Powers, W. T., Clark, R. K. & McFarland, R. L. (1960). Perceptual and Motor Skills 11(3):309–323. doi:10.2466/pms.1960.11.3.309. Reprinted in Living Control Systems: Selected Papers of William T. Powers (Control Systems Group, 1989), pp. 25–45.
Read in full in NOTE-959, from the reprint. Page numbers are the reprint’s.
Key takeaways
- Perception gives form to behaviour. “it is perception which gives form to behavior. Behavior will make sense to E only if E knows what perceptual variables the behavior is maintaining at some reference-level” (p. 26).
- Six orders, each a self-evident class of perception: first-order spinal loops, elementary sensations, configurations, sequences, relationships, and (vaguely) system concepts (pp. 28–35). “Eliminating higher-order perceptions leaves the lower; blocking the lower partially or totally eliminates the higher” (p. 27).
- Reorganization reshapes perception, not content. The N-system “never adds new information to the system … It merely alters the properties of a system, thus changing the transformations applied to existing information” (p. 37). Insight is “a new organization of perception” (p. 36). Its intrinsic variables may include “average stimulus input rate, or mean error signal in the hierarchy” (p. 35).
- Consciousness “is the state of the feedback-function in a subsystem in the hierarchy which is being affected by the output of the N-system” (p. 37).
- Conflict. Two higher systems demanding contradictory settings of a shared lower one leave it at a “virtual reference-level” (p. 40). One way out is “switching one system entirely to the imagination-connection [the mechanism (See Part I) of fantasy, wish-fulfillment, and closure]” (p. 41).
Standing in the record
Filed with the reading of Part I (LIT-1188, NOTE-954), because Part I’s question for the record, whether anything in the theory reshapes the perceptual functions, is answered mainly here. It does: the N-system alters the transformations that perceptual functions apply. But what drives it is error in inherited intrinsic variables, not a judgement: “insights which are not useful are quickly discarded because they create conflict or tend to increase intrinsic errors” (p. 36). So for CLAIM-341 it is a rival with the claim’s shape (the observer reshaped) and a different engine. For QUESTION-121 it offers “mean error signal in the hierarchy” as an intrinsic variable, a scoring signal about how the hierarchy is doing that the hierarchy does not set. It carries no instruction for machine-learning practice.