Can There Be a Science of Education?

Edward George Rozycki, Dissertation, Ed.D.
Temple University, 1973

edited 9/15/10

See related articles:
The Functional Analysis of Behavior: theoretical and ethical limits ;
Measurability and Educational Concerns ; also
The Philosophical Foundations of Human Cognition

Introduction: Can Science Help?

In 1773 Denis Diderot visited the Imperial Court at St. Petersburg to personally thank Catherine for the patronage which was to spare him an impoverished old age. His lively atheistic polemics were both the delight and the consternation of the Court. So that he not remain unremonished by his Christian hosts, it was arranged that he be present at a "mathematical" demonstration of the existence of God. The mathematician Leonhard Euler was to make the initial statement in what was hoped would be another entertaining debate. Euler, addressing himself to Diderot, declared with a tone of perfect conviction, "Monsieur, (a+b)n/n donc Dieu existe: respondez!" Diderot, totally ignorant of mathematics, was dumbfounded. He was granted permission immediately to return to France (DeMorgan).

In its broadest conception, this dissertation bears on the authority of the expert in matters beyond the immediate domain of his expertise. As the anecdote above illustrates, the range of that authority may well depend upon factors other than the rigor of his arguments.

The authority of putative experts in Education rests substantially on a host of untenable philosophical assumptions. Technical terms from the sciences are pressed into service and are misleadingly overextended, metaphors at best. The misuse of such terms as reinforcer, measure, function and probability -- to mention but a few -- has fostered a false belief in the relevance of certain research and theory to educational concerns. The presently popular demand for technical precision in educational endeavors assumes the applicability of quantificational techniques to educational outcomes. But such an assumption is highly questionable.

Spurious philosophy alone is the foundation of many important educational claims. For example, Skinner (1971, p.31) asserts that the nature of reward has been exhaustively studied in the experimental laboratory. J. Cohen (1969) identifies operantly conditionable behavior as voluntary behavior (Schleifer, p. 163). Stake (p. 97) would make it the responsibility of curriculum evaluators to "transform the behavior of ... teacher ... and ... student into data." Whitmore claims that only measurable educational goals are assessable (Mayer, p. 3).

The plausibility of such claims rest on unexamined philosophical assumptions of the sort advanced by Thorndike: "Whatever exists, exists in some amount" (Joncich, p.151). Other theorists allude to a world of observables (Nunnaly, p.97); terms such as "anxiety" are mere symbols having at most heuristic value (Nunnaly, p.97) or denoting fictions (Skinner, 1953, p. 27). Observables are pre-theoretically accessible to observation, i.e. they are merely seen (Anderson, p.7). Science can begin with our everyday concepts because in order to live efficiently "our ordinary language must take account of the causal forces at work in the world" (Stinchcombe, p.41). Description is a pre-theoretical activity by which we identify observables (Anderson, p.7). These and others like them are not merely ideological excrescences on otherwise viable theory; rather such assumptions are appealed to to justify restrictions on what is presented as properly scientific endeavor.

Skinner provides the most succinct statement of the conception of Science that I will be concerned with. Since the arguments of this dissertation are not meant to be restricted to a critique of Behaviorism, the word "external" has been removed from the following quotation -- three asterisks have been inserted to indicate its original positions:

The *** variables of which behavior is a function provide for what may be called a causal or functional analysis. We undertake to predict and control the behavior of the individual organism. This is our "dependent variable -the effect for which we are to find the cause. Our "independent variables" -- the causes of behavior -- are the *** conditions of which behavior is a function. Relations between the two -- the cause-and-effect relationships" in behavior -- are the laws of a science. Synthesis of these laws expressed in quantitative terms yields a comprehensive picture of the organism in a behaving system. (Skinner, 1953, p.35).

The main argument of this dissertation is this: one cannot define a behavior-variable for which the distinction between voluntary and involuntary behavior can be preserved. Thus there can be no functional analysis of voluntary behavior. But the voluntary-involuntary distinction is necessary to provide the normative base for educationally -- and, I might add, legally -- relevant decisions. Therefore, there can be no Science of Education which deals with individual behavior.

Perspectives: The Identification of Behavior.

In this chapter we are concerned with the activities of description and identification. It in often the case that a theorist's covert philosophical commitments seriously bias his notion of description so that certain terms are supposed to be more properly descriptive than others. We reject as a model for the activity of describing the following: one matches a word with a picture, selecting those terms which refer to properties inherent in the pictured objects. There is so much untenable theory assumed in this notion that in using it to deal with human action we begin having "stacked the deck" -- so to speak. But this notion of description enjoys unquestioned acceptance in psychology and the social sciences and goes hand-in-hand with a notion of identification according to which identifying is -- ideally, at least -- applying a definition.

