# Universal Error Curve

**URL:** <http://discourse.iapct.org/t/universal-error-curve/11916>\
**Category:** CSG2001\
**Created:** [April 17, 2001, 2:36pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916 "2001-04-17T14:36:41Z")\
**Posts on this page:** 20\
**Page:** 1

<div class="post-metadata">

**Author:** ![Bill\_Powers1](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/b/8e7dd6/32.png) [@Bill\_Powers1](http://discourse.iapct.org/u/Bill_Powers1)\
**Post date:** [April 17, 2001, 2:36pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/1 "2001-04-17T14:36:41Z")

</div>

[From Bill Powers (2001.04.17.0826 MDT)]

Dag Forssell shouldn't blame Rick Marken for the "Universal error curve,"  
but me. My suggestion is somewhere back there in the archives. The problem  
is that many control systems do not behave in the expected way when errors  
are large. In the vicinity of zero error, the error signal behaves  
normally: a small error gives a small error signal, and so on. But when  
you're very far from the goal, the error can be enormous. I want to be in  
San Francisco, but here I am in Colorado. The error signal ought to be big  
enough to vaporize my muscles. Of course it's not that big; in fact, it's  
smaller than it would be if I were just outside S.F. This sort of gradient  
is observed everywhere, even in animals. When a hungry animal is allowed to  
approach food pulling against a tether, the pulling force increases as the  
animal gets closer to the food -- just the opposite of what we would expect  
of a linear control system. Of course within a certain radius of the food,  
the force declines very rapidly as the animal gets close to it, implying a  
shrinking error signal. This seems to be a very general phenomenon,  
although I can't claim to have done any research on it.

Best,

Bill P.

---

<div class="post-metadata">

**Author:** ![Bruce\_Gregory](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/b/4bbf92/32.png) [@Bruce\_Gregory](http://discourse.iapct.org/u/Bruce_Gregory)\
**Post date:** [April 17, 2001, 5:05pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/2 "2001-04-17T17:05:39Z")

</div>

[From Bruce Gregory (2001.0417.1305)]

> Bill Powers (2001.04.17.0826 MDT)
> 
> Dag Forssell shouldn't blame Rick Marken for the "Universal error curve,"  
> but me. My suggestion is somewhere back there in the archives. The problem  
> is that many control systems do not behave in the expected way when errors  
> are large. In the vicinity of zero error, the error signal behaves  
> normally: a small error gives a small error signal, and so on. But when  
> you're very far from the goal, the error can be enormous. I want to be in  
> San Francisco, but here I am in Colorado. The error signal ought to be big  
> enough to vaporize my muscles. Of course it's not that big; in fact, it's  
> smaller than it would be if I were just outside S.F. This sort of gradient  
> is observed everywhere, even in animals. When a hungry animal is allowed to  
> approach food pulling against a tether, the pulling force increases as the  
> animal gets closer to the food -- just the opposite of what we would expect  
> of a linear control system. Of course within a certain radius of the food,  
> the force declines very rapidly as the animal gets close to it, implying a  
> shrinking error signal. This seems to be a very general phenomenon,  
> although I can't claim to have done any research on it.

It seems to me that there is another, equally plausible, alternative. If  
you are in Colorado, and want to be in San Francisco, you can implement a  
plan (a controlled sequence of perceptions). So long as you are  
implementing this plan the error associated with the implementation can be  
kept arbitrarily small. The other alternative is to set the gain on the "I  
am in San Francisco" control loop very low. (Alternatively, you can set the  
reference level for this perception to zero). Neither of these alternatives  
requires the ad hoc introduction of a "universal error curve."

BG

---

<div class="post-metadata">

**Author:** ![Bill\_Curry](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/b/13edae/32.png) [@Bill\_Curry](http://discourse.iapct.org/u/Bill_Curry)\
**Post date:** [April 30, 2002, 12:37pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/3 "2002-04-30T12:37:20Z")

</div>

[From Bill Curry (2002.04.30.0815 )]

> Rick Marken (2002.04.29.1710)
> 
> ....The phenomenon that the UEC explains is the phenomenon I describe above: where  
> a  
> controlled variable is pushed or kept far from its reference state with the  
> result that control of that variable stops or never starts. Like the knife that  
> is not retrieved, the Napoleon that isn't eaten, etc. This kind of thing happens  
> all the time. I think the "universal" part of the name means that this  
> non-linearity of the error curve might be a feature of \_all\_ comparator functions  
> in living control loops.
> 
> \> Might not "giving up" be  
> \> just as well explained by reorganization, which several objectors suggested?
> 
> Sure. But the situations I'm thinking of seem to involve something more like a  
> "truncation of" rather than a change in the control process. The bakery case may  
> be the most obvious. I want a Napoleon and it's easy for me to get it if I go  
> into the bakery; but I just pass by and am not "pulled into" the Napoleon  
> purchasing control process by getting close to the Napoleon dispenser. This is  
> the phenomenon that I think the UEC explains. It doesn't look or feel like  
> reorganization because I still like Napoleons and will control for one if doing  
> so is made easy (I get into the "control" range of the UEC).

The UEC may be a possible formulation to account for the phenomenon of "giving up"  
but can't normal hierarchical control provide a more parsimonious explanation? For  
example, you appear to have a "pass by the bakery" program level control system  
informed by memories of over-indulging in Napoleons. I would venture that you have  
another program control system for eating a Napoleon if you enter the bakery door.  
The conflict gets resolved one way or the other depending on relative gain at moment  
of the event.

My best,

Bill C.

