental processes such as perception and
motor control. Early on, the likes of Santiago Ramón y Cajal and German anatomist Korminian
Brodmann believed that the diversity in cytoarchitecture was sufficient to sustain the complexities
in mental abilities. Using this rationale, Cajal studied the intricate morphology of single neuron us-
ing the Golgi method [ 8 ] 2 and suggested that neurons communicate with each other via specialized
synaptic junctions [ 9 ]. 3 This hypothesis became the basis of the neuron doctrine, which states that
the individual unit of the nervous system is the single neuron. Broadmann later also classified
52 cortical areas based upon detailed analysis of cytoarchitecture through histological analysis.
However, with the advent of new electrophysiological experimental paradigms, it became clear that
the structural differences did not reflect the diversity in functional specialization. In other words,
the anatomically distinguishable layers of the cortex did not seem to map to all of the information-
processing capabilities of neuronal networks for the following reasons. First, the columnar structure
of neural assemblies is similar throughout the cortex [ 10 ]. The comparison of local processing in
each of the cortices revealed similar functions: signal integration and broadcasting.
2.2 CoNNECTIoNISM
The apparent differences of function seem to be consequences of diverse larger scale interconnectiv-
ity and communication, which are implicitly defined in the structure. Hence, structure defines func-
tion. From a global perspective of connectivity, the main pathways among cortical and subcortical
structures are primarily specified by genetics, but the fine details of the connections are influenced
by the interactions of the brain with the body and environment. The fundamental architecture of
the pathways consists of multiple recurrent connections of neurons within a nested hierarchy of
parallel circuits. If one were to attempt to simulate the connectivity of a real neural system, to link
each of the neurons in the brain to each other would require an exceeding large number of con-
nections with impractical volume 4 and energetic requirements. The problem then becomes one of
studying the minimum number of connections needed for neural networks. This problem has faced
neuroscientists for many years [ 11 ], and the general solution was derived by mathematicians Erdos
2 Cassic staining techniques respond to different cells or parts of cells: Nissl, color cell body; Golgi, nerve cells and
fibers; Weigert, myelin, to name a few.
3 The term synapse was coined by Sherrington in 1897.
4 Axons occupy more volume in the brain than the combined contributions of cell bodies and dendrites.