There is a real problem in attacking a viewpoint that is believed to have behind it the weight of both a scientific and a philosophical tradition: individual arguments, as, for example, that about the impossibility of defining everything, be they ever so clear, sound and persuasive, rarely damage the general creditability of that viewpoint. Problems, seen as unrelated, are relegated to the status of anomaly or treated as philosophic entertainment. Practically, this means they can be ignored. Thus an alternative viewpoint must be provided -- at least in sketch -- to frame the individual arguments within a whole, hopefully giving them a coherence and cogency they might otherwise lack. That is the point of this chapter.

Think of describing human behavior as something like bidding bridge hands. Describing, like bidding, may be done according to different conventions. John, in uttering something, may -- he need not -- be saying something. In saying something, he may again be telling Harry something. In telling Harry something, he may be informing him of something. And in informing him, he may be frightening him also. If we simple-mindedly take all of these as descriptions of things which we can see -- according to some notion of seeing as visually intaking -- then it is no wonder that theorists like Nunnaly and Skinner are wanting adequate notions of human action. Although, for example, Skinner sloganizes that "physical criteria" are necessary to identity behavior (1953, p.36), he says within the same page both that physical characteristics of the response and similarity in consequences define similar responses (1953, p.65); Nunnaly relies on intuition (p.88) to sort out otherwise arbitrary collections of observables into constructs. But we do describe John in saying that he is uttering, saying, telling, informing or frightening Harry. And we see him do it. But it is not without some substantial theory that we sort out his behavior into acts.

A second analogy is offered: a purpose is to an act as a sentence to a word. Chomsky's critique (1957, pp. 21-25) of a Markov-process-sentence-generator as adequate to provide a grammatical model for the syntax of a sentence can be applied -- mutatis mutandis -- to Markov-process models of behavior, e.g. Amidon-Flanders Interaction Analysis (1967). Adequate grammars need not enable us to determine individual words; why need a science of behavior enable us to determine individual acts?

The activity of identifying is mistakenly conceived as analogous to that of applying a definition: one checks to see if a particular object under consideration has those characteristics sufficient to identify it as a member of the class named by the definiendum. But not everything can be defined; at some point one merely recognizes some X to be a Y, either performatively by acknowledging it to be so - as when one learned what it was in acquiring the common language -- or by recal1ing it to be such. Idiosyncracies in identification may not matter in particular contexts: it is normally of no interest how we came to identify some as a Y; but only that we came to identify it thus. Such considerations indicate that the move from "X identifiable in context" to "X, identifiable in general" may be considerably more problematic than we are wont to suppose. Idiosyncrasy in identificational activity is compatible with community in language: we cannot infer from a commonality of linguistic usage, e.g. "intersubject-reliability," a commonality of criteria.

Hart 's notion of defeasible concept identifies a class of terms that are -- so to speak ~ permanent working hypotheses. Their application is warranted upon certain presumption; there are no sufficient conditions which entail that something is an X if 'X' is defeasible. It is argued in chapter four that voluntary act-types are defeasible, thus such acts are not caused vis-a-vis an even more general notion of cause than that offered by Skinner.


In contradistinction to Skinner's pretense to theoretical innocence, Coombs writes,

Our conclusions, even at the level of measurement and scaling (which seems such a firm foundation for theory building), are already a consequence of theory. A measurement or scaling model is actually a theory about behavior, admittedly on a miniature level, but nonetheless theory. (p. 5)

But it is at this miniature level that a theory of behavior is most significant.

An aspect of behavior is measurable if it can be conceived in such a manner as to provide a set of sets upon which a measure function can be defined. An aspect of behavior is probabilizable only if it is measurable. The minimal requirement for behavior to be able to figure in any kind of functional relation is that we be able to identify and sort out contrastive act-types, i.e. we must be able to define a behavior-partition - hereafter, BP. This is a set of act-types which are mutually exclusive and, in general, jointly exhaustive of all instances of behavior.

We will define a behavior-set, b, which is adequate to describe the voluntary and involuntary behavior of a model agent, A. It will then be demonstrated that b cannot be a BP.

A verb, 'X' -- with modifiers or not -- is an act-type if and only if, say, "John is X-ing" can serve as an answer to the question, "What is John doing?" By this criterion are excluded, for example, "owe', "require" and "own". Relative to our model agent, A, the following is true of b:

1. From the fact that 'A is X-ing' is sometimes true, it follows that 'X' is a member of b.

2. A can X if and only if 'X' is in b.

3. For some 'X' and 'Y' in b, A can X by Y-ing. For example, John can scare Mary by shouting.

4. For some 'X' and 'Y' in b, A can X in Y-ing, as when John asserts he is rich in saying, "I am rich."

5. For some 'X' in b, A can try to X, i.e. 'try to X' is in b.

6. For some 'X' in b, A can refrain from X-ing, i.e. 'refrain from X' is in b.

It is important to note that one cannot try to try to X, i.e. the language does not permit that formulation. Also, one cannot refrain from refraining from X.

