Hierarchical Brain

Afferent processing examples

This page contains simplified examples of afferent processing that aim to give a feel for the way that data, particularly sense data, may be processed in the brain using only very basic coincidence detection. To make the examples understandable, I have not used real neurons but model neurons that I call ABCD neurons. The examples should be read in order, because the first example gives more details that are skipped over in later examples.

These examples span all levels of afferent processing in my hierarchical structure of levels of description. The inevitable end result of this processing is large numbers of symbol schemas that represent concepts in the brain, as well as the connections between them that represent connections and relationships between those concepts in the real world, and efferent connections back towards where the data came in that are used by many high-level functions. Together, these symbol schemas and their connections form a model of my world.

Introduction

• Please see diagram information for general information about the diagrams used on this page. In particular, please note that the neurons shown in these examples are not real neurons but simplified model neurons that I call ABCD neurons.
  • ABCD neurons carry out the same coincidence detection on their inputs that real neurons can perform, but with only two inputs and one output. Real neurons can have many hundreds of inputs and outputs and can therefore potentially detect many millions of different coincidences.
    • An ABCD neuron will detect a coincidence of its two connected input ABCD neurons firing (raising an action potential) at the same time, or very nearly the same time, and will fire (raise an action potential) as a result, and pass this signal on to one other ABCD neuron.
    • A real neuron can potentially detect many millions of different coincidences created by the firing of two or more of its inputs at the same time, or very nearly the same time, and will pass on this signal to many other neurons.
  • In the detailed description of ABCD neurons, I have shown how many of them connected together can successfully model the relevant coincidence detection functionality of real neurons.
  • I use these simplified neurons because it would be very difficult to describe, or create diagrams that show, the coincidence detections carried out by real neurons with so many inputs and outputs.
  • The end result is that these examples are more easily understood, but many more ABCD neurons are needed than the number of real neurons in the human brain, by a factor of millions.
• The only functionality needed for this simplified processing is memory-enhanced coincidence detection, which uses the ability of neurons to add up the incoming signals of two or more other neurons, and the memory ability of synapses to strengthen or weaken the connections between neurons depending on their usage.
  • These examples ignore the many more intricate processes that are clearly possible, not only within neurons but also with the functionality provided by inhibitory synapses.
  • These examples therefore do not cover the selection process required for attention.
  • These examples also do not explicitly cover the effects on populations of neurons of neuromodulation, but, as explained on the page about ABCD neurons, these examples can cover the effects by treating them simply as other inputs that just happen to come from a remote place.
• These examples do not assume any pre-existing large-scale structured connections, although many do in fact exist and are created as the brain develops.
  • For example, a new-born baby can recognise a face.
  • There are also many low-level connections created from DNA blueprints that, for example, allow the visual processing areas in the cortex to recognise basic shapes such as lines, curves and edges (see afferent processing for more details on this).

Example 1 - Seeing a red frisbee

• Step 1. Imagine two neurons A and B (see diagram) which have afferent (incoming) connections from two different cone cells in the retina of my eye.
  • If my eye happens to be directed towards the red frisbee, rays of light reflected from two places close together on the frisbee could cause the two cone cells to fire, and these in turn could cause the two neurons A and B to fire at the same time as each other (or extremely close together in time).
  • If A and B fire together, because of what can be considered a “coincidence” in the sense data, another neuron C, which is afferently connected to both of them, may also fire using the functionality of coincidence detection.
  • These connections from A and B to C may strengthen because of usage. This is a low-level form of memory that may later be used for low-level prediction, because any strengthening of a synapse connection means that it may be more likely to be activated and/or to have a stronger effect in the near future.
  • Any connections (in either direction) between A and B will also strengthen, for the same reason (these connections are not shown in the diagram). This is a second low-level memory that may similarly be used for future prediction.
  • If there are efferent connections back from C to A and B (shown by the red arrows in the diagram), those connections will also be strengthened using the functionality of coincidence detection, because C is firing at the same time as A and B (or very close together in time). This is a third low-level area of memory that may also later be used for prediction.
  • A new mini-circuit has been formed, so that in future, if A and/or B fire, then C fires, which causes both A and B to fire, which causes C to fire, and so on. This is also a form of memory that may later be used for prediction, but at a slightly higher level than the previous three.
  
