A circuit variant of the inverse video memory

Part 2.14. A circuit variant of the inverse video memory

This monograph focuses primarily on the magnocellular striosome system and the parvocellular matrix system of the basal ganglia and the use of the climbing fiber signals emerging from both systems for the ability to store and recognize signals and signal sequences.

The parvo-cellular matrix system enabled the re-imprinting of already learned signals and was intensively involved in the realization of an "inner canvas", perhaps a first stage of consciousness. The question is whether we could extend the system of inner mapping of imagined - but not real - signals to signal sequences.

Would it be possible to play back the signals stored in a cerebellar video memory in such a way that we experience this video sequence as reality? Can we name a circuit that generates "dreams"?

The author believes that this is possible. He does not vouch for the fact that the circuit he has explained is exactly the one that actually causes the phenomena depicted. Other circuits would also be conceivable, which are also functional. But we do not want to name all the circuits for this purpose, but are already satisfied with a circuit example.

We remember the process of re-marking and the video memory. The latter consists of a chain of Purkinj groups in which the video signals were stored frame by frame one after the other.

We assume that each of these Purkinj groups has already completed the described substitution of its striosomal climbing fiber signal by the parvocellular climbing fiber signal of the matrix system. Then the images stored in the Purkinje cells are part of the long-term memory. Each image corresponds to a signal neuron in the responsible secondary cortex. When this is activated, the corresponding image appears before our inner eye. The Purkinje cells of the inverse cerebellum take care of this.

But we need a circuit in which these images, starting with the first one, appear one after the other as a film in front of our inner eye. Each picture for about 1/10 second. How can this be realized?

We remember the re-minting. When a self-signal - here a picture - has been recognized, the responsible Purkinje group excites itself with a climbing fibre signal generated by itself via the matrix system. So far, so good!

We have also described how the positive nuclear neuron sends out its axon, thereby picking up the existing marker of the site of origin, making its way to the thalamus, on to the cortex, to the matrix system (i.e. via the substantia nigra to the striatum, further to the globus pallidus interna, to the globus pallidus externa, to the substantia nigra pars reticulata), to the nucleus ruber, to the olive and from there to the Purkinje group, which had generated the original climbing fibre signal. This signal generated by the Purkinje group returns to the same Purkinje group via the matrix system as a climbing fibre and also docks to the nuclear neurons there. Marker substances enable the necessary pathfinding.

Now the marker-controlled path is not infinitely precise. It is therefore to be expected that the climbing fibre signal coming from the matrix system will not only dock to the nuclear neurons of the Purkinj group, whose replication it is intended to cause. Let us therefore imagine - purely theoretically - that a neighbouring group (or both) would also be contacted by this climbing fibre.

Neurologists have already discovered that one climbing fibre can contact several Purkinj groups.

And now we imagine a phase of signal poverty. The last stronger signal was (visually) a dog. Coincidentally there is a "short film" about the dog in the video memory - because we were (as every day) just walking him.

In the signal-free cortex, the complex cell "dog" is still active, while the other cortex cells are "sleeping".

This "dog cell" activates the corresponding Purkinje group with the first image of the "dog film" via its moss fibre. The inverse cerebellum recognizes this activity and sends the original elementary signals of the first dog image to the thalamus. There the illusion is created that we see a dog (picture 1). The thalamus signals ascend to the cortex, then they are sent via the bridge nuclei to the primary cerebellum. A Purkinje group - the "dog cell" - recognizes its own signal and now fires (even stronger than before). Their parvocellular climbing fibre signal is reproduced by the self-generated climbing fibre excitation.

But: This climbing fibre dedicated to re-marking should now not only re-mark (i.e. excite) its original Purkinj group, but also contact the neighbouring group(s). The marker situation of the neighbouring group(s) is not so different from the original group, so that docking there seems to be possible. Therefore, this neighbouring Purkinje group is now also excited. We want to call this excitation a climbing fiber secondary excitation.

It would be rather audacious to attribute the climbing fibre secondary excitation only to the fact that the marker-controlled growth process of axons is imprecise. A much more plausible reason would be a similarity of the signals. If neighbouring Purkinj groups have common partial signals in their imprinting signal, then at least part of the excitation of the moss fibres originates from the same cortical signal neurons. According to this (unconfirmed) theory, these neurons transmit the local marker combination to the connected moss fibres and Purkinj groups. Such Purkinj groups with partially identical partial signals in the intrinsic signals are called similarity chains. Thus, similarity concatenated Purkinj groups inevitably have similar marker situations, so that a docking of the re-imprinted climbing fiber axon becomes plausible.

If we consider that, according to theorem 2.20, similarity concatenated Purkinj groups are subject to an additional side excitation via common moss fibers, two additional excitations come into play: the side excitation of the intrinsic signal detectors because of the signal similarity and the side excitation via the repopulation because of the marker similarity. The sum of both can already be well above the trigger threshold of the Purkinje group: The neighboring Purkinje group will then recognize its own signal.

Another fact facilitates the development of a new algorithm. If we imagine the excitation of the neurons in the cerebellum realistically, we will see that a first excitation with a constant input has a higher effect than a second excitation. The excitation sensitivity of a neuron decreases with continued excitation. A kind of saturation or dullness occurs. Perhaps the required transmitter becomes scarce because its production cannot keep up. Or the cell makes little pauses in between. Perhaps it simply "tires". Either way. Such excitation characteristics are well known.

