Biomedical Engineering Reference
In-Depth Information
the loop via specific thalamic nuclei to premotor areas, is thought to participate
in planning the movement or in establishing a motor program that has to be seen
as a set of muscle commands, with its sequence planned prior to the beginning
of the movement without interaction with the periphery.
In the context of our current experimental approaches to motor learning,
there is a commonly accepted concept known as the Brindley-Marr-Albus-Ito
model. This actually comprises four distinct hypotheses, and we will focus on
these only. Brindley (13) was the first to suggest that the cerebellum as a site of
motor learning and a principal agent in the learning of motor skills. Marr (48)
proposed a detailed theory in which the cerebellum learns motor skill (action)
and learns to maintain posture and balance (reflexes). Accordingly, the Purkinje
cell learns to execute elemental movements due to the activity patterns provided
by parallel fiber synapses contacting the Purkinje cell dendrites at a given time.
The instruction arises in the cerebral cortex, and is conveyed to inferior olive
neurons, which in turn activate the corresponding Purkinje cell via the climbing
fibers, and thus determine the contexts given by the parallel fiber activity pat-
tern. After the Purkinje cell has learned to recognize the corresponding context,
the context alone is sufficient to activate the Purkinje cell, causing the next ele-
mental movement. The crucial step in learning is assumed to take place at the
only modifiable parallel fiber/Purkinje cell synapse, due to a facilitation of the
simultaneous firing of climbing fibers and presynaptic parallel fiber activity,
similar to a Hebbian synapse (27). In the case of eyeblink conditioning, Thomp-
son and colleagues (70) proposed that CS-information (tone) reaches the cere-
bellar cortex and nuclei via the mossy fibers, and US-information (air puff) via
the climbing fibers. The cerebellar cortex and the cerebellar nuclei are assumed
to be possible sites of association and plasticity in this type of learning with the
interposed nuclei of particular importance.
Albus' theory (2) is similar to, but extends, that of Marr (48). As Marr, he
assumes the mossy fiber-granule cell-parallel fiber system to be a pattern rec-
ognition data processing system; however, as more like the classical Perceptron
(60), performing a fan-out operation facilitating pattern discrimination and
learning speed. Moreover, the modifiability of synapses is not restricted but in-
cludes synapses of stellate and basket cells and, in contrast to Marr, with the
pattern storage accomplished by weakening synaptic weights. The learning
process is assumed to be associative, as in classical conditioning. The post-
climbing pause (or inactivation pause) in the Purkinje cell is regarded as an un-
conditioned response to the unconditioned stimulus represented by a climbing
fiber activation. The conditioning stimulus is the current parallel fiber activity at
the time of climbing fiber activation. After learning, the conditioning stimulus
alone should be able to evoke an inactivation pause, similar to that evoked pre-
viously by climbing fibers. Such a pause should thus be accomplished by a
weakened parallel fiber contact rather than by a strengthened one (2).
Search WWH ::




Custom Search