The Brain - a Simple Functional Model

This is the only paper to my knowledge that explains the brain. The model is simple and should be comprehensible. If you have problems to understand, ask. I am in the process to rewrite to make it more comprehensible, so feedback is appreciated. If you have facts that contradict that theory, I am eager to listen. If the logic is wrong, I'll correct. If you do not understand it, the problem is your bias.


The brain has the same elements as a computer, electric signals, wires and logical gates. But there are neither codes nor addresses. The search for was futile. Every brain has a different structure so would need different addresses. They would also need to be changed over time. A structure with addresses is to complicated to be inherited. For addresses are an agreement between sender and receiver that has to be given before addresses can work. Data encoding also has to be agreed before a computer starts to save memory. Then either every creation has to have its own code developed before its starting to work or the code was set in the beginning for all stuff ever to happen or be invented. Neither coding nor addressing can be invented by evolution.
Without addresses you cannot make an ordered set of bits that carries information. So coding as known in computers is not possible. Without addresses you cannot specify a location where information is sent to and where information is fetched from. It is therefore not possible to send two sets of information into two adjacent registers to compare them. Information once derived cannot be moved.
Information in the brain is not compared. It is evaluated through a filtering mechanism build of neural networks that delivers the same result in form a single pointer that a certain match was achieved. Not the pattern itself is stored in the neural network but an instruction working on patterns.
The cortex is a network of neural networks. With feedback connections, exciting and inhibiting connections it holds the memory and is responsible for recognition of sensory input and selection of motor output.
As information cannot be moved, long term and short term memory has to be kept in the same place. Long term memory is based on the topological structure of the neural network. Short term memory is based on signals flowing through that network.
The neural network in the cortex is set into learning mode by a simple signal from the hippocampus. This way the cortex can learn logical and spatial relationships. The basal ganglia form a loop of four elements: frontal cortex, striatum, pallidum and thalamus. This loop works as a flywheel to get the necessary momentum to trigger motor action. The amygdala controls that flywheel at the pallidum to decelerate and at the striatum to accelerate.
The amygdala sends neurotransmitter to certain brain regions to exhibit or inhibit general processing in these areas. Together with chemical signals to the body they trigger emotions.
The cortex cannot learn temporal relations. This is done in the cerebellum, it holds all temporal knowledge and all derived implicite memory. For any motor action and its feedback it sends a signal to the motor cortex to select the next motor action.


To start a materialistic approach to explain the brain there has to be a solid base. From there step by step a solid materialistic model is built. If something non-materialistic is needed to explain the brain, it for sure will be seen when our materialistic model shows a missing link.
All animals have brains. There are exceptions. There are exceptions to everything written in this paper, so this is the disclaimer for all of them, just to keep it concise. The brain comes with a huge selection of sizes from fly, mice, dolphins, men. Brains evolved in millions of years, they are built in every creature anew, they work all the same principles and all are individually different. They adapt to the world the creature is living in, deep sea, arctic, rain forest, farming, downtown, stock exchange, second live. Two requirements for the explanation can be concluded: The explanation has to be simple and it has to be independent of any content processed.