> **···**
>
> --  
> William J. Curry Capticom, Inc.  
> bill@powerseed.com 603.756.9933

---

<div class="post-metadata">

**Author:** ![Richard\_Marken2](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/r/edb3f5/32.png) [@Richard\_Marken2](http://discourse.iapct.org/u/Richard_Marken2)\
**Post date:** [April 30, 2002, 1:11pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/4 "2002-04-30T13:11:22Z")

</div>

[From Rick Marken (2002.04.30.0810)]

Bill Curry (2002.04.30.0815 ) --

> The UEC may be a possible formulation to account for the phenomenon of "giving up"  
> but can't normal hierarchical control provide a more parsimonious explanation?

Not really. In the situations I'm thinking of there is really no "giving up" in the  
sense of abandoning the goal. I still want the Napoleon and I will take a bite if it  
is placed next to me with a fork and cup of coffee. The UEC explains why I don't fight  
my way across town and into the bakery given the massive error that presumably exists  
when I want a Napoleon and there is none nearby. I think of "giving up" as abandoning  
an unachievable goal. Such "giving up" is done by a higher level system perceiving the  
futility (or other negative consequences) of the quest to achieve the lower level goal  
and resetting the hopeless goal to zero. This kind of hierarchical control would  
explain what happens when I'm sitting Napoleonless in my kitchen and conclude that "I  
don't really want a Napoleon after all" because producing that perception would just be  
too much of an inconvenience. The UEC explains why massive error doesn't drive massive  
output; why my Napoleon goal doesn't drive me bolting out the door, into my car (this  
is LA) and over to the bakery a block away.

Bruce Nevin (2002.04.30 08:46) --

> Rather than introducing the UEC as an explanatory principle and adducing a  
> property of comparators for which no physical property of neural systems  
> has been observed, it seems to me that another explanation is more  
> straightforward. Casting about for means to control the variable in  
> question (the knife lost overboard, etc.) and controlling by those means in  
> imagination, I remember or imagine undesired consequences. It is not  
> conflict that I avoid, but those undesired consequences...

Yes. This is the hierarchical control explanation of why the goal of getting the knife  
is abandoned (set to zero by a higher level system). The UEC explains why the dropped  
knife doesn't "pull" me overboard as I am reaching to control for the knife being in my  
hand.

Bruce Nevin (2002.04.30 10:25) --

> "I see that you have chosen not to retrieve the knife, Mr. Marken."  
> "That's a choice? What, you want me to drown, right?"

Yes, of course it's a choice. The UEC explains why I stop reaching for the knife  
\_before\_ I have chosen not to retrieve it (ie. before a higher level system has reset  
the reference for the "knife in hand" control system to zero).

Best regards

Rick

> **···**
>
> ---  
> Richard S. Marken, Ph.D.  
> The RAND Corporation  
> PO Box 2138  
> 1700 Main Street  
> Santa Monica, CA 90407-2138  
> Tel: 310-393-0411 x7971  
> Fax: 310-451-7018  
> E-mail: rmarken@rand.org

---

<div class="post-metadata">

**Author:** ![Richard\_Marken2](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/r/edb3f5/32.png) [@Richard\_Marken2](http://discourse.iapct.org/u/Richard_Marken2)\
**Post date:** [April 30, 2002, 2:05pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/5 "2002-04-30T14:05:50Z")

</div>

[From Rick Marken (2002.04.30.0900)]

Stephen O'Shaughnessy wrote:

> Is the goal the knife or is the goal to cut something?

The unstated goal in my example was "grasping the knife". The goal is set as  
the means of achieving another, higher level goal, cutting the bait, which is  
set as the means of achieving a still higher level goal, catching fish, etc.  
All these goals are being set -- and achieved, except in the case of grasping  
the knife -- simultaneously.

> Is the goal to taste pastries or to satisfy a hunger ( or higher, to restore  
> the body's reserve of energy)?

I was think of the case where I just felt like a nosh -- a technical yiddish  
term for controlling for the perception of good tasting stuff.

> Its much easier to pass a bakery on a full stomach than on an empty one.

Actually, for me a bakery is hard to pass even after a big meal.

> At what point do these 'disturbances' shift from being references in some  
> system to disturbances in this system? In other words, how do you partition  
> a system of interest from all the other systems?

I don't know what you mean be "these 'disturbances'". A disturbance is a  
variable that has an effect on a controlled variable that is independent of a  
system's own effects on that controlled variable. A disturbance never shifts  
from being a disturbance to being a reference. A disturbance to a controlled  
variable is always a disturbance to a controlled variable. Does that make  
sense?

Best regards

Rick

> **···**
>
> --  
> Richard S. Marken, Ph.D.  
> The RAND Corporation  
> PO Box 2138  
> 1700 Main Street  
> Santa Monica, CA 90407-2138  
> Tel: 310-393-0411 x7971  
> Fax: 310-451-7018  
> E-mail: rmarken@rand.org

---

<div class="post-metadata">

**Author:** ![bnhpct](http://discourse.iapct.org/user_avatar/discourse.iapct.org/bnhpct/32/34_2.png) [@bnhpct](http://discourse.iapct.org/u/bnhpct)\
**Post date:** [April 30, 2002, 2:26pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/6 "2002-04-30T14:26:02Z")

</div>

[From Bruce Nevin (2002.04.30 10:25)]

Bruce Nevin (2002.04.30 08:46)--

> **···**
>
> At 08:48 AM 4/30/2002 -0400, I wrote:
> 
> This relation of controlling in imagination to consequent control in act has been used, as I recall, to explain "planning" and "anticipation".
> 
> ... or, more pertinently, to explain "choosing". Bill Curry (2002.04.30.0815) is advancing the same notion.
> 
> Not that the mechanism underlying choice is well specified either, but this approach is a twofer, explaining both choice and "giving up". (For non-native speakers of English: "twofer" \< "two for [='fer'] the price of one".)
> 
> "I see that you have chosen not to retrieve the knife, Mr. Marken."  
> "That's a choice? What, you want me to drown, right?"
> 
> &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;/Bruce Nevin

---

<div class="post-metadata">

**Author:** ![Stephen\_O\_Shaughness](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/s/df705f/32.png) [@Stephen\_O\_Shaughness](http://discourse.iapct.org/u/Stephen_O_Shaughness)\
**Post date:** [April 30, 2002, 2:55pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/7 "2002-04-30T14:55:08Z")

</div>

Is the goal the knife or is the goal to cut something?

