AECT Handbook of Research

Table of Contents

31: Generative learning: past, present, and future

31.1 Introduction
31.2 Applied research
31.3 For the present: the generative model of teaching and implications for the design of instruction
31.4 The future
31.5 Conclusion
Search this Handbook for:

31. Generative Learning: Past, Present, And Future

Barbara L. Grabowski
Pennsylvania State University

31.1 Introduction

3 1. 1. 1 Generative Learning Defined

Over the past 20 years, attention has gradually shifted from investigating the effects of the external, physical form of instruction to examining what internal processes of learning are stimulated or induced by external stimuli. As a result, models and prescriptions for learning are founded on theoretical and empirical evidence about cognitive functioning, processes, and structure of memory (see 5.4). Using this foundation, designers develop a conception of what occurs within the learner, and use this conception to guide designs of learning rather than instructional environments. While instructional and learning environments both contain facts or data points (information with which learners interact), the key difference is who does what with that information. In a learning environment, the learner and his or her learning processes, styles, and activities take on prime importance. A learning environment is not devoid of instruction or an instructor, but rather the external stimuli simply take on a secondary role. In an instructional environment, the role of learner and instruction are reversed.

Generative learning theory and its companion model Of generative teaching is one such significant area of investigation whose theoretical foundation lies in neural research, research regarding the structure of knowledge and cognitive development, and whose focus is on the learner. 'Mis chapter defines generative learning and its foundation, reviews relevant research that tested the theory, describes the generative model of teaching and implications for instructional design, and concludes with a discussion of future directions for research.

Wittrock (1974a, 1974b) initially conceived of the model of generative learning that integrated several areas of cognitive psychology (see Chapter 5), including cognitive development, human learning, human abilities, information processing, and aptitude treatment interactions (see 22.3.3). His work stems from an attempt to explain and prescribe teaching strategies to maximize reading comprehension. While most of the original research deals specifically with reading comprehension, in theory there is much transferability to learning for understanding in general, regardless of the medium or form of the external stimuli. This article embraces the broader interpretation of this theory and model of learning.

From his initial conception to now, Wittrock emphasized one very significant and basic assumption: The learner is not a passive recipient of information; rather she or he is an active participant in the learning process, working to construct meaningful understanding of information found in the environment. The importance of asking the learner to generate his or her own meaning is clearly summarized by Wittrock's statement that "although a student may not understand sentences spoken to him by his teacher, it is highly likely that a student understands . sentences that he generates himself' (Wittrock, 1974b, P. 182). It is as Harlen and Osborne (1985) call it, "learning through the person" (p. 137).

While there are and have been four parts to his model, I find it interesting that in his current writing, Wittrock (1990, 1991, 1992, 1993) explains more thoroughly his second essential attribute, which he feels distinguishes his from other theories and models of learning. He elaborates on the importance of and difference between two types of learner-generated relationships: first, among the different parts of the information that are being perceived and, second, between that information and the learner's prior knowledge and other memory components. Comprehension occurs by formulating connections, rather than solely by the function of "placing" information or "transforming" information in memory. The subtle difference lies in the creation of new understanding of the information by the learner, rather than changing the presented information. (An analogy characterizing these important and subtle differences is presented in the next section in which generative learning is compared with other theories.)

These two combined attributes comprise only one of the four parts of his generative-leaming theory, namely, the processes of generation. The three other component processes that explain learning are motivational processes, learning processes, and knowledge creation processes. Metacognitive processes -also play a key role in his model, although in most cases he folds this idea into the learning processes component.

The concept maps portrayed in Figures 31-1 and 31-2 are illustrative of generative learning in action. These figures represent my comprehension of the ideas presented in Wittrock's (1974a, 1974b, 1985, 1990, 1991, 1992) writings regarding the progression of generative learning from neural brain processes research to models of thinking and teaching. The lines depict personally generated relationships between different concepts and ideas presented in his writings.

As shown in Figure 31-1, Wittrock conceptualized this model of generative learning based on a neural model of brain functioning and cognitive research on the process of knowing. From this foundation, the four components of the model are presented in shaded, rounded-comer rectangles with examples of each process presented as ovals under each. For example, "Attribution" is one example of motivational processes, and "Preconceptions" is one example of the knowledge creation processes. The process of generation is divided into the two types of possible relationship creations-between different parts of the information in the text and between the information in the text and prior learning and experience. This figure also implies a flow between the four processes of the model, with motivational processes activating learning processes, which in turn affects whether the process of generation will occur. The knowledge creation processes also affect the process of generation, but in a different way: Beliefs, preconception, prior concepts, and metacognition influence the quality and type of links that are created. Depending on the type of relationships generated, the four components converge for the purpose of learner-constructed reorganization, elaboration, or reconceptualization of the information to result ultimately in comprehension, as shown by the hexagons. Each of these processes will be discussed in detail later.

