18.3 CONTRIBUTIONS OF R. M. GAGNÉ

As R. M. Gagné is generally identified as the primary originator of a conditions-based model of instructional design, an understanding of his evolution of thought becomes foundational to understanding the theory that extends beyond his contribution.

18.3. 1 Precursors to Gagnes Conditions-Based Theory

In a review of factors that contribute to learning efficiency for a volume on programmed instruction sponsored by the Air Force Offi6e of Scientific Research, Gagné and Bolles noted that "the learning tasks that have been most intensively studied by psychologists have been of an artificial "laboratory" variety; relatively little is known about learning in real life situations" (Gagné & Bolles, 1959, pp. 13-14). In 1962, as one who worked as a researcher in an academic setting, then as a researcher and research director in a military setting, and finally back in the academic setting, Gagné reflected on military training research in an article entitled "Military Training and Principles of Learning."

Training research in the 1950s (see 1.10) put Gagné in touch with a wide variety of instructional problems, representing a wide variety of learning tasks. Illustrative studies in the literature are Gagné (1954), "An Analysis of Two Problem Solving Activities," involving troubleshooting and interpretation of aerial photographs, and Gagné, Baker, and Wylie (1951), "Effects of an Interfering Task on the Learning of a Complex Motor Skill," involving manipulations of controls similar to aircraft controls. In a review of problem solving and thinking, Gagné pointed out the relevance of troubleshooting studies to issues in concept formation (Gagné, 1959). Wide and vigorous participation in research on learning and instruction in the military environment, along with his thorough and rigorous background as a learning psychologist, may have created the dissonance that motivated Gagné to develop the concepts of types of learning outcomes, learning hierarchies and events of instruction, and conditions of learning.

18.3.2 Development of Types of Learning

In his chapter on problem solving for Melton's Categories of Human Learning, Gagné (1964) presents a table entitled "A Suggested Ordering of the Types of Human Learning," in which he proposed the following six types of learning: response learning, chaining, verbal learning (paired associates), concept learning, principle learning, problem solving (p. 312). He did not cite a previous publication of his here, so this may be the first appearance of his types-of-learning scheme. This is not to say that be had not engaged in much previous thought and writing on important differences between forms of learning. However, the pulling together of types of learning to form a totally inclusive scheme containing mutually exclusive elements appears to have taken place around the time the "Categories of Learning" symposium was taking place, early in 1962.

Gagné's thinking on types of learning is illustrated by his discussion of problem solving as a form of learning. In the following, he points out how problem solving, as a form of learning, differs from other forms of learning:

. . the learning situation for problem solving never includes performances which could, by simple summation, constitute the criterion performance. In conditioning and trial-and-error learning, the performance finally exhibited (blinking an eye, or tracing a path) occurs as part of the learning situation. In verbal learning, the syllables or words to be learned are included in the learning situation. In concept learning, however, this is not always so, and there is consequently a resemblance to problem solving in this respect. Although mediation experiments may present a concept during learning which is later a part of the criterion performance, many concept learning experiments do not use this procedure. Instead they require the S to respond with a performance scored in a way which was not directly given in leaming (the stating of an abstraction such as "round" or "long and rectangular"). Similarly, the "solution" of the problem is not presented within the learning situation for problem solving. Concept formation and problem solving are nonreproductive types of learning (Gagné, 1964, p. 311).

Perhaps the first full and complete statement of the types of learning conception appeared in the first edition of The Conditions ofLeaming (Gagné, 1965). In that work, Gagné began by reviewing learning theory and research, such as that by James, Dewey, Watson, Thomdike, Tolman, Ebbinghaus, Pavlov, and K6hler. To introduce the idea of types of learning, Gagné presents the notion of "learning prototypes": "Throughout the period of scientific investigation of learning there has been frequent recourse to certain typical experimental situations to serve as prototypes for learning" (p. 18). The differences in kinds of leaming among these prototypes is seen in the inability to "reduce" one variety to another, although many attempts have been made" (p. 18). To clarify how these distinctive forms of learning have come to be lumped together as one form, Gagné (pp. 18-19) pointed out:

These learning prototypes all have a similar history in this respect: each of them started to be a representative of a particular variety of learning situation. Thomdike wanted to study animal association. Pavlov was studying reflexes. Ebbinghaus studied the memorization of verbal lists. K6hler was studying the solving of problems by animals. By some peculiar semantic process, these examples became prototypes of learning, and thus were considered to represent the domain of learning as a whole, or at least in large part.