Is the goal to taste pastries or to satisfy a hunger ( or higher, to restore  
the body's reserve of energy)? Its much easier to pass a bakery on a full  
stomach than on an empty one.

At what point do these 'disturbances' shift from being references in some  
system to disturbances in this system? In other words, how do you partition  
a system of interest from all the other systems?

Steve O

> **···**
>
> -----Original Message-----  
> From: Bruce Nevin [[mailto:bnevin@CISCO.COM](mailto:bnevin@CISCO.COM)]  
> Sent: Tuesday, April 30, 2002 9:26 AM  
> To: CSGNET@listserv.uiuc.edu  
> Subject: Re: Universal Error Curve
> 
> [From Bruce Nevin (2002.04.30 10:25)]
> 
> Bruce Nevin (2002.04.30 08:46)--  
> At 08:48 AM 4/30/2002 -0400, I wrote:
> 
> This relation of controlling in imagination to consequent control in act  
> has been used, as I recall, to explain "planning" and "anticipation".
> 
> ... or, more pertinently, to explain "choosing". Bill Curry  
> (2002.04.30.0815) is advancing the same notion.
> 
> Not that the mechanism underlying choice is well specified either, but this  
> approach is a twofer, explaining both choice and "giving up". (For  
> non-native speakers of English: "twofer" \< "two for [='fer'] the price of  
> one".)
> 
> "I see that you have chosen not to retrieve the knife, Mr. Marken."  
> "That's a choice? What, you want me to drown, right?"
> 
> &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;/Bruce Nevin

---

<div class="post-metadata">

**Author:** ![Stephen\_O\_Shaughness](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/s/df705f/32.png) [@Stephen\_O\_Shaughness](http://discourse.iapct.org/u/Stephen_O_Shaughness)\
**Post date:** [April 30, 2002, 5:26pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/8 "2002-04-30T17:26:42Z")

</div>

I think the key to my misunderstanding was the simultaneity (is that a  
word?) of the situation. I saw the goal (grasping the knife) as a reference  
set by some higher goal (cutting bait). When the knife disappeared the goal  
disappeared with it (replaced by some other goal, searching for a missing  
knife?). I didn't see the need for a catch all (universal) error function.  
Goal achievement is simply interrupted by the higher level. A new goal is  
set and the old one ceases.

Suppose the boat pitches and throws the knife overboard. Is that different  
from pitching and throwing the hand off trajectory (but not moving the  
knife)? In the second case the knife remains the goal and the same set of  
system variables operate. The force of the pitch is a disturbance. In the  
first case several things can happen:

1. The hand keeps moving to where the knife was. The goal remains (in the  
imagination), although the environment has changed. The hand can still move  
to the same point in space. But what is the perception being controlled?  
Is it hand in space or is it the visual of hand over knife?

2. The hand immediately starts a search pattern (at some level the goal is  
the same, but what level?). This thought prompted me to ask how we partition  
the system of interest. It seems you can always go up a level to get an  
explanation. The original system no longer exists, but higher levels that  
set it up (and not included in the system of interest) do still exist.

3. I dive overboard after the knife. Now I have the knife but I can't cut  
the bait, it's back on the boat. The environment has so drastically changed  
that its not the same problem anymore. But at some level, outside the  
original system, the goal to cut bait still exists.

4. I 'regroup' and start looking for something else with which to cut bait  
(my teeth? 🙂 ) Again, the original problem is dropped and a new one(s) are  
created.

I guess what's giving me trouble is in how the system is being defined.

As far as disturbances, I was looking at any given system as being a subset  
of a larger system. Can't you always go up a level? Is there a practical  
upper limit? I know HPCT has set some useful levels. But are they the only  
ones possible? So, looking at the system of interest, actions outside the  
system that affect the system are disturbances. This would include  
reference signals at levels not currently recognized as part of the current  
system under study. That's what I was thinking anyhow, I not sure if it  
makes much sense.

Is the goal of cutting bait a higher level (vertical) or a different control  
system (horizontal)? In either case can't the control system with the goal  
of grasping the knife simply be dropped and the components reset for a  
different goal?

Respectfully befuddled,

Steve O

> **···**
>
> -----Original Message-----  
> From: Richard Marken [mailto:marken@MINDREADINGS.COM]  
> Sent: Tuesday, April 30, 2002 9:06 AM  
> To: CSGNET@listserv.uiuc.edu  
> Subject: Re: Universal Error Curve
> 
> [From Rick Marken (2002.04.30.0900)]
> 
> Stephen O'Shaughnessy wrote:
> 
> > Is the goal the knife or is the goal to cut something?
> 
> The unstated goal in my example was "grasping the knife". The goal is set as  
> the means of achieving another, higher level goal, cutting the bait, which  
> is  
> set as the means of achieving a still higher level goal, catching fish, etc.  
> All these goals are being set -- and achieved, except in the case of  
> grasping  
> the knife -- simultaneously.
> 
> > Is the goal to taste pastries or to satisfy a hunger ( or higher, to
> 
> restore
> 
> > the body's reserve of energy)?
> 
> I was think of the case where I just felt like a nosh -- a technical yiddish  
> term for controlling for the perception of good tasting stuff.
> 
> > Its much easier to pass a bakery on a full stomach than on an empty one.
> 
> Actually, for me a bakery is hard to pass even after a big meal.
> 
> > At what point do these 'disturbances' shift from being references in some  
> > system to disturbances in this system? In other words, how do you
> 
> partition
> 
> > a system of interest from all the other systems?
> 
> I don't know what you mean be "these 'disturbances'". A disturbance is a  
> variable that has an effect on a controlled variable that is independent of  
> a  
> system's own effects on that controlled variable. A disturbance never shifts  
> from being a disturbance to being a reference. A disturbance to a controlled  
> variable is always a disturbance to a controlled variable. Does that make  
> sense?
> 
> Best regards
> 
> Rick  
> --  
> Richard S. Marken, Ph.D.  
> The RAND Corporation  
> PO Box 2138  
> 1700 Main Street  
> Santa Monica, CA 90407-2138  
> Tel: 310-393-0411 x7971  
> Fax: 310-451-7018  
> E-mail: rmarken@rand.org

