NS – X: On the Mathematics of Brain Modeling
Peter A. Putnam
June 1974
Note
This is a rough draft of an attempt to draw together the themes of NS - I to IX. Themes often re-done before are foreshortened or indicated by notes. A bit stronger use of the context dependence of the act and the two ends theme to help explain the dynamics may be present — and hopefully a start on a better integration of old themes.
Note II
Some NS Themes that seem important at this re-reading, (Nov. 15, 1977) are: old brain, tropistic excess, anti-repetition devices, the impossibility of repetition, over-individuation of the act and the context dependent act, interaction via the SE, "site” (equivalent situation) replaces "thing", three factor interactions, learning via positive linkages, the self-engagement of the act, learning cycles (as filters), the quasi-repetition of the X, nextness and orientation independence and the causal and the objective, the end points, some end points and averaging, the two parts of the SRS, sensory side linkages, the reticular formation as act linker & the poker chip structure, the GS, learning not CR but added selectivity on CR from their interaction, brain structure needed for functional placement involves statistical factors and successful emergence, control via forms of prediction, - foreshortened (hence no paradox), the completeness of the S-R (R-S) frame & interior perspective.
1.Preamble
The engineering class and outlook, with physicists providing its broadest systematization (and mathematicians generating needed tools) is taking over an ever larger portion of the population at an exponentially expanding rate. This process has reached the point of qualitative change, where the expanding class begins to approach the 1/10th level needed for a takeover, where the exponential increase tapers off, and the new outlook develops into an inclusive or controlling one.
These engineering methods work well and provide a basis for close collaboration, at first only in relatively narrow niches involving the implementation of at least relatively well defined tasks. As techniques and guiding principles improve, the life sphere in which they apply gradually widens. Tasks need be less and less narrowly defined.
Now these engineering problem solving compulsions and procedures, as they build up confidence and momentum, gradually spread out, and attempt an attack on the whole of life. A mathematical calculator is made to reveal its secrets by doing calculations. The mathematical nature of matter (up to a certain level of precision) has been made to reveal its own secrets in this way. Now the brain and society are being made to reveal their secrets too, in the same way. As we make a piece of copper calculate equations for us, and clarify its own rules of calculation, so, too, we make brains reveal themselves.
But in the broadest sphere, where engineering methods are turned loose on problems in politics and sex, they emerge as a compulsion neurosis that people run from in fear. If the decision making heuristics of the broader life sphere are analyzed into their component factors, and latent principles pushed on to consequences, with ruthless persistence, as science now regularly does in more limited applied issues, the result is a disease, an uprooting with no trace of a solid road through. This panic has given rise to many attempts to isolate off the scientific method. But attempts to formalize such ways when themselves probed, turn out to but compound the complications and absurdities. There seems to be no clear way to protect ourselves from the inundation, nor any way to consummate it.
It is our contention here that a relative consummation is possible, and that these methods, seemingly so destructive when let loose on the "personal" or "political" or "artistic" spheres, can be harnessed. This consummation appears to require a very special kind of mathematics involving one person game theory, which the present review paper will be especially concerned with outlining in its broadest aspects. It is helpful to keep in mind the depth of the practical and related psychic forces that drive on or force this formalization of the group and individual decision process. This helps to make a bit more plausible the sweep of certain mathematical themes. Before the inner self-grinding of the problem-solving method let loose upon our daily rules of conduct, everything is dissolved.
Now even the suggestion that engineering methods can be turned loose to formalize the overall decision process, somewhat as it now formalizes parts of the decision process in the sphere of bridge construction, meets with powerful objections claiming it is an impossibility.
Suppose, for example, that one did predict his own decision process, how could this fail to have feedbacks effecting the decision, and how could it hope to incorporate a model of such feedbacks of its own role.
This kind of argument pre-supposes, however, a type of logical relation between the decision process, and the causal elements of the engineering explanation of that decision, which is not necessarily true.
It assumes that the decision process is something defined in a way that is independent of causal insight into the necessity of that decision process.
Now, in a sense, behavior takes care of itself, but the resolving of conflicts in latent rules of behavior does not. Such resolution involves some implicit causal insight to be stable, or successful. Perhaps this "naive" kind of implicit causal insight refines in steps, into the full engineering one. Perhaps, the "predictive" function of causal self-insight is what makes possible decision at all (via the isolation of what we will call the "repeating path"). Perhaps, the laws of matter, or structure of matter, cannot be known in itself, but only in and as it forms part of ways to reconcile rules of choice. Then the causal character of a rule becomes related to a kind of invariance needed for the emergence of stable condition reflex formations, or stable synaptic links in complex statistical interaction. Perhaps causal law should be basically interpreted as laws of emergence of knowledge, and only secondarily as law of matter (known about). (As we will see, physics can be interpreted as in fact encoded as the search "to isolate repeating" paths, or "to decide at all", and so involves knowledge and will.)
This sort of theme suggests avenues for facing into the many seeming paradoxes involved in "predicting" one’s own decisions. But it also involves a radical new epistemological reorientation and related world view. Hopefully, if this can be consummated, it will be adequate to integrate the deep psychic pressures of the engineering method that has slowly let mathematical problem-solving loose on life. Hopefully, a much deeper mutual cooperation and freedom may also, thereby, emerge in a world view which orients the human adventure in relation to the transcendental opening up of the laws of physics themselves.
2. Some Special Aspects of Brain Modeling
a. General
Experience with the kind of mathematical models which were so successful in physics, and science up to now, have very misleading aspects when brought to bear on the special problems involved in brain modeling. For one thing, they are oriented towards understanding situations which have been characterized "as" "organized simplicity” and in the more recent statistical physics, "disorganized complexity," whereas life involves "organized complexity.” There is some kind of shift of gears, or of the basic nature of the model building concern, which occurs as we reach the "organized complexity" level, which is not yet understood. Right answers may be staring us in the face, which we refuse to see (as Eskimo children solve certain mathematical problems easily that our education often prevents us from unraveling).
For one thing, existing models are oriented towards systematizing the deduction process. But behind a deduction process is a question that motivates it, and a more fragile direct inductive approach that can only be partly organized by the deductive machinery. (Poincare says of this deductive level of organization that its chief function is mnemonic, to help recall or invoke the deeper one.)
Could there be some indirect way to model or orient the induction process, as we do deductions And if so, might such a type of model help penetrate the interactions of a "life" type called "organized complexity," as models of deduction help penetrate the areas that are today’s science?
In any case, the kind of thinking involved in penetrating the organized complexity which characterizes brain operation, is masked by a variety of peculiar properties or aspects, some of which will be briefly suggested below. Later we will show how these hints can be integrated in a unified approach.
These "hints" have a deeply orienting character of their own, which will help to understand and coordinate later developments. Hence, they are, perhaps, worth some independent treatment, as a preparation for the more unified and concrete one person game theory approach to be developed later.
b. Why No "Key" New Predictions?
New causal insights must inevitably lead to new predictions. In the context of organized simplicity and disorganized complexity certain of these play a key role in testing new theories. A new insight is confirmed, basically, by the new elaborations that can be squeezed out of it. The specific contradictions that motivate the new causal insight are relatively isolated specialized issues. It is new consequences that test the depth of the resolution, and guide its formulation.
In the brain model context, by contrast, one hardly knows how to begin to begin. Available facts that need integrating are relatively endless, and such hints for or starts at brain models as exist, are obviously totally inadequate on their very surface. The key act here is not to come up with significant new predictions, but merely to show signs of seriously beginning to integrate the main pattern of facts already available. If a suggestion even avoids falling into instant absurdity, when considered as an aspect or property of overall brain functioning, it is sure to end up a powerful guide.
This external difference, corresponds to a much deeper internal one. Brain modeling can make use of a radically different kind of thinking than that in existing physics models, in part because the kind of information available for use in prediction is so much richer, and of so different a type. (We are "inside” this machine.) In addition, the kinds of prediction with which we are in practice concerned is also different, (e. g. behavior takes care of itself, it is only conflicts in rules over how to behave that is our concern).
Brain modeling may turn out to involve the "whole" man, an integration of insights from many (even social) fields, in a sense which would be absurd if applied to modeling of "organized simplicity.” And this "strange" wholeness may further reflect itself in the special way in which brain models emerge, as an almost "moral" struggle to get at the self-evident or to integrate in some harder symbolic way the many facets of practical living. Perhaps the exciting new results derived from experiments on the brain will not so much suggest new specialized devices, as give people the needed guidance to carry through on the consequences of old ideas long present, and tighten them into systematic forms, which, in showing the latent self-completeness of certain ideas, serve to exclude others.
Habits of model building derived from the existing physics are very strong, once learned. They tend to cut off concern for broader issues, in order to implement the cracks we can handle, a modest policy that has paid off fantastically in the past. But perhaps brain modeling is precisely a sphere in which the seemingly useless (long avoided) claptrap of broader issues, that collapses into such disgusting and slimy ambiguity as it approaches practice, is forced into harder forms that can actually take the gaff, and be pushed through to mean something specific at the causal or mechanical level.
c. The Effect of Statistics
At first glance statistical situations seem complicated. But later we see that the statistical aspect introduces a great simplification because it allows us to ignore details, and concentrate instead upon the invariance properties that allow interactions to add up at all (instead of being self-canceling). The formulation of these invariances cuts across the merely local groupings of "organized simplicity" in fresh new ways and brings global properties (relative volumes of phase space) into local relevance (what number we will read).
In addition, the formulation of the dominant or self-emergent forms, may in a sense be "contrary to the laws of physics,” Thus, statistical mechanics tells us that if we have a group of molecules up in one corner of a box, they will spread out uniformly over the box, and assume a close approximation to a class of possibilities that have the characteristics of equilibrium conditions (uniform temperature and pressure). In fact, however, there is a finite chance that the particles will all return to the original corner of the box where they were when we started our hypothetical experiment, and, given enough time, there is a certainty that they will return to that comer.
In what sense, then, are the laws of thermodynamics and statistical mechanics "laws", when we know that, given enough time, things will go back to where they started? Mrs. Ehrenfest and others were much concerned with this issue back in the 20s, and though the "problem” seems to have died, or gone out of vogue as a problem, it has morals to teach that may be of importance in guiding the formulation of brain models.
Experience with organized simplicity leads people to assume that the basic law of brain operation "must not violate the laws of physics.” However, this is not the case. It may very well be that, as with statistical mechanics, the basic overall laws may have "violation,” and yet be overwhelmingly correct in practice. It may, for example, be trivially easy to show cases where these overall laws do not hold, as we can see that the particles must return to the comer of the box. Yet these laws may remain overwhelmingly adequate in practice (we would have no sleepless night if told we will be shot when the particles do so return).
The laws of brain operation must be gotten at by formulating the kind of "invariance" properties that CR configurations have to have to hold up under the complex interactions to which they are subject in the brain. We may expect that the basic laws of brain operation, or "real" learning, are not based on the CR (conditioned reflex) but upon the interaction among CR’s, which interactions have an added selective effect (on CR’s, which go every which way).
Now the formulation of invariance properties may serve to indirectly define an emergency that cannot occur in specific ("artificial") situations. As in thermodynamics, the formulations may pre-suppose properties of the range of possibilities opened in practice, and cease to be true if these ranges are "artificially" closed, (as by "inadequate" time to spread into them, etc.) Given the enormous wealth of cue types, given the great redundancy of inputs from nature, given the variety of roads through to "solution" in such a maze, and the large number of searching probes, emergency of solutions of a certain general type may be so reliable in practice as to allow of the formulation of global orienting laws (like those of thermodynamics) that describe a solution process. Yet, despite this, it may be trivially easy (as in thermodynamics) to set up specific cases that violate these global laws.
Nevertheless the global law may hold in practice, and, because it holds, in effect allow us to introduce what can seem too many conditions. The conditions of "functional non-interference" in brain operation may seem to restrict too narrowly the range of operation of a part being considered, more narrowly than "non-interference" in physics narrows things. But so long as there is a de facto through path, along the sequence defined by the invariance conditions, we can get away with this simplification, as we get away with simplifications defining equilibrium, or limiting us to them.
d. What is the Question?
Statistical mechanics takes advantage of the very special nature of the questions asked in thermodynamics (what is equilibrium?) to introduce certain simplifications adapted to that limited question type. Is there any corresponding limit on the type of question which is asked regarding our behavior, which might serve as a key to simplifying brain models?
In a sense behavior takes care of itself. We are not in fact concerned over behavior, only over resolving conflicts over how to behave, a very different and more limited thing. Of course, one can, in a formal sense, ask for a prediction of his behavior, but this is not an existential concern, merely a formal one. The existential concern has this very deep limitation built into it.
The existential concern, as conflict or contradiction over how to behave, and center of attention, is the basic orienting feature of the "eternal present," the subjective world of immediate apprehension. In a way, it is itself a statistical sort of emergent. The muddiness of the present, as it gradually comes into self-conscious focus, ever recreates this form when one gives it the time. Short of this focus, the present is more nearly phenomena, than life or reason, anyway. One may live along "mechanically” without much explicit awareness of the contradiction as such, until the old framework breaks down. Then one is forced back to the contradiction as center of attention and conscious starting point.
It is a little like the Rosetta stone, in that it stands in three realms of discourse at once, and is the key for translating between them. The conflict or contradiction has a formal aspect, as statements that something is true and false at once, a related external aspect as a conflict over how to behave, and an existential aspect, as the felt center of attention and concern.
"The question" involves conflict over rules defining how to behave, and this limited question type we resolve as the "functional" path that leads to enactment. The struggle to predict merges into the struggle to decide at all, and the struggle to resolve (objective) rules of choice (of act).
This resolution process harbors for the group (if not for the individual) an overwhelming adequacy of latent answers in the richness of logical types searched over, (answers or resolutions that seem to require a model building process). This apparent adequacy (like the apparent dominance of equilibrium) gives to the special narrowing point of view involved, ultimately, an all-inclusive claim.
e. On the Logical Status of Brain Structures
In organized simplicity, and its disorganized multiplication, we conceive of the controlling structures as defined locally. But there is much evidence that basic brain structures are a function of interaction with the environment in a very much deeper sense than might be supposed at first, because such pre structuring as is present, can, for certain purposes, be treated as a genetically guided formation of the type of link that interaction with the environment would tend to force up anyway. Even the continued stability of some of these genetically defined links depends upon this parallel, or upon a reinforcement that derives from interaction with the environment.
Computer analogies, which start with a verbal processing stage that the brain ends in, support compulsion to pre-define far more local structures than may be in fact needed. The word is a parasite on a preverbal model, whose emergence may derive more from special properties of the class of correlations fed into the brain, in the context of internal "repetition” related interactions, than any pre-set mechanisms.
The "usual" computer program can often be thought of as defined in a way that is independent of the properties of the data fed in or acted on. But this is not necessarily the case. "Basic laws" in organized simplicity are locally defined in this independent way, but "basic laws" of brain operation, involving the formulation of guiding invariants, may not be. The deep dependence of brain structure on environment may allow the basic laws of brain operation to depend in essential ways upon properties of the data or correlations fed into the brain. These properties may not only shape local structure, but also be essential to global interactions. (E.g., such things as the existence of an "objective” world of orientation independent linkages, etc. may be essential to the simple formulation of the general purpose heuristic that underlies brain functioning, yet be a property of correlations fed into the brain, rather than some aspect of an independently or internally defined program.) Experience with organized simplicity tends to exclude possibilities of a logical type in which basic structure and global program are not locally pre-given, though the language of values and policy formation abounds in such.
f. The Adequacy of Neural Circuits
Exponentially opening mazes rapidly generate such astronomic possibilities as to jump us outside the sphere of common sense. Neural linkages seem quite special at first. But it has been shown (by McCulloch and Pitts) that any calculation that can be defined mathematically, can be easily set up in neural terms. Thus, the logical problem faced by the model builder is switched, as it were, from "whether or not" neural interactions are adequate, to "which" of the endless variety of ways to set up a given function, is the way the brain uses. The vast richness of logical types of available cues (from internal as well as external sources), the local flexibility (which even allows of all Boolean functions of simple inputs to a nerve if pre-synaptic inhibition is included), and the richness of inter-connection, which links every nerve to every nerve in about two stages, serve to open paths to a new kind of thinking, new at least for the engineer.
g. Separation of the Basic or Functional Aspect of Brain Operation
There is a need for a special type of mathematics to get at the invariants of the CR (conditioned reflex) interaction. Phase space serves to take advantage of the very special type of question asked in thermodynamics ("what are the characteristics of equilibrium?"), to introduce a radical simplification that allows us to separate out or ignore incidental self-canceling variations, and concentrate on aspects which add up.
Similarly, the behavioral emissions controlled by the brain have endless aspects. There is a great need to find some way to separate off those aspects which reflect overall brain operation, from those that derive from more local incidental aspects of that operation.
Separation off of what is essential to the typewriter’s function, from what is incidental, presents a somewhat related type of problem. Many factors affect the print; — pressure, what was previously hit, etc. There are so many ways to act on the typewriter, some ways bend the keys, some jam them, and this leads to further effects on the print, etc. There is only a small core of ways to act on a typewriter that is included in its "proper” operation, yet, even within this there are endless variations (changed pressure changes ink density, etc.) that have no relevance to function, (as well as the wide range of "interferences"). How get at the function-related characteristics of the human repertoire, and separate them off (as we might the typewriter’s function)?
Basic human brain function presents separation problems which are especially acute. The behavioral repertoire has many dimensions, many of which "float,” in the sense that they are susceptible of continuous variation, especially in frequency and intensity dimensions. These floating aspects (as distinguished from those locked in to specific values, as by conditioning) are easily effected by "almost" anything. It is only too easy to pick up influences, study their limits, and get formulas for interesting interaction effects that one hopes may lead somewhere. Similarly, with our typewriter, room temperature, and the previously hit letters, etc., effect the density of ink in the letter E. But is such a recognition a step towards isolating overall function?
How begin to get at the separation of the logical character of "functionality,” or the overall brain operation, from the endless other quasi-irrelevant effects (that often bear some indirect relation to functionality, or the machinery by which it is effected).
A key to this is the recognition that "thought" is concerned with ordering effects, with the "yes-no" of whether or not to do a certain thing next. The sum of a set of "crude" yes-no’s becomes very rapidly refined.
It is these yes-no aspects that involve the global ones, and it is via this distinction that they can be separated off.
Clearly, the brain has many analogue aspects, but if this theme is correct, these can be best oriented as secondary effects that derive from aspects of the machinery of yes-no encoding. They are best treated after the global aspect has been separated out, and in relation to this.