Gagné (1965) presented eight types of learning in the first edition, in a strict hierarchical relationship. All types but the first, signal learning (classical conditioning), have prerequisite relationships with one another. The eight types of learning, with corresponding researcher links, were:

1. Signal learning (Pavlov, 1927)
2. Stimulus-response learning (Thomdike, 1989; Skinner, 1938; Kimball, 1961)
3. Chaining (Skinner, 1938; Gilbert, 1962)
4. Verbal association (Underwood, 1964)
5. Multiple discrimination (Postman, 1961)
6. Concept learning (Kendler, 1964)
7. Principle learning (Gagné, 1964)
8. Problem solving (Katona, 1940; Maier, 1930)

Regarding the distinctions between these types, Gagné describes support for some of the distinctions (Gagné, 1965, p. 59). Table 18-1 summarizes that discussion.

Later editions of Conditions of Learning modified the list of types of learning considerably. Although the second edition (Gagné, 1970) reflected no change in the number or labeling of the eight types of learning, by the third edition (Gagné, 1977) information-processing theories were added to the treatment of learning prototypes, and a large section was added on information processing along with recasting the types of learning (see 5.4.1). The information-processing perspective, present in the third edition, was not part of the first or second edition, even though earlier work reflected a strong information-processing background(Gagné 1962c). Surprisingly, while Gagné's primary base was shifting from behavioral to cognitive in the third edition, task characteristics, rather than psychological processes, begin to guide the form and content of the types of learning. In Gagné's fourth edition (1985), a hierarchical, prerequisite relationship is limited to four subcategories of one major category, intellectual skills. The types of learning in the fourth edition are:

1. Intellectual skills

   discriminations

     concepts

     rules

   problem solving

2. Cognitive strategies
3. Verbal information
4. Motor skills
5. Attitudes

Gagné's descriptions of the categories of problem solving and cognitive strategies have continued to evolve in recent years. For example, in Gagné and Glaser (1987), "problem solving" was combined into one category along with cognitive strategies. Inspection of the text reveals that, in fact, domain-specific problem solving was meant here, along with strategies for learning and strategies for remembering (see pp. 66-67). The evaluation of Gagné's problemsolving category can also be noted in his fourth edition of Conditions of Learning, in which problem solving was moved out of the intellectual skills category as higher-order rules and appears to have become a category separate from both the rule-based learning of intellectual skills and the domain-general category of cognitive strategy.

Gagné and Merrill (1990) described an approach to the integration of multiple learning objectives for larger, longer-term efforts that are unified through "pursuit of a comprehensive purpose in which the leamer is engaged, called an enterprise" (p. 23). A learning enterprise may be defined as "a purposive activity that may depend for its execution on some combination of verbal information, intellectual skills, and cognitive strategies, all related by their involvement in the common goal" (p. 25). The storage of enterprises is discussed in terms of mental models (Gentner & Stevens, 1983), schemata (Rummelhart & Norman, 1978), and work models (Bunderson, Gibbons, Olsen & Kearsley, 1981). Three kinds of enterprise schemata are described: denoting, manifesting, and discovering. Disappointingly, all of their examples are of individual learning, not of sets of them.

What do these categories of learning represent? Gagné described the types of learning outcomes as "learned dispositions," "capabilities," or "long-term memory states" (p. 245), qualities that reside within the learner. He further described two of these categories, verbal information and intellectual skills, as having distinctly different memory storage systems. Gagné and White (1978) provided an empirical basis for the ,'verbal information" knowledge to be stored as propositional networks. They further described rule using as stored in hierarchical skill, structures, which they at that time called intellectual skills.

More recently, Gagné (1985) has described verbal information learning as stored as propositional networks or schemata. He describes rules, including defining rules-or concepts, as stored as "If ... then" productions. He does not suggest how problem-solving capabilities themselves are stored, although he implied that they are interconnections of schemata and productions. Nor does he explicitly conjecture regarding the storage mechanisms of attitudes, motor skills, or cognitive strategies.

As the concept of types of learning evolved from its neobehaviorist beginnings to a more cognitive orientation seen in the fourth edition of Conditions of Learning, the research basis for differences in conditi ons for their achievement appears to have been largely lost. Although the concept remains as intuitively valid as ever to many instructional technologists, direct support in the literature is shockingly absent. Kyllonen and Schute (1989) describe Gagné's types of learning as a "rational taxonomy," being developed via proposing "task categories in terms of characteristics that will foster or inhibit learned performance" (p. 120). The drawback to such an approach is that its basis does not lie in psychological processes and, therefore, such processes are unsysternatically considered.

