RIFs (was Re: Hierarchical Control (was Re: Evangelizing PCT))

[MartinT]

[Martin Taylor 2014.02.25.10.26]

Many years ago, Bill and I had a discussion on Reference Input

Functions (RIFs). The point we were considering was that the
reference value for any elementary control unit (ECU) at level N
depends on the output of many, possibly very many, ECUs at the next
higher level, N+1. The way these outputs combine to generate a
reference value for an ECU is a “Reference Input Function” (RIF), and, as Rick suggested in [From Rick Marken (2014.02.23.1440)] it is
very likely that the form of an RIF varies from level to level, as
do the forms of the Perceptual Input Functions (PIFs). At low
levels, an RIF might well be a weighted summation, while at some
high level it might be the kind of associative memory implied by
B:CP figure 15.3. At other levels of the hierarchy an RIF could well
have completely different form, such as a shift register with
weighted input values, or even (again as Rick suggested) a complete
program as in a computer.
The concept of a Meme is not that of a scalar signal value. It is a
complex structure. If it is relevant to the RIF at all, the meme
would have to be the particular pattern of input values, not any one
of them. However, is it more likely that a meme is a perception
rather than the patterning that generates a reference value?
Martin

···

Warren Mansell (apparently 2014.02.25.07.51) said

    Hi Rick, I was only restricting my reply to the perceptual

hierarchy under its immediate working properties. I recognise
that intrinsic reference values are a separate issue. However,
reorganisation, as I understand it would not lead to the same
kind of ‘variation’ in reference signals in the uppermost system
as the variation that all the (relatively) lower order systems
experience. As reorganisation alters the parameters of the
control unit rather than the reference signal directly, then one
might get a change in the impact of the reference signal on the
outputs that are generated, but not a time varied change in
reference signal that all the lower order systems would
experience.

    I like Kent's focus on attention, his MOL link makes sense,

and I think it would be fascinating if a meme could be
considered as the current uppermost system in attention at any
one time, which could be inferred by viewing quite automatically
and passed on from agent to agent, regardless of the potentially
different lower order systems and signals in those systems used
on the basis of different environments and between individuals.

Warren

[wmansell]

Hi Martin, good point about the convergence of multiple higher level systems within a function to determine the reference values at lower levels. I call have been calling them output functions but this might confuse them with the output in functions of the lowest level. However it is a bit confusing to call them 'input' as they are travelling from inside the organism outwards. Also, I guess this explanation couldn't apply to the currently uppermost system in awareness as it was currently operating in automatic mode - but it could apply within controlled mode to set the memory it uses in automatic mode. It definitely couldn't apply to the uppermost systems latent in the hierarchy as they would not have levels above them. Right?
Warren

···

Sent from my iPhone

On 25 Feb 2014, at 20:12, Martin Taylor <mmt-csg@MMTAYLOR.NET> wrote:

[Martin Taylor 2014.02.25.10.26]

Warren Mansell (apparently 2014.02.25.07.51) said

Hi Rick, I was only restricting my reply to the perceptual hierarchy under its immediate working properties. I recognise that intrinsic reference values are a separate issue. However, reorganisation, as I understand it would not lead to the same kind of 'variation' in reference signals in the uppermost system as the variation that all the (relatively) lower order systems experience. As reorganisation alters the parameters of the control unit rather than the reference signal directly, then one might get a change in the impact of the reference signal on the outputs that are generated, but not a time varied change in reference signal that all the lower order systems would experience.

Many years ago, Bill and I had a discussion on Reference Input Functions (RIFs). The point we were considering was that the reference value for any elementary control unit (ECU) at level N depends on the output of many, possibly very many, ECUs at the next higher level, N+1. The way these outputs combine to generate a reference value for an ECU is a "Reference Input Function" (RIF),

<mime-attachment.jpg>

and, as Rick suggested in [From Rick Marken (2014.02.23.1440)] it is very likely that the form of an RIF varies from level to level, as do the forms of the Perceptual Input Functions (PIFs). At low levels, an RIF might well be a weighted summation, while at some high level it might be the kind of associative memory implied by B:CP figure 15.3. At other levels of the hierarchy an RIF could well have completely different form, such as a shift register with weighted input values, or even (again as Rick suggested) a complete program as in a computer.

