[JerryVan]

Curt, I think your suggestion is right on the mark: the inputs to the lower (in the sense used in the Hawkins book) edge of the motor cortex are, in fact, the signals that the lower brain generates to activate the muscles. With that detail (which is missing from the book), it becomes clear that the same memory-prediction algorithm is at work in both the sensory cortex and the motor cortex.

Thus, motor commands -- the downward signals in the motor cortex -- are simply predictions of movement. For example, the first time that a baby touches a hot iron on Mom's ironing board, its hand jerks away before it is even aware of the pain. This movement is purely a reflex action, but it leaves a memory behind. A few days later, if the baby again finds itself reaching for the iron, the prediction that the hand will jerk away can fire down the motor cortex before the hand actually touches the iron, causing the hand to pull away in time to avoid being burned. In this instance, the prediction of movement triggers the movement.

Regarding the way in which pleasure and pain drive behavior, I would suggest that the cortex is, in some sense, indifferent to the daily struggles of the lower brain to pursue pleasure and avoid pain. However, the cortex does observe the actions of the lower brain, and, over time, learns to predict its behavior. The invariant that eventually forms at the top of the cortical heirarchy is a prediction that the lower brain will pursue pleasure and avoid pain. This prediction becomes an increasingly self-fulfilling prophecy as the memory-prediction framework operating in the cortex learns to aid and abet the lower brain in achieving its aims. The key point is that there is no need to impose a mechanism from above that somehow uses pleasure and pain to motivate the top of the cortical heirarchy to produce this behavior.

[Lawrence Phillia]

[JerryVan]

Good catch. I was (belatedly) replying to Curt's message, in which he states:

"The neocortex is a device which monitors what the lower brain is doing, and modifies it's behavior as required, though a learning process. So, all the inputs to the lower brain, are passed up to the sensor cortex for analysis, and all the outputs of the lower brain, are passed up to the motor cortex for analysis - just like the sensory side."

But I didn't read his explanation carefully enough. I misinterpreted it to mean that the motor cortex inserts its connections into the intermediate pathways (not the output, as Curt states) in the part of the lower brain that generates motor signals. For this to work, the motor cortex would have to somehow influence the lower brain to cause it to generate the desired motor signals to the muscles. I don't know if this scheme could be made to work in some fashion, but the lack of direct control would be unwieldy, and it would probably slow down response times and make fine motor control impossible.

Biology is more direct, as Hawkins explains on page 46:

"The motor cortex is...hierarchically organized. The lowest area, M1, sends connections to the spinal cord and directly drives muscles."

I'd like to think that I would have eventually discovered my mistake even if you hadn't pointed it out. But...thanks!

That said, I found what Curt had to say about the upward flow of information through the motor cortex very helpful. I noticed a missing (rather, a conspicuously absent) paragraph in the book and was happy to find this discussion in the forum. Namely, on page 120, Hawkins describes the downward flow of information through the motor cortex in a paragraph that ends with:

"...the way the cortex processes downward-flowing sesory predictions is similar to how it processes downward-flowing motor commands."

I expected the very next paragraph to discuss the upward-flowing information, but it wasn't there. (Perhaps it will appear in the next revision.)

[Lawrence Phillia]

Hope i didnt appear to be nitpicking that wasnt my intention, was simply asking a question. My personal research is in the area of cognitive motor control and in particular adaptive CPG's using hardware relaxation oscillators with parameters adjusted and stored using software...a digital-analog hybred , if you will.

Jeff's and Curt's theories and/or postulates look good in a broad or abstract sense (as do all theories). But when it comes down to the actual nut-n-bolts of the "mechanics" its easy to get confused and lost in the details.

On that note, heres a clipping from one of my bookmarks i'd like to post:

"The lateral zone of the cerebellum is involved in the control of independent limb movements, especially rapid, skilled movements. Such movements are initiated by neurons in the frontal association cortex, which control neurons in the primary motor cortex. But although the frontal cortex can plan and initiate movements, it does not contain the neural circuitry needed to calculate the complex, closely timed sequences of muscular contractions that are needed for rapid, skilled movements. That task falls to the lateral zone of the cerebellum.