---

<div class="post-metadata">

**Author:** ![Richard\_Marken2](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/r/edb3f5/32.png) [@Richard\_Marken2](http://discourse.iapct.org/u/Richard_Marken2)\
**Post date:** [April 30, 2002, 6:56pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/9 "2002-04-30T18:56:32Z")

</div>

[From Rick Marken (2002.04.30.1400)]

Stephen O'Shaughnessy wrote:

> I think the key to my misunderstanding was the simultaneity (is that a  
> word?) of the situation. I saw the goal (grasping the knife) as a reference  
> set by some higher goal (cutting bait). When the knife disappeared the goal  
> disappeared with it (replaced by some other goal, searching for a missing  
> knife?).

It's great that you are asking these questions. How else can any of us learn!

I would say that the goal (the reference signal specifying "grasped knife") was  
still there in the knife example; what changed was the \_perception\_ (of the  
position of the knife relative to the grasp). If the reference for "knife  
grasped" (r) is, say a neural current value of 100, then when the knife slides  
into the deep the perception of "knife relative to grasp" goes from, say, a  
neural current value of 90 to a neural current value of 0; i.e., error gets  
huge. Such an error would cause the reaching hand to go off into the deep, too,  
were it not (I think) for the UEC.

> I didn't see the need for a catch all (universal) error function.  
> Goal achievement is simply interrupted by the higher level. A new goal is  
> set and the old one ceases.

This does happen in many cases. A good example is in my "Levels of Intention"  
demo which is accessible from [http://home.earthlink.net/~rmarken/demos.html](http://home.earthlink.net/~rmarken/demos.html).  
When the error starts increasing exponentially (after a polarity reversal) a  
higher level system eventually steps in and changes the polarity of the  
relationship between hand and cursor movement. But if this did not happen -- if  
a higher level system in you were not able to change the polarity -- you would  
still eventually stop moving your hand (stop controlling the cursor) even  
though the error is continually increasing. This is the phenomenon that is  
being explained by the UEC.

Bill Powers makes the same points as me in his recent post (Bill Powers  
(2002.04.30.0853 MDT)). That (and this) post also suggests some simple tracking  
studies that could be done to map out the UEC. I think the development of such  
studies would be a very good exercise for those who \_don't\_ like the idea of a  
UEC. Who better to do tests of a model than those who are most skeptical of its  
merits.

Best regards

Rick

> **···**
>
> ---  
> Richard S. Marken, Ph.D.  
> The RAND Corporation  
> PO Box 2138  
> 1700 Main Street  
> Santa Monica, CA 90407-2138  
> Tel: 310-393-0411 x7971  
> Fax: 310-451-7018  
> E-mail: rmarken@rand.org

---

<div class="post-metadata">

**Author:** ![Stephen\_O\_Shaughness](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/s/df705f/32.png) [@Stephen\_O\_Shaughness](http://discourse.iapct.org/u/Stephen_O_Shaughness)\
**Post date:** [May 1, 2002, 12:11pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/10 "2002-05-01T12:11:30Z")

</div>

Thanks for being patient with a newbie.

Steve O

> **···**
>
> -----Original Message-----  
> From: Richard Marken [[mailto:marken@MINDREADINGS.COM](mailto:marken@MINDREADINGS.COM)]  
> Sent: Tuesday, April 30, 2002 1:57 PM  
> To: CSGNET@listserv.uiuc.edu  
> Subject: Re: Universal Error Curve
> 
> It's great that you are asking these questions. How else can any of us  
> learn!

---

<div class="post-metadata">

**Author:** ![Richard\_Marken2](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/r/edb3f5/32.png) [@Richard\_Marken2](http://discourse.iapct.org/u/Richard_Marken2)\
**Post date:** [May 1, 2002, 8:01pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/11 "2002-05-01T20:01:42Z")

</div>

[From Rick Marken (2002.05.01.1500)]

Bruce Nevin (2002.05.01 16:02) --

> In my case the disturbance can be described as defense of my notion  
> of how to do science and in a particular an aversion to Explanatory

Principles.

> The UEC is a description.

I would describe the UEC as a functional component of a working model of  
control.

> As I understand it, the UEC says if an elementary control system  
> produces its maximum output without correcting error it "gives up".

The UEC says that increases in error beyond a certain point lead to  
\_decreases\_ rather than further increases in output. That is, there is a  
non-monotonic relationship between error and output.

Now that you know what the UEC is, is the disturbance removed?

Best regards

Rick

> **···**
>
> --  
> Richard S. Marken, Ph.D.  
> The RAND Corporation  
> PO Box 2138  
> 1700 Main Street  
> Santa Monica, CA 90407-2138  
> Tel: 310-393-0411 x7971  
> Fax: 310-451-7018  
> E-mail: rmarken@rand.org

---

<div class="post-metadata">

**Author:** ![bnhpct](http://discourse.iapct.org/user_avatar/discourse.iapct.org/bnhpct/32/34_2.png) [@bnhpct](http://discourse.iapct.org/u/bnhpct)\
**Post date:** [May 1, 2002, 8:03pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/12 "2002-05-01T20:03:03Z")

</div>

[From Bruce Nevin (2002.05.01 16:02)]  
In my case the disturbance can be described as defense of my notion of  
how to do science and in a particular an aversion to Explanatory  
Principles.  
The UEC is a description. The only evidence for the UEC is observed  
changes in behavioral outputs for which it seems to me there is an  
alternative explanation without adding any new Explanatory  
Principles.  
As I understand it, the UEC says if an elementary control system produces  
its maximum output without correcting error it “gives up”. This  
is generalized (in some unspecified way) to cases where a hierarchical  
control system “gives up” because it _imagines_ that it  
would be unable to resist a disturbance no matter how hard it tried, by  
no matter what (imagined) means.