I like Kent's focus on attention, his MOL link makes sense, and I think it would be fascinating if a meme could be considered as the current uppermost system in attention at any one time, which could be inferred by viewing quite automatically and passed on from agent to agent, regardless of the potentially different lower order systems and signals in those systems used on the basis of different environments and between individuals.
Warren

The concept of a Meme is not that of a scalar signal value. It is a complex structure. If it is relevant to the RIF at all, the meme would have to be the particular pattern of input values, not any one of them. However, is it more likely that a meme is a perception rather than the patterning that generates a reference value?

Martin

[MartinT]

[Martin Taylor 2014.02.26.09.17]

Hi Martin, good point about the convergence of multiple higher level systems within a function to determine the reference values at lower levels. I call have been calling them output functions but this might confuse them with the output in functions of the lowest level.

I think that to call them output functions would confuse them with the output function of each Elementary Control Unit (ECU). They are input functions for an ECU. Each control loop has two points where variable from outside the loop influence the signal values within the loop -- the disturbance and the reference. Both are inputs to the loop. There are two outputs from the loop, the downward signal emerging from the ECUs Output Function and the upward Perceptual Signal that goes up to higher levels. The "input" and "output" to and from the organism are the only externally visible manifestations of what goes on inside the organism.

  However it is a bit confusing to call them 'input' as they are travelling from inside the organism outwards.

Every connection and function has its own input and output. It would be very confusing to say that the signal source that feeds a wire is an input if the wire faces upward and an output if the wire faced downward. The Perceptual Input Function and the Reference Input Function both accept input from many sources, the perceptual function from lower levels, the RIF from higher levels. I don't think I would be happy calling the inputs to a function "outputs", no matter where the function exists in relation to a larger structure.

Also, I guess this explanation couldn't apply to the currently uppermost system in awareness as it was currently operating in automatic mode - but it could apply within controlled mode to set the memory it uses in automatic mode.

I have no particular opinion about the function of awareness (well, I do, but it's irrelevant here), but if an RIF were such that an input from one particular source could override the effect of other inputs, then if that source were active the reference value would be set by that alone, whereas if it were not, all the other inputs would jointly determine the reference value. A function such as "If (X > 0) then Y else Sum(a*A+b*B+c*C....)" would have that effect. The reference value would be Y if it was switched on by the value of X (if X referes to awareness. the implication is that X>0 means "I am aware of setting this reference value to be Y"), but the reference value would be the weighted sum of next-level-up output values if Y was switched off (X<0 means "I don't care what that reference value is").

  It definitely couldn't apply to the uppermost systems latent in the hierarchy as they would not have levels above them. Right?

If we imagine a maturing system that is continually generating new ECUs and ever-higher levels of control, part of its maturation would necessarily be the development of new perceptual functions and RIFs. One could imagine a currently top-level ECU with a developing RIF that initially had no inputs, but as a new level began to develop above it, it would gain inputs. However, I suspect that the RIF would be more likely to develop and be reorganized in parallel with the new next-level connections.

Martin

···

On 2014/02/26 2:59 AM, Warren Mansell wrote:

Warren

Sent from my iPhone

On 25 Feb 2014, at 20:12, Martin Taylor <mmt-csg@MMTAYLOR.NET> wrote:

[Martin Taylor 2014.02.25.10.26]

Warren Mansell (apparently 2014.02.25.07.51) said

Hi Rick, I was only restricting my reply to the perceptual hierarchy under its immediate working properties. I recognise that intrinsic reference values are a separate issue. However, reorganisation, as I understand it would not lead to the same kind of 'variation' in reference signals in the uppermost system as the variation that all the (relatively) lower order systems experience. As reorganisation alters the parameters of the control unit rather than the reference signal directly, then one might get a change in the impact of the reference signal on the outputs that are generated, but not a time varied change in reference signal that all the lower order systems would experience.

Many years ago, Bill and I had a discussion on Reference Input Functions (RIFs). The point we were considering was that the reference value for any elementary control unit (ECU) at level N depends on the output of many, possibly very many, ECUs at the next higher level, N+1. The way these outputs combine to generate a reference value for an ECU is a "Reference Input Function" (RIF),

<mime-attachment.jpg>

and, as Rick suggested in [From Rick Marken (2014.02.23.1440)] it is very likely that the form of an RIF varies from level to level, as do the forms of the Perceptual Input Functions (PIFs). At low levels, an RIF might well be a weighted summation, while at some high level it might be the kind of associative memory implied by B:CP figure 15.3. At other levels of the hierarchy an RIF could well have completely different form, such as a shift register with weighted input values, or even (again as Rick suggested) a complete program as in a computer.

