Just to echo some of Martin Taylor's comments, the concept of
multiplexing is a broad one and includes the notion of transmitting
several different signals simultaneously (through whatever means).
Neurally, one of the great advantages of temporal pattern
coding of sensory information is that it permits multiplexing of
different types of information within a single spike train. For example,
if one looks at the timing of spikes in a single auditory nerve fiber
responding to a complex stimulus, one will find interspike intervals
corresponding to several different stimulus frequency components. These
are all interleaved with each other, so that one finds many different
intervals in any one segment of the spike train (say over 20-50 ms).
Thus a single response spike train potentially tells the brain about
much more than what is going on in the frequency regions nearest the
fiber's best response frequency -- it simultanously conveys
multidimensional information about "complex" pitch and gross spectral
shape (timbre).
In many other sensory modalities (color vision, taste, smell,
flutter-vibration) different temporal discharge patterns are evoked by
different stimulus qualities. I realize that it is hard to think about
sensory coding as being multidimensional and temporally multiplexed,
when all of the textbooks and standard models assume labelled lines of
one sort or another, but I think it is an issue of considerable
potential relevance for perceptual control theory. If neural signals
are multiplexed via various kinds of temporal pattern and
time-of-arrival codes, then many kinds of seemingly complex perceptual
operations (such as the extraction of the low pitch of complex stimuli)
suddenly become very simple.
The possibility of temporal multiplexing of neural signals
radically changes the topological constraints that are imposed on
neural networks by the assumption of "one neuron, one signal."
In effect, temporally coded, multiplexing neural networks have
additional degrees of freedom that their scalar counterparts do
not have .....qualitatively different types of input are carried via
qualitatively different types of temporal discharge patterns, so
there can be an inherent orthogonality to the various signals
coming from different sensory modalities.
Although I don't think temporal coding alters the basic
functional and explanatory structure of PCT (so in a sense it is
orthogonal to your most basic concerns), it does have
implications for how the percept-action loops are realized neurally,
and it may lead us to coding mechanisms that
drastically simplify (seemingly) complex sensory-motor tasks.
Peter Cariani
eplunix!peter@eddie.mit.edu