This is construed as a way of resolving external conflict by avoiding  
it.

One alternative explanation involves internal conflict. The system is  
controlling something else with higher gain. For example, it could be  
controlling aversion to imagined consequences, or it could simply be  
controlling something else at the same time. Putting maximum output into  
this conflict prevents control of much of anything else.

Another alternative explanation involves going up a level from the  
conflict. It appears to me that when we cast about for alternative means  
to the same end we have done something like going up a level. At this  
level of reconsideration we may decide that something else that we are  
controlling is more important. We may decide this for any number of  
reasons, including the perception that we aren’t getting control of this  
conflicted variable in any way that we can try or imagine.

I don’t think the processes of deciding and choosing are well understood  
in PCT, but when we do understand them we will have an explanation for  
everything that the UEC describes. Conversely, the UEC provides no  
insight into deciding and choosing. We decide and choose things all the  
time without being confronted with a disturbance that we cannot  
resist.

```
    /Bruce

```

Nevin

---

<div class="post-metadata">

**Author:** ![bnhpct](http://discourse.iapct.org/user_avatar/discourse.iapct.org/bnhpct/32/34_2.png) [@bnhpct](http://discourse.iapct.org/u/bnhpct)\
**Post date:** [May 1, 2002, 11:14pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/13 "2002-05-01T23:14:41Z")

</div>

[Bruce Nevin (2002.05.01 16:02)]

Rick Marken (2002.05.01.1500)--

> \> The UEC is a description.
> 
> I would describe the UEC as a functional component of a working model of  
> control.
> 
> \> As I understand it, the UEC says if an elementary control system  
> \> produces its maximum output without correcting error it "gives up".
> 
> The UEC says that increases in error beyond a certain point lead to  
> \_decreases\_ rather than further increases in output. That is, there is a  
> non-monotonic relationship between error and output.
> 
> Now that you know what the UEC is, is the disturbance removed?

No. This is a description. It is not an explanation. Nor does it say how the UEC (a descriptive statement about the error signal) is a functional component of a working model of control.

A working model of control should show the observed behavior (the non-monotonic relationship between error and output) as an effect of its working.

Methodologically, the importation of observed results as their own causes could be called "the method of adducing dormitive principles".

&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;/Bruce Nevin

> **···**
>
> At 03:01 PM 5/1/2002 -0500, Richard Marken wrote:

---

<div class="post-metadata">

**Author:** ![Fred\_Nickols2](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/f/53a042/32.png) [@Fred\_Nickols2](http://discourse.iapct.org/u/Fred_Nickols2)\
**Post date:** [May 2, 2002, 10:52am UTC](http://discourse.iapct.org/t/universal-error-curve/11916/14 "2002-05-02T10:52:34Z")

</div>

[From Fred Nickols (2002.05.02.0645)] --

> Bill Powers (2002.05.01.1840 MDT)]

\<snip\>

> I do believe that a process like decision-making or choosing does occur in  
> some kinds of behavior (I experience it when designing electronic  
> systems). But the only explanation of what happens during that process is  
> given by PCT -- I don't know of any other explanation at all. Perhaps this  
> is just a matter of my ignorance.

Could you say some more about what you've just said or point me to the  
relevant reading? More particularly, what is it you experience when  
designing electronic systems? How does PCT explain it?

Regards,

Fred Nickols  
740.397.2363  
nickols@safe-t.net  
"Assistance at a Distance"  
[http://home.att.net/~nickols/articles.htm](http://home.att.net/~nickols/articles.htm)

---

<div class="post-metadata">

**Author:** ![Richard\_Marken2](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/r/edb3f5/32.png) [@Richard\_Marken2](http://discourse.iapct.org/u/Richard_Marken2)\
**Post date:** [May 2, 2002, 1:16pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/15 "2002-05-02T13:16:48Z")

</div>

[From Rick Marken (2002.05.02.0815)]

Me:

> Now that you know what the UEC is, is the disturbance removed?

Bruce Nevin (2002.05.01 16:02) --

> No...A working model of control should show the observed behavior (the  
> non-monotonic relationship between error and output) as an effect of its  
> working...Methodologically, the importation of observed results as their own  
> causes could be called "the method of adducing dormitive principles"

Bill Powers (2002.05.01.1840 MDT)

> The attached Figure from a book by McMahon shows the observed relationship  
> between stretch and force in a frog muscle fiber...

> This curve explains why it is that when muscles are used to exert very  
> large forces, as in arm-wrestling, there comes a point where further stress  
> applied to the muscle causes a \_decrease\_ in muscle force, usually followed  
> by a collapse of the ability to generate any muscle force. ...

> This phenomenon occurs at the lowest level of behavioral organization. It's  
> not unreasonable to suppose that the pattern is repeated at higher levels,  
> though confirmation would be needed. No dormitive principles here.

Ok, Bruce (and others who think the UEC is PCT apostasy), now that you know what  
the UEC is and what phenomenon it explains, is the disturbance removed? (My  
guess is that it's not).