There are a great many qualitative aspects of brain operation that tend to support this theme. Different amplitudes of sensory input tend either to go into distinct channels (as in the auditory cortex) or to be fully compensated out (or a combination of both). Also, the cortex appears to be divided into quasi-independent vertical units of information processing, each responding in its own logical way. A spread in space may be viewed as a spread over a range of logical types, rather than an analogue spread. In addition, the neural circuits are highly non-linear. They are of a flip-flop character that engages all the way, or not at all. The list could be lengthened.
It is in part for such reasons that cybernetics has been so disappointing in its results. Beyond translating old insights into a pseudo-vocabulary of circuits and feedbacks, to show how easily it could be done, there have been little new results. Cybernetics was inspired by the theory of servo-mechanisms developed in World War II. It has so little applicability because the core of brain refinements is a compounding of yes-no's, not a linear or even near linear feedback of analogue factors, (as in servo-mechanisms).
This compound yes-no character of the brain's operation is further reflected in the jump-like character of changes in subjective awareness. There is a great deal of experimental evidence which goes to show that we feel a difference, only in and as it effects the ordering of emission of unit-like behaviors and words. The assertion of act A seems ultimately felt or defined only negatively, with respect to other acts rejected. This, as we shall see, is an opening wedge to an absolute idealism, which views life as a problem solving process in pure mathematics.
This yes-no separation, as a step in articulating a functional separation, involves a degree of "non-interference” in the definition of the parts, which is far greater than that in physics proper. The "functional" channels seem very special. But they still afford means to define and search out wide ranges of possibility, and use vast richness of available cues. Hence, such restriction on the form of the dominant channels may turn out to be overwhelmingly adequate, a form that "we", as a group, never need or do push beyond in our problem-solving. When we as individuals in a sense do push beyond, it is also a step that is outside conscious awareness, an expression of something ultimately incidental (like death).
h. On Locating the Functional
What is functional is something hard or impossible to get at systematically in a closed mathematical sense, but perhaps relatively easy to search out quasi-randomly. It is something very different from its mechanical embodiment. Stories are told about the radar design expert who could not recognize the function of a circuit he had himself designed ten years earlier, though once told the functional key to what it was about, he had no trouble verifying that it could be nothing else. A computer was delivered to Stanford without its handbook, so they tried to figure out how it worked from looking inside, pushing buttons and guessing. "They" were expert computer operators with experience in computer design, but they could not even begin to unravel the secret.
Such examples suggest that an understanding of brain operation might be, similarly, almost in our grasp, with all the pieces understood and in, without our realizing it. A small set of special simple tricks can compound to fantastic results, which seem impenetrable till the tricks, and their relations, are known. To find a way of penetrating and using a maze of logical possibilities (as distinguished from analogue possibilities) will require a radical new type of thinking.
In this connection it may be worth noticing how rapidly a functional configuration reaches a relative uniqueness, and the related isolation of distinct functional channels (even when they involve identical behaviors, e. g. People are often surprised to recognize as "the same,” a word well-known to mean different things in different contexts). Could new logical possibilities be developed orienting the formalism in relation to these quasi-isolated functional niches directly? In the dynamics of emergence of new CR configurations, the role of functional isolation and the role of certain invariance properties of certain correlations, must somehow contribute to the self-emergence of a latent model of the world. How?
But first, can we take these hints or clues and harden in the so vague notion of "functionality” in a causal form, which is also conceptual?
3. The Game Theory Framework (Life as Problem Solving)
Broadly speaking we are going to be viewing life as a problem-solving process in pure mathematics, and interpreting the subjective as "theoretic," (relating this to the theme that we can only feel a difference in and as it effects the ordering of emission of behaviors). The world will be treated as a big computer, with its more digital parts oriented in the brain, and its more analogue parts, largely invoked upon command (we look where we please), organized as the environment.
Can we take this loose point of view and harden it up, so as to individuate workably to the causal level? We will now begin to outline an avenue for doing so. It will involve a different sort of logical way for building a correspondence between causal and conceptual categories than now used in science. In the traditional view a physics formalism relates to the world via a type of observation called measurement.
Instead, we are suggesting that the link be made at the policy level, policy being defined more concretely as rules for how to behave. The link is made via a very special way of analyzing policy into generative components. A mechanical-conceptual parallel can then be traced that allows us to relate ("causally rooted") components of policy to corresponding synapses playing similar logical role in engaging behavior because of the conditions of its activation and what it in turn helps activate.
As a starter along this road, we want to see if we can further define the category "problem," so as to tie it down in an existential way in felt categories that may finally lead to a causal level.
As a starter we may say that every problem exists for us as a conflict over how to behave. This is far from obvious, as our methods of talking about problems are endless, and not, in general, behavior related. But when the categories defining problems become ambiguous we are forced back to behavioral categories. Concepts in mathematics are ultimately defined for us in relation to how word systems can be manipulated, and do not have an independent status, as the controversy over the question, "Do parallel lines meet?" forces us to recognize. The problem of simultaneity at a distance similarly forces us back to operational terms to clarify the conceptual breakdown.
Thus, every problem, insofar as it is a problem "for us," or an existential issue, has its aspect as a conflict over how to behave. This can be made explicit insofar as we need to, that is insofar as the defining categories, in terms of which problems are formulated in foreshortened ways, break down.
It is to be noted that the category problem, as this further articulated, embodies the yes-no separation theme, in that it is conceived in relation to the yes-no of the ordering of word or act-like units. This yes-no quality has a primordial character. Feeling does not start vague or "confused." The young don’t feel "confused,” you have to know a lot to. Awareness or feeling as felt has a clearly structured character, if examined carefully, of which so-called "confusion" is a complex late variety.
The category problem has an existential root. The search for a through or resolving path is also the process of solving the problem. Somehow we must connect problem solving categories with reinforcement type categories in some basic way to carry through on this theme. But how?
4. Game Theory
The concept of a problem can be gotten at more concretely via game theory. Life can be treated as a one- person game, somewhat as equilibrium properties of interacting bodies can be treated via the properties of a related phase space.
a.The "Move”
Basic to the game theory approach is the concept of a move. Moves have certain mathematical properties that closely parallel properties of the nervous system. They are all or nothing units. (You cannot half emit pawn to Queen four.) Somewhat similarly the circuits in the brain are highly non-linear, of a flip-flop-like character, tending to be engaged all the way, or not at all. Game moves mutually exclude. (You can’t emit pawn to Queen four and Knight to Bishop four at once.) Reciprocal enervation, which is so central a feature of the organization of the spinal cord, has counterparts in the central nervous system, (e. g., strong facilitation of the cortex at any point, tends to inhibit other areas ). Game moves are oriented by before-after relation, which is a type of distinction closely related to the basic form of intuitions in mathematics, according to certain mathematicians.
These parallels between aspects of brain operation and the mathematical or conceptual properties of game moves, will be indirectly exploited. There is no simple direct relation involved. It is merely that the logical and mechanical similarities suggest that a game-like framework may be a useful one.
The abstract move-like unit will play a very peculiar logical role in the kind of formalism we are going to be outlining here. It is not directly observable, any more than the hydrogen atom is. The sounds associated with words are no more move-like units, than the yellow flash associated with sodium atoms placed in a flame "is” the sodium atom.
Meaning is, in general, something that can be tied down via a naming process. What is not itself a name of something, can be, roughly speaking, treated as an incomplete symbol whose meaning is tied down via the way it enters into the determination of patterns of named factors. Although the notion of a move is no exception, it has also a status "prior” to the theory of names.
There are two basic channels, variously articulated, via which meaning is tied down. There is the naming process, and there is the effect on order of emission of move-like units. Naming makes use of endless categories. Any abstraction from the "eternal present " that is widely shared and communicable, is fair game. (Generally these are organized in clusters individuated to a certain relative uniqueness, for the purpose in hand.) But these endless categories (colors, shapes, etc.) have to be translated back into a common element, in the context of the value problem. They are there all ultimately defined as abstractions from policy, i.e., defined by the way they effect the ordering of move-like elements.
This gives to “move” a status prior to denotation, or naming, and so a status prior to consistency (in the named abstractions from policy that root the naming process).
Felt categories are going to be defined via their role in making word-like units different from one another, and in generating their orderings. The experienced world can be treated as made "real” (or feelable), in somewhat the same sense that a set of rules for a game are made "real,” by our agreeing to subscribe to them. A wall, for example, is felt as part of generators of a policy of not walking that way. Such policies (and related feelings) involve learning. When policies or motor coordination are broken down, corresponding parts of felt awareness are, too.
The felt qualities in terms of which naming is carried out, are intermediate categories defining policy. They can in principle be further defined by their effect on the ordering of behavior, insofar as we need to resolve a given conflict in the policy they generate. But to try to do this in general would be inconceivable. (Similarly, to try to work problems in mathematics, using the basic axioms directly, would often take more paper than there is matter in the universe, and more time than the age of the universe. In practice, in both mathematics and life, intermediate categories are redefined in terms of the basic ones, only insofar as this is needed to establish or to clarify their properties.)
When the system of named categories breaks down, when we cease to be able to root policy via the naming process in a consistent way, one falls back on the move-like units, as providing a root prior to consistency. "Consistency" is not needed for belief, as belief derives its reality from enactment. And the lack of consistency in the policies generated by felt and named factors does not mean that they provide no grounding. In physics we often work usefully with inconsistent systems, avoiding branches of deduction that cause trouble. There are parts of the organization of named categories, (e.g., those related to the unchangeable character of past history), that hold up, despite lost "belief" in old ways of identifying the theoretic content of the felt in mathematical terms (because of inconsistencies to which the old ways give rise.)
The abstract move may be regarded, then, as a kind of ultimate building block of the felt reality. It is an existential unit, as is an atom, or a steel beam. Physics is a great pile of number dust that can be grouped in endlessly different ways depending on the problem in hand. If our concerns are geological, or bridge building, or, etc., we use correspondingly different groupings. (The move-like grouping will turn out to be optimal in building a causal model of felt categories, as tensile strength, density, etc., are optimal for bridge design (all being ultimately groupings of the number dust of physics, e. g., field strengths as functions of position, etc.)
The category move is not oriented by the space sense. Rather, the space sense derives from the combination properties of moves. Behaviorism and operationalism (insofar as we attempt to develop a metaphysics, or carefully systematize the categories of the decision process as articulated in their terms), are apt to stop at a half-way house that orients moves spacially. This leaves us with all the old conundrums in a modified form. It is not space time, but the exponentially opening tree of possible choices that serves as the ultimate basis of orientation.
b. Context Dependence
The move, or act, as here conceived, involves an essential context dependence (a context that is ultimately defined via relations to other moves). It is not the "abstract" act, but the functional role that counts. The same act in different contexts, serving different functions, is engaged by very different nerve paths. Indeed, the abstract equivalence may never be noticed, till some accident (or joke) draws it forth as a surprise.
This context dependence of the basic orienting unit has in it, even at this early stage, the seed of important simplifications of "explanation" in the learning process. When learning is oriented in relation to the context dependent act, then a degree of separation appears as pre-given (not needing independent explanation). This prevents a premature interaction among rules of choice, until they have been individuated and elaborated in combination a certain distance to the point of contradiction in a "same” context. This pre-given separation (which will depend upon genetically given characteristics of neural connections) is essential to the unified view we are here developing. (If one orients in relation to an abstract act, the underlying simplicity is obscured.)
The notion of "equivalent situations" becomes, then, the basic mechanical one via which change is oriented, rather than "thing" or "(abstract) act.” "Equivalent situation" is, in a sense, much more flexible than "thing" or "act.” It has a flexibility that lends itself to a view of learning as additive. The equivalent situation can be returned to and further qualified, in a way in which "things" cannot be. To learn we must act. In reverse, if we don’t enact certain act types, we don’t reshape related circuits.
The context dependence of the act gives a basic orienting role to "equivalent situations" and simplifies the mechanical-conceptual parallel.
c. The Goal Function
To complete our abstract framework we need to isolate some goal function for the "life game.” This involves a kind of metaphysical tour de force. It must have a foot in the world of mechanical properties yet at the same time be the form under which we experience our own will.
We will try to show that the category "repetition” serves as such an orienting goal function, but this needs a good bit of added explanation. Repetition is here meant in the external institutional sense. Life is a complex cycle. We work to eat and eat to work. Each part of this complex cycle is tied in by relations to other parts which give rise to it, and to which it gives rise. The definition of repetition is "superficial,” and in that lies its power and strength, its ease in application. Malinowski calls behavioral factors as oriented in terms of their conditions of repetition, "institutional elements,” and regards this as the basic concrete isolate of cultural anthropology.
A form of repetition (or life style) in his institutional sense does not last long in time. The repetition has to be continually recreated.
Repetition in this sense is closely related to what some people call functionality. As McCulloch notes, a signal is distinguished from noise only via functionality. Eddington suggests that repetition is also the key to meaning in mathematics and life, and can be used to clarify a variety of basic epistemological issues.
How can we get out of so simple a theme the rich variety of phenomena that the brain manifests? How does it relate to the details of brain operation?
d. Anti-Repetition Devices
Overall brain operation will be based on the selective effect of a condition which we will be calling "repetition.” This repetition is of an "institutional" character, and opposed to most repetitions of a simplistic form that are passing and do not lend themselves to integration into a large whole. Indeed, the brain has a variety of powerful anti-repetition devices. Locally, there is a phenomena of negative induction which inhibits surrounding areas of any area that is strongly facilitated, thus preventing a spread or resonance. There is also the phenomena of inhibitive recurrent collaterals, which turn off output cells after they fire. They serve to extend the muscle firing uniformly over all muscle cells, for example, and prevent the excess use of any part. They help keep us from being stuck in ruts, or building up repetition on incidental features, by constant shifting of responding areas.
Then, there are also global anti-repetition devices. If any closed loop "takes off,” as a result of internal feedbacks, this leads to the activation of suppressor bands, and a general inhibition that knocks out the configuration in question. What does repeat tends to recruit, augmenting with successive repetitions, till it pushes to some extreme that gives rise to regulation by involving other effects. Feedbacks doubtless go in endless circular ways, but these cannot add up to stable recruiting patterns, unless they have a very special logical character.
Broadly speaking, this character depends upon a certain type of relation to the motor pathways, via which the major sources of reinforcement are isolated. But this "repetition" is ultimately of an external character. Circuits that recruit directly into effecting "external" change must be involved. The repetition depends upon a dynamic return into "equivalent situations.” It is an inherently dynamic type of repetition that can build up at all only because of certain remarkable symmetry properties in the class of correlations fed into the brain. Generally speaking, dynamic repetitions (which the anti-repetition devices help filter out) are "impossible" to find. Most "dynamics" is just never twice the same.
e. The "Impossibility" of Repetition
This "impossibility” of repetition has important orienting counterparts on several levels. It means that what slips by must satisfy very specialized conditions whose satisfaction can seem miraculous. (Einstein refers to the miracle that nature is understandable at all.)
Yet in the most commonly articulated overall perspective, repetition, rather than appearing a miracle, which is only with greatest difficulty periodically re-created, appears as a trivial taken for granted starting point. In the perspective of most present day treatment of the value problem, what one seems to have is an excess of repeating possibilities, and a need to choose which form of repetition (from among a smorgasbord of competitive ones) will dominate. The art of life seems the art of smelling out which will win now, and betting on it, if you can get there first.
Fundamentally this view is an illusion that is associated with the middle class role. A centralization of the group decision process is essential to survival. When there are no answers as yet, or rather when there is a set of conflicting answers, someone has to decide which are "right” and which "wrong.” Someone has to suppress certain insights and enhance others (at their expense). The middle class help do this. Leadership has to help fill in the arbitrary for the group, and where theory is inadequate, this arbitrary also involves choosing sides in unresolved issues, to keep things going.
The ability to choose sides at all depends upon a certain position of relative power. It is not available to the lower class role. There, the struggle is not to choose sides but to resolve in new vision. The bottom is involved in dreaming new dreams, the top in implementing old ones.
The above themes need a lot of added explanation. They are briefly suggested here because they have counterparts of a very basic epistemological character. What on its surface appears as a difference in social role, has counterparts at the level of the basic categories in terms of which the decision process is experienced as causally rooted. The middle class has the illusion of life as a one-sided choosing. This illusion finds its way into every corner of written thought, from the foundations of mathematics to psychology and physics.
In formulating an overall law of brain operation, we will find that properties of the correlations fed into the brain play an essential role in the formulation of that law. This seems strange from the perspective of organized simplicity. The way a camera works, the basic laws of its operation, have nothing to do with the character of the image focused. But as we will see in very broad outline, this is not the case in brain operation. Indeed, what seem like external properties of social interactions provide powerful indirect techniques for specifying certain properties of the class of correlates fed to the brain. These "socially" formulated properties are essential to the emergency of basic laws in overall processing that are the content of the laws of brain operation.
In this overall perspective the middle class serves to help act out contradictions by side choosing to avoid social paralysis. As a result they have a one-sided view of the world. They conceive of it as competitive when it is, in fact, cooperative. They conceive of the problem as one of choosing sides, when it is, in fact, one of integrating visions or, more concretely, integrating complementary conditions that serve to help define the repeating path. Choosing sides is not wrong. It is essential to break paralysis and to help generate the needed data via which synthesis is achieved. But it is a function that is subordinated in the larger picture.
The adequacy of repetition as an orienting criteria appears absurd from the point of view of the middle class function. This deep rooted illusion of the seeming absence of a common unifying goal function (repetition) must be dealt with explicitly, if we are to get at the meaning of the theme as here outlined. Of course, so brief a presentation of so central an issue, with so many ramifications, at best only suggests a direction, or approach for later filling in. But presenting a type of mathematical framework is all we aim at doing. If, as we believe, the basic roadblock is the lack of a synthesizing framework, not the lack of data, such an outline may be useful, even when left in so incomplete a state.
f. "Problem" as "Heuristic Contradiction"
In the light of the game theory framework, the category problem, which we saw had existential aspects as conflicts over how to behave, can be further defined as a conflict of heuristics of the common goal function, "search for the repeating path,” in the exponentially opening field of possible choices of act or move. Thus, in chess, the heuristic "avoid double pawns" can collide with "avoid pins,” in a particular game. This leads to a problem to be solved (putting in qualifying variables) because they are both heuristics of a common goal function.
We are saying that the mathematics of life is of this same logical form. Life itself is a one person game, and life problems are ultimately imbedded as conflicts among heuristics of the common goal function.
Now the heuristic concept is related to the drive concept, which serves as a source of selectivity in what is ultimately a genetically defined field of possibles. This field is opened slowly from the "crib," as the success of old solutions pushes us beyond their context of origin, opening out new regions where they no longer hold up, but fall into mutual contradiction leading to further qualification.