I like Kent's focus on attention, his MOL link makes sense, and I think it would be fascinating if a meme could be considered as the current uppermost system in attention at any one time, which could be inferred by viewing quite automatically and passed on from agent to agent, regardless of the potentially different lower order systems and signals in those systems used on the basis of different environments and between individuals.
Warren

The concept of a Meme is not that of a scalar signal value. It is a complex structure. If it is relevant to the RIF at all, the meme would have to be the particular pattern of input values, not any one of them. However, is it more likely that a meme is a perception rather than the patterning that generates a reference value?

Martin

[wmansell]

Thanks Martin, I take your point and you have expanded my understanding of a PCT hierarchy somewhat!
Warren

···

On Wed, Feb 26, 2014 at 3:13 PM, Martin Taylor mmt-csg@mmtaylor.net wrote:

[Martin Taylor 2014.02.26.09.17]

On 2014/02/26 2:59 AM, Warren Mansell wrote:

Hi Martin, good point about the convergence of multiple higher level systems within a function to determine the reference values at lower levels. I call have been calling them output functions but this might confuse them with the output in functions of the lowest level.

I think that to call them output functions would confuse them with the output function of each Elementary Control Unit (ECU). They are input functions for an ECU. Each control loop has two points where variable from outside the loop influence the signal values within the loop – the disturbance and the reference. Both are inputs to the loop. There are two outputs from the loop, the downward signal emerging from the ECUs Output Function and the upward Perceptual Signal that goes up to higher levels. The “input” and “output” to and from the organism are the only externally visible manifestations of what goes on inside the organism.

However it is a bit confusing to call them ‘input’ as they are travelling from inside the organism outwards.

Every connection and function has its own input and output. It would be very confusing to say that the signal source that feeds a wire is an input if the wire faces upward and an output if the wire faced downward. The Perceptual Input Function and the Reference Input Function both accept input from many sources, the perceptual function from lower levels, the RIF from higher levels. I don’t think I would be happy calling the inputs to a function “outputs”, no matter where the function exists in relation to a larger structure.

Also, I guess this explanation couldn’t apply to the currently uppermost system in awareness as it was currently operating in automatic mode - but it could apply within controlled mode to set the memory it uses in automatic mode.

I have no particular opinion about the function of awareness (well, I do, but it’s irrelevant here), but if an RIF were such that an input from one particular source could override the effect of other inputs, then if that source were active the reference value would be set by that alone, whereas if it were not, all the other inputs would jointly determine the reference value. A function such as “If (X > 0) then Y else Sum(aA+bB+c*C…)” would have that effect. The reference value would be Y if it was switched on by the value of X (if X referes to awareness. the implication is that X>0 means “I am aware of setting this reference value to be Y”), but the reference value would be the weighted sum of next-level-up output values if Y was switched off (X<0 means “I don’t care what that reference value is”).

It definitely couldn’t apply to the uppermost systems latent in the hierarchy as they would not have levels above them. Right?

If we imagine a maturing system that is continually generating new ECUs and ever-higher levels of control, part of its maturation would necessarily be the development of new perceptual functions and RIFs. One could imagine a currently top-level ECU with a developing RIF that initially had no inputs, but as a new level began to develop above it, it would gain inputs. However, I suspect that the RIF would be more likely to develop and be reorganized in parallel with the new next-level connections.

Martin

Warren

Sent from my iPhone

On 25 Feb 2014, at 20:12, Martin Taylor mmt-csg@MMTAYLOR.NET wrote:

[Martin Taylor 2014.02.25.10.26]

Warren Mansell (apparently 2014.02.25.07.51) said

Hi Rick, I was only restricting my reply to the perceptual hierarchy under its immediate working properties. I recognise that intrinsic reference values are a separate issue. However, reorganisation, as I understand it would not lead to the same kind of ‘variation’ in reference signals in the uppermost system as the variation that all the (relatively) lower order systems experience. As reorganisation alters the parameters of the control unit rather than the reference signal directly, then one might get a change in the impact of the reference signal on the outputs that are generated, but not a time varied change in reference signal that all the lower order systems would experience.
Many years ago, Bill and I had a discussion on Reference Input Functions (RIFs). The point we were considering was that the reference value for any elementary control unit (ECU) at level N depends on the output of many, possibly very many, ECUs at the next higher level, N+1. The way these outputs combine to generate a reference value for an ECU is a “Reference Input Function” (RIF),