Best regards

Rick

> **···**
>
> --  
> Richard S. Marken, Ph.D.  
> The RAND Corporation  
> PO Box 2138  
> 1700 Main Street  
> Santa Monica, CA 90407-2138  
> Tel: 310-393-0411 x7971  
> Fax: 310-451-7018  
> E-mail: rmarken@rand.org

---

<div class="post-metadata">

**Author:** ![Stephen\_O\_Shaughness](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/s/df705f/32.png) [@Stephen\_O\_Shaughness](http://discourse.iapct.org/u/Stephen_O_Shaughness)\
**Post date:** [May 2, 2002, 4:38pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/16 "2002-05-02T16:38:22Z")

</div>

Would you really expect the disturbance to be removed? Your (or anyone's)  
explanation of something does not automatically provide others with  
acceptance or agreement. PCT applies to single organisms. You've (Rick)  
told me PCT can be applied to larger entities. Would this still be PCT or  
something using PCT-like tools and methods? In this case I'm seeing  
multiple entities with PCT loops (I'm not sure of how to apply the syntax  
yet). One is you in the act of transmitting something (UEC ideas) into the  
environment. Two is me (and everyone else) in the act of receiving  
(pulling) something from the environment (UEC ideas). I can see how PCT can  
be used to study culture or economics, but I'm going to limit myself to  
single organisms (me in particular) for now.

Somewhere, further up the hierarchy, we process these ideas. Isn't this  
where the disturbance will enter in? I guess question one: Do disturbances  
enter the system at level one or can they enter at any level? I'm trying to  
understand this debate in the context of PCT. It seems to me agreement or  
disagreement with the idea is largely immaterial, unless the receiver  
already holds exactly the same idea. I'm wondering if the current thread is  
really a disturbance or just normal operation of PCT systems.

By placing an idea (thought, comment, or whatever) into the environment  
aren't we trying to cause that thought to be duplicated in another organism?  
Creation?!? Or this trying to take PCT too far?

Isn't conflict the result of two (or more) systems? Your perception, from  
the environment, that your idea wasn't created in me. My transmission, into  
the environment, that I'm having trouble processing the idea (either I can't  
decode it or it clashes with existing loops in my systems). Thus I've not  
(re)created the idea in my systems.

I'll be the first to admit I'm way to green on the topic of PCT and probably  
stepping all over myself. I really feel like PCT is being applied too  
broadly. It looks like it's being applied to the whole of the debate when  
it's really only valid as applied separately to the individual engaged in  
the debate. It also looks like one PCT system is being applied when there  
are really several distinct systems operating. Specifically I see ego  
maintenance systems, learning systems, and socialization systems.

I hope I've not offended anyone.

Steve O

> **···**
>
> -----Original Message-----  
> From: Richard Marken [mailto:marken@MINDREADINGS.COM]  
> Sent: Thursday, May 02, 2002 8:17 AM  
> To: CSGNET@listserv.uiuc.edu  
> Subject: Re: Universal Error Curve
> 
> [From Rick Marken (2002.05.02.0815)]
> 
> Me:
> 
> > Now that you know what the UEC is, is the disturbance removed?
> 
> Bruce Nevin (2002.05.01 16:02) --
> 
> > No...A working model of control should show the observed behavior (the  
> > non-monotonic relationship between error and output) as an effect of its  
> > working...Methodologically, the importation of observed results as their
> 
> own
> 
> > causes could be called "the method of adducing dormitive principles"
> 
> Bill Powers (2002.05.01.1840 MDT)
> 
> > The attached Figure from a book by McMahon shows the observed relationship  
> > between stretch and force in a frog muscle fiber...
> 
> > This curve explains why it is that when muscles are used to exert very  
> > large forces, as in arm-wrestling, there comes a point where further
> 
> stress
> 
> > applied to the muscle causes a \_decrease\_ in muscle force, usually
> 
> followed
> 
> > by a collapse of the ability to generate any muscle force. ...
> 
> > This phenomenon occurs at the lowest level of behavioral organization.
> 
> It's
> 
> > not unreasonable to suppose that the pattern is repeated at higher levels,  
> > though confirmation would be needed. No dormitive principles here.
> 
> Ok, Bruce (and others who think the UEC is PCT apostasy), now that you know  
> what  
> the UEC is and what phenomenon it explains, is the disturbance removed? (My  
> guess is that it's not).
> 
> Best regards
> 
> Rick  
> --  
> Richard S. Marken, Ph.D.  
> The RAND Corporation  
> PO Box 2138  
> 1700 Main Street  
> Santa Monica, CA 90407-2138  
> Tel: 310-393-0411 x7971  
> Fax: 310-451-7018  
> E-mail: rmarken@rand.org

---

<div class="post-metadata">

**Author:** ![Richard\_Marken2](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/r/edb3f5/32.png) [@Richard\_Marken2](http://discourse.iapct.org/u/Richard_Marken2)\
**Post date:** [May 2, 2002, 5:02pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/17 "2002-05-02T17:02:23Z")

</div>

[From Rick Marken (2002.05.02.1200)]

Stephen O'Shaughnessy wrote:

> Would you really expect the disturbance to be removed? Your (or anyone's)  
> explanation of something does not automatically provide others with  
> acceptance or agreement.

Exactly. The discussion of the UEC is a disturbance to a perception that some  
individuals are controlling for. We know this because people are pushing back  
against the idea. Indeed, one of these people, Dag Forssell, actually said that  
talk about the UEC created an error for him. The implication (from a PCT  
perspective) is that talk of the UEC pushed some variable Dag is controlling  
away from its reference. I don't expect my (or Bill's) explanations of the UEC  
to remove this disturbance. What \_might\_ result from these explanations,  
however, is a change in the variable controlled -- to something more like what  
Bill and I are controlling for -- so that talk about the UEC will no longer be  
a disturbance.