The notion of "success" as the source of the breakdown of old forms is a key one that is easily obscured. "Interaction" with the environment is, in this view, not a source of anything. All it does is draw out different latent possibilities from a pre-defined repertoire. This gives rise to the confusion that it is a source of change, (a mistake that Skinner makes very explicit), because there is no basis for separating out significant new development, from mere difference in the manifestation of old themes.
The new development comes not from interaction with the environment, but from interaction with other heuristics which the successful opening of new environments draws out. Interaction with the environment, as such, is too crude to be anything but destructive. Only in the balance of interaction between two heuristics, each successful in some niche, do we get forms open to a potentially progressive qualification.
This opening via success, via the contradictions (X’s) in old solutions, closes the world picture in pure problem-solving terms.
This problem solving view of life is oriented ultimately as an opening of the laws of physics themselves, which serve as the ultimate systematization of our heuristics. A causal orientation can seem to debunk man. If man’s reactions are independently predictable, this would have consequences that are not too ego flattering. But as the conflicts that define the center of attention in adults are a function of the inadequacies in our best available causal models, the outcome cannot also be predicted by our model. Hence, the causal view, in the context of a transcendental opening of the laws of physics, escapes such "debunking” of human effort.
g. Induction and Deduction Space
The theory of the operation of the brain is closely related to the problem of the meaning of meaning. There are two great avenues, each all inclusive, into the rooting of meaning. On the one hand, meaning is rooted via the naming process, in a picture gallery world of qualities, laid out in a space-time-like set of nextness relations. This is what we call deduction space. The rules of deduction allow us to infer certain patterns from others. The categories that enter in are endless, any abstraction that is communicable, or easily invoked in others, is fair game.
By contrast, what we will call induction space, is oriented in relation to the exponentially opening tree of possible moves. All meaning can be defined via the way it effects decision, as all meaning can also be defined via the way it enters into determining named patterns. The latter views the world as built out of things that are the carriers of properties, and interrelated by spacial relations. The former (induction space) views the world as built out of moves, that are interrelated by conditional correlations, or rules of choice (the world being built out of heuristics of the common goal function). Change in deduction space is understood as motion, in induction space as contradiction. In the overall the two views merge as space time and plan, respectively.
Actual problem solving starts in some inconsistent elaboration process in "deduction space." (Wild notes that only the critical faculty is creative, imagination is purely imitative.) This inconsistency forces us to analyze the effect of its defining categories on behavior in order to resolve the inconsistency, i.e., it forces us into the more detailed induction space view. Once the X is resolved, it leads to a modified deduction space. The categories of deduction space are ultimately definable as abstraction from policy. When these break down, we are forced back into explicit policy terms.
The isolates of deduction space have a high degree of orientation independence, and attitude independence. They are organized by nextness relations. Via them, or in their foreshortened terms, endless responses can be indirectly generated, as the numbers representing the "objective” situation (weight, etc.) can enter into the determination of endless specific measurements.
Actual living is, of course, handled only via the foreshortened categories of deduction space. Induction space is only used, indirectly, to resolve an isolated X by clarifying the categories of deduction space. It is this limited indirection that makes the use of induction space possible at all. It could never be used directly to solve a problem, as the procedures would be too clumsy and long.
This view of the deduction or problem-solving process, or this way of analyzing it, is one which will lend itself to a gradual hardening up, to a point where mechanical conceptual parallels can be isolated. We will now show (via a model progressively defined in stages) how we may use this sort of general game theoretic framework to develop a causal brain model.
5. The S-R (R-S) Framework
Facts can be grouped and abstracted any which way. So too can the "number dust” of physics. How one groups is a function of the task in hand. A grouping which is ideal for one purpose, can be totally useless for another. And the purpose need not be "practical," in the superficial sense. What is good for one practical purpose may be no good for another practical purpose, true, but what is good for the "ideal" purpose of isolating basic laws may also be good for almost no practical purpose at all. (For example, putting equations in tensor form can be of crucial help in isolating basic laws, as in general relativity. Yet the solution of most actual problems that involve experimental results, as the advance of the perihelion of mercury, involves going to a very special frame, and dropping the invariant formulation, which would be absurdly clumsy to work with in practice. Similarly, in quantum theory Dirac’s methods (using the commutation relations directly to solve) though marvelous in exposing basic principles, are in practice impossibly clumsy, and, except for certain very special cases, almost impossible to use directly to solve actual problems.)
Now the world can be analyzed in stimulus-response (S-R) type units. The response engages other stimuli, so that such units form a potentially complete system. The concept of stimulus and response are here generalized to include internal as well as external factors. "Response" of a nerve may involve stimulating other nerve cells directly, or rotating the eye via the eye muscle which in turn changes response of eye cells. Such (R-S) coupling of the response to other stimuli can be treated as the "source" of the stimulus (we look where we please).
This kind of a view treats the world as built up of S-R units (with their (R-S) couplings). At the heart of such units is the nerve cell, but the S-R unit may also often incorporate into that cell certain aspects of the environment. How we choose to incorporate response and environment into the cell unit to produce an S-R unit is full of arbitrary aspects. We can "extend" things as much or little as we please. But in addition, the kind of extension implicit in an S-R (R-S) analysis that is all inclusive, involves a sort of fluid lumping that cannot possibly be written down directly. (Looking yields different couplings in different rooms.) What, then, is the point of invoking a type of decomposition which is totally useless and undefinable in practice?
The reason for this derives from the very special character of the concerns for and perspective on human computers which we have, from inside, that we do not have from outside for, e. g., ordinary computers, or radios. Our concern, insofar as it is self-conscious, is not deciding what to do, but resolving conflicts that impede this process. Mere doing is an automatic consequence of former learning. Our concern is for the learning process. Insofar as we define "being" as the stuff of subjectivity, or as involving this, then the learning process is also the process of creation itself (as later examples will help make clear).
Now all learning may be ultimately oriented as synapse change. If all other social change can be treated as flowing "automatically" from learned synapse change, then the S-R (R-S) framework is a potentially all-inclusive one. However, in practice our understanding of learning does not seem at all oriented in terms easily translated into synapse change, so that such an assertion can seem artificial and useless. But, perhaps, this is due to the limited character of most analysis of learning (far short of the step by step Skinner box analysis).
In addition, in the more limited perspective of "learning from others" ideas seem to enter in from outside, relatively pre-formed, and this is even more the case in recall of learned material. Such a perspective, far short of creation, does not involve analysis back into the more detailed S-R factors, and the deep, indeed absolute, fluidity of the felt, as something built up, ex nihilo, is totally obscured. But when we are concerned with the full detail of first creation of an idea, then, perhaps, a full Skinner box type analysis (with natural counterparts at the level of synaptic change)may be needed and relevant.
In building overall causal brain models it is this perspective of creation (not mere local or "relative" prediction, relative to given social forms) that is the essential orienting one. The world is treated as a big computer, with its more analogue aspects in the environment, and its more digital aspects concentrated in the brain. Learning involves certain special changes in (S. -R) linkage, at synapses in the brain. In our need to position this, we, who are inside our computers, have available as given most (not all) of what people outside are concerned to predict. Our limited inside task is to reconcile predictions, not predict. Hence, S-R (R-S) linkages are to insiders simple in practice, as they, in effect, represent themselves. The limited aspects of them that we are concerned with, needed to orient the generating principles of brain operation (implicit in reconciliation procedures) are easily available. Most details of the way the generating machinery selects specific acts takes care of itself. Behavior takes care of itself.
The S-R framework can orient the process of synapse selection which emerges in CR interaction, because only very limited aspects of this largely inaccessible decomposition are relevant from inside. But the fact that these aspects are orienting has a deep latent content. It reflects the self-contained view of the world that emerges in relation to creation. "Everything" that bears on a given X is internalized in process of solution (that "everything" being quite limited). In the solution process, creation appears internal to the self and can be oriented as CR selection or synapse placement, the building of a special type of correlations (CR’s) embodying resolving causal insight. This placement is latently objective to us, because of the need to build up synapses in person independent and orientation independent ways, which leads to the generalized search form — points to be discussed later (sec. 11). In addition, the adequacy of the S-R frame reflects our ability to trace out the resolution process to a relatively causal level (where emergence of resolving correlations occurs).
Meaning depends on correlation. Those that are "causal" relative to a given range of variation and data are those that cannot be further improved by adding available qualifying data. The causal laws can also be viewed as laws or conditions of CR emergence or of CR stability under complex interactions. This theme will take on much deeper content later as brain models are developed in a concrete way. Causal laws are encoded as systematizations of heuristics, enacted by us. In this way they can be viewed as built up in CR terms out of a wealth of latent S-R (R-S) material.
The potential symbolic completeness of the individual standing alone (with the group (insofar as he can know it) as his own outer projection, and means of acting out his personal X) is reflected in the one person game theme. (The individual relates to the world by imitative coupling to sub-aspects of the self). What resolves for the individual easily carries the group. (It is much harder to convince yourself than the group.) This interior self-completeness is reflected in the closed self-contained nature of the S-R view. It is success, or self-contradiction that generates openings. (Interaction with the outside just draws out various latent implications of already established insights, not lasting change of a progressive character.) Hence, CR change is a category in terms of which all creation (insofar as felt) can be traced. The S-R (R-S) decomposition emerges, as a result of such themes, as self-complete practical and orienting.
6. The "Old Brain"
There is a conjecture regarding brain functioning which we will call the "old brain.” We will assume that it acts directly on the cells themselves (rather than the synapse). If firing is correlated with area reinforcement, then responsiveness is enhanced and the related motor responses, or tropisms in which the cell participates, are selectively strengthened. In reverse, a correlation with area inhibition, leads to reduced response, and even degeneration.
There is a crude ambiguity and total inadequacy of these themes in contexts that involve anything like spacial orientation or "finding" and "returning.” In addition, any given effect of this type can be interpreted as involving conditioning of a "new brain" type, hence it is hard to infer from given data. (The best argument derives from its simplification of the overall scheme of explanation.) From all these reasons the possibility of such effects in animal brains is largely ignored.
Yet when one deals with the repertoire of, say, worms, the limited learning of which they are capable falls easily enough into such "old brain" themes. And, even in the human brain there appears to be present effects of this type, though there they can also be easily interpreted in a more complex way. If correlations are sufficiently crude, there may be, in this view, direct counterparts at the cell level.
Signals not used, or regularly inhibited, cease to be noticeable, and related cell types even degenerate. Animals raised in an environment of horizontal stripes (no verticals) have lots of cells that respond to horizontals, few to verticals, as adults. Although conditioning is also involved, there appears to be a cell as well as synapse effect.
This "old brain" theme, although hopelessly inadequate, may nevertheless play an important role. It may help provide a primary shaping of cell types and response levels, not only in the overall, but also in small sub-niches that higher forms of synapse conditioning and their interaction give rise to. These old brain themes fall into ambiguity or mutual contradiction very quickly, but it is precisely within this area of contradiction in old brain themes that the new brain would, in this view, take hold. Vast though this area of contradiction is, it represents a significant narrowing which makes the work of the new brain, in qualifying this area of contradiction, more manageable.
Any ’’learning, ’’ so-called, connected with the old brain has no subjective counterpart.
The old brain shapes the genetically defined set of "worm” like tropisms in which the cells effected participate. It involves genetically defined drives that activate certain areas and then provide a source of selectivity. Thus, we have "beginnings” of learning. But what we call learning (which includes all learning with any trace of consciousness) involves synapse learning and the tying together of formerly activated tropisms in ways that reflect individual past history, and their internal interaction via these interlinkages.
7. The "New Brain"
a. General
The "new brain" is the synapse brain. When, in certain areas, the firing of the axons of one cell is correlated with the firing of another cell it comes into near contact with, and if this correlation is itself correlated with a local positive reinforcement, then some change occurs (whether a growth, or chemical change, or both is irrelevant to the overall function) such that the firing of the given axon leads to the facilitation and a more probable firing of the other cell. This generalized conditioned reflex principle begins a translation of concepts into mechanical categories.
There is evidence inhibition usually goes via specialized inhibitor cells, and specialized anti-drive centers, so that all learning remains positive (the inhibitory links being genetically laid down).
The synapse allows of a whole new class of interactions. It gives rise to interlinkages that are a function of very specialized aspects of past history, which, in turn, interact with each other. We have now to start developing, in stages, methods of analyzing these interactions, which, on their surface, present such a maze or jungle of seemingly impenetrable possibility. These are shaped by the genetic pre-structuring of connections, the many sources of drive-like selectivity as well as past history. The Chain Model, etc.
The kind of computers that man builds are what might be called series computers. They can carry out a very large number of steps per second (in the millions) each of which steps is determined by a relatively small number of factors. If there is an error anywhere in what may be billions of steps, the final answer is apt to be totally absurd. (Although some self-correction can be built in, it is relatively secondary.)
The human computer, by contrast, is a "parallel” computer (a phrase of Von Neuman’s). A step involves of the order of a second in time, so that only a relatively small number are carried out, but a large number of factors compounded in parallel serve to shape each step. This parallel aspect of the human computer allows for massive built-in redundancy, and self-correction as you go.
Parallel computers should tend to be optimal for picking up new correlations of basic resolving significance. Series computers should be more optimal for deriving the implications of already established correlations or principles. The first might be said to be more nearly optimal for induction, the latter for deduction, but the line is not absolute.
Now the brain is a picker-up of correlations and can do so every which way. To a crude first approximation all parts of the brain are said to be interconnected, and most cells reach to most other cells in two stages. There are just endless latent correlations to pick up. Is there any class of these that have a special orienting significance?
The motor pathways play a central organizing role in the nervous system. They define the channel that has a potential of entering into complex forms of repetition. To enter into such repetition, a convergence of factors is needed which must integrate in act-related terms to hold up. Characteristic rhythms related to the present act are widely distributed and help in this integration.
Because of the central role of the motor pathway, correlations that are directly act-related are especially important. One may view the brain as serving to tie together past acts of types that have had reinforcement somewhere. Once reinforced, an act has a quasi-independent life of its own, with a small secondary control of reinforcement. As a result, if act B follows act A, then when A is facilitated, so is B. The fact that B followed A in the past, makes it more likely that it will do so in future. Although the original occurrence of B involved contextual factors, the latter facilitation of B via A occurs in a way independent of this. Thus sensory factors in effect take certain motor links into transference, when the motor correlation goes via them.
This "chain model" of the brain as a linker of past reinforced context dependent acts involves a paralleling action. If in the chain of acts, B Q D and E all follow A, then they are in effect brought into parallel or excited in parallel by A.
Now the basic operation of the brain as a parallel computer may be regarded as the establishment of a relative dominance as among some sub-set of a set of factors brought into parallel. If any other than the dominant act of the sub-set are engaged, they will engage the dominant act, which, in turn, inhibits out the other acts. The sub-set has something of the character of a quasi-random (external) search (RS) for the relatively dominant act which serves to stop that search, and establish a, at least temporarily, stable relative dominance (RD). These notions are, of course, merely broadly functional forms of a certain type that need a lot of filling in to reach a causal level. They serve to point out, however, a type of link, and a type of selectivity, in whose terms we will be analyzing the brain’s overall operation.
The notion of an RS will be related more concretely to the notion of a drive. What satisfies the drive, stops the RS and determines the RD.
The notion of an external RS has internal counterparts. We can call the series of acts, following a given act, which that act facilitates before its own emission (roughly the set of things we expect to do next) the series elaboration (SE) of that act. As RD's are established by interaction with the environment, the SE becomes longer, and incorporates a wider range of possibility.
Now the SE of a given act can give rise to a modified RS or drive, such that the act one was about to start in on is no longer the RD, and this before the emission of that act. When this occurs one has the beginnings of an internal RS, which harbors the embryo of what might be called thought. The internal RS continues until it finds an SE that does not undercut itself before the emission of the act. Once this is found a steady state (SS) results, and an internal RD is established. The internal RD is a function of the SS and leads to the emission of an act. The external RD is a function of what stops the drive and leads to an end of acts in that RS.
Now the notion of an RS and RD are idealizations, like the two-body problem in physics. In fact, drives (or RS) interact. An RD relative to one drive may break up RD relative to another. Thus, the RD gives rise to a logical feedback that reshapes other RS as well as its own. Interaction between RD can be itself treated as a type of compound RS.
The RS-RD theme is a very flexible one, into which a rich variety of compound interactions can be analyzed, as we analyze many body problems into a compound of two body problems. There is, in a sense, only a two body physics, "everything” being treated a compounding of two body interactions. So, too, though there is no ideal isolated RS and RD, we will be treating the whole brain operation in their terms. Of course, this is a very flexible form. However, even at this early stage, there are already certain basic properties of the nervous system that tend to confirm the central role of these themes or functional forms in the patterning of its organization.
The central role of the RS-RD theme is seen, for example, in the properties of "caught in stop.” A rat can be trained to push a button to escape stimulation in a central part of the nervous system that is experienced as very painful. As the source of this basic RS has no connection with the stability of the organism, one might expect that the RD to which it gives rise would not be such as to permit survival, if the RS-RD theme is, indeed, central to the total organization of the nervous system as here suggested. This is, indeed, the case. Animals "caught in stop" for even a few hours undergo irreversible rewiring that leads to death. On the other hand, animals "caught in start,” (that press buttons to receive pleasurable stimulation) may keep this up indefinitely, so long as time out is given to eat, with no ill effects. Pleasure, as an arrest of RS, only confirms existing connections.
The centrality of the SE theme is seen in certain paradoxical experiments. In general, the SE associated with a given drive, leads to the vision of its satisfaction, which reduces the drive to the minimum necessary to activate the needed act. The baby cries and then is fed. Next time it cries less, provided it is still fed.
The effect of inhibition in operant conditioning is superficially inconsistent. It intensifies the behavior at first, rather than inhibiting it, as might be expected from the point of view of classical conditioning. This can be explained, in our model, as due to the fact that the drive is reduced by the SE. The first effect of the inhibition is to break down the SE that reduces the drive, thus (paradoxically) upping it. (In classical conditioning there are no SE effects, because the linkages are more direct. The autonomic regulation does not have the double layer system that is present on the motor side, allowing internal SE.)
A much more extended discussion of this "Functional Model of the Nervous System” is found in the author’s paper with the above title (1963).
The drive concept, as here further articulated, may be treated as a pre-engaged act of the type that serves to stop the drive. The drive is engaged somewhat as is any other act, and itself serves to engage other acts both in the general sense of activating act types, and in the sense of activating those specific acts that lead into acts which satisfy the drive or lead to the “pre-engaged” drive stopping acts.
Drives, like acts, mutually exclude. What engages the drive via SE, becomes a source in its own right of selectivity, via the drive it controls.
