NOTE-959: Powers, Clark & McFarland — General Feedback Theory, Part II

Contribution

Part II applies Part I’s model to people. It names six orders of perceptual variable and gives a tabletop “Portable Demonstrator” for the first five, read off reaction times. It works out the N-system (reorganization) as the source of insight, defines consciousness by it, and gives a theory of conflict between control systems from which it derives regression, resistance and the psychoanalytic defences. It closes with a statement of values. The one experiment is “still in process” (p. 36).

Key insight

The hierarchy fixes what a person can perceive, and only reorganization changes that. “it is perception which gives form to behavior” (p. 26). The N-system “never adds new information to the system in the sense of providing a specific answer to a problem. It merely alters the properties of a system, thus changing the transformations applied to existing information” (p. 37). Learning, in this paper, is a change of perceptual transformation, experienced as insight, driven by intrinsic error and filtered by imagined conflict as well as by behaviour.

Assumptions

• All people are organized in the six orders, “for the time being” (p. 26). The authors cannot say why there are six: perhaps reality, the brain, the culture, “or even … the authors’ microcosm!” (p. 26).
• The observer is a system like the observed. “One must never forget that the person observing human behavior is a system like the one he is observing” (pp. 25–26).
• Order is not complexity. “Complexity is not a factor in determining relative order” (p. 27).
• Part I’s model, with its recording functions, imagination connection and N-system, is taken as given.

Key results

• Rules of order (p. 27). “In order to perceive and control nth-order variables, one must simultaneously perceive and control lower-order variables (except for n = 1), but the reverse is not true. Eliminating higher-order perceptions leaves the lower; blocking the lower partially or totally eliminates the higher.” Each system is “controlled to the extent that it does not choose which of its past experiences are to serve as goal-perceptions” (p. 27). “The highest order of reference-signal is set by noise or by random action of the N-system” (p. 27).
• The orders and demonstrations (pp. 28–35).
  • First: spinal loops; an arm pushed down rebounds “as if on a spring”, an active correction by electromyograph (p. 28).
  • Second: elementary sensations, “identity signals” (p. 29); a pushed arm first returns, then swings down as instructed (p. 29).
  • Third: configurations; a finger-pointing task shows a slower third phase of correction (p. 30).
  • Fourth: sequences; tracking a circling finger, S “continues to ‘track’ for nearly half a second” after it stops, and “His reaction time does not shorten significantly with practice” (p. 31).
  • Fifth: relationships, shown by a rule over sequences (“increasing the number of L sequences by one after every U sequence”, p. 32), with a reaction time “considerably slower” than fourth order (p. 33). “it is the act of changing interpretations that reveals the presence of a fifth-order system” (p. 33).
  • Sixth: “rather vague”; system concepts, such as personality, a government, “a self-concept” (pp. 34–35). No sixth-order reaction time was found (p. 35).
• Facts and sixth-order error (p. 35). “A ‘fact’ may be partially defined as a perception which does not cause an error in a sixth-order system”. A perception that does causes the person to look “for added information which will give a different fifth-order or lower perception, thus correcting the sixth-order error … Sometimes the required information is found, sometimes not. The nature of one’s sixth-order systems and the activity of one’s N-system will then determine whether the required shift in sixth-order perception and reference-level will occur.”
• The N-system (pp. 35–37). Its intrinsic variables are “probably the signals associated with drives, while others may represent more subtle conditions, such as average stimulus input rate, or mean error signal in the hierarchy” (p. 35). Its specificity is unknown: random, “localized in regions where error-signals exist”, or by some better rule (pp. 35–36). During reorganization skilled behaviour deteriorates; in the authors’ learning experiment, reaction time “begins to vary and becomes longer” just before each drop to a new plateau, which they “take to be evidence of N-system activity” (p. 36).
• Insight, and its veto by imagining (pp. 36–37). The N-system is responsible for “insight”, “the ‘aha’ reaction when one suddenly perceives a pattern in lower-order information which he has never seen before”; “it is merely a new organization of perception.” “insights which are not useful are quickly discarded because they create conflict or tend to increase intrinsic errors”. The golfer who thinks of using his putter like a billiard cue: “a moment spent imagining doing so would reveal the fifth-order and sixth-order (‘That’s not golf!’) conflicts which he could expect. Therefore this behavior pattern is not selected as a reference-level on the golf course.” A new transformation may “lie dormant” until useful (p. 37).
• Consciousness (p. 37). “Consciousness, we propose, is the state of the feedback-function in a subsystem in the hierarchy which is being affected by the output of the N-system. Thus the same subsystem can perceive and control a variable either consciously or unconsciously, depending on whether the N-system is actively connected to it.” Stated “merely” as a definition, with reasons withheld.
• Reduction of order (pp. 38–39). Subjects “attempt to use the lowest order of information available that will suffice”: asked to tell “left, left” from “right, left”, nearly all showed third-order reaction times (“0.3 sec., approximately”) instead of fourth (“0.4-0.45 sec., approximately”), having seen the sequence as an arrangement (p. 38). Symbols are “order-reduced representations of higher-order perceptions”; “most language is an order-reduced representation of experience” (p. 39).
• Conflict (pp. 39–42). When two order-n systems demand contradictory settings of a shared lower system, it holds a “virtual reference-level” and the two higher systems are “removed from the organism’s set of environment-controlling systems” (p. 40). Driven to their limits they oscillate (“vacillation”) or hand control to a cruder older system (“regression”, pp. 40–41). Remedies: altering a controlling system, altering a reference, “switching one system entirely to the imagination-connection [the mechanism (See Part I) of fantasy, wish-fulfillment, and closure]”, or suppression (p. 41). Anything other than a higher-order change of reference “or a realistic change in properties via N-system action” is “the ‘defense of the ego’” (p. 41). The rigidly maintained behaviour is one order below the conflict (pp. 41–42). “putting both or all the conflicted systems into the state of consciousness is the only way to start the N-system to work changing them” (p. 42).
• Values (pp. 43–44). The optimum hierarchy has no significant conflict and a self-concept that includes “knowledge of the properties of the N-system and the signs of its action” (p. 43). A person “would even have to learn, if he could, just what constitutes his own intrinsic state, because this information is not built into his learned systems” (p. 44). “A person must beware of setting up any ideal in his mind as sacred and untouchable, for by so doing, he sets up an automatic mechanism to counter the effect of his N-system, and guarantees a halt in his development” (p. 44).