• Step 2. Now imagine another triplet of neurons D, E and F (see diagram) connected in the same way as A, B and C but with afferent connections from different retinal cone cells.
  • At the same time as that portrayed in the previous diagram, light rays reflected from other near-by places on the red frisbee can cause other cone cells to fire, which will then cause sensory neurons D and E to fire, and so also F.
  • If there is another neuron G that has afferent connections from both C and F, and if C and F fire at the same time, connections from C and F to G will strengthen.
  • Efferent connections (the red arrows) from G to C and F will also strengthen, forming mini-circuits between them.
  • Since C and F have fired at the same time, this means that A, B, D and E must also have all fired at the same time, so any efferent connections from G to A, B, D and E will also strengthen, forming larger circuits made up of larger numbers of neurons, which means those circuits or oscillations are more likely to propagate for longer.
  • Neuron G only fires when A, B, D and E all fire, so it is now “representing” in some sense a short segment of the curve of a red object (which could be a frisbee or something else red and round).
  • The strengthened connections and the extra circuits creates in this step can all be regarded as memories that may later be used for prediction.
  • Neuron G could also take part in a higher-level prediction process if the incoming data is not complete.
    • For example, if, in the near future, only B and D fire (because the light to the cells connected to A and E is blocked), the newly-created afferent connections from B and D to G (the black dotted lines) might cause G to fire.
    • This could cause the efferent (red) connections from G back to A and E to be activated and might cause them to fire.
  
• Step 3. The two steps shown in the previous two diagrams can be regarded as the first two of many, part of a recursive and hierarchical pattern recognition and compression system that takes place when sense data is processed.
  • When many cone cells in the retina receive light from the whole of the red frisbee (see diagram), and many hierarchical levels have been built up, neuron X fires, and now “represents” this view of the whole frisbee.
  • Since this is as far as the hierarchical afferent processing can go for this view of the frisbee, neuron X is then considered to be part of a symbol schema that represents the whole concept of a frisbee, or, if this is my first viewing of a frisbee, the start of a new symbol schema. This is represented in a simple animation in this diagram.
  • The many connections that are built up within and across many hierarchical levels become not only a memory of the frisbee in my brain but also a piece of intelligence or knowledge about the frisbee that can later be used for prediction.
    • When I recall this memory (when neuron X fires, in this example), the activation of efferent connections (the red arrows) is the mechanism by which I recall visual aspects of the frisbee; this is called reinstatement.
    • The same connections can be used for simple predictions, either when part of the object I am looking at is obscured or when the incoming sense data is incomplete for other reasons.
    • The circuits that are formed comprising the afferent and efferent connections will mean that oscillations will occur (circuits of propagating signals) whenever neuron X is triggered. This type of circuit is used for higher-level functions such as perception, attention and thought.

Notes for example 1

• This example is only considering the input from one eye. Example 4 below gives an idea of the how two different views, either at different times, or from the two eyes at the same time, are likely to be dealt with.
• This example would only happen in real life if a small child saw a frisbee for the first time and had never seen any similar (e.g. round) object before, which is clearly very unlikely, but this is intended to be a very simplified starting point. Example 6 below gives an idea of the more likely scenario when other round or red or plastic objects have been seen before.
• It is known that many of these types of hierarchical connections do already exist at birth (see coincidence detection), but even if they didn’t they may well form over time, because of plasticity of synapse connections.
• Not all connections that are created will exist for ever; if connections are not used, they may be pruned (see coincidence detection for how this happens).
• This example is about the recognition by the sense of vision of just one object at one angle.
  • In real life there are obviously very many objects to be potentially viewed at many angles, many other things to be detected and processed via other senses and many non-concrete concepts and ideas as well.
  • Some of these are covered in further examples below, but these examples are intended to give an idea of how simple coincidence detection can potentially be at the root of all data processing by the brain.

Example 2 - Throwing a frisbee

• Now let’s look at sense data coming from the arm and wrist as the frisbee is thrown. The processing of this type of data is known as proprioception.
  • The same principles apply as in Example 1 above; if two neurons fire close together in time then the connections will be strengthened, a higher level neuron will fire, and efferent connections (the red dotted lines) will be strengthened, forming circuits.
  • One difference between this and the previous example is that the exact order and timing of the signals are likely to be more important.
  • Another difference with processing of proprioception sense data is that it will arrive via the self symbol schema, which means that any efferent connections will be back via the self symbol schema.
• Assuming the hierarchical processing for this event finishes with neuron G, then this neuron will be part of the symbol schema that represents the whole concept of a frisbee, shown in this animation. The joining of neuron X from example 1 above with this neuron G is shown in example 3 below.
• Obviously, if I am learning to throw a frisbee for the first time, the muscle movements I attempt initially may not have the desired end result, but with practice the results will get better, and the correct movements will be remembered.
  • Old incorrect connections that are no longer used will get pruned and replaced by new correct ones.
• This example shows how a memory of how to perform the correct action is established, because the efferent connections are used, not only for reinstatement and prediction but also for performing actions.