Drives rapidly come into interaction via the SE and the many ways in which they are posited on environmental cues. There is a degree of isolation of the RD via context dependence, so that SE is needed to draw out extensive interaction. Also, except in very simple situations a factor that represents the drive is an essential part of this context dependent RD separation.
The interaction of drives drawn out by the SE, forces the encoding into certain special forms invariant enough to hold up. It thereby slowly gives rise to a unified view from within which the various genetically defined drives may be treated as heuristics of a common goal function, and as special cases of a generalized notion of heuristic "contradiction.” Let us now see how this occurs.
8. The X as Generalized Drive
The drive theme is not directly related to repetition. In fact, the drives usually do contribute to homeostasis or survival. But homeostasis is a category that is clear only where there is no argument. We all agree it is best not to die of starvation or heat. But, especially in contexts where there is an argument, behavior seems often such as to upset homeostasis and prevent survival. Some way has to be found to take the ambiguity out of these terms, so as to penetrate the region of conflict.
The key to doing this is the category repetition. To a first approximation it parallels what is meant by homeostasis, in regions which do not involve conflict. But where conflict emerges, it allows us to penetrate its structure. The logical form of conflict derives from the elaboration of what may be regarded as heuristics of repeating paths. Conflict derives from the elaboration of heuristics, (which occurs when they succeed), beyond the contexts of their origin into regions where their latent over-generalizations lead to contradiction (X). The elaboration process even leads to the self-destruction of the repetition, something quite contrary to the homeostasis theme. New forms of repetition often are formed in the collision, but not always. There are many dead ends in nature, and in the personal perspective, though, perhaps, not in the overall.
The isolated drives do not cry out their latent ideality which allows us to interpret them within the framework of a pure mathematical game theory as heuristics of the repeating path. The thirst drive, or the mechanisms of temperature regulation, have many incidental features which make "nonsense” of the heuristic if it is drawn outside a limited context of origin.
But outside this limited context, the selectivities imposed by the many drives are in conflict with one another. The drives, in effect, mutually suspend each other. It is this mutual suspension that allows room for further qualification, and in so doing preserves their latent ideality. As we will see in more detail as we go along, it is the category repetition which enters explicitly into this gap, where the drive factors collide, as the logical form of the source of resolution. The elaboration of the drives, by SE, and their related control by endless environmental cues, breaks down any kind of CR link, except those with very special logical invariance properties.
Via the forces that push or winnow the synapse linkages into this hard-to-find invariant form (to be discussed in more detail later) the CR principle takes on a far deeper meaning. The process of reconciliation of the many CR’s with one another (as we will see) relates an objective consistency in the defining of the repeating path, to the conditions of CR stability in its act-selecting and related information processing role.
As a result of these interactions, "the contradiction,” existentially defined as center of attention, emerges as a generalized representative of all the drives. It takes one form here, and another form there, but is everywhere interpretable as a contradiction among overgeneralized heuristics of the common goal function. The process of reconciling drives emerges as the process of filling in a predictive ambiguity as regards what will happen. The X becomes an objective representation of the drive.
(For example, an itch may engage a variety of different scratching actions till we find the "right" spot to scratch, then we keep repeating the same one. The unsuccessful scratches can be viewed as wrong guesses or wrong predictions, that contradict each other, regarding how to alleviate the itch, etc. In addition the X may be oriented in many different ways. It always has factual counterparts but they may derive from rules of extension at a meta-level or at a basic descriptive level, etc. For example, the meta-theme that being continues to exist, despite "theories,” leads to contradiction with the collapse theme in general relativity. The meta-theme that when two simple things are identical this should be derivable from the theory, helped motivate general relativity (as inertial and gravitational mass were identical for no reason in earlier theories). The X is rooted as X in rules of extension in endless ways — most of which lead "nowhere,” or merely motivate routine filling in.)
The process of alternation that the X represents, involves successive relative inhibition of the competitive branches by one another. This alternation is effective in rejecting the branches involved if a through path is found that resolves the X. Thus, the inhibition is, in effect, only relative to the through path that resolves and so is reinforced.
Characteristics of what generally stops the objectivized X are strongly correlated to primitive drive stopping, too, and so to reinforcement. Characteristics of the X itself get tied to a relative inhibition. In this way repetition slowly subordinates and "uses" the genetically defined drives.
At first there appear to be two sources of reinforcement, repetition and satisfying the drive functions, and two sources of inhibition, contradiction or the mutual inhibition that derives from exclusion or an SE that keeps undercutting itself before emission, and drive related inhibitions as in activation of a drive without satisfying it. As a result of SE and CR interaction effects, the former themes subordinate the latter, making the latter special genetically shaped mechanisms carry out the former themes in foreshortened form.
X'S, like drives, are highly individual and can be individually conditioned. X’s function as narrower source terms, and sources of signal, much as do drives (which can be viewed as special genetically defined cases of the generalized X.) The dominant terror is the inability to decide. As generalized X it subordinates other fears as special aspects of itself.
The X has peculiar properties which make of this concept a central unifying theme via which to attack the problem of emergence of the new in a systematic causal manner. In terms of the X we can say something about the way in which the new emerges, in the light of which a powerful attack upon the value problem can be mounted. The new grows out of quite specific X's in old rules of choice. An X involves a situation where the SE keeps undercutting itself before emission. It is (only) in this kind of situation that a significant or lasting change has a chance of emerging.
The X is the source of the new in a very literal and objective way. All regulation depends upon specific X. The CR principle has no internal source of regulation. When it serves to link Act A to B it tends to make it come faster, harder, and more often after A until some specific source of X gives rise to a counter factor. For example, we may ring a bell of a certain pitch, and use it to condition a salivary response. The generalization of this response that is present when first learned (a response to higher and lower pitches, etc.) is not narrowed by further reinforced repetition of the same pitched bell. It is only narrowed by ringing a bell of higher or lower pitch and not following it by reinforcement, i.e., it is only narrowed by explicit X.
All regulation depends upon such specific X. Consequently emergence of refinements can be followed in a step by step fashion, each step being related to specific X’s.
In addition, the X itself may be treated as generating the field of possibles associated with compound RD. Here Murphy’s law is deeply orienting. "If it can go wrong it will go wrong." What is over general, will gives rise to X’s that generate a field of possibles within which a repeating path lies, that resolves. Repetition is hard to find, and is found only in "invariant " forms closely related to what is called "Truth,” (Kierkegaard). This role of the X as itself the generator of the field of possibles is a deeply unifying recognition. There is no stable repetition until there is resolution.
The through path, that reaches emission, incorporates more conditions in earlier solutions. These more complex conditions are then used to condition simpler ones (not so simple as the earlier start), by a foreshortening process, using the more complex form of control to give rise to new simple coincidence effects.
The branches in an X all grow up in distinct contexts that are kept isolated by their context dependence until SE draws out the latent X. The resolving path too has context dependent features that are reinforced with it. It is partly for this reason that learning turns out to be additive. The through path is a way to decide at all in the face of the X. To reinforce that through branch is to draw in a further enlarged context. Then the successful elaboration of the new qualifications lead to new X’s.
The overall theory via which heuristics are resolved, and its implied simples, jump about. But the concrete techniques themselves expand in a continuous additive way as an expansion of the context needed to coordinate these techniques.
We have now to consider what this additive integrating context is, and how it builds up out of the few "invariant” CR's that allow of a stable passage from stimulus and drive to response.
9. Invariants of the Compounding Process
We have now to try to get at some key orienting characteristics of the compounding process, so as to be able to get a hold on this jungle of interacting possibility. The game theory framework, with its categories of contradiction and context dependence, RS, RD, SE, etc., provide an orienting functional framework in terms of which to build an attack. With respect to an unconfined field of possibles, it has many powerful confining aspects. But the forms are still abstract and functional with a great deal of ambiguity as regards the how (and whether or not) we can fill in the concrete details.
a. Context Dependence and Uniqueness
Through channels from stimulus to response are few and hard to find. The acts, or "moves,” defined more concretely as that compounding of factors needed to activate the given act, do not interact with one another. They are all too well separated for the most part, and such interaction as occurs is of a trivial nature, a forgetting, or not having bothered to learn, which quickly eliminates itself when it does momentarily raise its head here and there.
The interaction occurs in the processes by which these move-like function-related units are interlinked. Stability of memory traces is, in certain theories, related to an individuation of the correlations involved to uniqueness. Once they reach a unique level they have a kind of immunity against being over-ridden.
Uniqueness isolates the response. Whence comes this isolating uniqueness? The linkage of the move-like units gives rise to X's, and this gives rise, in resolution, to added factors in the compounding of the through path. Earlier simpler forms are now drawn into parallel by SE and mutually suspend, till a through path is found. Added inhibitions are then incorporated in finding of a more limited through path. The resolution process, as among the rules linking acts, thereby forces upon the moves themselves an excess of isolating uniqueness.
This successful isolation of the move, and its related context dependence, implies a rich pre-given complexity of compounding which is, in effect, winnowed by the X. In a sense early simpler forms mask a self-concealing complexity. The many competitive moves keep channels relatively narrow from the start. The pre-given separations, and the relative uniqueness of the many factors, allow of certain special characteristics in their compounding which we will be getting into later.
b. The Genetically Defined Field of Possibles
Inputs have a latent tropistic bias. Eyes move towards higher spots and along lines. Touch engages related movement, etc. This tropistic content is very permanent in lower animal forms. In higher forms, however, where there are "competitive tropisms,” the word is no longer used. An excess of tropistic response among factors that mutually collide, has what might still be called its tropistic aspect but obscured. The "competitive tropisms" mutually suspend each other, so that there is no "tropism" in the usual simple sense, or only weak remnants, that come to light only under special circumstances, in children, or near sleep, etc.
In part, this extended tropism theme is a matter of definition. But the notion of a direct link of the sensory to the act, so that a context dependence is taken up in the act from the first, remains a useful orienting one. These may be linked up and winnowed over, and then their compounded linkages further winnowed over. Thus, the initial tropisms can be viewed as a field of context dependent possibles that interaction with the environment compounds and selects over.
(As sensory cues have a motor bias, so, too, the place of origin of the cell dendrites determines the place where the axons will end, as experiments with transplanted limbs show.)
The sensory (or place of origin) is never neutral, but involved in a motor bias from the first. This helps to complete the world picture in move-like units, and indicates how literally and simply the unification pushes through in these terms. Of course, at this stage, the point is in part a matter of definition. The sensory can be treated as independent, and "conditioning" the motor response, so that one has, from the start two types of categories. But the taking over of motor factors by sensory cues can also be treated as a motor linkage, followed by a foreshortening that drops out the need for an explicit undergoing of certain latent motor factors. As there is already a logical operation of the "take over" form definable in purely motor terms, the need for two types of orienting categories is less compelling, and their introduction may serve to hide the simplicity and sweep on power of a latent unifying view.
c. Logical Spread
Different input intensities go into distinct channels insofar as they carry a meaning. Output differences in intensity which are of importance are locked in by specific X, the resolving intensity level being stabilized by a set of yes-no feedbacks from cues responding to "too much” and "too little.” Of course, intensity differences effect floating aspects of response not locked in. But these are secondary effects, that depend upon local details of the encoding mechanism, rather than global interactions.
The "spread" of a given signal is basically a spread over ranges of logical possibility. Different vertical groupings in the cortex, different layers within these groupings, different control areas, etc., all manifesting forms of exclusion that can be activated by signal types available there.
d. The Positive CR Link
Conditioning seems to take a positive character. Inhibition involves special cells, and anti-drives have their own areas. Inhibitions appear to be genetically laid down, as the machinery of a pre-given structuring of the field of possibles.
Thus, the inhibition of act A in a given context K goes via the assertion of B that suspends it and may be itself suspended by A. Then, some other factor enters that further qualifies K, and uses the old engagements of A and B, rather than A and B directly, to specially facilitate one in a way that now excludes the other in a more qualified way. Inhibition only goes via other positive alternatives, with their related associations (at least in the conscious or general case).
The positive CR link implies that changes take place in different places when facilitating and inhibiting. This distinct area theme fits with an additive view of development. As counter-factors don’t go in the same area, then can't "in effect" just wash out the old.
The inhibitions appear more labile than the positive links, as Pavlov notes, in part because they serve a separation function that is stabilized only relative to a positive through path. The learned content is the positive, and the negative is a shifting implication. Positive links can be subordinated, but not eliminated.
e. Success Generates the X
Basically, the X is ever regenerated by success as the success opens into new contexts from a standard crib position. This opening core of positive resolved material, is "held in,” as it were, by a generalized fear of the unknown. This fear confines, but the core remains positively organized about believed themes. The fear covers mere messiness for the most part, but hidden in that is the small sliver of the new. The child opens an internalized field of possibles (built out of the more general genetically defined one) "adiabatically" from the crib, as far as she can integrate, and a bit further. What can’t be integrated has to be rejected or avoided. This gives an existence versus non-existence character to an overall view of issues.
f. Sensory Side Linkage
Motor factors appear to be linked together via related sensory cues, without need of feedbacks. The scratch reflex and the swallowing sequence, etc., as well as other well established sequences, appear to hold up after related motor afferents are cut. The number of mechanical stages is only the number of the logical stages.
The motor links appear in a two-layer form. A secondary motor area close to the temporal side appears related to thinking about or planning acts, which in turn engage the actual acts. The two layered motor regulation is to be contrasted to the one-layer automatic regulation. Internal SE depends on this two-layer arrangement. (There is also a third motor area possibly related to long terms sets, or purpose, which is close to the limbic system.)
Sensory links are separated on the basis of functional role, and are only open to requalification when we return to a same or equivalent situation. The sensory links engage ahead of time circuits that represent what is to follow, which is released by crude cues related to timing. But any approach to the actual machinery of this requires other kinds of considerations not yet developed, to harden in these still too general themes.
g. Three Factor Analysis
The interaction at the cell level involves an input and output and a factor that takes the output over, as a result of a certain past history. The take-over can bring into parallel factors that were formally sequential, resulting in a compound form. A bringing into parallel by resolving the X is the form of all learning.
Generators of the X come in pairs that are tied down in context dependent ways in the resolution process. Inherent in a bringing into parallel process are new forms of responsiveness, whose "meaning” can be clarified by going back to the X’s out of whose resolution they emerged, and tracing out the three-factor analysis, or take-overs that underlie it (as Skinner begins to do in his learning box method).
10. The Mechanical Conceptual Parallel
a. General
Synapse placement has to be oriented via a mechanical-conceptual parallel. It is only in this indirect way that a simple placement procedure can be coordinated. This is an all-inclusive orienting theme. If part of the synapses did not fit such a theme, it might be hard to see how they could be integrated with those that did. Or rather, insofar as the linkages might seem not to fit into a mechanical conceptual parallel, then the phenomena is detached and subordinated via a model of itself, so that it fits in after all.
For example, rage over one issue can easily transfer itself to another where it is not appropriate. This transference is not within the mechanical conceptual parallel. But this leads to confusion which leads to a model of the transference mechanism, and counter compensation, taking into account the limitations on concept due to its embodiment. This, in effect, returns us to a treatment where the mechanical conceptual parallel orients.
The mechanical conceptual (M-C) parallel involves the theme that, insofar as the configuration does not function as pure concept, it gives rise to X's which generate the needed qualifications to re-subordinate to a conceptual role when needed. Programming computers to simulate complex human behavior (as in chess playing machines), we find that the method of encoding the computers closely parallels the human explanation of her own thinking. The "push” to a near conceptual form is already seen at the multi-purpose computer level.
We have already seen a variety of applications of the M-C parallel theme, to orient basic properties of neural interactions, or provide unified techniques of positioning significant synapse changes. Most primitive and basic of these is perhaps the relation between meaning and correlation. All meaning depends on correlation. Where there is no correlation, there is no meaning. The CR principle may be regarded as a mechanical reflection of this conceptual theme, rather than a special kind of limiting mechanism. The brain may be most broadly viewed as a picker-up of correlations, that minimal isolate implicit in all meaning, or the very possibility of meaning.
Another basic M-C theme is that all learning, or regulation goes via the X. The theme is essential to the self-completeness of the MC parallel. It traces the origin of conceptual changes back to concept (via the X), rather than to something outside concept. It also begins the broader theme that the elements that enter into concepts have to be analyzed and oriented, via the dynamics of their emergence in successive X’s. Every concept has associated with it the X’s it serves to help resolve, which are specialized foreshortened forms of its own origin.
The game theoretic formulation begins the interpretation of the drive, in relation to the mathematics of heuristics of a common goal function. The drive's activation component is related to the X's the heuristic resolves, and what stops the drive to the genetically pre-encoded root heuristic.
The X is a kind of Rosetta stone, linking three distinct bodies of syntax, as it is at once syntactic, behavioral, and felt. As existential, it provides the mechanical entry point for the category "problem.” It also serves as a unified representation of all the drives, or gradually emerges in such a role, with maturity. The inhibition or negation associated with drives becomes related to mutual suspension in the X, and facilitation or affirmation to X resolution. What at first seemed two, emerges as one (as inertial and gravitational mass, or gravitation field and acceleration , emerge as essentially the same in relativity).
To push on the MC parallel we need the working recognition that repetition is essentially "impossible," (even in relation to equivalent situation). It can be built up at all, only via the most elementary links, which have very special invariance properties, step by step.
b. Orientation Independence
What kind of invariance properties can a correlation or related CR have, which might contribute to its chance of survival in a complex CR interaction? We leave to one side, for the moment, just how this invariance is used, we just look for any invariance that might be used to help compound a stable through path from stimulus type categories to response categories.
One central type of invariance is orientation independence. There is considerable indication that response is built up in orientation independent ways. For example, if a goat is trained to lift its left front leg off a yellow spot when it hears a buzzer to avoid a shock, one might expect that what would be learned would be a link between the buzzer and a lifted front leg. In fact, what is learned is the orientation independent avoidance of the yellow spot. If another leg is put on the yellow spot, that other leg is lifted without further training when the buzzer is sounded. Upside down with its head on the yellow spot, the goat extends its legs and lefts its head. There is also evidence that the EEG associated with observation of the same object in different locations in the visual field remains much the same.
How can conditioning be built up in orientation independent ways? Are there any CR links that have these properties, from the first, to build on? Or do we need fancy internal machinery to achieve this orientation independence (by representing an object by the whole group of its related transformations)?
The key isolate here is nextness. The nextness of a color to another holds up in a wide range of orientations, till the two colors cease to be seen. (Nextness is also the only local isolate of point set topology, the only basic local deformation independent link.) Nextness type isolates have also a very close relation to what is called "objectivity.” A correlation rooted in nextness has a person-independent quality and comes as part of related sets, of a type known as things in space. Now objective links have a near or relatively "causal" status also. "Things staying where they were" is a start on Newton’s first law. The nextness theme is a good start on isolating causal invariants, too, or rather opens naturally into a more general theme of the self-emergence of causal configurations.