Basic Building Blocks

To dig down to that solid base we put aside everything that has no clear definition. How far we have to go down, we can see in computer science. There are basic elements with precisely defined logical behaviour. These elements can be built with solid-state physics and for all further design and programming it is sufficient to use the logic abstraction and let the engineers handle the mundane stuff as power supply, heat dissipation and interferences.
It is known that the brain works with electric wiring and there are knots called soma and synapses. This is a good starting point for a computer engineer - as I am - and I will postpone chemistry - not my thing - for later.
To make up a model of a computing system, that evaluates patterns of incoming signals and produce an outgoing signal requires wires, signals and logical gates. The brain has them all. It is a hugh mesh of neurons. Every neuron has a thick knot called soma with incoming lines called dendrites and an outbound line called axon. Dendrites and axons can branch of like trees. The ends of axons connect to dendrites or the soma of other neurons with synapses. The synapses pass the electric signal with chemistry.
The signal consists of small spikes of electric charges that travel along the axons and dendrites. The spikes have all about the same strength and length. There is also a lot of chemistry involved, but to explain the logic, it is sufficient to know that it works. And if we really need it, there are many ways to implement.
The first gate we look at is the exciting synapse. It has two tasks. It passes the signal in form of spikes from the axons of one neuron to the dendrites of another one. The other much more important task is, to not let a signal pass the other way. The second gate is the inhibitory synapse, it docks somewhere on a dendrite and weakens any spikes running there. There are many kinds of synapses. Exciting or inhibiting is for now the most important feature.
The next gate is the soma. It sends out a signal when it receives 'enough' incoming signals through all the dendrites of that neuron. The gates in a computer send a signal when they receive a signal on at least one incoming line or when they receive a signal on every incoming line. Which way depends on the type of gate used. In a mesh where the gate has thousands of incoming lines, a threshold for 'enough' is a reasonable implementation. The second task of the soma is to keep the signal at a predefined level. All spikes send get the full length and strength. All elements required for a logical circuit are present.
One spike alone is not sufficient as a signal. To excite the soma to send a signal - and it usually sends a burst with up to 1000 spikes a second - there are a lot of incoming spikes required. Even a burst received on one dendrite is usually insufficient to pass the threshold. Compared with the way the soma works, it can be assumed that a signal is a burst with 'enough' spikes. If enough is below the threshold it is considered noise. The signal could be in encoded in the form of the spikes or in their rhythm but that would require a complex mechanism to encode and decode - and nothing indicates that.

No Addresses

The brain is not a Computer. It is a huge hard wired circuit, not a computer. It does not use programming, addresses, or coding. It can be compared to a pre-computer integrated circuit, individually build and wired for one special purpose.
The most important difference to a computer, the brain does not use addresses. Neurons are neither labeled, named, numbered, ordered. A Neuron has only one output axon, it cannot select a receiver. If more neurons would work together, they would need the same input.
Addresses have not been found.They would be different for every brain as it has a different structure and would also need to be changed over time.
A structure using addresses is too complicated to be inherited. Before addresses can work they have to be agreed between sender and receiver. Such a function cannot be invented by evolution.
Without addresses there is no ordered set of bits that carries information - no byte, no record. Coding as known in computers is not possible. No location can be given to send or fetch information. Information cannot be moved. Information cannot be put side by side and compared. With these limitations we have to accept that information must be gained, stored and used at the very same location.


The cortex is the seat of the explicite memory. It is one huge homogeneous memory array and every memory location has its own processor. Every node and connection has its dedicated purpose. Information is encoded by its location in the circuit and thus cannot be moved. Sensory information are located in the back, motor information on top, objects at the left, relations and locations at the right and higher abstractions in the front.

Cortex Memory Map

Cortex Memory Map

Long term memory is implemented as the structure of the wiring. Short term memory or working memory is implemented by signals running the wiring. Learning is done by rewiring.
The cortex is build as a net of neural networks. All information is in the weights of the neural networks, the strenght of the synapses. Regardless of the origin or level of processing there is no difference how the neural networks gain, store and use information.
Neural Networks search patterns in input set and return output pointers. They do not compare data, do not store input patterns. They process patterns of concurrent input signals. One remembers when the processing yields the same output pointer as last time. The output pointer point to the location of information on the next processing level. There the output pointers form a new input set to be searched for patterns at the next level.
The neural networks may not only evaluate discrete patterns, the kind of signal strong, focussed, diffuse, vanishing, weak, ambiguous, oscillating, alternating may be processed and recognized also.
As the patterns recognized have to be of concurrent signals the cortex stores only logical and spatial context, information labeled explicite memory. The cortex does not store temporal context, so cortical constructs (inventions, dreams, constructions, vehicles and mythical creatures) prefere to hover, fly, roll or rotate instead of sway, pitch or walk.

The net of neural networks forms a semantical net. It is working sequential, parallel, bidirectional, looped, using exciting and inhibiting connections.
Forward connections in the sensory cortex areas implement passive memory. They are used for perception and recognition. Backward connections implement active memory. They are used for reasoning, planning, searching, dreaming. Backward connection are required for pattern completion and to fill the blind spot.
Forward connections in the motor cortex areas compose information for moves ( limb, target, direction, force... ). The backward connections select concurrent moves ( body, arm, hand, feet... ).
Forward and backward connections form feedback loops. This enables to amplify signal contrast, reduce noise, and keep a signal in a flip flop for very short time memory.
The feedback loops will start to oscillate. The observed frequencies indicate the size of loops involved. The observed wave patterns can be irregular driven by external perception. Regular waves indicate a common internal origin when the brain is planning, dreaming or meditating.