An environmental variable (such as discussion of the UEC) is usually called a  
"disturbance" only if it tends to push a controlled variable away from its  
reference. For example, rain is a disturbance if you are having a picnic  
outdoors. If you are not having a picnic, then rain may simply be rain. The  
same is the case with talk of a UEC. Talk of a UEC is a disturbance if you are  
controlling for, say, a perception of PCT as only what is written in B:CP. It's  
just talk about a UEC (not a disturbance) if you are controlling for, say,  
understanding how people control.

> PCT applies to single organisms.

Yes. PCT applies to one organism at a time. The people who find discussion of  
the UEC disturbing may all be controlling different perceptions. But each of  
these people is controlling a perception that is disturbed by discussion of the  
UEC.

> You've (Rick) told me PCT can be applied to larger entities. Would this  
> still be PCT or  
> something using PCT-like tools and methods?

Yes, I (and Kent McClelland, for example) have used PCT to model aggregate  
behavior. I think it's still PCT.

> Somewhere, further up the hierarchy, we process these ideas. Isn't this  
> where the disturbance will enter in?

No. Disturbances are in the environment and they only act in the environment.  
Disturbances have effects on the environmental variables that are perceived in  
different ways at different levels of the hierarchy. So a disturbance might  
affect a perceptual representation of the disturbed environmental variable at  
one level but not at another. For example, a color filter is a disturbance to  
perceptions at the sensation level but it might not be a disturbance to  
perceptions of the same environmental variable at the configuration level. For  
example, through the filter a bear might look blue instead of brown but it's  
still shaped like a bear.

> I guess question one: Do disturbances  
> enter the system at level one or can they enter at any level?

Again, disturbances don't enter the system. Only sensory measures of  
environmental variables enter the system. Disturbances are environmental  
variables that act on variables in the environment that are perceived by the  
system. The UEC discussion exists in the environment of those who are offended  
by it. They are presumably offended by it because this discussion (the textual  
symbols that make up the discussion) results in a perception of something (say,  
of PCT) that differs from the person's reference for that perception (for, say,  
PCT as a particular, fixed entity).

> I'm trying to  
> understand this debate in the context of PCT.

I am too. I hope my comments above help you see how I understand it from a PCT  
perspective.

> I hope I've not offended anyone.

Certainly not me. I think it's great that, as a newcomer, you have had courage  
and interest enough to ask questions and describe what you think. This is a  
good way to learn, especially if it is supplemented with observation and work  
with the model (which I believe you are doing since you are part of the  
Robotics project). Keep up the good work.

Best regards

Rick

> **···**
>
> --  
> Richard S. Marken, Ph.D.  
> The RAND Corporation  
> PO Box 2138  
> 1700 Main Street  
> Santa Monica, CA 90407-2138  
> Tel: 310-393-0411 x7971  
> Fax: 310-451-7018  
> E-mail: rmarken@rand.org

---

<div class="post-metadata">

**Author:** ![Isaac\_Kurtzer1](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/i/439d5e/32.png) [@Isaac\_Kurtzer1](http://discourse.iapct.org/u/Isaac_Kurtzer1)\
**Post date:** [May 2, 2002, 6:17pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/18 "2002-05-02T18:17:35Z")

</div>

i.kurtzer(2002.05.02.1350)

[From Rick Marken (2002.05.02.0815)]

> Me:
> 
> \>Now that you know what the UEC is, is the disturbance removed?
> 
> Bruce Nevin (2002.05.01 16:02) --
> 
> \> No...A working model of control should show the observed behavior  
> (the  
> \> non-monotonic relationship between error and output) as an effect  
> of its  
> \> working...Methodologically, the importation of observed results  
> as their own  
> \> causes could be called "the method of adducing dormitive  
> principles"
> 
> Bill Powers (2002.05.01.1840 MDT)
> 
> \> The attached Figure from a book by McMahon shows the observed  
> relationship  
> \> between stretch and force in a frog muscle fiber...
> 
> \> This curve explains why it is that when muscles are used to exert  
> very  
> \> large forces, as in arm-wrestling, there comes a point where  
> further stress  
> \> applied to the muscle causes a \_decrease\_ in muscle force,  
> usually followed  
> \> by a collapse of the ability to generate any muscle force. ...
> 
> \> This phenomenon occurs at the lowest level of behavioral  
> organization. It's  
> \> not unreasonable to suppose that the pattern is repeated at  
> higher levels,  
> \> though confirmation would be needed. No dormitive principles  
> here.
> 
> Ok, Bruce (and others who think the UEC is PCT apostasy), now that  
> you know what  
> the UEC is and what phenomenon it explains, is the disturbance  
> removed? (My  
> guess is that it's not).

I don't think UEC is PCT apostacy. However, I do think it has not been  
sufficiently tested to consider it "universal". That precious few other  
PCT concepts that have been suffiently tested to be considered  
scientifically solid facts (see hierarchy and reorganization) does not  
weigh in favor of another minimally tested idea.  
The McMahon figure (introduced to give weight to the UEC) shows the  
nonlinear force/distance relation seen in an isolated muscle fiber.  
This is a spring. There are no reference signals or error signals or  
perceptual signals in that denervated fiber any more than in a rubber  
band. I can also stretch some rubber band until there is plastic flow and  
it has a different elastic constant. Have I proven a general principle  
about living control systems? I think the idea of nonlinear error  
functions is an justifiably explorable topic but currently its a "just so"  
story.

Isaac

> Best regards
> 
> Rick  
> --  
> Richard S. Marken, Ph.D.  
> The RAND Corporation  
> PO Box 2138  
> 1700 Main Street  
> Santa Monica, CA 90407-2138  
> Tel: 310-393-0411 x7971  
> Fax: 310-451-7018  
> E-mail: rmarken@rand.org

Isaac Kurtzer  
Ashton Graybiel Lab of Spatial Orientation  
Brandeis University

---

<div class="post-metadata">

**Author:** ![Richard\_Marken2](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/r/edb3f5/32.png) [@Richard\_Marken2](http://discourse.iapct.org/u/Richard_Marken2)\
**Post date:** [May 2, 2002, 8:29pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/19 "2002-05-02T20:29:27Z")

</div>

[From Rick Marken (2002.05.02.1530)]

i.kurtzer(2002.05.02.1350).