The Somatotopic layered organization of the human computer is optimal for the central role that nextness plays. Visual and auditory and somatotopic nextness is converted in the nervous system into mechanical nextness, and related nextness types summed to remove in part position dependence. The close parallels between somatotopic and visual or auditory nextness is strongly made use of. Children blind at birth, who come to see after they can talk as a result of cataract removal, first speak of discrimination in the visual field as if it were a somatotopic discrimination in sensitive skin areas.
And children’s art first experiences the world as of a set of qualitative blop organized by nextness relations. Perspective is a late sophistication that only half emerges out of a refined compounding of nextness factors.
Interaction between drives or subroutines is in a certain sense, hard to find. Inside chess, or inside golf there are radical simplifications that can be introduced. But how relate chess to golf? How find points of interaction? "Reaching for bread" and "Closing the window,” how do they interact? The details of nextness relations are involved via which goal functions are objectified and different move types, position types, and drive types inter-related.
To chart the breakdown edges of which tactic wins were, whether to reach or to walk, then reach, which direction to start in, all these kinds of issues that inter-relate types of functions, involve properties of nextness.
Nextness has a spacial and temporal aspect. The temporal aspect has to be made spacial (in the brain) to be caught, and, in reverse, the spacial nextness has to effect a temporal nextness to be relevant, or be part of a reinforced configuration that defines a through path.
The nextness factors, insofar as they compound to define a past location, have an absolute invariance. Insofar as they define a now present configuration, the invariance is only relative. Both are, of course, aspects of the "eternal present,” as is our sense of the future. Nextness orients a field of possibles actively opened from here and now by us in endless ways, a step or two. (A great chess player thinks two moves ahead, as does a great painter two brushstrokes. More is impossible. Two is genius.)
The genetic pre-encoding is optimal in many ways to gather in nextness type factors. Cells that sum similar nextness factors over a wide angle have been found in the optic areas of the cortex. And once local nextness types are compounded to uniqueness, more general associations among such compoundings can take on an invariant or objective significance. What one needs is some read-in, read-out mechanisms to and from invariant nextness themes. It is in their terms, as intermediate variables, that responses are most simply formulated.
Ninety percent of the cortex is associational. Levi-Strauss sees in the nextness of space and time, plus the jump of similarity that links distinct contexts, the basic building blocks of all thought. They are equally central to all cultures, and show no signs of development. The neurons that serve as abstractors of spacial and temporal nextness appear to involve short, medium and long time constants and spacial averagings. A great richness of types of cues is thus generated.
The way in which nextness factors compound, via the category repetition, has to be traced out slowly step by step, in relation to its self-emergence. The full why of their importance has yet to be justified or traced out.
One starting theme in this direction begins with the recognition that, except in very elementary situations (e.g., jerk back from what hurts) the definition of the act depends upon drive type, which must enter into the compounding in some form, as a kind of source term. In addition, to define an act in a simple invariant way, it must be defined via its end point, in relation to our present position. This end point, in turn, must be set up via its connection to the objectivized drive factor, and the components that enter into this must all be invariants, which tends to mean nextness defined.
That eye motions are controlled by the image of the endpoint can be easily verified. One can follow an oscillating finger with his eye, but it is impossible to make his eyes oscillate smoothly without the finger to lock on. Instead they jump. Gross body motions are controlled in a related way, via end point images, although refined shapings within a given function make use of special cues that come from anywhere. It is the jump between function types that, especially, tends to force us back on the basic invariant isolate of nextness.
There are other kinds of invariance, in addition to those related to spacial or orientation independence, that need to be considered. Drive independence is one. Some factors must enter into the shaping of many drives. Also, the linkages have to work over a wide range of moods and a wide range of levels of responsiveness. Hyperventilation can alter thresholds by fifty percent. Sleepiness, alcohol, etc., also have major effects. Yet basic meanings and spacial orientations must be unaffected by these big changes. This points up the need to use simple eliminating correlations in all encoding stages, correlations that are locked in by X's they engage, and not dependent upon the level or manner of first engagement to stabilize their final output. A few simple factors, so locked in, quickly compound to a complex result. In reverse, well locked in simple factors which develop in this way a lot of invariance, can be built on, or compounded with others, in a way in which less invariant ones cannot. One can build a building no higher than the (invariance of the) foundations will take.
We have developed a variety of basic themes. Can they be compounded or integrated in a concrete way?
11. The Generalized Search (G-S)
a. General
The concrete form under which the many factors we have been considering, integrate, is what we will be calling a generalized space-oriented search. The squirrel that searches for characteristics of the place where he hid a nut, within a field of possible run and dig sequences, and the mathematician who searches for characteristics of a solution, within a field of allowed or possible substitutions in a mathematical system, both are defining their problems more concretely as types of space oriented search within a field of objectivized possibles, actively opened out.
The generalized search (GS) as the form under which the interacting CR's integrate, is not something that "emerges.” Rather it is there from the first. The many aspects of genetic pre-encoding predefine subroutines and secondary sources of reinforcement, which lend themselves to integration as part of no other form. From the first, the structural features to be found in the brain are better oriented functionally in relation to the generalized search. For example, genetically encoded eye movements that follow lines, to points of maximal complexity, or lead to bright spots, or grasp motions, or expectations to re-see where objects vanish, involve secondary drives and tropisms best oriented or integrated functionally, as aspects of the GS, and only fully oriented in this more concrete frame.
b. The Break Up of Drive Stopping Factors
The SRS (stop random search, a generalization of satisfy drive) conditions break up naturally into two types of components. There is the qualitative component, posited at various "locations,” which indicates that specific drive or RS stopping procedures can be successfully or routinely implemented from there, and then there is the process of locating the places from which solution or SRS is relatively automatic or routine.
Thus, thirst, as a drive, has a related set of qualitative components (the glinty surface that means water, or the shape that means water faucet, etc.) which once located, may be used as cues to release relatively well defined procedures for thirst satisfaction. There is also the problem of "finding” the water, or water faucet.
This space-time search looks, at first, like a special kind of problem solving. But perhaps all problem solving, can be, and in fact is, fitted into its terms, insofar as they have an existential or felt status for us. It must be kept in mind that the space-time ordering ultimately defines or confines a field of possibles, openable from now (by travel, libraries, etc.), by available means. Location is in that field of possibles, something vastly richer than a (mere) rigid surface of a photographic now.
This interpretation of all problems within the forms of a generalized search (GS), by the split of the SRS into two components, is a theme that further specializes the game-theoretic frame into which all problems can be read. But although this is a further specialization, it is not a "restriction" on what can be said. If a framework contains set theory, or if set theory can be read into it, it can contain everything. The search theme has its content not as a restriction on problem type so much as a restriction (or guide) defining how any given problem type is, in fact, imbedded as a sub-aspect of a general problem-solving machinery. It does not restrict the problem, but it does help define how it is encoded, or imbedded in the larger picture, or related to other problems.
c1. The Emergence of the Generalized Search (GS)
(A Hodgepodge Scratch Sheet)
It is as framework for the coordination of the search form that the category "world" emerges. Broadly speaking, the GS provides a set of intermediate variables in whose terms response can be conditioned, or the potential effect on a latter response of new information encoded, most economically. This is, of course, not a reason for emergence, but it lends itself to being formed into such reasons.
Understanding the general form of emergence requires an appreciation of the "impossible" character of the search for repetition, or a through path from stimulus to response. The child starts from his crib with a certain repertoire that serves to satisfy a set of drives. With success, contexts broaden, and mutual suspension of responses set in, requiring further qualifying variables to define a through path.
Drives are organized as part of the GS from the start. Drives posit a control of reinforcement upon various mutually excluding signal types, which, in turn, determine orientation responses, go to, and grasp, and taste, etc. that satisfy the drives. Grasp as a subroutine reinforced by certain contact characteristics of objects, and guided by related visual cues, defines a type of isolated functional niche that has its own quasi-independent evolution. As the given object appears in more positions, qualifying variables derived from hand position and object position compound. A few extremums (e. g., left and right) are first solved. Refinements in these emerge filling in the middle. The solution of a few specific grasp responses, carries with it the intermediate, or a big start on the intermediate forms (as Taylor points out in "The Behavioral Basis of Perception"). Working solutions involve a variety of quite specific solutions, which mutually exclude, with intermediates as analogue compoundings of these once the set is adequate for that. A general procedure for orientation independent choice of act can be "simpler" set up in this quasi-digital way, than in a more usual analogue way using servo-mechanisms, where special calculations are needed. Where these calculations are of a limited routine nature, they can often be better effected via a set of specific answers, plus elementary ways to average between. When re-learning the visual field, we see how quasi-independent the same behaviors are when relearned in different positions, using different parts of the visual field, etc. (See Taylor’s book, cited above.)
The process of concrete explanation of any given behavior involves a cascade of specific tricks carefully isolated, to provide the needed regulation. There is no general formula. The basic drives (hunger related) activate different search related cues, that take into transference locomotion, grasp and taste, etc. The actual act emerges as a function of present position and cue position, in ways that can be set up in terms of simple coincidence effects in stages.
There is (perhaps most elementary) the "self-engagement" of acts, via related signals that always accompany the acts. Then the engagement of these acts via their self-engaged inputs becomes possible. As objects searched for change position, and as these objects are themselves changed as a function of drive, the resultant X’s mutually suspend all simplistic links that do not involve both end points. The nextness relations of these end points provide additional needed cues as to direction, etc. Nextness relations among cue types that can be objects of search, or regulate this search, are of a limited character.
At first it might seem that nextness and potential coincidence was just endless. But few of the signals responded to are such as to be able to enter into repetition. These few must form part of the control of learning cycles to and from a standard (e. g. crib) position. All learning can be viewed as a compounding of cycles. Children pick up and put down. Then pick up and put down — all starting from some standard position. Opening doors and closing doors, turning lights on and off, etc., are all parts of cycles of return to a (relatively) same state. More complex learning can be viewed as an indirect compounding of allowed or latent cycles no longer undergone explicitly. As a consequence, in advanced learning the basic role of the learning cycle theme for all learning is easily obscured. The parts of such cycles mutually engage. Cues used to help select a through response are themselves engaged, via the CR principle, by what precedes them in the cycle. The learning cycle is the basic unit of mutual engagement on the sensory side.
The interaction of cycles in a given location gives rise to a locational gestalt of nextness related factors. These nextness links allow of foreshortened forms of mutual exclusion of acts, via a mutual exclusion of cue types in a given nextness situation. The mutual exclusion of the act is basic, but it makes use of exclusions as among cue types to implement this in foreshortened form.
The dynamics of CR interaction, broadly speaking, undercuts all links that do not have a causal status. Links that form part of repetition in some context, provide starting places, but these quickly mutually suspend, till a compounding of factors in and out of the search form is, in effect, relatively causal.
The mutual engagement of factors contribute to the solving process, in that they bring to the fore the field of once relevant resolving cues for that field of possibles in that location. There are not that many.
For the cue link to emerge at all it has to be part of a learning cycle. The cue that engages a through path is also preceded by what leads to the act in the learning cycle, which engages (via the CR principle) that cue. This represents, in effect, an internalization of the cue. Once it is internalized, with a stable niche in the cycle, then it can be linked to other factors as they mutually suspend each other in the X.
The gestalt is formed of factors that have a potential for shaping a given field of possibles. These mutually imply via coincidence effects that basic brain rhythms, (as foreshortened compound learning cycles), help give rise to. Actual motor links tend to X, unless rooted in causal factors. The act types and related gestalts are vastly over-separated in process of reconciling the rules of extension of the mutually engaging cues (that help engage the acts). The "causal" is relative to a given field of possibles, and the degree of act separation that field effects is a function, in the end, of the degree of selectivity in isolating repeating paths that is defined in its terms.
The X becomes itself objectified as inconsistent nextness. We feel the wet water where it is not, when thirsty, and try to re-arrange things so as to remove this inconsistency. The elaboration process, the rules of extension as among nextness factors, becomes a supersensitive source of latent objectified X. Such themes as that two objects can’t occupy the same place, give rise in elaboration to objectivized X as mutual exclusion among object-like configurations (which are sets of mutually inducing nextness factors that go together).
The X opens back to the spacial sense, to nextness. In general, we foreshorten greatly on this full latent nextness. In the X this is reopened, as an objectivization of the field of possibles. It is opened up till the factors generating the X can be resolved within its field. The nextness-formed "bottom" of the objectivized X is ever adequate in practice to harbor answers.
The field of possibles keeps enlarging, as generated by the X until a through, path is found, as is ever possible, via a systematization of which wins where. As each branch of the X wins somewhere, the issue is solvable. The excess uniqueness of separation provides plenty of clues to winnow over. In reverse, once a through C (correlation) is found, there are plenty of coincidence effects to objectivize it or define it independently ahead of time (if and when there are X’s deep enough to make this objectivization relevant).
The first stage of objectivization is in terms of the body’s own movements, treated with some detachment. How get a hand to the mouth from different positions? To answer even so simple a question involves setting up a "frame" of distinct body positions, in relation to which "now" and "end point" are located. This detached treatment of the body's own displacements provides separated elements that can be independently engaged. These are then taken into transference by more distant cues in anthropomorphic ways.
The X's force (individuate) back to invariant (always holding) C's. A nextness cannot (relatively) be further qualified. The pair of defining act types or configurations that enter into the X, are more than adequately separated. What is reinforced in an objectivized choice of branch, is some rule of extension of nextness rooted factors.
The X generates the branches of the historical style cycle. A more stable choice of branch involves a finding of the qualitative component of the given SRS somewhere in this cycle. Then there follows a foreshortening. The X pushes open to the space sense via which the behavioral issues are made as self-conscious as need be. The space sense leaves room for endless opening ambiguity. A stop marks not the stop of ambiguity but a resolution of the X.
Development remains an additive process in part because the pieces are over-individuated. They interconnect of themselves. The struggle is to find places in the nextness field. This struggle, incidentally, preserves an over-individuation.
The bringing into parallel can be viewed as "the aim.” When there is an X, the finding of its place, or the place of the corresponding qualitative aspect of the SRS leads to a foreshortening that constitutes a spanning of related fields of possibles, or a bringing of them into parallel.
The foreshortening later generates other X’s. This tends to obscure the latent additive root. It is the foreshortenings that enter into X, e. g., a way of reading out history as principle. All we, in fact, "do" has its aspect as principle. But all this should not obscure the additive root, at the history level, from which the principles (in X) are abstracted. What is a change of principle to resolve X at one level, is an added perspective at the history level, extending the field of possibles, (as functional refinements).
An RD makes possible other coincidence effects. The RD itself is a foreshortening that depends on the through path. Not much gets a chance to enter into it, but all that shapes the then available techniques may.
Through paths are rules of selection conditioned in terms of well individuated acts. These rules are reconciled only via pure theory, involving a theory of the field of possibles. The push towards a full objectivization of these rules, lets loose a bevy of X’s in the linking rules, which, in their secondary objectivization via words, pushes us to pure theory.
The push is a push to CR’s that "always" hold. For example, even statistical rules have to be objectivized in terms of links that ever hold, etc., but only insofar as we are forced to by the X. In clear sailing things foreshorten out, so you don’t see the latent objective roots. The foreshortening regeneralizes as it were but to something very different from the naive over-general start.
The orientation is ever relative to the now room (short term memory) with extensions opening from there. The nextness factors orient a first order field of possibles, and are the needed cue for spanning it. In all this an arbitrary component plays a big role, often as first choice, or as a quasi-cyclic randomizing if one choice won't do.
It is not the CR that constitutes learning, but a selective aspect of their interaction that forces the few remaining through paths into a "relatively” causal form. Except for relatively isolated trivial situations, there is no simple path from stimulus to response. The road through involves an independent "prediction" of the end point, via which the act is selected and engaged. The components that enter into this "prediction" have a relatively causal character in their capacity of orienting what effects SRS, or constitutes the repeating path. The path from stimulus to response has to go a circuitous way. It is foreshortened out of factors that predict it. This involves a relative objectivization via the bringing into parallel of a set of intermediate variables (our sense of the relevant parts of nature) in whose terms the variety of path from stimulus to response or act selection can be finally conditioned. The set of factors that are involved in a mutual engagement are a function of the branches of the learning cycles in X in that given context.
The X is objectivized as inconsistent nextness, or inconsistent extensional rules. The X is an essential source component that determines the associational extension of the present, that may become activated via the end-back theme. It also serves to determine the aspects of the present inputs especially reinforced, or gone forward from.
The extension rules are what generates the basic X as the acts are, in general, over individuated to be "accidentally” engaged at once. The acting out fills in our sense of nature with over-individuated factors till one locks in the qualitative parts of the SRS to the space sense, after which solution is relatively routine. The X's among extension rules generate related motor X's, or related reactivated learning cycles, via which relevant parts of nature are opened and the foreshortened rules of extension are reconciled. The pattern of all this involves theory.
The additive adiabatic opening of learning with success is built up on the only non-self-cancelling combinations. The additive aspect has history as one side, and methodology or technique as the other. It is a two-pronged thing. The rules of extension as they now exist, involve a simplification or re-generalization that depends upon a kind of foreshortening of history. But this simplification is not the same as the naive beginning, and involves much complexity to support it, a complexity that the X reveals or re-opens (a paradoxical point related to the initial intensification of operant responses by inhibition).
There is a great deal of intermediate foreshortening that is so powerful that one is apt to lose sight of the more basic general themes that emerge only in breakdown. Mutual exclusion of acts is objectified as a factual inconsistency that implies inconsistency among acts engaged as a special case of the exclusion of two objects in the same place. (This point Kant calls self-evident as it does not involve elaboration to bring the point home.)
The requirement of logical form becomes independently imposed. If a CR set does not fit within space-time type elaboration, or fact-assertion, then it may engage disruptive responses long before details are filled in. Cues that correlate with disruption become sources of inhibition. Foreshortenings go every which way. The read-in read-out must be of the general form of the through path, if not there is no chance for reinforcement.
Cues that limit form prior to content are no less "conditioned," but their relation to the act is more obscure. Unless tracing through to points of total breakdown of old form, there is a big residue of these cues that seem to have an a priori character that reflect special internal mechanisms.
In the solution process there is a breakdown of felt form, and its recrystallization. But the root in history holds up.
At first, caught in the X, we can only use part of available consistency criteria. Imagination can be viewed as such a partial acting out. In the end one can, however, face into the full model with its existential roots (autobiography). The push is, in a sense, to eliminate imagination, tying imagination back into the real via a "why it concerns us now.”