18.3.3 Development of the Learning Hierarchies Concept

A study by Gagné and Brown (1961) revealed thinking that led directly to Gagné's conceptions of learning hierarchies and types of learning. Here, in the context of programmed instruction, Gagné and Brown were concerned with the acquisition of meaningful "conceptual" learning, as compared with the rote memorization or association learning that characterized the work of Holland and Skinner:

. . . from an examination of representative published examples of programs (e.g., Holland, 1959; Skinner, 1958), it is not immediately apparent that they are conveying "understanding" in the sense of capability for inducing transfer to new problem situations. They appear to be concerned primarily with the usages of words in. a variety of stimulus contexts" (p. 174).

The phenomenon of transfer appears to have been central to Gagné and Brown's concerns, both of transfer from prerequisite learnings to higher-level outcomes (sometimes termed vertical transfer) and in terms of transfer from the learning situation to later application (sometimes termed lateral transfer). Although a great deal of attention is given to the study's programmed instruction forinat in the report, it is clear that the authors' interests were focused on a question of vertical transfer to problem solving (the particular learning task would now be considered relational rule use). In Gagné and Brown (1961), the authors described a study with a programmed instruction lesson teaching concepts (see 2.3.4) related to number series: the terms value and number After a common introduction to the fundamental concepts, the study employed three treatment methods to teach application of the concepts to finding the key to number series problems: (1) rule & example (R&E), (2) discovery (D), and (3) guided discovery (GD). The authors considered issues such as "size of step" and others of interest in programmed-instruction research of the day. However, they concluded that "some aspect of what has been learned ... is of greater effect than how it has been leamed7 (p. 181). The difference in "what" as supplied by the three treatments was that the guided discovery method required use of previously learned concepts in a new context.

Although all three methods were effective in teaching learners to solve numerical series problems, the GD and D methods were superior to the R&E method, with the GD method-being the most effective of all. The inferiority of the R&E method was attributed to the fact that it did not require learners to practice the application of concepts to a problem situation. In other conditions, learners could make the application but were believed to have, in general, not applied the concepts to the problem situation.

A postscript: It is ironic perhaps that in this early study, one that employed programmed instruction methods and reflects GagnCs thinking at that time very much as a neobehaviorist, the instructional strategies labeled "discovery" and "guided discovery" were found to provide superior instruction. It should be noted that the "discovery" method used was more structured than what many today might construct: A good amount of supplantive instruction on prerequisites preceded the "discovery" condition.

Gagné's first references to "learning hierarchies" appears in articles published in 1962: a report of a study, "Factors in Acquiring Knowledge of a Mathematical Task," and another study, "The Acquisition of Knowledge," which involved similar learning tasks. These reports were preceded by a study by Gagné and Paradise (1961) which formed a foundation for the latter studies. In 1961, Gagné and Paradise found support for the proposition that transfer of learning from subordinate sets of learning tasks could account for performance in a terminal learning task. In a subsequent study, Gagné, Mayor, Garstens, and Paradise (1962) sought to extend and confirm the validity of the idea of the "learning hierarchy."

Gagné, Mayor, Garstens, and Paradise (1962) sought to test effects of three factors that should mediate the effectiveness of learning hierarchies: (a) identifiability, which roughly translates into "acquisition of prerequisite concepts"; (b) recallability, stimulated in the study by cueing and repetition of prerequisite concepts; and (c) integration, in this study provided by what Gagné and Briggs later termed "provision of learning guidance," which was directed toward assisting the learner in applying concepts to problem situations. Two variables, used in various combinations, served to modify a basic learning program: repetition (high and low) and guidance (high and low). The posttest supplied information about achievement of not only the terminal task (adding integers) but also the 12 prerequisite learning sets, each scored as "pass" or "fail." These data were analyzed to supply evidence of the effects of the treatments on transfer. Success in final task achievement correlated highly with the number of subordinate tasks successfully achieved for both of the two terminal learning tasks (.87 and .88). Patterns of transfer amo ng the subordinate tasks also conformed to theoretical predictions.

In "The Acquisition of Knowledge," Gagné began by explicating the concept of a "class of tasks," differentiating the idea from "a response" by noting that in acquiring useful knowledge, it is inadequate to consider knowledge as a set of responses, since, when applied, it is impossible to identify from each specific response which skills, such as multiplication or punctuating compound sentences, the responses imply: "Any of an infinite number of distinguishable stimulus situations and an equal number of responses may be involved" (1962, p. 229).