As the location and context of a neuron determines the information encoded, different location encodes different information. Every information is stored only once, this implies only in one hemisphere. If moves are perceived the same they are recognized by the very same Mirror Neuron. Parallel processing is standard, but only possible for different information.

Dreaming is a conscious experience in cortex without real world input. It has a common internal origin and produces observable regular waves. Dreaming amplifies connected information, searches for common patterns. It eliminates unconnected information (noise, garbage, speckles). Dreams are perceived as strange. They use associative logic but miss temporal order.

The thalamus distributes preprocessed information to cortex, hippocampus and amygdala. The information is then evaluated on independent paths. A connection is kept only by coincidence.


The hippocampus triggers explicite learning. It receives input from Thalamus and Amygdala and is bidirectionally connected to the cortex to identify the areas for learning. The hippocampus sends an unspecific 'write' signal that sets the cortical neural network from perception into learning mode.
The hippocampus receives its information on a different path from the thalamus, so the information learned ( in the cortex ) and information that causes learning (in the hippocampus) is independent.
Learning is implemented by change of synaptic weights of the cortical neural networks. LTP, NMDA receptors and Hebb synapses are used to establish new connections.
When emotions set the cortex into learning mode, remembering can alter the memory.

Different hippocampus regions implement different kinds of learning. [1]
The CA1 region is necessary to remember at all. The bidirectional connections between cortex and hippocampus project back to the cortex area of origin.
The CA3 is required for pattern completion and to get the complete picture out if incomplete information. It projects back to cortex areas that are on a lower processing level, where the missing components of the (to be stored) information are located.
The dendate gyrus is required to distinguish between two different events, when the same thing has happened to you twice at different times. It does this by projecting to higher cortex areas that provide further context to distinguish the events.

Adult Neurogenesis is limited to specific areas. [2]
New neurons in a neural network would change the context of adjacent old neurons and thus the information they encode. You would forget more than you learn. Information cannot be stripped out of the connections and learning must only gradually change the strenght of the connections.
After brain damage the information is lost anyway. You have to relearn with new neurons.
Initial learning is bottom up only, adult learning is top down too. Recovered people get more pedantic, orderly, tidier, religious.
The olfactory cortex learns new scents with new neurons. The evaluating network stays unaltered. For adults there are no new colors, tone pitches, limbs they could learn.
Learning takes place by rewiring the cortex. After completion the connected counterpart in the hippocampus must forget by complete rewiring with new neurons. Otherwise you could not learn new things.

Brain Control Structures

Control Structures


Emotions adapt the body to peak performance. They distribute limited resources efficiently and sustainable. This includes the brain as a huge consumer of resources.
The Amygdala controls the emotions. It gets fast raw input from the thalamus and slower refined input from most cortex regions. This input is either processed to initiate different emotions or points directly to the memory location of specific emotions in the amygdala.
Output from the amygdala goes to hippocampus to initiate learning, to the striatum in the basal ganglia (nucleus accumbens) to initiate volitional action, and to frontal gyrus cinguli to initiate predefined instinctive reactions.
Further output goes to the brain stem to emit neuromodulators to put the brain into appropriate moods, to hypothalamus and hypophyse for hormonally initiated visceral reaction and to gyrus cingulus for neuronal initiated visceral reaction.
Neuromodulators are released to cortex areas to change the properties of neuronal signal processing. This changes the behaviour and we perceive it as different emotions.

Serotonin is released by the raphe nuclei to unspecifically inhibit large areas. Its primary other use in the body is related to digestion. It indicates the body is fully supplied and no further ingestion or any other action is required. This is a relaxed and positive mood and also perceived as happy.

Noradrenalin is released by the locus ceruleus unspecific to most brain regions. Emitted in tonic mode (on high level) the brain goes into exploration mode to find nourishment in the environment. Every new perception is interesting and distracting as it may be a hint to pursue. Emitted in phasic mode (on low level) some action is already pursued. A new perception has to be much stronger to cause a distraction ( and raise the noradrenalin level ).