The X with which we identify has roots in correlations that begin to bring into parallel (by loving what we hate) branches of X's else acted out, and so intensifying the X or internalizing it. The rest (the push to science) is a tying of this latent correlation into the main body of causal insight. Science slowly winnows out the unqualified (causal) universals latent in the particular start on bringing into parallel. In the political perspective this involves internalizing the style cycle, and then isolating the structure of the field of possibles internally generated.
The word pushes this internal acting out via its parasitic relation to objectivized factors. Only by using pools of internal association as stepping stones can a verbal conditioning be set up (as Skinner sees).
The word involves a kind of second abstraction process, the isolation of what affects the verbal order. This parasites on the first, the isolation of what affects the motor order.
The abstract act or word can’t be made use of in a formal way at first as a basis for organizing the decision process. That is to say, we cannot push through a world model adequate to coordinate practical decisions in its terms. With meaning still tied down via the false category "observation,” one is, in effect, limited to elaboration, and a related choice of sides (the middle class role). If one looks to observation, in the usual working sense, to ground the abstract word, it (the abstract word) has no impact on the present decision process, and seems meaningless. Only as one stops treating observation as basic, and goes on to the struggle for a pure game theoretic interpretation of named or observed categories (subordinating the side choosing middle class) does the abstract word emerge as a working concept.
This involves pushing all qualitative factors to the point of their re-definition in ordering terms, and exposing the latent theoretic character of the felt via the generalized search and the role of repetition. This latter involves building verbal models of the preverbal role of different factors, and then meta models of these models. In doing this, the felt or real becomes referred back, not to observation, but to places in patterns of abstract words that generate an orienting field of possibles, in which all qualitative aspects can be treated as types of relatedness. In this indirect way the existential word ultimately effects a model of itself. This process, and the symmetries or harmonies related thereto, give to the dynamics of CR interaction depth and fascination. The basic X is within ourselves.
From the perspective of the over-all laws of the dynamics of the GS, once understood, the GS emerges as a search for new laws of physics. The category adventure itself is oriented in relation to the opening up of the laws of physics. It is via the place in this adventure that the individual now finds herself. Until she can define terms in the one person game she now keeps repeating, "what do you mean? " — that so terrible question.
c2. The Emergence of the Generalized Search (GS) (Redone)
CR Interaction and Control Variables
Learning depends, not on the CR, but upon their interaction which serves to determine how or whether a CR will be used to shape acts. Other CR’s are the most important factors that controls both the emergence of CR's, (whether they will be used), and the role they play once they take hold.
This sort of statement can seem rather vague. How can you go about defining more clearly what controls what, in a context dealing with human behavior where a change in anything anywhere changes everything everywhere, after a short period of time? A key is to distinguish lasting change, of an institutional character, from "incidental” change, which, though it may take your and my life, leads to no lasting change in the life pattern of the group. Environmental factors, of course, effect CR’s, and change things, but they have no lasting effect, as they are too crude. They merely draw out this or that possibility from a pre-formed potential. Lasting change depends on causal insight which derives from CR interactions with other CR’s. The separation of lasting change from change in general can be made more precise via the notion of a control variable. What is the nature of the control variable of lasting change, and how does it differ from, e. g., hitting the nail, instead of shopping, etc., which gives direct rise to merely incidental largely self-cancelling long term effects? We have said that the control variable lies in the CR interactions — but where, and in what form?
For reasons that we will briefly re-outline later in this section, the control of institutional change takes the form of causal insight, encoded as self-prediction. This is the only control variable. CR configurations that do not take the form of causal insight come as part of contradictory (heuristic) sets. Murphy's law states that if it can go wrong, it will. Only relatively causally rooted CR's "can't” go wrong (relative, i.e., to existing tasks and available means). The situation is structured by our method of control which is the method of coming to a decision at all, or finding repetition at all. Only what is causally rooted can repeat. (The illusion of the inadequacy of repetition, and of the need for choice of sides, is a middle class illusion. Those are most zombie-like who think themselves most free.)
We will now briefly re-outline a sequence of qualitative causally rooted considerations filling in the patterns of CR interaction that lead to a self-emergent regulatory role for heuristics taking the form of causal self-model insight.
The Isolating Individuation of the Act
Basic to an understanding of the brain's dynamics is the recognition of the extreme narrowness and unitlike character of potential parts of through paths. They are isolated from one another by vast regions of potential combinations of movement that are rejected out of hand as "meaningless," etc., a separation which the context dependence of acts vastly increases.
This allows of a (relative) non-interference of act-like units with one another, or at the level of linkage of past acts. The latent redundancy, and the excess individuation of the acts themselves for the most part keep channels from interacting, the channel being defined in a context dependent way. Of course, there are momentary sleepy confusions, and occasional over-simplifications, as when distracted, but these are of a trivial nature, and rapidly regenerate a more than adequate separation.
Significant X derives from the extension of acts into new contextual regions. It derives from latent heuristics generating past acts, which, in new contexts, brought on by success, give rise to SE that undercut themselves in seemingly inescapable ways before emission.
The branches of the X form part of local solutions that keep apart as long as they can. They only lead to X’s when this is inescapable. Similarly the X’s (which are related to pre-engaged acts) keep as well separated from each other as they can. All this contributes to an excess individuation on many levels. The old gets very complex before it breaks down, more than complex enough to contain the new as a sub-aspect of the over-elaborated old.
This excess individuation in emittable acts is neither trivial nor a matter of convention in the definition of acts. It depends in essential ways upon the enormous redundancy with which facts are recorded, but it involves more than that. It is hard to form an emittable combination, as "most” SE undercut themselves. But once a particular combination is formed, it tends to have a permanent aspect. Of course, as contexts broaden, the role it plays may be transformed. A drive or X type that engaged it may be resolved, so that it is not emitted again, etc.
If acts were abstractly defined, there would be plenty of "interference" at the act level. So it might seem that what we are speaking of is partly a question of the semantics of what is meant by an act. And, of course, it is that in part, as all statements of substance are. But it is not this in whole. Different acts are engaged in different ways. What we learn when drunk we forget when sober. The context dependence is real and deep. The same act in different functional contexts uses different circuits.
Whence this over-individuation? It derives in a sense from the process of generating X’s in new contexts. The set of factors that must compound before latent emission becomes possible, despite constant elaboration to new contexts, is high, high enough so that there is great uniqueness in each way.
The uniqueness or individuation is built up out of "wholegram"-like configurations of mutually reinforcing cells, or hierarchies of these, as we will see in more detail. These are such, in their "decay channel"-like separation, that the mere set of coincidence effects, or nextness related associations between two wholegrams (and related acts) can tie one to the next uniquely. The uniqueness makes a locking in of sequences via elementary linkages, easy, a phenomena that any book on mnemonics shows us how to make use of.
This relative over-individuation of the act, so that the concern is (only) with extension into new areas, depends, in part, upon the nature of the correlations fed into the brain, their redundancy, etc. The X comes from the SE, not the acts which are too isolated to interact. The struggle is to get to emission at all (the external RS being relatively routinely internalized and subordinated to the internal RS where basic change or learning occurs).
This characterization of the overall pattern is ultimately justified by the rest of the picture. Keeping it in mind helps to clarify or orient the meaning of other pieces to be filled in.
The Learning Cycle
An initial positing of SRS criteria on the environment is pre-given genetically. The qualitative part of the SRS "searched for" is constantly changing. The mutual exclusion of drives is partly objectivized as mutual exclusion in the now searched for qualitative part of the SRS. In the absence of more specific drive positing, we may find ourselves possessed by a quasi-random investigation called "curiosity” with touch and taste as the carriers of SRS. Orientation reflexes, gasp, etc., are largely given genetically, as in the frog. But in man, there is an excess of competitive tropistic possibility which tends to mask the genetic root.
The first setting up of an internal mutual reinforcement, despite all the anti-repetition devices, depends upon the learning cycle. Basic infant learning is of a cyclic character, and all adult learning merely takes these partly pre-formed cycles over in a variety of ways. The child starts her world from a position in a crib. A yellow rattle hangs up to the left. If a posited touch or taste drive strays to a yellow factor, it can be regularly satisfied by certain leftward movements. Other movements return us to the starting position. A motor cycle is internalized as a cycle of associations.
"Return to home base" combined with "search out yellow spot" generate a cycle. A cycle is not a repetition in the sense in which we are using the word, but it isolates material out of which repetition may be built. It isolates an element that is present in the field of motor possibles to be spanned.
The learning cycle provides a powerful filter on the kind of cue types that have a chance to enter into internal compounding. To do that, cues must be internally engaged as part of the motor learning cycle as internalized. A cue selecting a following act, is itself engaged by a proceeding act so that the cue is internally engaged without any actual occurrence. This gives a starting handle. Cues not part of such cycles, have not the needed proper timing (even if internally engaged) to lock in with others in compoundings to be discussed later. Despite the foreshortening made possible by the bringing into parallel, all learning has explicit motor counterparts. These may seem relatively "distant,” as in mathematical insights, but it is only within the decision process itself that significant changes take place. Coincidences not serving to shape existing decision (covert or overt), don’t cumulate or coordinate with others, and are washed out.
There is a tendency to discuss the association of related visual cues, as if this could be built up as to apart from a cue’s relation/ the selection of drive stopping acts. In this sort of approach possibilities are just too vast with no way to regulate them. The learning cycle theme serves as an essential filter on cues.
This motor learning theme integrates cues in ways that are already prediction and control related. It also begins the theme of control by the end point via its relation to now. The learning cycles, into which all learning can be analyzed, serve as basic units of internalization and a first filter for relevance.
Via this repetition which involves the drive factor objectivized as pre-engaged act, sensory factors appear as engaged as well as engaging. This learning cycle pre-repetition also defines a class of competitive possibles that one starts with in the conscious sense. (The large genetically set class of possibles is largely subconscious).
There are what might be called sub-repetitions within learning cycles, branches that get engaged together at a given cycle stage. Sometimes one may be first, sometimes another, and so they end up engaged together. This mutual engagement is a kind of sub-cycle that enhances a given stage in a larger cycle, rather than an independent cycle of its own.
Series Elaboration (SE) and Internal Random Search (RS)
Interaction of the isolated act-like units comes via SE, that touch new sources of RS that stops "present" emission before it occurs. Thus SE leads to paralyzing multiple determinations of the present act that stops emission, and lead to added sources of selectivity in finding a path through. The SE makes emission or repetition even harder to find. It involves or generates ever rising levels of inhibition, or added sources of selectivity.
What gets rid of a multiple response, the SE that is not undercut, gets relatively reinforced. Thus one has the beginnings of a relation of selective inhibition in addition to the mutual exclusion) of competitive acts. What seem to be two distinct phenomena operate in related ways via the SE. In the end, the genetically pre-defined drives are subordinated to the removal of multiple response theme. Where they collide, this theme takes over, and brings an added content in.
Now the removal of multiple response pushes links into a relatively causal form. What is not relatively causally rooted, generates multiple responses under SE. It requires a very special form of CR placement to be motor related in origin, yet remain neutral or objective enough not to generate inconsistent acts.
The SE undercutting itself by elaboration provides a more refined concrete model of what, in other writers, is left as the vague catchall of "stochastic process,” something far too unstructured and vast to ever succeed at much of anything. The stochastic process is here seen as generated and shaped by the X itself, and more specifically still by the SE that undercuts itself.
The "paralyzed” or "suspended” state associated with an SE that undercuts itself, plays a central orienting role in a wide range of situations that do not involve actual motor suspension. Once into these states, we learn to handle them by a variety of devices. The most primitive device is just to up the response level of the channels in X, so that one wins out. This leaves us with a related muscular tension, and a suspension (or avoidance) of one side of the X. Another device is to begin to objectivize the X. The underlying X configuration is left suspended, but it serves as motive to undergo acts of "self-expression" via which the X is represented as problem and (hopefully) solved (with group help).
A middle ground is the use of the group to organize people with a related paralysis. Often a way to put in a choice of sides at the group level can be more stable, and less damaging, than one introduced at the personal level. In practice a combination of all three of these devices are used.
The paralyzed state is thus a much more significant orienting state than might be supposed at first, as there are a rich variety of ways to live with and use it. By means of them we can face into or invoke X's it would be impossible to handle directly by primitive devices.
Its Relation to External RS
The basic generator of significant change is the SE that undercuts itself or the internal RS. It draws out only latent X among overly separated elements by extending them in phantasy into new areas. The removal of resultant multiple determinations by finding an internal through path is the form of the basic group learning process. Alternations in the X are never twice the same (’’repetition is found only in the truth") so that an adequate field of possibles is generated by the X itself.
The external RS at first appears central and orienting, with the drive factors fixed, and genetically given. As the internal RS first emerges, it appears as a parasitic subsidiary of the external RS, providing secondary corrections. (Similarly, the word, in its first emergence appears as a mere secondary parasite on the preverbal gestalt). But as the internal RS advances in importance, an inversion takes place. At first the internal RS is oriented as helping reconcile external RS ahead of time. Later, the external RS is oriented as helping fill in gaps in the internal RS. The internal RS becomes dominant, and the external RS is oriented as a relatively routine externalization process to help fill in regions where no through path can be found, which are then internalized and winnowed internally.
(A second inversion process occurs later in our orientation when the verbal part of the internal RS subordinates the non-verbal part. At first, the word merely parasites on the preverbal (name magic). To understand is to name. But later, to understand becomes to identify the name's place in a big verbal processing system. Facts on modern research are oriented by the way they effect words, not by a "feel" as they are for the hunter.)
The difference is one of orientation, or the sort of grouping that best simplifies information processing. At first the internal RS only emerges when the external RS fails, and its function is oriented by the external RS. Later, it is the failure of the internal RS that generates a return, to fill in its gaps. (Still later it is the failure of the verbal part of the internal RS that orients.)
The Breakup of the SRS Criteria Into Two Components
The guiding SRS criteria can be usefully separated into two types of components, a qualitative one, whose isolation insures a relatively automatic realization of a related drive reduction, and the search, in a spacially oriented field of possibles generated by our own acts, for the appropriate qualitative criteria. Via this qualitative criteria we get access to drive stopping subroutines. These reduce the remaining solution (drive stopping) process to what is (relatively speaking) an algorithm, that can be easily set up or conditioned using available cues (and the help of imitation).
The basic qualitative aspects separate out in process of reconciling the many learning cycles with one another. The encoding is pushed into orientation independent form by this. The basic type of link that holds up under change of drive, interlinking distinct types of functions, is nextness. It is the basic compoundable CR with an orientation independent root that has a chance of not leading to multiple acts if compounded in a "proper” form. Nextness organizes the basic intermediate qualitative variables in whose terms read-in read-out processes are defined.
The end point, in conjunction with the new position of the body, provides a set of cues often adequate to select a response type. The first objectivization stage of these is in terms of the body’s own displacements. At first the coordination of the two cue types is hardly separated. But the mouth is reached by the hand in different ways from different positions, etc. The itch on the foot, handled differently than one on the neck. Once these primitive RS only involving the body begin to get solved, more distant receptors begin to play a role. The rattle in the crib is up, and the visual cue may take over an independently set up motor sequence. The visual gestalt opens hand in hand with related motor coordination. The visual cue defining the end point may work with non-visual cues (the unseen hand) to define a reaching.
Bringing Into Parallel (Wholegrams)
The factors of the learning cycle brought into parallel in the X or by the SE in any given situation mutually engage each other. The cues defining the through path are compound cues made up of factors brought into parallel. Control variables not locked in by a near pure nextness, will give rise to X.
The methods for bringing into parallel involve finding a through path. It is only in this way that competitive branches take up a definite relation, engaging each other mutually, and providing an adequate set of spanning cues. The way the spanning is effected step by step in terms of those cues involve a bag of specialized tricks each adapted to specific steps or X’s. Like the two-body problem in physics, such X related groupings and RD’s form an infinitely flexible tool into which all we need can be fitted by the right tricks.
The X branches when brought into parallel involve a qualification of earlier rules of extension of qualitative factors. The "acts” appear as already more than adequately separated, a separation incidentally pushed on by the bringing into parallel, as a by-product. Hence, the "only" concern, in practice, appears to be the reconciliation of rules of extension of qualitative factors, objectivized via their nextness relations.
The learning cycle forms an essential first filter. What it preserves must be relevant cues to guide finding SRS in some situation. When situations are related, these may mutually engage. A wholegrams effect takes hold, as mutual support makes the whole more easily excitable than a part. If sub-parts are facilitated, the closest "wholegram" takes hold. These wholegrams in effect constitute a sub-field of possibles, with associated spanning cues, from which to compound a guiding selection. In this indirect way a vast field of motor possibilities, which could never be faced into directly, can be considered.
The wholegrams in part mutually exclude, as do acts, and in part compound in larger complementary ways. Where similarity forces interaction, separation is made more detailed.
Context dependence of a complex conscious type becomes possible building with wholegrams. Over-individuation of acts prevents extensive interaction except via the SE which helps keep the interaction at a conscious level.
In this indirect way, using wholegram interactions, the context dependence may become "conscious," via cue components that form part of complex learning cycles, (and so are independently engaged ”at once” by motor related factors as well as contributing to their engagement).
In addition, there is always an unconscious component of context dependence, that keeps factors separated in the external RS before any significant internalization occurs.
The compounding in parallel prior to emission depends upon isolating some working form of independent "prediction" of the resolving act, via a sequence of nextness or similarity type links. Once one has a configuration that does that, to use it to condition a needed next act, or to use it to help isolate the relevant "meaning” to us of a signal, is often a relatively easy matter.
The conscious gestalt or wholegram (that merges into Augustine's Eternal Present) need define the act only in a qualitative way. There always remains a routine ambiguity removal of a semi-subconscious nature, in the final pass to action, which involves a second or different layer of circuitry.
The act must be defined as part of one global system. This is seen in the fact that signals applied to the context do not enter consciousness until they engage a correlated EEG component that is wide spread, with systematic delay to distant areas, such as meaningful systems have. (See F. Morrell's experiment, reported at NRP, March 25, 1975). Signals applied to the cortex require high amplitude levels and a larger time (roughly one second) for recognition. Because they must indirectly engage via some part of a system built up in other ways, such signals are inefficient at shaping any conscious response.
This "system" that any cue must form a part of to enter consciousness we will call the generalized search (GS).
The Generalized Search (GS)
The process of X resolution takes on the character of a generalized search. At first, the squirrel's search for the nut it buried last year looks like a very different kind of problem solving from that in mathematics, say. But mathematical problems, to be well formulated, have to be defined as a search within a spacetime oriented field of possibles (e. g., allow substitutions in a word system). All problem solving fits within the forms of the GS. In order to get two heuristics to collide in a conscious way you need competitive SRS criteria that can be tied down via nextness relations, in terms of which the heuristics are formulated. I.e., heuristics must be tied down as part of the GS.
