18.2 EVOLUTION OF THE CONDITIONS-BASED THEORY

The first full statement of a conditions-based theory of instruction appears to have been by R. M. Gagné in the early 1960s.2. However, there was a considerable amount of conjecture within this paradigm by a variety of researchers prior to Gagné. In addition, Gagné and others have developed a conditions-based theory along a variety of lines of thought until the present day. In this section, we will review work leading to the conditions model, discuss Gagné's early and evolving conceptions, and review various lines of research in the conditions-based tradition which have appeared subsequent to Gagné's first work (see also 1.6, 5.2).

18.2.1 Early Work Leading to Conditions-Based Thinking

Among the earliest writing that specifically addresses the need to beware of overgeneralization of knowledge about learning, Carr (1933) cautioned that conclusions that are valid for one category of learning may not be valid for others. The categories of which Carr spoke were not within a formally defined taxonomy or system, but rather they were reflected in the different experimental tasks, research procedures, and measures employed in different studies.

Interest in devising useful categories of learning persisted over the decade, and Melton wrote in the learning theory chapter in the 1941 Encyclopedia of Educational Research (Melton, 1941) of efforts to develop a psychologically based taxonomy of learning outcomes. During this same period, Tolman (1949) described six categories of learning, and Woodworth (1958) described five categories.

The behaviorist movement lent a rigor and precision to the study of learning that is perhaps difficult to appreciate today (see 1.4.4,1.4.5,2.2.2.3). When one looks at the work of some behaviorist learning researchers, one finds compelling (if not esoteric) evidence that there are different kinds of learning with different conditions for their attainment. Wickens (1962) described Spence's studies of animal learning involving both aversive conditioning and approach behaviors:

Spence (1956) has used the same approach in differentiation of the instrumental avoidance situation of the type represented by the eyelid conditioning from an approach learning represented by an animal scurrying down a runway for its daily pellet. Spence is quite specific: The antecedent of H (the intervening variable leading to the running response) in the latter case is a function only of n and not of incentive magnitude; in the former, it-the intervening variable H-is a function of n and also of magnitude of the UCS. This conclusion leads him to describe the excitatory component of behavior, E, as being a function of two intervening variables, H and D, insofar as classical aversive conditioning is concerned, while the excitatory component of runway behavior requires, for him, three intervening variables, H, D, and K (Wickens, 1962, p. 81).

Another specification of differences in learning tasks is seen in Bloom's taxonomy (Bloom, Englehart, Furst, Hill & Krathwohl, 1956). This group's thinking about the need for a taxonomy of educational outcomes originated at an informal meeting of college examiners at the 1948 meeting of the American Psychological Association, at which "interest was expressed in a theoretical framework that could be used to facilitate communication among examiners" (Bloom et al., 1956, p. 4). The taxonomy arose not from a synthesis of research but in response to a collective need for standardization of terminology.

Applications of Bloom's taxonomy, however, have frequently assumed a stature similar to that of psychologically based approaches. For example, a study by Kunen, Cohen & Solman (1981) investigated the cumulative hierarchical assumption of Bloom's taxonomy. The study concluded that there is "moderately strong support for the assumption that the taxonomy represents a cumulative hierarchy of categories of cognitive operations" (p. 207). The tasks used in the study involved recall of knowledge, recall of applications, recall of words related to a synthesis task, and so forth ("the dependent variable was the number of critical words correctly free recalled," p. 207). As all of the tasks appear to involve recall, we have some doubt about the validity of conclusions supporting a hierarchy of cognitive operations. Furst's (1981) review of research on Bloom's taxonomy leveled a great deal of criticism of the taxonomy in terms of its lack of cumulative hierarchical structure. It seems clear that the taxonomy's uses have exceeded its original design and purpose.

18.2.2 Military and Industry Training Researchers

In the 1950s and 1960s (and continuing to the present), a substantial amount of research and development related to learning and instruction has been conducted by the military services and industry (see also 1.10). Edling and associates pointed out that this group of scientists in military and industrial settings was large for its work to be so unfamiliar to many educators. Indeed, in 1963 the Army's HumRR0 employed 100 "training psychologists" of whom 65 were Ph.D.s, and the Air Force Training and Research Center at one time employed 168 psychologists of whom 100 held Ph.D.s (Edling et al., 1972, p. 94).

Among the contributions of these researchers were some "relatively sophisticated taxonomies of learner tasks," such as those developed by Cotterman (1959), Demaree (1961), Lumsdaine (1960), Miller (1962), Parker and Downs (1961), Stolurow (1963), and Willis (1961). Gagné's work, also perceived of evolving within this context, was seen to be "particularly powerful" (Edling, 1972, p. 95).

Of these taxonomies, Miller's treatment of learning types illuminates the idea of "task analysis" as it was viewed in the 1950s and early 1960s (Miller, 1953, 1954, 1956, 1962). Miller, employed by IBM, proposed that an "equipment task analysis" description "should include analysis of perceptual, short-term recall, long-term recall, decision-making and motor processes implied by the initial equipment task analysis" (Smode, 1962, p. 435). Miller reflected the mainstream approach by focusing on job tasks, although it is clear that consideration of cognitive processes greatly influenced much of his analysis scheme's structure and content.