In the commentary to section 2.01 of the Model Penal Code (Morris, p.112) it is stated that

The law cannot hope to deter involuntary movement or to stimulate action that cannot physically be performed.

If, for our purposes, we think of a legal prohibition as a command to refrain from doing something, there must be a presumption that one can so refrain. Similarly, to require that one do something must presume that it is something one can try to do. In light of these considerations, I will give two criteria for the identification of voluntary behavior:

V1: 'X' is a voluntary act-type if and only if 'X' is in b and 'refrain from X' is in b.

V2: If 'X' is voluntary, then 'try to X' is in b.

It follows from V1 and the description of b that some of A's acts are voluntary.

Let us continue with the description of b:

7. A believes he is trying to X, if he is trying to X.

8. 'X' is voluntary if and only if A is X--ing by trying to Y, where 'X' and 'Y' may be identical.

If, for 8., 'X' = 'Y', then 'X' is intentional; if 'X' is not the same as'Y',

then 'X' is inadvertent.

9. For every 'refrain from Xi' in b, there is a Yj different from 'refrain from X' such that A refrains from X by Y-ing.

Assume b to be partitioned and to contain at least one voluntary act-type, b'. Thus 'refrain from b' ' is in b. But by 9. there is some Yj such that A refrains from b'-ing by Y-ing. Since b is assumed partitioned, 'Y' and 'refrain from b" are identical, which is false. Also, since b' is voluntary, by 8. some instance of b' is also an instance of 'try to Y'' where 'Y' may be identical with b'. Thus 'b'' and 'try to b'' are identical, which is false. We cannot have both the partition and the voluntary behavior.

A second independent argument is this: assumed b partitioned and containing at least one voluntary act-type, b'. Now, depending on the beliefs and knowledge of A, the formula

'A is trying to b' by Y-ing'

need never be false for any 'Y' different from 'b" . For any 'Y' for which the formula is true, 'b' and 'Y' are not mutually exclusive, which contradicts our assumption of partitioning.

A refutation of the preceding arguments would require a demonstration that none of the conditions 1 through 9 is necessary to any voluntary act.

No functional analysis of voluntary behavior is possible.

Turner would maintain that

scientific explanations of behavior are in principle reducible to the more basic discipline of neurophysiology. Reduction implies that the theoretical; mediating constructs of one science are ultimately realizable in the language of an adjacent science. (p.23)

But from the preceding arguments it follows that concepts of voluntary behavior either cannot figure in a scientific explanation of behavior, or that such reduction is ultimately impossible. Because refraining and trying are not just another type of behavior, because one can -- so to speak kill two birds with one stone, so are voluntary act-types not reducible to combinations of "single events" (Skinner, 1953, p. 15) which are mutually exclusive and jointly exhaustive of all possible behavior. Thus is voluntary behavior neither measurable nor probabilizable. It cannot be operantly conditioned. Given the conception of Science we have been dealing with (Skinner, 1953, p.35), a Science of Education is impossible.

The Causes of Behavior

If 'cause' and 'effect' are conceived of as indicating, respectively, the independent and dependent variable in a functional relation, then since there is no general voluntary-BP, voluntary behavior is not caused.

If 'cause' is conceived as that through which something else is controlled or affected, then depending upon whether or not we attempt explanation in terms of human agency, we might get quite odd and incompatible causal analyses. If, for example, following Stinchcombe (p.31) we expect an effect to covary unidirectiona1ly with its cause, i.e. we can produce a change in the effect by manipulating the cause, but not vice versa, we find that we might be able to move our forearm muscles only by moving our fingers; but looked at in another way, it is the movement of the forearm muscles that causes the finger movement.

Another oddity is that we can control the flight of a golf ball perhaps only by controlling our follow-through. But, physical1y speaking, the follow-through can have no effect on the flight of the ball.

A singer may be able to control the pitch of his voice only by thinking the tone to originate outside his head. Can fictional causes have effects?

If 'cause' is conceived of as some state of affairs which, when obtaining, produces some other contingently related state of affairs, then since voluntary-act-types are defeasible, i.e. they are not predicated on the basis of sufficient conditions, they are not caused. Agents do not cause their acts.

A more subtle behavioral analysis is proposed by Ryle with his notion of disposition. To say, for example, that John is boasting from vanity is to say that his boasting satisfies the law-like proposition that whenever he finds a chance of securing the admiration and envy of others, he does whatever he thinks will produce this admiration and envy"(p.89). Such an analysis is clearly inadequate because neither the antecedent nor consequent conditions are unique to vanity. John is vain if his purpose in boasting is merely to secure the admiration and envy of others; but if he has other purposes, then although he, finding a chance to secure the admiration and envy of others, does what he thinks will secure this admiration and envy, he need not be vain. He might, for example, be trying to alleviate an acute feeling of inferiority.