Heuristics become rules of extension of the spacetime gallery, via their interpretation as part of the GS. The qualitative components, detached via nextness systems, serve as intermediate variable via which all "conscious" deliberate control goes.
The conversion of quantitative distinctions into distinct channels goes both ways. The final conversion of a nextness relations into a choice of act involves a stage in which distinct channels lead to quantitative differences in the ballistic movement. It is perhaps in these conversions that the systematic visual distortions (that e. g., can occur in sickness) may be located. The felt involves all aspects shaping systematic motor coordination.
The GS has a deep set genetic basis. The drives effect behavior in part via selective reinforcement of cues to be located. The funneling of drive related responses to searched for cues, (in part organized by genetically given relations of mutual exclusion), precedes a drive's actual successful positing. In the absence of specific drive activation, this genetically set search-like behavior proceeds in a quasi-random way, filling in latent ambiguities or multiple determinations by acting them out.
The X becomes slowly objectivized as inconsistent nextness. Taste is the dominant source of SRS in infancy. Touch is a key secondary source term that ends one stage (e. g., reach) and leads into the next (e. g., grasp). Infant tasks get broken up into pieces, the "search for till touched,” and "what to do then.” Once infant search and grasp problems are solved, the resultant insights allows of a more abstract approach to search, in which the search-like roots may be progressively ignored.
The X serves as a generalized source term. When no superceding drive is activated, any latent SE that undercuts itself becomes a source of quasi-repetition, that activates a needed acting out to fill in the ambiguity, as involvement builds up in "both" branches.
In a sense curiosity is basic, and hypothalamic drives that seem primary, are secondary, serving to help define what our curiosity is about. Of course, without these hypothalamic sources to give curiosity a direction, it would rapidly destroy itself. There are only a limited number of niches that can be usefully opened. These ultimately all open out of the child’s mouth stage, as a symbol of the centrality of economics (feeding). Ultimately, the genetic drives are qualified out into a pure ideality, as heuristics of repetition.
There is a kind of generalized fear that shapes and limits the search. We avoid areas where one can’t (yet?) decide, as children avoid pain or fear. Someday these, too, will be opened. But, meantime, they help confine us to a positive core, where plenty of useful new search still goes on.
The parasiting of words on nextness type factors are special types of channels of repetition of social origin. Once search and find, grasp and eat, etc., are solved, imitative coupling opens the basic root gestalt of the GS. Imitation opens only certain of the endless latent X along special guided lines. The social interaction narrows latent channels of repetition which allows of pushing the X far further.
The biggest part of the trouble, in practice, is not to solve the X, but to find the X. In practice the X branches are not well separated. There is too much ambiguity to focus the X. The latent X merges into the general avoidance of the unknown. It is very hard to find an X. It requires a hardening in of the branches of the SS such as only the written word effects to generate the added class of X’s that underlie modern thought.
There is a hierarchy among latent X, as well as a mutual exclusion that allows of step by step resolution. Solving one activates another. The "building" up of X resolutions can be raised only as high as the separation of the pieces allows you to focus things. Hence, the dominance of the heart over the head. What draws in deeper demands, seems at first an impediment, as it generates a paralyzing separation associated with "excess" conditions or a SE constantly undercutting itself. The fineness of separation is a function of the number of conditions on repetition that are integrated. Taking on "too" many, is ever the avenue into the future. It allows of the isolation of new basic X's as one works out the more routine ambiguity. At first the excess conditions can seem counterproductive and what is squeezed out, a vanity. At times it is. But in the end, life goes that way.
The Additive Dynamics of Opening
The global dynamics of gradual narrowing of repeating forms, takes a very peculiar logical form because of the properties of the correlations fed into the brain.
The opening begins "adiabatically" from the crib. A SE depends upon the successful resolution of primitive external RS to begin to elaborate. It can elaborate only a step here and there further than can be successfully resolved, enough to generate an X. Then it stops as it were to digest that.
Any resolution depends upon a latent causal content. Nextness, as our basic isolate, depends for its successful taking hold upon a parallel objective counterpart. This, in turn, depends upon the continuity of identity, the relative constance with time of thing-like configurations. This is a causally rooted fact, related to what one calls belief in an objective world.
Thus the central role of nextness generalizes naturally into the relation of causal insight to emergence. It reflects more specifically the theme that one must build in terms of relatively causally rooted invariant correlations or CR's from the first.
One can integrate X's or series elaborate only as far as one can resolve, plus a step more here and there. The resolution process makes use of old pieces. Color and shape and place, etc., are a vocabulary in terms of which anything to be said, can be said. The dictionary is adequate, and ever a few steps ahead of need. The task of resolving X’s can be adequately formulated in its terms.
Root forms that ground the naming process are ever more than adequate, and refine automatically more than a step ahead of need. The conscious issues lie ever in their combination. Brain capacity is never an issue in evolution. As soon as heuristics are isolated powerful enough to use it, it is easily built in.
This adequacy of basic forms is related to the over-individuation of the act, and to the theme (in Freud and James) that learning is additive, or that there exists a perspective from within which it can be so viewed. To resolve is to locate the resolving behavior in nextness terms, via which it is also selected. The rules of extension into new areas, or more precisely, the rules for resolution of X's in such extensions, undergo radical transformations. A reason for this is that the resolution of X’s requires a global organization to be locally consistent. A global systematization of the field of possibles is a key step in this, and such systematizations undergo radical jumps.
But such "jumps" are also a consequence of the additive opening up of new fields of possibles. Identification of old SRS criteria in a broader perspective often leads to very different rules of choice. Theory has beneath it a nextness rooted geometry or vision, organizing a hierarchy of SRS conditions, coupled variously into the present gestalt, via word processing, etc.
History and methodology open additively. The "equivalent occasion” is susceptible of a continuous opening, in a sense in which "thing" or "idea" cannot. What counts here is that there is a perspective from within which development is additive. That there are many from within which it is not, is beside the point.
The enormous foreshortening of actual practice tends to obscure this additive root, as it also obscures the relation of the felt to behavior.
The richness and variety of these foreshortening criteria, so essential to the removal of multiple response, are endless. There are the grammatical compulsions that limit form as apart from content. There are criteria related to relaxation, etc., etc. These many foreshortenings are only broken through in what is called personal or cultural breakdown. Only then is the need to trace back to additive and behavioral criteria made necessary.
"Plan"
When a plan or decision making is routine, it can be analyzed into what are called a collection or quasi-hierarchy of TOTE units that control the sequence of operations to be planned by delimiting alternate paths or choosing one (Arbib). Such TOTE units involve some test (T) to see if a condition is satisfied. If yes, exit (E), if not, perform some operation (O). Then test (T) again to see if the condition is satisfied.
The TOTE units determine behavior by testing a field of possibles, and then selecting some act as a function of this. The tests and operations and exits to new units, represent the basic kinds of conditioning within the GS. It uses the additive objectivization as intermediate variables.
In the SE and related positing of drives on cues, what leads to an RS or SRS carries the corresponding properties. As the positing moves a step along, the logical character of the act that satisfies the related sub-RS and sub-SRS vary enormously. The distinctions T, O, E, are helpful in orienting such logical distinctions.
The process of testing (T) allows of the generation of cues by our own acts in highly selective ways, thus vastly increasing the range of logical types of available cues. "Testing" is a way of generating an SRS, when the latter has been posited in a basic cue, and secondarily in a variety of cues that lead by an "Operation" to the basic cue.
Operations are concerned with external changes. The separation between information gathering and operation is not, of course, absolute, but it has a deep practical status. The simplest forms of TOTE units involve tests of discrepancy from some image, with negative feedback to remove this discrepancy.
The building up of TOTE units uses a present body position’s relation to where the SRS is located, to engage intermediate acts. Imitative coupling is crucial in the generation of more complex forms. The take-over of RS-SRS criteria by others is first effected by imitation and later by word.
Imitation of our own past is a key starting and ending point. Via imitation we use the present to represent a wider arena. This makes possible the planning process with its breaking up of the whole task into subtasks. A plan’s hierarchic organization roots a problem to be solved in cascades of solved sub-routines. (We lean forward, and use the reflex to catch ourselves already there, etc.) Higher order routines use lower ones, and vice versa at times. Such separations always involve some rigification. If "catching yourself" is left automatic, it interferes in the tumbler’s art, but is a useful simplification for the average person walking.
The key invariants are verbally formulated. By setting up rules in words a CR is given an absolute or objective basis that cannot (for practical purposes) be undercut. Words allow of the use of the present to model a complex situation, via a modeling of now present verbal interactions with it.
X’s can be oriented by means of TOTES when they are routine or basically already solved, and as the reformation of TOTES, when pushing into the unknown. The act-like units, the RS and SRS, are more primitive functionally than TOTES, orienting elements out of which TOTES are, in turn, defined. Plan is a kind of middle category, viewable as a composite of TOTE units. Description involves an overview with categories in whose terms the interaction of plans can be treated. The verbal function that underlies description involves meta forms at least one stage more detached in their logical status. Nextness isolates which enter into endless TOTE units
are also used to orient meta and meta meta levels, indeed, are as orienting at all levels, reflecting the secondary nature of the hierarchy concept.
The TOTE groupings can often be further usefully analyzed in terms of what has been called the control problem. This involves the breakup of nextness groupings into things and background. The object has its own sub-field of possibles, (understood imitatively, the stone "stands still,” faucets turn two ways, etc.). We may first be able to decide what choice in the objects related field of possibles would effect an SRS. Then we make a choice from within the body’s field of possibles to effect the object in the previously decided way.
The verbal coupling hardens in separations as much as needed, giving the word an absolute or mathematical character, so that the brain emerges as a latent general purpose computer. The rules of logic, etc., reflect conditions that emerge in enactment, via processes of pre-encoded self-correction, and need only be partly present in preverbal encodings. Insofar as links violating logic tend to generate RS that leads to their suspension before emission, the logic is built in preverbally.
Imitative coupling makes of TOTE reformation or X resolution a social dynamic. Where X's cannot yet be resolved internally, their branches are projected out as different people (or different moods of the same person), thereby generating (in style cycles) the needed field of possibles. Characteristics of the style cycle, once isolated and internalized, lead on to resolution as a reformed hierarchy of TOTE units. The verbal function builds models of the breakdown edges of TOTE interactions, or heuristic interactions, via which, consistent resolution process emerges. The breakdown edges of interactions are objectivized in a hierarchical way, via models of models. The process of building models of models, maximizes the separation of sub-X’s of the basic X, via which, step by step, a gradual return is achieved. That this works depends on the themes that the basic conflict is with ourselves, that simplicity is related to truth, and that a dominant X can always be isolated whose resolution breaks the style cycle, etc.
The Overall Dynamics
The problem solving dynamics can be interpreted as an hypothesis testing process at each step. The branches of the X tend to be read out as the hypothesis that they lead to SRS — which turn out wrong, etc.
The encoding, in pushing to orientation independent form forces on the word’s control of ourselves a certain "read-in read-out" type indirection via an internal nextness built model. A change of signal is not directly interpreted as changed scene (things seem to stand still as eyes move), but acts rather via the read in. Indeed, the scene can remain the same in the eye, yet seem to jump (as when the eye muscle is paralyzed and one "tries" to move his eyes).
The X is ultimately revealed as self-X. The world is a cooperative place. War is a part of education, a branch of experimentation, soon to be outmoded as not violent enough. In war a crude external field of possibles, that clarifies which heuristic wins where on very primitive questions, is generated. People would rather fight (act out outside) than think (act out inside). But soon we will make them think, despite.
There is no way to consistently resolve X’s except via a theory of the field of possibles. This is generated by the local distinctions and their correlations that enter into the systematization of all branches of the style cycle.
In this broadest perspective it is a question of linking rules of extension, making them part of one inclusive system, not by "selecting", but by adding in qualifications to old root forms.
The acting out process involves generating a style cycle via which one can learn to predict his way from one branch of the X to the other, thereby internalizing it. Then the fight to objectivize this internalization begins. Of course, the branches of the X are "objective” as past history. But the categories that enter into the X's subjective generation as heuristics of the common goal function are not objective, as they collide. The struggle is to reground the value problem in relatively objective terms, which alone can resolve, and which are also relatively causal (relative, of course, to the existing technological level).
The field of possibles opens to the full combinatorial one (limited only by a taking of the simpler first). There is no a priori limitation, one must face into the full field, as Dirac says, evaluating each guide on the evidence, with a preference for the simple.
The struggle to internalize begins with the alternation derived from loving what one hates, a love made possible by the ability to act through the style cycle. A swarm of paralytic effects are thereby generated, or specific situations where one can’t decide. These have to be cultivated at an interior level. This paralysis comes to afflict much of the old problem solving machinery on all levels. Where old models break down, one builds models of them, treating them at second hand as phenomena. Where these breakdown, meta meta models come in. The number of such layers is very small in practice, although potentially unlimited in principle. (The second order functional calculus is enough for "about” anything.)
The X falls apart in this way into many sub-X's. Methodology, art forms, science, etc., all become infected with the "disease" (of paralysis) in different forms.
But after the sub-sub-X's are acted out far enough, they begin to close off and recombine. The way a particular sub-X in the arts yields a stable form of repetition, contributes to a return.
But the return is in a sense a way to live with and use the paralysis, not get beyond it. The stable way through a well-formed X involves theory formation, and a more refined self-conscious decision replacing relatively blind over-general heuristics. The "paralysis" is in a sense used. Old "instincts" don’t work anymore, and the paralysis opens us to a more self-conscious individual and group decision making.
Beginning steps at an art level push on old forms to intensify the X. Art helps draw people into the paralysis where there are latent inner beginnings of a resonance, or stable way to handle old methods that break down. The improvisational parts of art begin a bringing into parallel. Imagination abstracts sub-aspects of a resolved X. Gradually these recombine as a new sense of the real.
The theme that these do recombine is related to the additive nature of learning, and the objectivization theme. The field of possibles in the style cycle can be charted, and reveals not chaos (as the middle class believe), but a definite structure. Improvisational art catches the first bits of structure. As one stands back from this improvised catching of repetition, one discovers principles and via these, deliberate composition becomes possible.
That the improvised pieces slowly combine into a cascaded method of objectivizing the full decision process, is related to what one means by the doctrine of simplicity. The central issue is catching or reforming (further qualifying) the basic generators of the field of possibles (the root invariants (of physics)), not elaborating ingeniously, de facto. The ingenious elaboration bit is relatively rapid and easy, once an adequate theory is found. The new simplicity is itself first found as a sub-aspect of the self-defeat of this complexity. The first roads into the simple are crooked. But once found, the simple reaches a wider context possessively.
At the personal level, this re-simplification is related to being able to face the full group (or full picture) again. The social X, first faced abstractly, in sub-bits, or in "imagination", finally recombines to full autobiography. The full X faced as real, becomes science. The many analogies with methods of X resolution in other areas, or in the past, catalyze or force on a compounding of present X formations in many ways (asking, e. g., "What do you mean? ") till a way through is found.
12. Set Theory and Money (Capping the GS)
Broadly speaking this way through involves building models of models till all qualitative differences are reduced to types of relatedness. The whole process never breaks the surface of history, wherein the answer always lies. But the process of manipulating symbolic models of models of ---- brings in pieces of history at quite a distance from practice.
Set theory is to be interpreted not as sets of things, but as sets of pure labels, all intrinsically equivalent, via which all qualities are reduced to types of relatedness. The process of X formation, and its break up into a hierarchy of many quasi-independent forms, divides and pushes back in an existential way till the pure label is pulled out, via physics. Then all qualities are ultimately … [Line of text here is unreadable]… in "pure label" terms.
The field of real now possibles involves the acts themselves. These are ultimately systematized via physics. Imagination is vastly greater than the real world, the real world being defined as a sub-aspect of it via the category repetition.
Set theory articulates an ultimate field of possibles, by giving plenty of room to define the qualitative categories that generate and span decisions as types of relatedness. At the other extreme, money values, or more precisely socially necessary labor time (Marx and Ricardo) provide an ultimate formalization or quantification of the value problem at the global level. The drives are ultimately objectivized in a unified way in relation to the freeing of labor time. Of course, the main game lies in the X, and in the related reforming of the economic game.
Nevertheless, insofar as things are solved out, a $ sign can be put on them. (Even the dynamics of opening of the X can be gotten at in dollars indirectly, as Jane Jacobs shows.) Dollars define a basis of compromise that is objective, the compromise of the engineer. The art of life is the art of compromise. Our constitution is a pile of compromises. But compromise in this sense, is a matter of truth, not with truth. The money theme is, indeed, a key one in extending and making precise the honesty theme in ambiguous areas (as Lincoln saw). There is a depth and energy in middle class money values, that has not yet been fully dug out (see e. g. Ann Rand). The trouble with middle class values is not the money theme, but their failure to carry through on it. Liberals, in attacking money values, end up in an impotent quagmire. Money, payments for your role, is the formal condition that activates the role’s heuristics. Money goes to the ultimate depth of motivation, de facto. A role that does not have social reinforcement, i.e., money, (or direct spin-offs from it) — can’t survive.
The new must individuate itself till it can make its way in money terms, to take hold at all. Money, move and set element form a transcendental trinity for orienting the dynamics of the differentiation process, a trinity not tied to any specific level of technicalization .
The money theme, in its "outrage” side, correctly pushes the X to its consequences, all the way. This is the function of the centralization process. It is there to generate outrage or make dream (by elaboration for its own sake). It is ideal as farce (as Peter, Potter, and Parkinson see).
Ultimately the life adventure is oriented as the opening of the laws of physics and physics emerges as life-style as well as fact. (The cultural jigsaw puzzle falls together when one relates cultural insight to causal self model insight needed for a related task.) The dissolving (destructive? ) power of money thinking on old cultural forms is such that there is no stopping place short of breakdown, and a new primitivism, or the laws of physics themselves. It is the emergence of science as powerful enough to provide a global synthesis and self-models (not just a bag of odd, isolated facts of destructive power when tied to money) that has saved us from the destructive breakdown that money thinking has meant in the past.
13. Some Misguided Computer Analogies
Analogies between brains and conventional computers have been an important source of guidance in developing brain models. But they are also sources of misguidance, as the design of human computers is different from that of computers we build in important ways.