Claims

id	claim	strength	support
C1	Human perception is organized in six orders (sensation, configuration, sequence, relationship, system concept above the spinal loops)	weak to moderate	introspective definitions; reaction-time demonstrations for orders 1–5, informal, no numbers except order 3–4
C2	Reaction time lengthens with order: about 0.3 s at third, 0.4–0.45 s at fourth	moderate	the left/right sequence task, p. 38; sample size not given
C3	Subjects use the lowest order of information that suffices	moderate	”invariably found”, p. 38; one task described
C4	Reorganization changes perceptual transformations, not information, and is felt as insight	weak	asserted, pp. 36–37; the plateau-noise observation, p. 36
C5	A new perceptual organization can be discarded by imagining its use, before any action	weak	the golfer example, pp. 36–37
C6	Consciousness is a feedback function under N-system action	weak	proposed as a definition, reasons withheld, p. 37
C7	Conflict between higher systems freezes the shared lower system at a virtual reference; the defences are ways around it, one of them a switch to imagination	weak	derivation from the model, pp. 39–42; a tabletop demonstration, pp. 42–43

Concepts

• FBCS. “externally fed-back Feedback Control Systems” (p. 25).
• Multiordinality. The same behaviour describable at many levels, “Both” an artefact of observation and a property of the system (p. 25).
• Order n. A class of perception self-evidently different from the class it is built from (p. 27).
• Identity signals. Second-order elementary sensations (p. 29).
• Fact. “a perception which does not cause an error in a sixth-order system” (p. 35).
• Insight. A new organization of perception produced by the N-system (p. 36).
• Consciousness. As above (p. 37).
• Reduction of order. Using lower-order information, or symbols, in place of higher (pp. 38–39).
• Virtual reference-level. The compromise value a conflicted shared subsystem holds (p. 40).