<mime-attachment.jpg>

and, as Rick suggested in [From Rick Marken (2014.02.23.1440)] it is very likely that the form of an RIF varies from level to level, as do the forms of the Perceptual Input Functions (PIFs). At low levels, an RIF might well be a weighted summation, while at some high level it might be the kind of associative memory implied by B:CP figure 15.3. At other levels of the hierarchy an RIF could well have completely different form, such as a shift register with weighted input values, or even (again as Rick suggested) a complete program as in a computer.

I like Kent’s focus on attention, his MOL link makes sense, and I think it would be fascinating if a meme could be considered as the current uppermost system in attention at any one time, which could be inferred by viewing quite automatically and passed on from agent to agent, regardless of the potentially different lower order systems and signals in those systems used on the basis of different environments and between individuals.

Warren
The concept of a Meme is not that of a scalar signal value. It is a complex structure. If it is relevant to the RIF at all, the meme would have to be the particular pattern of input values, not any one of them. However, is it more likely that a meme is a perception rather than the patterning that generates a reference value?

Martin

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences

Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

[Bob_Hintz]

bob hintz 2014.02.26

I have often wondered why the term is reorganization system rather than an organization system when a reference is made to changes in the hierarchical control system. The phrase “maturing system” certainly points at the fact the ecu’s are always being created and learning and unlearning is always occurring even if PCT does not chose to examine this process very often.

Is the organization creating system a control system with perceptions, references and outputs that result in the formations of ECU’s? Babies don’t control much outside themselves for several years and are quite dependent on others for survival.

bob

It seems to me that the whole hierarchy is created to

···

On Wed, Feb 26, 2014 at 10:43 AM, Warren Mansell wmansell@gmail.com wrote:

Thanks Martin, I take your point and you have expanded my understanding of a PCT hierarchy somewhat!
Warren

On Wed, Feb 26, 2014 at 3:13 PM, Martin Taylor mmt-csg@mmtaylor.net wrote:

[Martin Taylor 2014.02.26.09.17]

On 2014/02/26 2:59 AM, Warren Mansell wrote:

Hi Martin, good point about the convergence of multiple higher level systems within a function to determine the reference values at lower levels. I call have been calling them output functions but this might confuse them with the output in functions of the lowest level.

I think that to call them output functions would confuse them with the output function of each Elementary Control Unit (ECU). They are input functions for an ECU. Each control loop has two points where variable from outside the loop influence the signal values within the loop – the disturbance and the reference. Both are inputs to the loop. There are two outputs from the loop, the downward signal emerging from the ECUs Output Function and the upward Perceptual Signal that goes up to higher levels. The “input” and “output” to and from the organism are the only externally visible manifestations of what goes on inside the organism.

However it is a bit confusing to call them ‘input’ as they are travelling from inside the organism outwards.

Every connection and function has its own input and output. It would be very confusing to say that the signal source that feeds a wire is an input if the wire faces upward and an output if the wire faced downward. The Perceptual Input Function and the Reference Input Function both accept input from many sources, the perceptual function from lower levels, the RIF from higher levels. I don’t think I would be happy calling the inputs to a function “outputs”, no matter where the function exists in relation to a larger structure.

Also, I guess this explanation couldn’t apply to the currently uppermost system in awareness as it was currently operating in automatic mode - but it could apply within controlled mode to set the memory it uses in automatic mode.

I have no particular opinion about the function of awareness (well, I do, but it’s irrelevant here), but if an RIF were such that an input from one particular source could override the effect of other inputs, then if that source were active the reference value would be set by that alone, whereas if it were not, all the other inputs would jointly determine the reference value. A function such as “If (X > 0) then Y else Sum(aA+bB+c*C…)” would have that effect. The reference value would be Y if it was switched on by the value of X (if X referes to awareness. the implication is that X>0 means “I am aware of setting this reference value to be Y”), but the reference value would be the weighted sum of next-level-up output values if Y was switched off (X<0 means “I don’t care what that reference value is”).

It definitely couldn’t apply to the uppermost systems latent in the hierarchy as they would not have levels above them. Right?