Automation starts with a verbal or symbolic type of encoding, using lists for possibles, etc., which is a very late emergent in living computers. As a result we find ourselves led into simulating non-verbal models with verbal ones, which can be very inefficient. There is no need to represent symbolically what is already available existentially in the analogue or digital parts of the human computer. The act takes care of itself, so there is no need for the symbol-processing parts of the brain to provide a determination of acts in any general way. In the brain these verbal aspects parasite on pre-verbal functions of an elaborate model-like character, such as is implicit in the GS. To attempt to set this up in verbal terms, though abstractly possible, is in practice unthinkably clumsy.
Yet it is precisely this sort of thing that is so often suggested when people first try to bring computer analogies to bear on problems of how the brain functions. For example, orientation independence is approached by representing the object via the whole group of transformations to which it may be subjected (Arbib-78), rather than by building on invariant correlations from the first, and then filling in from a few extremum cases.
Skinner's beautiful book on verbal behavior has been sharply criticized for what, from this point of view, is its great strength. The narrowly verbal aspects of brain function are limited and parasitic. Because of this, Skinner's simple account is vastly deeper and more complete than seems at first possible. To have attempted to deal with "deeper" issues in narrow verbal terms would have been to miss them altogether.
Mutual exclusion is a theme that runs through neural organization at many levels, the level of acts, drives, wholegrams, etc. That any part may take control, or the distributed nature of the control, is related to the mutual exclusion theme. It does not require another independent mechanism to centralize decision (Arbib 193). Mutual exclusion depends upon the fact that only certain well separated paths get a chance to enter into repetition. The centralization theme derives from properties of the correlations fed into the brain, and not from independently encoded mechanisms, although the genetic pre-encoding does support this theme.
Similarly the role of thing and background can be treated as pre-given in the form of the memory units. It can also be treated as emerging as a result of the properties of the correlations fed in, the role of imitative coupling, etc. Of course, there is a genetically pre-given class of correlations that is such as to accelerate this type of emergence. But this is very different from a rigid pre-given form. The GS theme also depends on the type of correlations fed in.
The compulsion to read pre-givens into the brain is reinforced by philosophies of mathematics that read special properties related to brain functioning into the foundation of logic. When this is done, there is a tendency to view corresponding aspects of the brain’s information processing as pre-given.
The moral is to keep out of the foundations of logic insights into optimal ways to process information in our world (e. g., the analysis into things that carry properties, or the space-time frame that carries ingressing universals, etc.). This point will be discussed later in relation to the foundation of mathematics.
14. On the Relation of Neuro-modeling Language to Automata Language
Broadly speaking, automata language is too rigid. It seems to imply some pre-given goal that cannot be changed, and pre-given structural units, to bring its language to bear on decision making. Yet it can simulate a rich variety of meaningful human endeavors (chess playing, problem solving in Euclidian Geometry, etc.).
On the other hand, neuro-modeling appears too local and flexible, leading off into too vast a swarm of possibilities. It can be used to simulate certain surprisingly refined local interaction effects (See S. Grossberg), and a variety of "separating out" processes when reinforcements are "properly tailored,” but is absurdly clumsy if you try to simulate real computer programs (chess playing) in its terms because no "proper tailoring" can be found.
The theory of games provides a way to translate between these vocabularies. It provides a rigid pre-given goal function (repetition) and structural units (the move) via which a translation to automata type language can proceed. It gets at local effects via nextness themes and the end point theme, so as to give to the conditions of reinforcement a kind of proper tailoring to allow of the needed emergence.
One person game theory also helps understand the relation of structure to learning. Which is more important or basic is an empty question. Genetic emphasis on the type of correlation that will be useful, and the genetically pre-encoded heuristics of repetition, are essential to make emergence of learning possible at all. On the other hand, learning, in the broader sense of an emergent ordering of behavior that depends upon repetition, appears all inclusive. But this ties down what one means by learning in a very special rigid way that could be called structural.
If a field of possibles can be treated as genetically pre-given, it is also so vast that it might "almost" as well be open. If it were ever not vast enough, it could easily be enlarged. Learning and structure are complementary yet each all inclusive in the one person game framework that orients emergence.
15. Some Mechanical Counterparts
a. Excision experiments (as in Functional Model of the Nervous System (FM)
b. Emergence of visual gestalt (as in FM and re: Taylor)
c. Caught in stop and start (as in FM and NS VII - 119).
Pain, as harboring old brain remnants, may involve primitive forms of mutual exclusion that are inadequately under higher control. A more lasting activation of branches may help support a lasting avoidance. Acupuncture might then be viewed as acting via primitive mechanisms of lower area (medulla) exclusion.
d. The reticular formation
The reticular formation is the oldest, most basic part of the brain. It can perhaps be thought of as linking together present acts. The so-called "poker chip" structure of the reticular formation is ideal for this purpose, linking as it does "everything to everything,” now.
This simple function may be the most basic one. Its massive reduction of the range of combinatorial possibility, begins to bring things within a range where more specialized processing can be applied. Some 90% of the sensory inputs to the spine stop in the lower levels, here on the way up.
This interlinkage function has a central coordinating role. On the one hand, it is highly specific, as the representative of the present act, and the means of spreading its characteristic rhythms. These depend, of course, on the function in which they occur, as well as what they are in external terms. Timing effects play a vital role in the build-up of coincidence effects via which factors compound to uniqueness. Via the many unique incidentalities of selection, higher order controls may enter in. They enter in where there is a relative ambiguity or mutual suspension.
Simpler linkages between sensory input and response do not seem to depend as much on reticular activation as was once supposed. The activation of basic drive factors in the hypothalamus are what is most rudimentary in wakefulness. But the reticular formation must also be activated if learning is to take place or if the repertoire is to exhibit the sort of normal flexibility which is akin to learning. Its activation is essential to habituation. Interlinkage (via the reticular formation) may help lead to habituation, by "acting out" the latent motor tendencies of the given input to rejection.
The generalized tropism theme helps provide a more concrete model of the habituation process. It serves to define a class of context dependent motor links, whose inhibition will inhibit the sensory source term. The reticular formation may induce the needed dependent self-cancelling linkages where they lead to a relatively high level of multiple response. It is only those paths that stay confined get through. If they don’t, then some branch of the self-cancelling compounding will be enhanced to the point needed to get through.
e. Built-in compensation
The many forms of built-in compensation or constancies serve as helpful or essential steps in building up invariant conditioning (as in FM). Not crawling at birth leads to later reading problems. The reason for this is as follows. The more invariant the encoding at its roots, the higher the elaboration it can take without breaking down. What is more attitude dependent breaks down more easily. (Poets are slow language learners.)
f. Pre-encoding
There is a big pre-encoding that gives a preference to signals that just changed. This helps break through functional sets (See Stephen Grossberg’s articles), keeping us alert to the new, while doing other things.
g. Mid-brain
The mid-brain organizes the body frame of reference, and begins a "where" that is quasi-independent of externals. It helps set up a functional motor root later taken into transference by specific cues.
The hypothalamus is the basic drive source, and, as such, the source of wakefulness (not the reticular formation). Via the limbic system, it posits its relevance on cues.
Attention is organized by relations of mutual exclusion. Aspects of it are independent of drive factors, but taken over by them. Sometimes the center of attention is outside the external now present, leading back in jumps by chained elaboration to this present, or from this present on till a picture is completed enough for the act to be selected. The quasi-repetition associated with the X is objectivized as many cycling stages of picture filling and attention jumps.
h. The temporal areas
The temporal areas harbor the highest order of compounding of factors from all modalities. It is at this highest level that we often start selecting a through path for satisfying a drive first activated in lower centers. The actual act is filled in via objectification steps. Frontal areas help activate the future or plan needed for drive satisfaction. They integrate delayed responses (without being the source of delay), delays needed to fill in plans that may often begin with temporal cues. The occipital often provides fine corrections to coordinate the present act. The alpha rhythm "ends" here in the occipital, and is strongest here, starting below. All areas interact in forming the over-all gestalt.
The secondary motor area in the temporal lobes are related to thought about acts, via which acts are defined more broadly and functionally. There is another motor related area in the longitudinal fissure that seems related to fixing long term motor attitudes that support present plan, via the limbic system. Eye movements play an important catalytic role. They may act as a guiding filter that converts nextness to body related data. Eyes turn first, then needed head and body reorientations are often easy functions of eye movements.
i. The "self-engagement” of motor acts
The "self-engagement” of motor acts plays an important role. The many visual and peculiar kinesthetic cues that are correlated to and so come to engage the act itself, provide important "bases" from which to extend the act or series elaborate.
That only cues which form parts of learning cycles have a chance to enter the internal compounding, is a vital filter that makes the remaining SE more manageable and more relevant. The fact that CR's depend upon prior facilitation may mean that the self-engagement factors tend to be first independently caught as factors on the sensory side that accompany the act, and then used to condition the act, when they are pre-engaged by some SE. The pre-engagement helps prepare the ensuing act, limiting the role of the present to timing it, or some refinement of details. There is a preceding alpha rise in the spine, before the act. So, too, there is a two layer structure inside the brain with a preceding rise that facilitates the act type that can be seen in the EEG. The lack of such a two layer system in the autonomic side makes its responses less subject to imaginative evocation or regulation, and directly attached to the conditioning factors.
j. The cerebellum
The cerebellum has a variety of delays to smooth the present act, with the same sensory inputs as go to the cortex. Given an already independently defined type of act, little added refinement is needed in the cerebellum, to shape timing and reinforcement levels. Not much additional compounding is involved, as the cerebellum can start with signals correlated to the act itself. A link has merely to be relatively reinforced, in the short time constant context of the coordination of the given act. The coordination does not depend on feedback, as there is not time enough. Acts are often "ballistically" engaged, in a first approximation. There is an initial muscular tension, and a delayed "stop" to the tension, that are a function of the distance to be "reached,” etc. This is smoothed in with sensory side linkages. Because no selection among discreet isolated channels is involved, a method of steepest descent may be used to guide a large part of cerebellar smoothing action. The role of the motor area in the longitudinal fissure may be related at long time constant averaging to that of the cerebellum at short time constants, picking out, by a method of steepest descent, attention functions that minimize random search, or maximize an overall reinforcement. Small area or timing or other variations about a present response could easily lead slowly into major changes, when certain variations are reinforced more than others. It should be remembered that each set of "equivalent situations" has its own quasi-independent shaping.
k. Special nextness type cells
The special nextness type cell is important (the generalized somatotopic theme) till the decision is made. These can involve specialized abstractors over narrow or wide areas, and their linkage. (The possibly often inhibitory role of horizontal cells help introduce selective patterning.) Patterns in the frontal (and amygdala) areas, where decision is made or act type selected, are more diffuse and perhaps less structured in specialized ways, as all functional combinations of acts need some availability.
The narrowing of neural groupings with age, the progressive differentiation is related to progressive myelination which makes responses more independent locally. This increases to the age of fifty. Details must be coordinated by then, to keep aggressive, or the falling apart just paralyzes. Hospitalism is a related phenomena from early years. There, rising levels of inhibition require a greater integration of parts, or the individual dies.
l. Information types
Information types further purify at higher levels, over what they are at the reception centers. This just repeats the theme that separation is, ultimately, not just a function of inputs, but also of the selective role of the internal dynamics. The inputs are a kind of pre-acceptance of this.
In addition, there are many special responses or release mechanisms that involve combinations of learned responses and genetic pre-encoding. These special responses must fit into the over-all dynamics, for if once countercompensated, they are often so encoded that they never re-emerge.
16. Links to the Broader Picture and Values
(Briefly Suggested or as outline notes that refer to, or remind of, earlier versions).
a. Quantum theory and Everett-Wheeler
The theme that quantum theory deals with knowledge about fact, not facts (Heisenberg) is not implemented in the present physics formalism. This NS model provides a way to do this. Discuss the Everett Wheeler model as a way to make explicit the orienting role of the word, and of the field of possibles (ultimately the Universal model).
b. No problem with self-reference
There is no problem with self-reference, in part because subjective models control acts directly, as we enact our theory of the repeating path. Correlation or CR’s that serve to resolve a theoretic contradiction are also effective and "automatically" do take hold at the behavioral level.
c. The act or move is basic
The abstract, context dependent move or act (a dependence defined functionally via learning cycles) is basic. The act is made real by enactment or policy, not observability (which it does not have in a direct sense). There are no "simples,” the contradiction telling how far down to go.
The dynamics of X discharge involves its breakup into a hierarchy of quasi-independent sub-X’s. As certain of these sub-X’s are resolved, their answers combine, and lead in the end to the resolution of the originating social X. The pattern of break-up of the X involves a going back into history, where adequate resolving cues are found. The over-all implicit analysis involves the break-up of all qualitative categories into types of relatedness, via often indirect identification in a field of possibles (ultimately physics defined).
In this indirect way the generating factors in competitive heuristics are objectified resolving the X. That this process works is related to the guiding role of simplicity. The basic X is with ourselves.
Another perspective on the over-all dynamics is afforded by the theory of control variables of lasting change or institutional change. Such control variables only take the form of causal (self-model) insight. Only CR's that constitute causal insight (relative to the given field of possibles) are effective in X resolution.
Only by making use of such themes as these, that allow of an integrated overview of the over-all dynamics, can defining everything via its relation to the abstract (context dependent) act be used in a working way to organize practical decision making. It is hard to carry through on the abstract act theme, consistently. That is why many toy with the theme only to, in effect, nullify it in the way they use it or complete the picture in its terms.
Insofar as the over-all goal function is applied directly, the decision making system is ultimately treated or implied to be uninterpreted. But sub-goals involve interpretation to foreshorten to workable preparations the uninterpreted system (that is merely enacted or indirectly implied, not described in explicit detail). Because the over-all goal is grounded by enactment rather than interpretation, meaning does not depend on consistency. Reconciled sub-goals "can” be consistently interpreted together when necessary to resolve X’s, and consistency is a sensitive test of adequate reconciliation. Sub-goals "can" be defined within the generalized search. But the over-all goal cannot, and must not be treated directly (as repetition) in act terms. There are thousands of latent consistency tests to push towards in breakdown, and repetition selects out the one of orienting concern. To push for greater consistency at random would just generate cascades of useless X's leading nowhere but disaster. It is enough that striving for greater consistency is a helpful guide in dealing with specific X’s already pushing on us.
d. The World as Turing machine
The world can be analogized to a Turing machine, with the world functioning as an undefined and implicitly infinite storage. (Defined or treated as finite storage introduces limitations that are not appropriate .) The infinite is implied as a way of representing of the felt as felt. Such representation (needing the infinite) is always a lie, because our felt world is incomplete, and to represent (or act on it) is to treat it as complete, ignoring breakdown edges, etc. The infinite is put in to be taken back out (as is the function that may serve as its own argument) when the full details are spelled out.
"World” is a sub-aspect of the uninterpreted system, related to the category repetition and the foreshortened methods of sub-goal coordination via the GS. "World" is not at all consistent. Limited aspects of policy are capable of being so made insofar as necessary, via adequate individuation.
e. Capacity is never an issue
Note that capacity is never an issue. Life is ahead of theory in the rediscovery of the self in new terms. Capacity needed for this is easily built in. What limits capacity, is the need for spanning heuristics. Too much is dangerous.
f. The pressure of “What do you mean?”
The New World is being driven up by the pressure of, "What do you mean?" We can now carry through when necessary (very indirectly at times) on defining concepts in ordering terms. Church’s theory of recursive functions, that side-steps the problem of interpretation (leaving it for after, and to go any which way), is ultimately justified by this as an adequate orientation. The world (as felt) emerges as an "emanation of the word,” the stability and form of the felt depending on verbal theory. This helps give a deeper meaning to the Church- Turing theme.
Intension, in the sense of the universal of consciousness, derives from an implied extension, by finding stable rules of choice, or repeating paths, among the context dependent acts. The act or move may be regarded as an ultimate basis of extensive representation, but in practice we find adequate representation far short of that, in the half-way house of world or physics.
All the felt involves extensive interaction with the environment. The genetically pre-encoded is too limited in itself to give more than subconscious responses.
g. Mathematics
The three math schools as in earlier papers. Start with set theory as motive. Re-define set element as pure label.
Math is interpreted via the act and physics much more concretely. But if the more concrete (existential and causal) interpretation (using game theory) contains set theory, it, of course, contains everything, (as part of our ideal physics representation of statistical possibilities, as the permutations of 20,000 words from Webster's contains all poems and books ). The role of the word can, of course, be gotten at with a kind of double indirection, as well as directly as enacted.
Logic has an eternal non-developing character outside progress (as Kant, Bochenski, and Levi Strauss note from their different points of view).
The West tends to confuse the process of calculation with all of mathematics. But even in calculating, induction provides a vital guide. In addition induction involves things, even inside explicitly defined mathematical systems, that cannot be interpreted as calculations, as Goedel shows. In practice this induction process is of crucial guidance, even when one can later support its results by calculation (which one often cannot, even in simple important results, such as, the four color problem).
h. The importance of the paradoxes
The paradoxes help force the recognition that there are no simples prior to repetition to justify one form of repetition over another. They force us back on the uninterpreted system, recognized as basic because existential or enacted.
The middle class role is the source of the compulsive illusion of a need to reify, or to get a foundation prior to repetition.
The "existential infinite" is a fine example of the absurdity to which such a reifying method of talking leads. It is also a fine example of a certain depth of honesty in the middle class position, setting forth its own absurdity with such unabashed directness. This absurdity is, of course, essential, as these categories are put to work and so must be such as to articulate the decision process as actually experienced in a working way. The middle class had to claim a logical basis for side choosing, at a time when their role in so doing was not yet understood or subordinated. Their honest metaphysics must reflect this claim.
Do all improvable theorems involve impredication? Could it be that one may at least conjecture that one can get at step-like substitution paths (via a theory of all possible related models) in all cases except where impredication outjumps us?
i. The social roots of problem solving
Problem solving has social roots. The personal X is bom of the social and discharges to it. Branches of the X are divided between people. The link to the world is by imitative coupling not observation. Nature as emanation of the word, and its relation to Buber’s themes.
j. Problem solving dynamics
The analysis of the problem solving dynamics can be broken up into three stages, separated by two dark nights (following John of the Cross). The Cross and three "empty” or "node" themes, etc.
The relation of sex to problem solving.
The relation of the X as objectified to factual error. The relation of the solution process to individualism. The not yet objective (often contradictory) factors in decision only gets qualified to the point of becoming objective inside the isolated individual. The withdrawal as for the return.
History as a history of physics. No debunking, as concern is a function of the X.
This is a cooperative world. The Tory Anarchist theme.
k. Money
Money and the total formalization of the value problem (Marx) (Beth suggested may be possible). Depth of the middle class money theme (Rand). Liberal error in rejecting money values.
The social and personal dynamics is oriented by the… [Source document ends]