From a discussion on the neurochemistry of consciousness, but some points about the stochastic nature of connectivity in some "higher" level neurons might be interesting to people who think about neural networks and AI.


Date: Wed, 22 Sep 1999 16:44:38 -0400
From: "M. S. AtKisson"
Reply-To: "PSYCHE Discussion Forum (Theoretical emphasis)"

To: [email protected]
Subject: Re: neurochemistry of consciousness

On Mon, 20 Sep 1999, John McCrone wrote:

Valerie, I too feel that once you have started to get some sort of grip on the "electrical circuit story" of how the brain is organised - the easy to see stuff of neuron spikes and neuron pathways - you have to then weave in the neurochemistry. The question is how to fit the neurochemistry into an "information processing" framework that is the standard way of thinking about what the brain does?
Although this is a theoretical list, I would like to point out two areas of molecular/physiological data that could inform this sort of discussion. George Dawson fairly neatly summed up part of the problem -- the plethora of specific transmitter receptor subtypes, the thousands of possible combinations of transmitter and receptor 'pairs' at the synapse. Still, I believe the following is of use to people who like to think about consciousness.

First is the work by C. F. Stevens, which indicates that central synapses are not so digital as one would like to think. The on/off connectivity used in many models is based on the best-studied synapse, the neuromuscular junction. That connection has fail-safes, such that an action potential is guaranteed to evoke transmitter release and subsequent muscle contraction. In contrast, Stevens' work with hippocampal neurons suggests strongly that in central connections, the probability of transmitter release in response to an action potential is about 1 in 3.

There are ways to both increase and decrease this probability. Part of the problem in discussing these mechanisms is that one has to think of excitatory neruotransmitters passing on the electrical signal and also modulatory neurotransmitters which can act on either the sending or receiving neuron.

For example, assume an excitatory set of connections from neuron A, to neuron B, and then to neuron C. Even assuming that one packet of neurotransmitter secreted from A is sufficient to induce an action potential in B, what is the probability of a single action potential in A resulting in a signal transmitted to C, given the probability of an action potential translating to transmitter secretion is 1 in 3 for both A and B? Simple probability says 1 in 9.

If at the junction between A and B there is another connection from neuron Y which enhances the probability of release (say via acetylcholine) to 1 in 2, the chances get better. However, there may also be a connection from neuron Z onto B which decreases the probability of transmission between B and C.

It's a mess.

It's a mess, especially considering that each of the hypothetical neurons A, B and C make significant numbers of excitatory input and output connections with other cells, as well as having modulatory and self-feedback contacts. Also, those modulatory connections at the junctions of A and B or B and C do not apply to connections between A and F or B and D.

Although I tend to think on the "presynaptic release and post synaptic receptor" scale of things, do remember that these processes produce the spikes measured by people like Patricia Goldman-Rakic. This brings me to the second line of data pertinent to this discussion. Goldman-Rakic records from single units in the prefrontal cortex of awake behaving monkeys.

The behavior paradigm in her experiments is that of a delayed response task. The stimulus is presented, then removed, and then at a second cue the animal performs the appropriate response. She has found cells where the firing pattern changes (either increases or decreases, depending on the cell) at specific points in this paradigm -- during the 'remembering stage', during the execution phase, etc. Several neurotransmitter types are involved in producing these correlated spikes, and she has also shown that the subtypes of transmitter receptor play specific and separate roles in shaping the spike pattern.

John McCrone:

Most attempts to tie neurochemistry to consciousness suffer from the fact that people attempt to make the connection at too high a level.
Yes, and too globally. Many of the modulatory neurotransmitters do have 'squishy' rather than point-to-point secretion profiles (norepinephrine from the locus coeruleus terminals in the hippocampus is a prime example), but there is also specificity of connection.

John McCrone:

It could be that the brain uses dopamine simply to turn a set of prescribed connections off and on - so in principle all transmitters are functionally equivalent and dopamine is used just to define a set of master circuits that can control a more general maze of circuitry.
I believe it is more complicated that this. Most people seem to regard that there is a fixed set of circuits in the brain, and for certain systems it is probably true -- the sorting of retinal input is a good example. However, I believe the evidence is growing that there may be at least some 'circuits' which are created only as need arises.

Peter Main's connectionist machine, where "each node is continuously responsive to its weighted inputs" is well worth considering, with the addition that the weight of the inputs is dynamic rather than static. The weights in this case are the probabilities of release, of synaptic transmission, which can be regulated by many different biochemical pathways, pathways initiated by electrical activity or neruotransmitters.

George Dawson:

There is practically no knowledge of the specifics of neuronal pathways (fiber counts, receptor counts, bandwidths, latencies) and arrays in humans that would allow for basic estimates of information flow between anatomic sites of interest.
Even with such information, the estimates are likely to be impossible. Serial electron microscopy was used to reconstruct the connections between and among the few hundred neurons of the tiny worm C. elegans. This labor was in part intended to allow modelers to 'crack' the worm's entire neural profile, giving architecture along with what is known about the neurochemistry, but I don't think it was yet sufficient.

I defend posting this discussion to the theoretical list because I think such details can inform theory, particularly when trying to integrate neurochemistry with behavior. Ketamine was mentioned as disrupting glutamate transmission, but it is crucial to the drug effect that there are only certain types of glutamate receptors affected. These types are NMDA receptors, which in some cells seem to be responsible for "coincidence detection." Mice overexpressing one NMDA subunit are recently touted to be "smart" mice in spatial memory tasks. Heavy recreational ketamine users report strong visual architectural hallucinations. The first two concepts jibe -- increased coincidence detection means faster learning -- but where does the second, the altered state of consciousness, fit in?

John Mikes pulled out the car analogy again, and I have to agree. Locomotion is not in the tires, the engine, the axles or the gas tank, yet without everything in working order, the car doesn't move normally. The moving parts of the car are predictably dynamic, and perhaps the plastic connections in the brain will one day be understood. Still, once can't understand a manual transmission without at least having the concept of the gear in mind. I consider the details of neurotransmission to have at least some utility in informing our attempts to understand the brain.

Peg.

References:

Goldman-Rakic PS. Annals of the New York Academy of Sciences. 868:13-26, 1999 Apr 30.

Allen C. Stevens CF. An evaluation of causes for unreliability of synaptic transmission. Proceedings of the National Academy of Sciences of the United States of America. 91(22):10380-3, 1994 Oct 25.

====
M. S. AtKisson
Department of Neuroscience, Tufts University

"The mental experience works well through the Internet,
but the brain life is blocked off by neurodynamics,
which is God's own firewall preventing philosophers
from accessing brain codes. The way of the hacker is hard."
Walter J. Freeman

Hosted by www.Geocities.ws

1