Both the frontal association cortex and the primary motor cortex send information about intended movements to the lateral zone of the cerebellum via the pontine nucleus. The lateral zone also receives information from the somatosensory system, which informs it about the current position and rate of movement of the limbs-information that is necessary for computing the details of a movement. When the cerebellum receives information that the motor cortex has begun to initiate a movement, it computes the contribution that various muscles will have to make to perform that movement. The results of this computation are sent to the dentate nucleus, one of dentate nucleus the deep cerebellar nuclei. Neurons in the dentate nucleus pass the information on to the ventrolateral thalamus, which projects to the primary motor cortex. The projection from the ventrolateral thalamus to the primary motor cortex enables the cerebellum to modify the ongoing movement that was initiated by the frontal cortex. The lateral zone of the cerebellum also sends efferents to the red nucleus (again, via the dentate nucleus); thus, it helps control independent limb movements through this system as well."

So there appears that a great deal is involved in a motor command. But even this can be questioned from the viewpoint of finer detail, e.g Its believe by some (like yours truely) that the cerebeller cortex is the actual storage center of fine-tuned motor programs. If this is true then there should be a physical path (or cerebeller-cortical loop) directly from the cerebeller cortex to the neocortex bypassing the dentate nuclei. A study from the Salk institute although from a primate seams to confirm this:

Salk Institute for Biological Studies, La Jolla, California 92037, USA.

We used transneuronal transport of neurotropic viruses to examine the topographic organization of circuits linking the cerebellar cortex with the arm area of the primary motor cortex (M1) and with area 46 in dorsolateral prefrontal cortex of monkeys. Retrograde transneuronal transport of the CVS-11 (challenge virus strain 11) strain of rabies virus in cerebello-thalamocortical pathways revealed that the arm area of M1 receives input from Purkinje cells located primarily in lobules IV-VI of the cerebellar cortex. In contrast, transneuronal transport of rabies from area 46 revealed that it receives input from Purkinje cells located primarily in Crus II of the ansiform lobule. Thus, both M1 and area 46 are the targets of output from the cerebellar cortex. However, the output to each area of the cerebral cortex originates from Purkinje cells in different regions of the cerebellar cortex. Anterograde transneuronal transport of the H129 strain of herpes simplex virus type 1 (HSV1) revealed that neurons in the arm area of M1 project via the pons to granule cells primarily in lobules IV-VI, whereas neurons in area 46 project to granule cells primarily in Crus II. Together, the findings from rabies and HSV1 experiments indicate that the regions of the cerebellar cortex that receive input from M1 are the same as those that project to M1. Similarly, the regions of the cerebellar cortex that receive input from area 46 are the same as those that project to area 46. Thus, our observations suggest that multiple closed-loop circuits represent a fundamental architectural feature of cerebrocerebellar interactions.

Hope this somewhat lengthy post ( and clips ) will spark some interest. :>)

[JerryVan]

[Lawrence Phillia]

[eightwings]

[JerryVan]

Lawrence Phillia wrote:

[Lawrence Phillia]

A very good rendering of your proposed ideas and thoughts on cortical column syncrony and sequence generation. IMO more members should adopt this posting technique, makes things alot easier to follow.

Some more thoughts:

In the diagram at the top left (columns A and B), if two pulses leave these points at different times ( no delay blocks for column B ?) im thinking a "gate keeper" pin could be added to give us a 3-input nand gate to provide the timing window you mentioned in your explanation, this could also provide for a "veto" inhibitory input .

I too have bin tackling the problem of sequence timing and interval memory in both digital an analog hardware for quite some time. My present project is in combining the two domains into a digital-analog threshold comparator array...dirty digital if you will, isnt that what neurons do ?

[JerryVan]

Lawrence Phillia said:

[wlorenz65]