Therefore, we assume that the nuclear neurons of the neighbouring group would be excited more strongly by the recurrent post-marking signal than those of the first Purkinj group, which is now somewhat "tired". The output, however, is, as has been shown, independent of whether the Purkinje group recognizes its own signal or is excited by the re-marking axon. The thalamus is immediately informed that the intrinsic signal has been recognized. The receptive neighbour inhibition in cerebellum nuclei - where there are plenty of inhibitory interneurons - weakens the weaker signal to the first dog image, the second dog image prevails in its activity. This is promoted by the additional inhibition of the climbing fibre signal by the active nuclear neuron, so that picture number two prevails.

This signal reaches the cortex via the thalamus. Now the inverse cerebellum becomes active. If a cortex neuron is active whose complex signal is the second image of the canine film, the inverse cerebellum produces exactly the elementary signals of the second canine image in the thalamus. Again we have the perfect illusion: We see the second dog image (of the film) in front of our inner eye.

Now the whole thing starts all over again, this time image 2 was created as an illusion. The replication axon of the Purkinje group to image 2 excites, among other things, the neighbouring image cell to image 3. If it should also excite the neighbouring Purkinje group to image 1, the result there is clearly lower because of "saturation", so that image 3 prevails because of the receptive neighbouring inhibition. So we dream the whole dog movie, which shows us how we walked our dog the day before.

But: It was assumed that there were no signals in the cortex. If an external stimulus - for example, the quiet scratching of our dog Djego at the bedroom door - affects the primary thalamus via our receptors, its activity inhibits the weaker dream signals via the numerous inhibitory interneurons present.

At this point we ask ourselves once again whether a Purkinje group also inhibits its active climbing fibre signal during its repopulation. If it did, the excitation to the first image would already end after a very short time. We estimated the time delay of the parvocellular climbing fiber signal without this inhibition to be 125 milliseconds. The same amount of time would then be required for image 1 to appear in front of our inner eye.

But while the climbing fiber signal during the post-marking process also excites the neighboring group, in case of a possible inhibition the output of the negative nuclear neuron would already be inhibiting on its way to the olive to inhibit exactly this climbing fiber signal. Now the path from the cerebellum nucleus to the olive is relatively short, so that a possible inhibition signal from the cerebellum would reach the olive very quickly. The author estimates this time to be under 5 milliseconds. Thus, the second image of the dog video would reach full activity faster (after 10 milliseconds) (5 ms there, 5 ms back). This would be 100 frames per second. Dreams would therefore be played back much faster if there was strong inhibition of the dog's own re-imprinting climbing fiber signal in the olive. Perhaps the future will bring new insights! In the extremely interesting book "Brain and Behaviour" by Monika Pritzel/Matthias Brand/Hans J. Markowitsch of Spektrum Akademischer Verlag Heidelberg 2009 we read the following regarding this problem on page 471:
(quotation begins)

"But the totality of our biography can also - due to special circumstances - be shifted as a whole into the "now". An example of such a time-lapse phenomenon (Frank & Plötz, 1952; Wagner, 1943) is the fall of a bricklayer who survived a 15-meter fall from a scaffold. He described that within the 2-3 seconds of the fall he saw important events from his life in hundreds of images that seemed to be like a film in chronological order. It was calculated that the bricklayer must have seen about 100 frames per second and the wealth of information was so overwhelming that the possibilities of verbalization could be ruled out."

(end of quote)

We remember that the output neurons of the cerebellum nuclei are single-signalling neurons that are inhibited by the Purkinje cells when they do not recognise their own signal. The one-signal strength is a mean value that is formed from the moss fiber input. It corresponds to the average signal strength. Certainly this single signal is extremely strong when there is a lot of stress. And stress at its maximum is when a bricklayer falls off the scaffolding. And if the single signal strength is extreme, then the possible inhibition of the climbing fibers in the olive is also particularly extreme, which explains the extremely fast image sequence in the above example. The only thing missing is the plausible explanation of the retrieval of the "visual history" by the sudden fear of death.

Since the stimuli acting in the described extreme situation must be assigned to a large extent to the external signals (because the bricklayer had not rehearsed the fall as a "stuntman"), their extreme strength leads to a shift of the working point in the associative matrix of the cerebellum. This shifts the working point of the cerebellum from the question mode with many answers to the storage mode with only a few answers. As a result of the stress, only a few Purkinj groups are activated. Only the most concise "pictures" of the past are recalled. This procedure must have proven itself evolutionarily. Thus, a living being on the run from its predators (due to stress) only retrieves the life-saving images of the possible escape route instead of dealing with thousands of earlier signals. Fast and targeted action makes more sense than helpless searching.

This example shows that psychology and its findings are also of great benefit for the development of a real neural circuit of the brain.

Especially with regard to the oscillation theorem still to come, a phase-shifted, time-delayed inhibition of the post-impression climbing fiber signal, e.g. in the nucleus olivaris inferior, would be useful.

Although the inverse video memory presented here is theoretical, it is quite suitable for thinking about what other neural circuits might exist.

It should be pointed out that the algorithm presented is only applicable to similarity chained visual signal sequences, because a secondary excitation of self-signal detectors only occurs if the neighboring Purkinj groups have stored similar signals. A prerequisite for this are "continuous" video sequences without jumps (e.g. scene changes, image editing, etc.). Similarly important is the temporal coherence of the visual images when storing a digitized time signal. The activation of other dream types is not explained here.

In this monograph, the binary and analogue systems of the brain were not presented. Nevertheless, the associated theory has been completed for a long time and is awaiting publication. The future will show whether the author will be able to publish these parts of his scientific research as well. May the readers decide whether the presented work is rather an improbable hypothesis or whether there is a reference to reality.

Theory of the neuronal circuity of the brain and analytical thinking

Part 2.14. A circuit variant of the inverse video memory