If we imagine a maturing system that is continually generating new ECUs and ever-higher levels of control, part of its maturation would necessarily be the development of new perceptual functions and RIFs. One could imagine a currently top-level ECU with a developing RIF that initially had no inputs, but as a new level began to develop above it, it would gain inputs. However, I suspect that the RIF would be more likely to develop and be reorganized in parallel with the new next-level connections.

Martin

Warren

Sent from my iPhone

On 25 Feb 2014, at 20:12, Martin Taylor mmt-csg@MMTAYLOR.NET wrote:

[Martin Taylor 2014.02.25.10.26]

Warren Mansell (apparently 2014.02.25.07.51) said

Hi Rick, I was only restricting my reply to the perceptual hierarchy under its immediate working properties. I recognise that intrinsic reference values are a separate issue. However, reorganisation, as I understand it would not lead to the same kind of ‘variation’ in reference signals in the uppermost system as the variation that all the (relatively) lower order systems experience. As reorganisation alters the parameters of the control unit rather than the reference signal directly, then one might get a change in the impact of the reference signal on the outputs that are generated, but not a time varied change in reference signal that all the lower order systems would experience.
Many years ago, Bill and I had a discussion on Reference Input Functions (RIFs). The point we were considering was that the reference value for any elementary control unit (ECU) at level N depends on the output of many, possibly very many, ECUs at the next higher level, N+1. The way these outputs combine to generate a reference value for an ECU is a “Reference Input Function” (RIF),

<mime-attachment.jpg>

and, as Rick suggested in [From Rick Marken (2014.02.23.1440)] it is very likely that the form of an RIF varies from level to level, as do the forms of the Perceptual Input Functions (PIFs). At low levels, an RIF might well be a weighted summation, while at some high level it might be the kind of associative memory implied by B:CP figure 15.3. At other levels of the hierarchy an RIF could well have completely different form, such as a shift register with weighted input values, or even (again as Rick suggested) a complete program as in a computer.

I like Kent’s focus on attention, his MOL link makes sense, and I think it would be fascinating if a meme could be considered as the current uppermost system in attention at any one time, which could be inferred by viewing quite automatically and passed on from agent to agent, regardless of the potentially different lower order systems and signals in those systems used on the basis of different environments and between individuals.

Warren
The concept of a Meme is not that of a scalar signal value. It is a complex structure. If it is relevant to the RIF at all, the meme would have to be the particular pattern of input values, not any one of them. However, is it more likely that a meme is a perception rather than the patterning that generates a reference value?

Martin

–
Dr Warren Mansell
Reader in Psychology
Cognitive Behavioural Therapist & Chartered Clinical Psychologist
School of Psychological Sciences

Coupland I
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

See teamstrial.net for further information on our trial of CBT for Bipolar Disorders in NW England

The highly acclaimed therapy manual on A Transdiagnostic Approach to CBT using Method of Levels is available now.

Check www.pctweb.org for further information on Perceptual Control Theory

[MartinT]

[Martin Taylor 201402.26.15.08]

bob hintz 2014.02.26

I have often wondered why the term is reorganization system rather than an organization system when a reference is made to changes in the hierarchical control system.

For me, if a system is initially organized one way, and that organization changes, I would tend to call the change process "reorganization" rather than "organization". To me "organization" sounds like what you do to a room in chaos, not to one in which you want to move a couple of chairs and a bookcase.

The phrase "maturing system" certainly points at the fact the ecu's are always being created and learning and unlearning is always occurring even if PCT does not chose to examine this process very often.

Is the organization creating system a control system with perceptions, references and outputs that result in the formations of ECU's?

Don't ask me for more than my opinion. Ask Nature for the facts. My opinion is that it is not, at least not as a coherent structure. Again just my opinion, though it agrees with what others have suggested -- within the reorganization process there are control loops such as you describe.

Babies don't control much outside themselves for several years and are quite dependent on others for survival.

Actually, most of them control quite well what they need to control. If they did not, they would die. How do they control? One way is to disturb perceptions controlled by mother. They make different kinds of noises and facial expressions and a good mother who winds up with a well-developing baby controls her disturbed perceptions in ways that affect the baby's perceptions. If the baby feels hungry (error in a satiation perception) it cries. If Mother controls a perception of baby crying at a reference value of zero, she tries different things that a baby might sense (changing diaper, rocking baby, etc.) finally winding up by feeding it. Baby stops crying, but its reorganization process also slows or stops at an organization "when hungry, cry". Mother may reorganize to have a stronger connection between "I want baby not to cry and she is crying" and "I feed baby".