18.3.4 Research Confirming Learning Hierarchies

In 1973, Gagné described the idea of learning hierarchies and noted that learning hierarchies have the following characteristics: (a) They describe "successively achievable intellectual skills, each of which is stated as a performance class"; (b) they do not include "verbal information, cognitive strategies, motivational factors, or performance sets",and (c) each step in the hierarchy describes "only those prerequisite skills that must be recalled at the moment of learning" to supply the necessary "internal" component of the total learning situation (pp. 21-22).

Gagné also described several studies on the validation of learning hierarchies. A fundamental way to accomplish this is to look at differences in transfer between groups that attain and groups that do not attain hypothesized prerequisites. Gagné, Mayor, Garstens, and Paradise (1962, Table 3, p. 9) is cited as an example providing positive evidence from such an approach. Other validation studies were reported, each looking in one way or another at the validity of a particular learning hierarchy-in other words, at the extent to which the hierarchy was a true description of prerequisite relationships among hypothesized subtasks. As a set, these studies can be seen to present evidence to the validity of the concept of learning hierarchies. The studies are summarized in Table 18-2.

In addition to the above, studies by Gagné and associates commonly cited to support the learning hierarchies hypothesis include: Gagné, 1962; Gagné and Paradise, 1961; Gagné, Mayor, Garstens, and Paradise, 1962; Gagné and Bassler, 1963; Gagné and Staff, University of Maryland Mathematics Project, 1965.

It should be noted that in "Factors in Acquiring Knowledge of a Mathematical Task" and in "The Acquisition of Knowledge," Gagné dealt primarily with learning hierarchies, not yet with the idea that different types of learning might require different instructional Conditions. The dirust of GagnCs ideas at this point was toward the organization and sequence of instruction, not toward the form of encounter.

18.3.5 Development of Events of Instruction and Conditions of Learning

18.3.5.1. Events of Instruction. In "The Acquisition of Knowledge" (Gagné, 1962), in addition to presenting the "learning hierarchies" concept, Gagné also introduced a precursor to the nine events of instruction. The description is of four functions that a theory of knowledge acquisition must account for:

1. Required terminal performance
2. Elements of the stimulus situation
3. High recallability of learning sets
4. Provision of "guidance of thinking"

Another foundation for the events of instruction was Gagné's thinking on the idea of internal and external conditions of learning, which is fundamental to the thesis of the first edition of Conditions of Learning in 1965. Internal and external conditions are defined (p. 21), and discussion of each of the types of learning is organized essentially along lines of internal and external conditions for achievement of that type of learning. To summarize his descriptions of these two types of conditions: Internal conditions were primarily described as learners' possession of prerequisite knowledge, and external conditions were viewed as instruction.

The first edition of Conditions of Learning (Gagnél 1965) did not have a discussion of the "events of instruction" in the same fashion as the term later came to be used-as a listing intended to be inclusive, reflecting events that must occur, and if not supplied by instruction, generated by learners. The treatment in Conditions of Learning, under the heading "External Events of Instruction," included discussion of (a) control of the stimulus situation (strategy prescriptions varied with types of learning), (b) verbally communicated "directions" (directing attention, conveying information about expected performance, inducing recall of previously learned entities, and guidance in learning by discovery), and (c) feedback from learning.

The "events of instruction" conception may be more directly attributable to L. J. Briggs's work than Gagné's, although the two collaborated extensively on it. For example~, Briggs, Campeau, Gagné, and May's (1967) handbook for "multimedia design of instruction" uses nearly all the elements of what was to become the "events of instruction" within its examples, but it does not present a list of the events (see Briggs, 1967, pp. 53-73; May & Briggs, 1967, pp. 74-138). In another chapter in that manual, Briggs, Gagné, and May noted (p. 45) as "instructional functions of stimuli," the following:

1. Set a goal in terms of performance desired.
2. Direct attention.
3. Present instructional content (also stimuli).
4. Elicit response.
5. Provide feedback.
6. Direct the next effort.
7. Help the student evaluate his performance.

Also noted here under "other special functions of stimuli" are: (a) providing the degree of cueing or prompting desired, (b) enhancing motivation, (c) aiding the student in recall of relevant concepts, (d) promoting transfer, and (e) inducing generalizing experiences (Briggs, Gagné & May, 1967, p. 45). Between the two lists, the events-of-instruction formulation appears to have been taking shape.