Example 3 - Seeing and throwing a frisbee

• In this example, the frisbee is being seen and thrown at the same time. Because of the “coincidence” of sense data coming in from the two events at the same time, neurons X and Y will fire at a similar time, which will cause a neuron Z, which is afferently linked to both of them, to fire. Any efferent connections back from Z to X and Y, and beyond, will also strengthen, forming circuits of various sizes.
• Neurons X, Y and Z in the diagram will all be part of the symbol schema that represents a frisbee in my brain, as you should see in the animation in this diagram.
• If I (as the owner of this brain) encounter or think about a frisbee in future, this symbol schema will be activated (which means the neurons in it will fire), and reinstatement of the efferent connections will cause me to remember what a frisbee looks like and also the feel of how to throw a frisbee.

Example 4 - Two views of a frisbee at different angles

• This example covers two scenarios:
  1. One eye seeing the frisbee from two different angles within a short period of time.
     • This could happen if, for example, I have never seen a frisbee before, or if I see a red circular object but don’t immediately realise that it is a frisbee.
     • I will then (probably with an unconscious action) adjust my view in order to see it from a different angle, or to see other features of the object, such as its curved edges, which will then confirm that it is a frisbee.
     • How does my brain know that these two views within, say, less than half a second of each other, are probably seeing the same frisbee?
  2. Two eyes seeing the same frisbee at slightly different angles at the same time.
     • This is a normal situation but is the basis of my judgement of depth and distance, and how I believe I see the world in three dimensions.
     • How does my brain know that these two things being seen my two eyes are the same object?
• In this upper part of this diagram, neuron X is created from the sight of the frisbee by one eye, in exactly the same way as in example 1 above.
• In the lower part of this diagram, neuron Y is created from the sight of the frisbee at a slightly different angle, either by the same eye a very short time later (the dashed lines in the top view indicate that the first view is no longer active), or by the other of my two eyes at the same time.
• In the first scenario, despite the first view is no longer active, neuron X can continue to be active for a short period of time after the view is changed, because of the circuit that has been created by the efferent (red arrows) and afferent (black arrows) connections, which means that loops of signals continue to flow between X and neurons nearer the sensory neurons that triggered it. If the second view happens within the time that X is still active, then X and Y will fire at the same time, and the connections to Z will be made or strengthened.
• In the second scenario, X and Y will be active at the same time, and so the connections to Z will be made or strengthened.
• As in the previous example, neurons X, Y and Z in this diagram will all be part of a symbol schema that represents a frisbee in my brain, as you should see in the animation in this diagram.

Notes for example 4

• For how the brain is able to recognise a frisbee at a viewing angle that it has never seen before, see example 6 below.

Example 5 - Connecting to other symbol schemas

• Before progressing to other examples, we need to look at how connections are made between symbol schemas.
• If you look again at example 1 above, the same view of the same frisbee with exactly the same processing, but involving different neurons, will create different symbol schemas for “properties” or “qualities” of the frisbee, such as “red” and “plastic”, if they do not already exist.
• Connections will then be created or strengthened between these symbol schemas representing properties of the frisbee and the symbol schema representing the frisbee. This will happen because of the “coincidence” of neurons in the two separate symbol schemas firing at the same time. Or, from a higher level of description, it could be said that the two symbol schemas are activated at the same time, so links between them, in both directions, will be created or strengthened.
• The result will be a structure as illustrated in this diagram. (As detailed in the page giving diagram information, lines between symbol schemas represent multiple connections in either or both directions.)
• Obviously there will be many other connections that are not shown here.
  • The symbol schema for the concept of “red”, for example, will be connected to many other symbol schemas that represent other red objects.
  • This is an example of what I call an elemental symbol schema, one that is created directly only from input from one sense.

Example 6 - Two views of a frisbee over a longer time

• Example 4 above gave details of how my brain can link two views of a frisbee within a very short space of time. But if I look at the same frisbee at a much later time and at a different angle from before, how does my brain know that these two views, perhaps days apart, are probably seeing the same frisbee?
• This diagram shows a simplified version of this event and one way that the desired result could be obtained, using symbol schemas that represent properties of the frisbee. The major differences between this example and previous examples 1-4 are that X and Y do not fire at the same time as each other, and that the desired result is that Z fires if either X or Y fires.
• When I first see the frisbee, neuron X fires, and is connected to symbol schemas that represent attributes of a frisbee, for example, “red”, “plastic” and “round”. When I see the frisbee again at a later time, from a different angle, neuron Y fires, and this is also connected to the three symbol schemas. Another neuron Z, which is also linked to all three symbol schemas, will fire in either case.
• When X and Z fire at the same time, connections will be made directly between them; similarly when X and Z fire at the same time, connections will be made between them.
• Once again, neurons X, Y and Z are now part of a symbol schema that represents a frisbee.

