Information integration and neural plasticity in sensory processing investigated at the levels of single neurons, networks, and perception

In this doctoral thesis, several aspects of information integration
and learning in neural systems are investigated at the levels of
single neurons, networks, and perception. In the first study
presented here, we asked the question of how contextual,
multiplicative interactions can be mediated in single neurons by
the physiological mechanisms available in the brain. Multiplicative
interactions are omnipresent in the nervous system and although a
wealth of possible mechanisms were proposed over the last decades,
the physiological origin of multiplicative interactions in the
brain remains an open question. We investigated permissive gating
as a possible multiplication mechanism. We proposed an
integrate-and-fire model neuron that incorporates a permissive
gating mechanism and investigated the model analytically and
numerically due to its abilities to realize multiplication between
two input streams. The applied gating mechanism realizes
multiplicative interactions of firing rates on a wide range of
parameters and thus provides a feasible model for the realization
of multiplicative interactions on the single neuron level. In the
second study we asked the question of how gaze-invariant
representations of visual space can develop in a self-organizing
network that incorporates the gating model neuron presented in the
first study. To achieve a stable representation of our visual
environment our brain needs to transform the representation of
visual stimuli from a retina-centered coordinate system to a frame
of reference that is independent of changes in gaze direction. In
the network presented here, receptive fields and gain fields
organized in overlayed topographic maps that reflected the
spatio-temporal statistics of the training input stream.
Topographic maps supported a gaze-invariant representation in an
output layer when the network was trained with natural input
statistics. Our results show that gaze-invariant representations of
visual space can be learned in an unsupervised way by a
biologically plausible network based on the spatio-temporal
statistics of visual stimulation and eye position signals under
natural viewing conditions. In the third study we investigated
psychophysically the effect of a three day meditative Zen retreat
on tactile abilities of the finger tips. Here, meditators strongly
altered the statistics of their attentional focus by focussing
sustained attention on their right index finger for hours. Our data
shows that sustained sensory focussing on a particular body part,
here the right index finger, significantly affects tactile acuity
indicating that merely changing the statistics of the attentional
focus without external stimulation or training can improve tactile
acuity. In the view of activity-dependent plasticity that is
outlined in this thesis, the main driving force for development and
alterations of neural representations is nothing more than neural
activity itself. Patterns of neural activity shape our brains
during development and significant changes in the patterns of
neural activity inevitably change mature neural representations. At
the same time, the patterns of neural activity are formed by
environmental sensory inputs as well as by contextual,
multiplicative inputs like gaze-direction or by internally
generated signals like the attentional focus. In this way, our
environments as well as our inner mental states shape our neural
representations and our perception at any time.