Figure 31-2 denotes the research by Luria (1973), as described by Wittrock (1992), on which generative learning was founded. As depicted here, Luria identified three functional units of the brain that are activated through the ascending and descending reticular activating systems and the frontal lobes of the cortex. In each of these, units, responsibility for cognitive functioning originates, which then activates or manages one of the processes of knowing, which then influences one of the four components of Wittrock's generative-leaming model-again depicted by the shaded, rounded rectangles.

The first unit, arousal and intention, influences an individual's learning processes and motivation. External stimuli arouse attention through the ascending reticular activating system. Without active, dynamic, and selective attending of environmental stimuli, it follows that meaning generation cannot occur regarding that environmental stimuli. The influence of arousal on attention flows from the environment outside of the learner, but interacts internally. Intention is activated by the descending reticular activating system, which stimulates attribution and interest. Attribution and interest influence the motivation of the learner. Attribution of effort, or the process of giving credit for success or failure to one's own effort, can influence whether or not the learner will exert the effort to be "attentive to the underlying structure of the information to be learned" (Wittrock, 1985, p. 123) and thereby become actively involved in generating understanding. If the learners attribute success to themselves, it follows that motivation to exert effort will be greater than if they attribute success to external forces (Weiner, 1979). The influence of intention on motivation for meaning generation flows from within the learner.

The second functional unit is the unit for receiving, analyzing, and storing information. The coding of information is managed by the frontal lobes of the cortex. The functions of the brain in this unit are influenced by the knowledge creation processes. Wittrock identifies many parts to the knowledge creation processes in several of his writings. Primarily, he includes beliefs, concepts, preconceptions, metacognitions, and experiences (see Fig. 31-1.). In other words, these are the components of memory. It is between these existing beliefs, concepts, preconceptions, etc., and environmental stimuli that relationships are formed and, thereby, understanding and comprehension are generated. According to Wittrock (1974a), "cognitive theory implies that learning can be predicted and understood in terms of what the learners bring to the learning situation, how they relate the stimuli to their memories, and what they generate from their previous experiences" (p. 93).

The third functional unit is the unit for planning, organizing, and regulating cognition and behavior. This unit operates through the frontal lobes of the cortex to coordinate learning and integrate information. These are the processes of metacognitive monitoring and generative processes-the heart and soul of generative-leaming theory. By generating relationships between parts of what the learners see and hear, and by integrating that information with what exists in memory, learners reorganize, elaborate, and/or reconceptualize information, not simply "stuff in more information." It is a process for which meaningful understanding and comprehension are predicted outcomes. (see fig. 3 1 - 1.)

Wittrock (1990) claimed that there are two types of activities that can be judged as generative. Activities that generate organizational relationships between different components of the environment include "titles, headings, questions, objectives, summaries, graphs, tables, and main ideas," while those that generate integrated relationships between the external stimuli and the memory components include "demonstrations, metaphors, analogies, examples, pictures, applications, interpretations, paraphrases, and inferences" (P. 354). In Figure 31-3, those examples are shown in the ovals connected to one of the two types of relationships. From other activities proposed by DiVesta (1989), Goetz (1983), and Jonassen (1986), concept maps, diagrams, outlines, and identifying scripts within narratives seem appropriate to be added to the organizational relationship list. Mnemonics, clarifying, and predicting seem appropriate for his second, the integrated relationship list, linking external stimuli to internal components of memory. Notetaking, diagrams, and concept maps could be appropriate for both lists, depending on which cognitive processes were used to create which type of link---organizational or integrative. That is, if learners were only relating different ideas extracted directly from a text passage, together it would be classified as organizational, whereas if they related the information to prior knowledge, it would qualify as integrative. Figures 31-1 to 31-3 represent organizational maps. Table 31-1, shown in the next section, portrays generation as integration by reconceptualization and elaboration.