Connections

Part II depends throughout on Part I (LIT-1188, NOTE-954), whose imagination connection it uses for closure, fantasy and the golfer’s veto. The psychoanalytic defences, regression and resistance are reinterpreted, not cited. Learning-machine experiments are mentioned without citation (p. 36). The six orders became nine in Powers’s 1973 book (LIT-1179, not read) and later eleven (my knowledge, not this paper’s).

Bearing on the record

• CLAIM-341 (thesis).
  • This is the closest the 1960 theory comes to the claim. Two passages have the claim’s loop, a judgement that changes what is perceived next, without action.
    • The golfer: an insight, a new organization of perception, is rejected because “a moment spent imagining doing so would reveal the fifth-order and sixth-order … conflicts” (pp. 36–37). The imagination connection here tests a change of the observer and selects against it, with no action on the world.
    • The sixth-order “fact”: a perception that conflicts with a system concept sends the person looking for information that “will give a different fifth-order or lower perception”; if none is found, the N-system decides “whether the required shift in sixth-order perception and reference-level will occur” (p. 35). That is a judgement at the top reshaping perception below.
  • The engine is still not judgement. What changes perception is the N-system, driven by intrinsic error. If its intrinsic variables include “mean error signal in the hierarchy” (p. 35), then persistent error in learned systems, which is what a failing judgement is in this model, would itself drive reorganization. That would close the claim’s loop inside PCT: judgement → error → reorganized perception → new observations. The paper offers the variable as a guess and does not draw the consequence (my observation).
  • Consciousness ties to the reshaped observer. Defining consciousness as “the state of the feedback-function … being affected by the output of the N-system” (p. 37) makes conscious perception just the perception whose transformation is being changed. That fits the claim’s “lucid agent” better than anything in the 1973 paper, though it is offered as a definition, without reasons.
• CLAIM-339 (counter). Part II supplies the counter’s failure mode by name. Switching a system “entirely to the imagination-connection” is “the mechanism … of fantasy, wish-fulfillment, and closure” and one of the ego’s defences (p. 41), and an ideal held “sacred and untouchable” “sets up an automatic mechanism to counter the effect of his N-system” (p. 44). The paper’s cure is not action but consciousness: putting the conflicted systems into consciousness “is the only way to start the N-system to work changing them” (p. 42).
• QUESTION-121.
  • Candidate 4, processing signals. “mean error signal in the hierarchy” (p. 35) is a signal about how the interpretation is going, not what it says, fed to a system the learned hierarchy does not set. That is candidate 4 placed inside candidate 2’s channel.
  • Candidate 1, unselected input. “average stimulus input rate” (p. 35) makes the amount of input, not its content, an intrinsic variable. In sensory deprivation it would be in error and drive reorganization (my observation). In locked-in syndrome, with input spared, it would not.
  • What is lost with action. “Eliminating higher-order perceptions leaves the lower; blocking the lower partially or totally eliminates the higher” (p. 27) is about perception. Nothing in it says that loss of first-order output removes higher perceptions. So on this model a locked-in person keeps all orders of perception, and loses only the control of them through movement (my observation).
• CASE-107. Nothing directly; the paper does not discuss paralysis.
• For the anthology. No instruction for machine-learning practice.

Limitations

• The six orders are introspective proposals; the authors say they may be peculiar “to the authors’ microcosm” (p. 26).
• The demonstrations are informal, with reaction times given only for orders 3 and 4, without sample sizes or variance. The complex-learning experiment is “still in process” (p. 36).
• The definition of consciousness is stated without its reasons.
• The N-system’s specificity and the intrinsic variables are guesses (pp. 35–36).
• Read from an OCR extraction of a reset reprint.

Open questions

• Whether “mean error signal in the hierarchy” is an intrinsic variable. If it is, error in learned judgement reshapes perception through reorganization, which would make CLAIM-341’s loop a part of PCT.
• Whether the golfer’s veto, imagined conflict selecting among new perceptual organizations, can keep a model in touch with the world, or only with the person’s existing system concepts. In the example it protects “That’s not golf!”, a sixth-order reference, which is the self-confirming case.