Babies do lots of things that may seem random to the parents. Those things alter perceptions. Most of the altered perceptions may be of no interest to the baby, but some of them might result in "I want to see that again" -- a reference condition. But as yet there's no connection between the current perception, the reference, and the action. Or perhaps it would be better to say that there are connections, but also there are lots of other connections and the appropriate set of them is not yet preferred. Doing random things when there are not a lot of different things you can do is not a huge search space, and if you can sense whether you are getting nearer to "seeing that again", the e-coli process can work to reduce the search space. Hebbian and anti-Hebbian synaptic modification could well be the underlying mechanism. I don't think it would take too long for a few non-genetically determined control units to form.

Martin

[Bob_Hintz]

bob hintz 2014.02.26

[Martin Taylor 201402.26.15.08]

I am still struggling to learn this form of communication and am pretty slow, but it seems to me that an infant does enter a room in chaos and has to learn to perceive a room and chairs before learning that chairs can be moved within a room or even taken out of a room. If we start at birth, I imagine that most infants are fed before they experience internal sensations that might be labeled “hunger” and certainly before they create an ECU that results in output “crying” in order to control that sensation. Actually I guess sucking is the output that controls that sensation if there is something in their mouths to suck on that provides nutrition. I suspect that infants and caregivers both learn what crying means (possibly a dangerous word, but I have started reading some of your material) in terms of the infant’s internal sensations by producing trial and error sequences that you described (changing, rocking, feeding) all of which require social interaction. I suspect “crying” is an output that initially controls the presence of a caregiver no matter what change in internal sensations produces an error. That presence is the essential condition for the improvement of any internal or intrinsic error.

I have started reading your Paris presentation, but have not gotten very far. Do you think that “crying” becomes a virtual message at some point rather than simply one of few behaviors that an infant can produce when experiencing error.

bob

···

On Wed, Feb 26, 2014 at 2:30 PM, Martin Taylor mmt-csg@mmtaylor.net wrote:

[Martin Taylor 201402.26.15.08]

bob hintz 2014.02.26

I have often wondered why the term is reorganization system rather than an organization system when a reference is made to changes in the hierarchical control system.

For me, if a system is initially organized one way, and that organization changes, I would tend to call the change process “reorganization” rather than “organization”. To me “organization” sounds like what you do to a room in chaos, not to one in which you want to move a couple of chairs and a bookcase.

The phrase “maturing system” certainly points at the fact the ecu’s are always being created and learning and unlearning is always occurring even if PCT does not chose to examine this process very often.

Is the organization creating system a control system with perceptions, references and outputs that result in the formations of ECU’s?

Don’t ask me for more than my opinion. Ask Nature for the facts. My opinion is that it is not, at least not as a coherent structure. Again just my opinion, though it agrees with what others have suggested – within the reorganization process there are control loops such as you describe.

Babies don’t control much outside themselves for several years and are quite dependent on others for survival.

Actually, most of them control quite well what they need to control. If they did not, they would die. How do they control? One way is to disturb perceptions controlled by mother. They make different kinds of noises and facial expressions and a good mother who winds up with a well-developing baby controls her disturbed perceptions in ways that affect the baby’s perceptions. If the baby feels hungry (error in a satiation perception) it cries. If Mother controls a perception of baby crying at a reference value of zero, she tries different things that a baby might sense (changing diaper, rocking baby, etc.) finally winding up by feeding it. Baby stops crying, but its reorganization process also slows or stops at an organization “when hungry, cry”. Mother may reorganize to have a stronger connection between “I want baby not to cry and she is crying” and “I feed baby”.

Babies do lots of things that may seem random to the parents. Those things alter perceptions. Most of the altered perceptions may be of no interest to the baby, but some of them might result in “I want to see that again” – a reference condition. But as yet there’s no connection between the current perception, the reference, and the action. Or perhaps it would be better to say that there are connections, but also there are lots of other connections and the appropriate set of them is not yet preferred. Doing random things when there are not a lot of different things you can do is not a huge search space, and if you can sense whether you are getting nearer to “seeing that again”, the e-coli process can work to reduce the search space. Hebbian and anti-Hebbian synaptic modification could well be the underlying mechanism. I don’t think it would take too long for a few non-genetically determined control units to form.

Martin