Following the formal analogy provided by the "laws" of motion (i.e., the "rules" for describing motion), we now look for the class of phenomena which are appropriately described as changes in zero learning (as "motion" describes change of position). These are the cases in which an entity gives at Time 2 a different response from what it gave at Time 1, and again we encounter a variety of cases variously related to experience, physiology, genetics, and mechanical process:
(a) There is the phenomenon of habituation—the change from responding to each occurrence of a repeated event to not overtly responding. There is also the extinction or loss of habituation, which may occur as a result of a more or less long gap or other interruption in the sequence of repetitions of the stimulus event. (Habituation is of especial interest. Specificity of response, which we are calling zero learning, is characteristic of all protoplasm, but it is interesting to note that "habituation" is perhaps the only form of Learning I which living things can achieve without a neural circuit.)
The most familiar and perhaps most studied case is that of the classical Pavlovian conditioning. At Time 2 the dog salivates in response to the buzzer; he did not do this at Time 1.
(b) There is the "learning" which occurs in contexts of instrumental reward and instrumental avoidance.
There is the phenomenon of rote learning, in which an item in the behavior of the organism becomes a stimulus for another item of behavior.
There is the disruption, extinction, or inhibition of "completed" learning which may follow change or absence of reinforcement.
In a word, the list of Learning I contains those items which are most commonly called "learning" in the psycho-logical laboratory.
Note that in all cases of Learning I, there is in our description an assumption about the "context." This assumption must be made explicit. The definition of Learning I assumes that the buzzer (the stimulus) is somehow the "same" at Time 1 and at Time 2. And this assumption of "sameness" must also delimit the "context," which must (theoretically) be the same at both times. It follows that the events which occurred at Time 1 are not, in our description; included in our definition of the context at Time 2, because to include them would at once create a gross difference between "con-text at Time 1" and "context at Time 2." (To paraphrase Heraclitus: "No man can go to bed with the same girl for the first time twice.")
The conventional assumption that context can be repeated, at least in some cases, is one which the writer adopts in this essay as a cornerstone of the thesis that the study of behavior must be ordered according to the Theory of Logical Types. Without the assumption of repeatable context (and the hypothesis that for the organisms which we study the sequence of experience is really somehow punctuated in this manner), it would follow that all "learning" would be of one type: namely, all would be zero learning. Of the Pavlovian experiment, we would simply say that the dog's neural circuits contain "soldered in" from the beginning such characteristics that in Context A at Time 1 he will not salivate, and that in the totally different Context B at Time 2 he will salivate. What previously we called "learning" we would now describe as "discrimination" between the events of Time 1 and the events of Time 1 plus Time 2. It would then follow logically that all questions of the type, "Is this behavior "learned' or "innate'?" should be answered in favor of genetics.
We would argue that without the assumption of repeat-able context, our thesis falls to the ground, together with the whole general concept of "learning." If, on the other hand, the assumption of repeatable context is accepted as somehow true of the organisms which we study, then the case for logical typing of the phenomena of learning necessarily stands, because the notion "context" is itself subject to logical typing.
Either we must discard the notion of "context," or we retain this notion and, with it, accept the hierarchic series—stimulus, context of stimulus, context of context of stimulus, etc. This series can be spelled out in the form of a hierarchy of logical types as follows:
Stimulus is an elementary signal, internal or external. Context of stimulus is a metamessage which classifies the elementary signal.
Context of context of stimulus is a meta-metamessage which classifies the metamessage.
And so oil.
The same hierarchy could have been built up from the notion of "response" or the notion of "reinforcement."
Alternatively, following up the hierarchic classification of errors to be corrected by stochastic process or "trial and error," we may regard "context" as a collective term for all those events which tell the organism among what set of alternatives he must make his next choice.
At this point it is convenient to introduce the term "con-text marker." An organism responds to the "same" stimulus differently in differing contexts, and we must therefore ask about the source of the organisms's information. From what percept does he know that Context A is different from Con-text B?
In many instances, there may be no specific signal or label which will classify and differentiate the two contexts, and the organism will be forced to get his information from the actual congeries of events that make up the context in each case. But, certainly in human life and probably in that of many other organisms, there occur signals whose major function is to classify contexts. It is not unreasonable to sup-pose that when the harness is placed upon the dog, who has had prolonged training in the psychological laboratory, he knows from this that he is now embarking upon a series of contexts of a certain sort. Such a source of information we shall call a "context marker," and note immediately that, at least at the human level, there are also "markers of contexts of contexts." For example: an audience is watching Hamlet on the stage, and hears the hero discuss suicide in the con-text of his relationship with his dead father, Ophelia, and the rest. The audience members do not immediately telephone for the police because they have received information about the context of Hamlet's context. They know that it is a "play" and have received this information from many "markers of context of context" — the playbills, the seating arrangements, the curtain, etc., etc. The "King," on the other hand, when he lets his conscience be pricked by the play within the play, is ignoring many "markers of context of context."