> I don't think UEC is PCT apostacy. However, I do think it has not been  
> sufficiently tested to consider it "universal".

That sounds much better Would you be willing to talk about it (or even test  
the idea) if we changed the name from UEC to HNMEC (hypothetical  
non-monotonic error curve)?

Best

Rick

> **···**
>
> --  
> Richard S. Marken, Ph.D.  
> The RAND Corporation  
> PO Box 2138  
> 1700 Main Street  
> Santa Monica, CA 90407-2138  
> Tel: 310-393-0411 x7971  
> Fax: 310-451-7018  
> E-mail: rmarken@rand.org

---

<div class="post-metadata">

**Author:** ![Bill\_Powers1](http://discourse.iapct.org/letter_avatar_proxy/v4/letter/b/8e7dd6/32.png) [@Bill\_Powers1](http://discourse.iapct.org/u/Bill_Powers1)\
**Post date:** [May 2, 2002, 9:31pm UTC](http://discourse.iapct.org/t/universal-error-curve/11916/20 "2002-05-02T21:31:31Z")

</div>

[From Bill Powers (2002.05.02.1448 MDT)]

Fred Nickols (2002.05.02.0645) --

> Could you say some more about what you've just said or point me to the  
> relevant reading? More particularly, what is it you experience when  
> designing electronic systems? How does PCT explain it?

Last question first. We're talking here about particular things a human  
system happens to become organized to do, not about anything fundamental.  
My proposals about levels of organization are meant to include things like  
performing rule-driven behaviors (the so-called "logic level"), but they  
don't have anything to say about \_which\_ rules will be in effect. For that,  
we simply have to observe what people do. PCT can't help us there.

We would use PCT to explain how it is that a rule can be put into effect,  
not a particular rule but any rule. Briefly, I would suppose that we  
perceive our actions and their effects as being examples of a rule, compare  
the perceived rule with the rule we intend to be in effect, and if there is  
a discrepancy,. alter the reference signals being sent to lower systems so  
as to reduce the discrepancy. "Was the move I just made (or imagined) a  
legal move in chess? " If not, retract it before letting go of the piece,  
and try a different move.

Designing electronic systems and probably most other kinds is a juggling  
act, in which one considers different components, materials, and  
relationships among parts while controlling for many variables. For  
example, in the robot design that Bruce and I are working on, we've  
considered servo motors of many sizes, weights, torque  
capacities, speeds, prices, degrees of durability, and frictional losses.  
We've looked at at least three ways of simulating a muscle's internal  
spring characteristics and viscous damping (which we've decided to emulate  
electronically rathen than use literal dashpots). We've considered plastic  
construction materials and model-maker's plywood. We've considered direct,  
cable, and chain drives. We've looked at electronic interfaces and  
breadboarding materials, and at pressure-sensitive rubber for detecting  
forces. We've looked at the way many other people have designed their  
hexapods, and indeed two such people have joined in the effort and are  
contributing their own extensive experience in building walking models.

The process is one of trying to meet many criteria (reference levels) as  
closely as possible and all at the same time, even though initially the  
reference conditions are highly uncertain. A great deal of it takes place  
in imagination, as we put mental models into action to see what problems  
arise. Each of the criteria involves its own controlled variable, and  
reference levels for each one that we would like the final model to match.  
For example, Bruce just worked out that a particular design would be able  
to support a total of a little over 2 kilograms of weight on its six legs.  
Is that enough? If it turns out that all the stuff we want to load on the  
body weighs more than that, the answer is no: We'd satisfy one set of  
criteria but make it impossible to satisfy others, and that would be a  
conflict. So we could, if necessary, consider putting less load on the body  
or increasing the power of the motors, or maybe other things we haven't  
thought of yet.

As far as I can see, no real decisions have been needed yet. Each possible  
design leaves a certain amount of overall error, but nothing has progressed  
far enough to call for a choice to be made. As the model takes a more  
definite shape, certain things catch the eye -- for example, in one drawing  
there is a servo motor mounted on the distal leg segment, whereas in  
biological systems the motive power for that segment would come from  
muscles in the proximal segment. We aren't explicitly trying to mimic any  
biological system, but the thought is always there in the background -- a  
goal of lesser priority, meaning we control for it with lower gain than for  
the others.

In my own design efforts over the years, I've always felt more disorganized  
than I thought other people were. I don't have a nice neat plan that leads  
uniformly toward the final design. I get a sense of the system gradually  
taking shape as a photographic print in the developer gradually begins to  
show forms . Many variables change as the design progresses, gradually  
converging toward one definite form rather than another, until in the end  
all other possibilities have fallen away, leaving the finished product. I'm  
controlling for a lot of variables, herding them toward goal states which  
themselves change during the development. There are tensions among the  
different control processes, and reducing or eliminating them is one of the  
means of moving toward the final design.

Needless to say, I see a great deal of reorganization going on, at least in  
the way I design systems. In engineering schools there is a strong tendency  
to try to solve problems by looking up solutions already worked out, but I  
always seem to see something to improve on, or do differenty anyhow, and  
probably waste a lot of time reinventing wheelss

Different people are different, is what it comes down to. But the hierarchy  
of perception and control follows the same principles in all of them, or so  
I think.

Best,

Bill P.

> **···**
>
> > Regards,
> > 
> > Fred Nickols  
> > 740.397.2363  
> > nickols@safe-t.net  
> > "Assistance at a Distance"  
> > [http://home.att.net/~nickols/articles.htm](http://home.att.net/~nickols/articles.htm)

[Next page](http://discourse.iapct.org/t/universal-error-curve/11916.md?page=2)