The first edition of The Conditions of Learning (Gagné, 1965) contained a section called "component functions of the instructional situation" that, except for the label, was to be virtually identical in conception and content to the "events of instruction" seen in later editions of The Conditions of Learning, as well as in Gagné & Briggs, Principles of Instructional Design (1974). The eight functions were: (a) presenting the stimulus, (b) directing attention and other learner activities, (c) providing a model for terminal performance, (d) furnishing external prompts, (e) guiding the direction of thinking, (f) inducing transfer of knowledge, (g) assessing learning attainments, and (h) providing feedback. With Principles of Instructional Design (Gagné & Briggs, 1974), the full Events of Instruction Model was first presented and fully discussed. Table 18.3 presents the Events of Instruction along with cross-reference to other sections in this handbook which relate to each section.

18.3.5.2. Conditions of Learning. Completing Gagné's contribution to conditions-based theory are his discussion of the internal and external conditions of learning that support each type of learning outcome. "Internal conditions" are those cognitive processes that support the acquisition of particular categories of learning outcomes. "External conditions" are those instructional conditions provided by teacher, materials, or other learners that can facilitate the internal conditions necessary for learning. These external conditions, too, vary according to type of learning. Not surprisingly, given Gagné's transition from behavioral to cognitive theory bases, be developed the external conditions model first.

As an instructional psychologist, Gagné was particularly interested in the external conditions that might occur or could be provided to "activate and support" the internal processing necessary for learning to occur (1985, p. 276). In fact, Gagnd defined the purpose of instructional theory as "to propose a rationally based relationship between' instructional events, their effects on learning processes, and the learning outcomes that are produced as a result of these processes" (1985, p. 244). Therefore, Gagné derived the external events from the internal events of information processing. (See Table 18.3.)

Gagné particularized the general external events, the "events of instruction," that begin to be described in his work in 1962 to specific prescriptions for external conditions for each type of learning, event by event, for each of the categories of learned capability. Much of these external conditions are logically derived from the intersection of the function of the external event (those cognitive processes that it supports) and the nature of the learned capability. In "Domains of Learning" (1972), Gagné argued very specifically for a conditions-based theory but did not present research directly on it; rather, he presented arguments about the nature of different learning domains, buttressed often in a general fashion by research. The five domainsmotor skills, verbal information, intellectual skills, cognitive strategies, and attitudes-are the level at which he argued that there is a difference in how they should be taught, particularly in terms of (a) the kind and amount of practice required and (b) the role of meaningful context. Additional criteria as means by which types of learning can be contrasted with regard to instructional concerns appear in a 1984 article, "Learning Outcomes." In Gagné and White's 1978 article, two general domains of learning outcome were discussed: (a) knowledge stating and (b) rule application. References used to support the distinctness of these two domains include Gagné (1972) and Olson and Bruner (1974).

In 1987, Gagné and Glaser (1987) developed a review that included a brief survey of Gagné's early work, learning as cognition, importance of short-term memory, learning complex performances, knowledge organization for problem solving, mental models, and self-regulation. Table 18-4 (on p. 550), reproduced from that review, provides an excellent summary of hypothesized differential learning conditions for types of learning (see 5.3.5.4).

18.3.5.3. Internal Conditions of Learning. Gagné suggested that, for each category or subcategory of learning capability to be acquired, certain internal conditions were necessary. By 1985, Gagné described these internal conditions as two kinds: (a) prerequisite knowledge, which is stored in long-term memory, and (b) particular cognitive processes that bring this old knowledge and new knowledge together and store it in a retrievable form. Gagné described these cognitive processes using an informationprocessing model: attention, selective perception, semantic encoding, retrieval, response organization, control processes, and expectancies.

It should be noted that in Gagné's detailing of the internal conditions of each type of learning, the major internal condition that he described was prerequisite knowledge. For example, Gagné specifies the internal conditions for rule learning to be knowledge of (the ability to classify previously unencountered instances and noninstances of) component concepts. This may be because the research base for the identification of the specific internal conditions for each learning capability was inadequate or because as an instructional theorist, his predominant interest was the external conditions that could support the generalized informationprocessing mechanism and those internal conditions necessary prior to the initiation of new learning.

Gagné (1984) suggested that the internal events that may differ most across learning capabilities are "(a) the substantive type of relevant prior knowledge, (b) manner of encoding into long-term storage, (c) requirement for retrieval and transfer to new situations" (p. 514). Therefore, in his 1985 edition of Conditions of Learning, he pointed out that the events that may differ most significantly from learning category to learning category are those corresponding to the above three internal events: These external events are (a) stimulating recall of prior knowledge, (b) providing learning guidance, and (c) enhancing retention and transfer.