At the human level, a very diverse set of events falls within the category of "context markers." A few examples are here listed:
(1) The Pope's throne from which he makes announcements ex cathedra, which announcements are there.. by endowed with a special order of validity.
(2) The placebo, by which the doctor sets the stage for a change in the patient's subjective experience.
(3) The shining object used by some hypnotists in "inducing trance."
(4) The air raid siren and the "all clear."
(5) The handshake of boxers before the fight.
(6) The observances of etiquette.
These, however, are examples from the social life of a highly complex organism, and it is more profitable at this stage to ask about the analogous phenomena at the pre-verbal level.
A dog may see the leash in his master's hand and act as if he knows that this indicates a walk; or he may get in-formation from the sound of the word "walk" that this type of context or sequence is coming.
When a rat starts a sequence of exploratory activities, does he do so in response to a "stimulus?" Or in response to a context? Or in response to a context marker?
These questions bring to the surface formal problems about the Theory of Logical Types which must be discussed. The theory in its original form deals only with rigorously digital communication, and it is doubtful how far it may be applied to analogue or iconic systems. What we are here calling "context markers" may be digital (e.g., the word "walk" mentioned above) ; or they may be analogue signals — a briskness in the master's movements may indicate that a walk is pending; or some part of the coming context may serve as a marker (the leash as a part of the walk) ; or in the extreme case, the walk itself in all its complexity may stand for itself, with no label or marker between the dog and the experience. The perceived event itself may communicate its own occurrence. In this case, of course, there can be no error of the "menu card" type. Moreover, no paradox can be generated because in purely analogue or iconic communication there is no signal for "not."
There is, in fact, almost no formal theory dealing with analogue communication and, in particular, no equivalent of Information Theory or Logical Type Theory. This gap in formal knowledge is inconvenient when we leave the rarified world of logic and mathematics and come face to face with the phenomena of natural history. In the natural world, communication is rarely either purely digital or purely analogic. Often discrete digital pips are combined together to make analogic pictures as in the printer's halftone block; and sometimes, as in the matter of context markers, there is a continuous gradation from the ostensive through the iconic to the purely digital. At the digital end of this scale all the theorems of information theory have their full force, but at the ostensive and analogic end they are meaningless.
It seems also that while much of the behavioral communication of even higher mammals remains ostensive or analogic, the internal mechanism of these creatures has become digitalized at least at the neuronal level. It would seem that analogic communication is in some sense more primitive than digital and that there is a broad evolutionary trend toward the substitution of digital for analogic mechanisms. This trend seems to operate faster in the evolution of internal mechanisms than in the evolution of external behavior.
Recapitulating and extending what was said above:
(a) The notion of repeatable context is a necessary premise for any theory which defines "learning" as change.
This notion is not a mere tool of our description but contains the implicit hypothesis that for the organisms which we study, the sequence of life experience, action, etc., is somehow segmented or punctuated into subsequences or "contexts" which may be equated or differentiated by the organism.
The distinction which is commonly drawn between perception and action, afferent and efferent, input and out-put, is for higher organisms in complex situations not valid. On the one hand, almost every item of action may be re-ported either by external sense or endoceptive mechanism to the C.N.S., and in this case the report of this item be-comes an input. And, on the other hand, in higher organisms, perception is not by any means a process of mere passive receptivity but is at least partly determined by efferent control from higher centers. Perception, notoriously, can be changed by experience. In principle, we must allow both for the possibility that every item of action or output may create an item of input; and that percepts may in some cases par-take of the nature of output. It is no accident that almost all sense organs are used for the emission of signals between organisms. Ants communicate by their antennae; dogs by the pricking of their ears; and so on.
In principle, even in zero learning, any item of experience or behavior may be regarded as either "stimulus" or "response" or as both, according to how the total sequence is punctuated. When the scientist says that the buzzer is the "stimulus" in a given sequence, his utterance implies an hypothesis about how the organism punctuates that sequence. In Learning I, every item of perception or behavior may be stimulus or response or reinforcement according to how the total sequence of interaction is punctuated.
Was this article helpful?