Ryle's analysis in terms of purposes and mental states might appear to be quite remote from the simple behaviorism of a Skinner. However, such an analysis is quite congenial to a Reductionist program on the assumption that purposes can be accounted for by functionalist explanations. Such an assumption is made by both Turner and Stinchcombe. A functionalist explanation would have it that certain consequences of behavior "feed-back" causally to modify that behavior. Thus to say that John has a purpose or a want is to specify some effect of his behavior. But even unlucky or stupid people have purposes and their behavior may never have its intended effect. Functional causation cannot account for motive or purpose, nor is it reducible itself to a special case of direct linear causation.

Although individual human acts may not be efficiently caused; character traits may well be. The autonomy of human action does not foreclose on the possibility of meaningful educational research.

Science and Education

B.F. Skinner is a main source of Scientism in Education. His educational claims rest upon any of the following: unsupported assertion, dubious philosophical assumptions or logical error. His claim that rewards are positive reinforcers is a bald assertion; but rewarding is an activity identified via intentional considerations, whereas reinforcers are identified solely by their consequences. Skinner believes that physical criteria are adequate to identify even voluntary act-types. He seems to believe further that physical criteria can identify responses in the absence of supportive theory as to the capability of the organism to perceive and respond. But for any response-class defined by physical characteristic c, there may exist a physical characteristic e, common to only a subset of that class defined by c, members of which alone have been reinforced. Without testing for the ability of the organism to discriminate between those c-responses that are or are not also e-responses, the identity of the reinforced response cannot be established.

Skinner's operant conditioning theory cannot be patched up by allowing consideration of other than external variables. For theorists who recognize the relevance of internal states (Blough and Millward, Lazarus) the task of defining a general behavior variable is even more problematic.


Amidon, Bdmund J. and Flanders, Ned A. The Role of the Teacher in the Classroom. Minneapolis: Association for Productive Teaching, Inc., 1967.

Anderson, Barry F. The Psychology Experiment. Belmont, Cal.: Brooks-Cole Publishing Co., 1971.

Blough, Dona1d S. and Millward, Richard B. "Learning: Operant Conditioning and Verbal Learning," Annual Review of Psychology Volume 16. Edited by Paul R. Farnsworth. Palo Alto, Cal.: Annua1 Reviews, Inc. 1965.

Chomsky, Noam. Syntactic Structures. Hague: Mouton & Co., 1957.

Cohen, J. Operant Behavior and Operant Conditioning. Eyewitness series in Psychology. In Michael Schleifer "Instrumental Conditioning and the Concept of the Voluntary". Ethics 82, 2 (January, 1972)

Coombs, Clyde H. Theory of Data. New York: John Wiley & Sons, 1964.

DeMorgen, Augustus. "Assorted Paradoxes," The World of Mathematics. Edited by James R. Newman. 4 vols. New York: Simon Schuster, 1956.

Joncich, Geraldine M. (ed.). Psychology and the Science of Education: selected writings of Edward L. Thorndike. New York: Teacher's College Press. 1962

Lazarus, Arnold A. "Behavior Therapy and Clinical Problems: A Critical Overview," Annua1 Review of Behavior Therapy: Theory and Practice. Edited by Cyril M. Franks and Terence G. WiIson. New York: Bruner/Mazel, 1973, 9 - 39.

Mager, Robert F. Preparing Instructional Objectives. Belmont, Cal.: Feardon Publishers, 1962.

Morris, Herbert (ed.). Freedom and Responsibility. Stanford: Stanford University Press, 1961.

Nunnaly, Jum C. Psychometric Theory. New York: McGraw-Hill, 1967.

Ryle, Gilbert. The Concept of Mind. University Paperbacks. New York: Barnes & NobIe, 1949.

Schleifer, Michael. "Instrumental Conditioning and the Concept of the Voluntary". Ethics, 82, 2 (January, 1972). 163-170.

Skinner, B.E. Science and Human Behavior. New York: Free Press.1953.Bantam/Vintage. New York: Random House, 1972.

Stake, Robert E. "The Countenance of Educational Evaluation," Readings in Curriculum Evaluation. Edited by Peter A.Taylor and Doris M. Cowley. Dubuque, Iowa: Wm. C.. Brown Co., 1972.

Stinchcombe, Arthur I. Constructing Social Theories. New York: Harcourt, Brace and World, 1968

Turner, Merle B. Realism and the Explanation of Behavior. New York: Appleton-Century-Crofts, 1971.