Notes for example 6

• The connections between symbol schemas are required in this example because of the following two issues that you might have realised when reading the earlier examples:
  1. Neuron X in the above diagrams above would fire when the relevant cells in the retina see any red circular object of the particular size at the particular angle, not just a frisbee.
  2. Neuron X will only fire when the frisbee (or other red circular object) is seen at exactly the right angle, at the right distance (so it is seen as exactly the correct size), and in perfect lighting.
• This is how my brain can recognise a frisbee at an angle that it has never seen a frisbee at before.
  • It can do it because it has experienced seeing other round things at many different angles, and knows that any round thing can look like various oval shapes depending on the viewing angle.
  • This does need a lot of possible connections into the symbol schema for “round”, but bear in mind that the neurons shown in these diagrams are very simple ABCD neurons, and real neurons are capable of detecting many coincidences at once, so in fact the number of real neurons needed is far less than it might seem.
  • These connections to other properties mean that the brain can use an effective equivalent of the duck test - “if it looks round like a frisbee, looks plastic like a frisbee and looks red like a frisbee, then it probably is a frisbee”!

Example 7 - Connecting to the self symbol schema

• As suggested by example 5 and example 6 above, afferent processing does not stop when a symbol schema has been created or updated; exactly the same process of coincidence detection is also responsible for linking different symbol schemas together.
  • This includes all properties or qualities of the symbol schema, such as “red” and “round”, as in the examples above, as well as other connections or relationships such as positions and movement.
• This also applies to the self symbol schema, as shown in this diagram.
  • The coincidence in this case is that neurons in the self symbol schema that receive signals about the position and activity of parts of the body (called proprioception) fire at the same time (or very soon after) neurons connected to the muscles of the hand and wrist.
  • This means that connections will strengthen in both directions, shown by the black and red arrows.
  • This is the way that young babies learn how to move the correct muscles.
    • I don’t mean that a tiny baby will be learning to throw a frisbee, but will just be slowly learning the result of movement commands. This is called motor babbling, and is a process that is known to start even before birth (see babies - action).
    • The word “babbling” is used because of the similarity to the way that babies are known to learn speech, by babbling, but this starts after birth, for obvious reasons.
• Eventually, links all the way from the self symbol schema back to near the sensory neurons will be created or strengthened, as shown by the green arrows.
  • These links will eventually be used for action, so that the brain has detailed control over what the muscles of the body do, and how they do it.
  • These links can be also be used for attention, so that I can be conscious of what I am doing, and for reinstatement, which means that, when I think about a frisbee at a later date, I remember what it is like to throw one.

Example 8 - Starting to create an attention schema

• This diagram is the same as in example 7 above, but with one extra neuron Z neuron shown that happens to have inputs from both the frisbee symbol schema and the self symbol schema.
  • When my attention is on throwing the frisbee, it will notice the coincidence of neurons in the frisbee symbol schema being active at the same time as neurons in the self symbol schema that are involved with attention.
  • Neuron Z is therefore the start of an attention schema, the start of a new symbol schema.
  • Many more neurons will be added to this symbol schema over time when attention is on other angles of view for the frisbee, and for other objects and concepts, and also using other senses.
  • Eventually, this attention symbol schema will become part of the self symbol schema, but if will remain a symbol schema in its own right. It will be a model, a compressed and simplified version of attention.

Example 9 - Hearing a regular pulse

• This example is different from the others in trying to show how sounds are processed by the brain. It is obviously more difficult to illustrate sound, but this is a specific example of how a regular beat may be processed.
• Incoming signals of light in the examples above have several possible properties such as position, brightness and colour each of which are likely to activate a different set of processing circuits; incoming sound signals also have different properties such as pitch, volume and timbre which are also likely to be processed by different circuits; this diagram only shows one path.
• The neuron marked A is activated first by the incoming sound signal of a certain pitch, and this signal is passed along a chain of neurons. If the incoming signals are a regular beat, then at some point, one neuron (marked E in this diagram) will receive a signal at the same time as neuron A receives another one. Neuron X will be able to spot this coincidence, and will be the start of a symbol schema that represents this one note repeating at a certain speed.

---

Go back to where you last came from or return to the top level summary.

These web pages have no adverts and no cookies, no data is collected or processed so there is no privacy policy, and access is not secure. You are free to use any material on these pages provided it is credited to this site and referenced to the correct page.

The views, opinions and statements in these web pages are my own personal opinions. I am an amateur neuroscientist with no formal training, and I have no accounts on any social media platform. Any comments, corrections or criticisms are welcome, although I don’t guarantee to reply - email me at info@hierarchicalbrain.com.

Page last uploaded Tue Feb 20 13:04:29 2024 MST