Dopamin is released by the substantia nigra to excite the striatum. In difference to noradrenalin it initiates action, whereas noradrenalin excites perception.
Dopamin is emitted when a goal worth to be pursued is identified. It initiates the required execute signal for any performed action. All the information required for an action is supplied by processing information in the cortex. Dopamin is the implementation of intent and volition. It is only indirectly related to reward, pleasure or addiction.

Adrenalin is emitted by the adrenal gland when a threat is encountered. It enables immediate action with all available resources. This condition is sustainable only for a very short time.


The frontal cortex memorizes a specific move in one specific area regardless if perceived, read, thought, planned, dreamed or executed. The basal ganglia make the difference. They provide an unspecific 'execute' flag to the cortex.
A loop connects cortex, striatum, pallidum, thalamus, cortex. The cortex exhibits the striatum, the striatum inhibits the pallidum, the pallidum inhibits the thalamus and the thalamus exhibits the cortex.
Two inhibiting steps in the loop enable to excite and to inhited the execute signal. Exhibiting the striatum (N.A.) the amygdala promotes action, exhibiting the pallidum inhibits action.

Implicite Memory

The cerebellum is the seat of implicite memory. The cerebellum projects to the cortex, not to the spine, nor to the muscles. When the signal comes back to the cortex, it is too late to alter or fine tune a move. Besides the motor cortex gives only a rough motion signal. Fine tuning is done in the spine anyway. The cerebellum can only select the sequentially next move.
The cerebellum stores temporal context, skills, implicite memory, sequences of moves, everything that has a chronological order. The cerebellum is required for walking, sports, bicycling, playing musical instruments, martial arts.
Implicite skills have to be learned by practising. The learning signal is delivered by successful executed moves. As errors trigger emotions explicite learning is not appropriate.
Skills are executed unconscious and can't be explained. Executed consciously (by cortex) movements become unstable and may even fail.
Cerebellum damage must be compensated by cortex and the chronological context is compensated by logical context.


Consciousness was not required here to explain the function of the brain.
It has fast access to a wide range of perceptual information that can't be moved. Abstract concepts, moves or planned actions are only perceived by (expected) input or given names.
My hypothesis is that consciousness is a feature of complex selfsustained oscillating circuits in cortical neural networks of a sufficient complexity and size. The human cortex has the sufficient size and structure.
When no or not enough signals are running consciousness can cease and it comes back when the signal level is higher.
Different senses create different topological structures [3]. Different volatile patterns induce different qualia. Discrete perception is based on discontinous topology (red, green, blue, yellow; slopes in 15 steps).


The brain has a simple mechanism. The complexity comes by size and recursion.
It does not use addresses and this implies that information must be gained, stored and used at the very same brain location and is never moved.
The brain is a hard wired circuit, a net of neural networks.
Neural Networks process patterns of concurrent input signals. One remembers when the processing yields the same output pointer (pointing to the same next neural network) as last time.

Cortex - Explicite Memory, stores logical and spatial information
(objects, categories, moves, relations, episodes, places, actions)
Long Term Memory - topology of wiring
Short Term Memory - signals running that wiring
Cerebellum - Implicite Memory, stores sequential information
(skills, bicycling, martial arts)
Hippocampus - provides unspecific 'write' or ' learn' signal to cortex areas
Basal Ganglia - provides unspecific 'execute' or 'action' signal to motor cortex
Amygdala - switches body and brain into emotions or moods
(learning, dozing, exploring, acting, distractable, committed)
Consciousness is not required to explain brain functions.
It is a feature of complex selfsustained oscillating circuits in cortical neural networks of a sufficient complexity and size. The human cortex has that sufficient size and structure.


[1] Susumu Tonegawa, Symposium "Vision of the Future", 2005, Picower Institute, MIT, (Hippocampus Regions)
[2] Gerd Kempermann, Adult Neurogenesis, 2006, Oxford University Press, New York
[3] Mriganka Sur, The Brain and Mind, MIT, June 13, 2004, (Consciousness)


Werner Seyfried
Hohenheimer Str. 54, 70184 Stuttgart, Germany,
Diplom-Informatiker, Universität Stuttgart
IBM Deutschland GmbH,

The content of this paper was presented as poster at the
Neurex/BCCN Annual Meeting, June 22nd & 23rd, 2007, Freiburg, Germany