Much of the progress in defining learning tasks made by the military and corporate researchers may be attributed to their employer's demands. Increasingly, technical training requirements in the military and in industry were placing high demands on the skills of training designers to develop instruction in problem solving, troubleshooting, and other expertise-related tasks. Bryan (1962) discussed the pertinence of troubleshooting studies to the topics of transfer, concept formation, problem solving, decision making, thinking, and learning. To perhaps exaggerate a bit, one can envision academic colleagues running rats in the laboratories, while their counterparts who were employed by the military and large corporations were struggling with issues of human learning and skilled performance. This pressure to describe complex learning (often felt by academics as well) forced behaviorally trained psychologists to consider cognitive issues long before the mainstream, and produced a unique blend of "neobehaviorism" with what we might call "precognitive" psychology. As we view Gagné's work and its evolution, this blend and transition will be clearly illustrated.

18.2.3 Acadernic Learning Psychologists

The thinking of academic psychologists about types of learning are well represented in a 1964 volume edited by A. W. Melton, Categories of Human Learning. In chapters by N. H. Anderson, E. J. Archer, G. E. Briggs, J. Deese, W. K. Estes, P. M. Fitts, R. M. Gagné, D. A. Grant, H. A. Kendler, T. S. Kendler, G. A. Kimble, A. W. Melton, L. Postman, B. J. Underwood, and D. D. Wickens, concerns and progress toward understanding varieties of human learning are discussed. Two of these contributions will be discussed here for the information they contain on state of the art and as illustrations of the categories defined within this period.

Underwood (1964) discussed possible approaches to a taxonomy of human learning, proposing how it would be possible "to express the relationships among research findings for all forms of human learning" (p. 48). Underwood noted that a single, grand unified theory did not yet exist in which a master set of statements and relationships could lead to deductions of findings in each of the various areas of interest in learning research. A second approach that Underwood suggested in the absence of a grand unified theory was to attempt to

express the continuity for all human learning ... in terms of phenomena produced by comparable operations. Thus, can the operations defining extinction in eyelid conditioning be duplicated in verbal learning, in motor learning, in concept formation, and so on, and, if so, do the same phenomena result from these operations? (p. 48).

Underwood (p. 48) noted that a difficulty in doing such cross-category research is that the differences among tasks make it physically impossible to manipulate them in comparable manners:

For example, it would seem difficult to manipulate meaningfulness on a pursuit rotor in the same sense that this variable is manipulated in verbal learning. Or, what operations in problem solving are comparable to variations in intensity of the conditioned stimulus in classical conditioning?

Underwood later described a technique for determining the similarity of learning in different situations. That technique is illustrated in work by Richardson (1958) in which "A descriptive difference between concept formation and rote verbal learning can be stated in terms of the number of identical responses to be associated with similar stimuli" (p. 49). The number of responses to a stimulus associated with a concept-learning task was different from the number of responses to the same stimulus when it was part of a rote-learning task, reflecting a lack of "continuity" between the two types of tasks.

In a seminal chapter in Categories of Human Learning, Melton (1964) pointed out that neither the physical structures of the human organism nor its cognitive processes themselves (such as motivational, perceptual, and performance processes) provide guidance for a taxonomy of learning, as such structures would provide in classifying physical attributes. Of the need for a conditions-based approach, Melton (1964, p. 327) bemoans the lack of articulation between training design questions and knowledge about learning:

When one is confronted with a decision to use massed or distributed practice, to insist on information feedback or not to insist on it, to arrange training so as to maximize or minimize requirements for contiguous stimulus differentiation, etc., and (one) discovers that the guidance received from experimental research and theory is different for rote learning, for skill learning, and for problem solving, taxonomic issues become critical and taxonomic ambiguities become frustrating, to say the least.

A strong element of formalism, in addition to the practical concerns noted in the quote above, seems to shape Melton's thinking, which is illustrative for the time in which he was writing. Melton wrote at length about a "taxonomy" of learning from the standpoint of taxonomies themselves and how they come about in science. A persistent theme is bow the "primitive categories" will end up being used--to what extent they will be used and how they can conceivably be modified. The primitive categories, reflected to a large extent by chapter topics in the book, are rather long-standing areas of research and theory interest in learning psychology: conditioning, rote learning, probability learning, skills learning, concept learning, and problem solving. Apparently, Melton expected that all of these topics should, as organized in some appropriate and meaningful way, be related to one another in a taxonomic, hierarchical structure. One may notice, coincidentally perhaps, that Gagné's first edition of Conditions of Learning included a set of learning types that were, unlike those found in more recent editions, in toto a taxonomic list, in which each category in the classification scheme was prerequisite to the others (with the exception of the first category, classical conditioning). In later versions of the types of learning, Gagné included many categories that he did not propose as being in hierarchical relationship (only the -learning types within the intellectual skills category are proposed as hierarchical in more current versions.)