Only those activities that involve the actual creation of relationships and meaning would be classified as examples of generative-learning strategies. Restructuring or manipulation of environmental information presented to the learner by definition requires him or her to generate either organizational or integrated relationships and constructed personal meaning, thereby qualifying as a generative activity. If this activity were simply tracing with no generation of relationships or meaning apparent, rather than involve new positioning, the activity would not qualify as a generative activity. Other controversial activities such as highlighting or underlining can be argued as not being generative, since the activity involves examining only single components, even though the learner may be selecting author-written main ideas. Even if the learner is integrating the sentences with prior knowledge, there is no covert evidence of that integration, as the focus of the activity is a selection task in which he or she is simply selecting from among many parts. An activity must involve meaning making in order to qualify as generative. An activity in which the learner simply selects sentences that someone else has already composed cannot be considered a generative activity. The generated main idea relates all or some of the ideas presented in the passage together. If learners are relating the textual information to their own prior experience, knowledge, or preconception, however, it could be argued that highlighting or underlining would be generative. As will be seen in the applied research section, Rickards (1979) would support this notion.

31.1.2 Relationship of Generative Learning to Other Schools of Thought

Wittrock (1991, 1992) often compares his own theory with other theories. These comparisons are quite useful for understanding the nuances of his teaching recommendations. In Table 31-1, by depicting an integrative restructured elaboration, generative-learning theory is compared with other contemporary schools of thought: behaviorism, connectionism, schema theory, information processing, and constructivism.

These schools of thought differ in many ways, the most significant being what unit of analysis is examined and explained, and how thinking and learning are defined and exemplified. These basic differences are often subtle, yet they contribute directly to the type of model that has been constructed and the implications that are drawn for instruction. The purpose of this section is not to describe each of these theories in detail (see other chapters for further description); rather, it is to discuss overall salient differences between the various models and generative-leaming theory and what these differences imply for instruction. The last two rows of Table 3 1 -1 depict those differences, one directly, and the other in an analogical reconceptualization.

Of all the theories, behaviorism (see 2.2; Skinner, 1990) presents the most extreme difference from generative learning. That difference lies in how the role of the learner is perceived and what this perception implies for learning. For generative learning, the learner is the key-the controller of whether information is learned or not. Understanding all of the neural processes that affect learning, from intention to components in memory to attribution, will aid the designer in selecting or creating appropriate activities that take these factors into account when encouraging the learner to code or integrate information. The learner must also be actively and consciously relating ideas. For behaviorism, the learner plays no role, except as a passive recipient of information. The behavioral design of instruction must center on creating a stimulating message that reinforces by positive or negative feedback. Higher-level coding or integration is irrelevant in the prescription.

Connectionisin (Wittrock, 1992) is similar to behaviorism, in that its intent is in strengthening associations. However, the network of individual memory is important, as in generative-learning theory. Connectionists, however, establish networks by strengthening associations by externally driven, repeated practice rather than creating personally drawn relationships between and among ideas. Understanding is internally created in generative-leaniing theory, making repetitions unnecessary.

Schema theory (Rummelhart, 1981; Rummelhart & Ortony, 1977) is similar to connectionism in that it deals with patterns of data points or schema. Basically, these data points form the knowledge units that are manipulated in generative-learning theory. Because of the way knowledge is stored, instructional and learning activities must connect new to existing knowledge so that it is easily retrievable.


Comparison Generative Learning Theory
(Wittrock, 1992)
(Skinner, 1990)
(Wittrock, 1992)
Schema Theory
(Rumelhart, 1981)
Information Processing
(Bell-Gredler, 1986)
(Jonassen, 1990)

A study of brain as controller neural connerctions memory associations knowledge representation in memory stages and levels of processing philosophy of constructed meaning
Learning defined by: learner-generated relationships behavioral change associations creation of, addition to, restructuring of, or fine-tuning of schema process of encoding information for retrieval individually constructed understanding
Type of thinking: brain as model builder by controlling the 4 processes automatic paired response in S-R chain neurally induced schemata construction and reconstruction transference of external stimuli to memory so that it may be retreived building understanding from experiences
Levels of thinking comprehension/ understanding - coding, elaboration, reorganization, reconceptualization unnecessary unit of analysis conceptual comprehension rehersal, coding, organization, conceptualization, integration, and translation unspecified
Type of model: neurally controlled learning stimulus-response chains subconceptual network model of memory structural (networked) knowledge representation of memory representation of the sequence of emtal operation and form of stored knowledge n/a
Components of theory: 4 processes: motivation, learning, knowledge creation, and generation operant conditioning, response formation, shaping, reinforcement schedules networks of patterns and weights of nodes and connections schema, schemata, scripts, and plans stages: sensory receptors, short-term store, working memory, long-term memory; levels: deep, and surface processing source of reality, learner as builder
Implications for instruction: activities that guide (induce) mental processes relating information careful construction of the physical form of emssages with repeated, rewarded practice activities that strenghten connections through repetition activities that relate new to existing knowledge so that it is retreivable activities that activate attention, facilitate processing in working memory, and facilitate transfer into long-term memory creation of contextualized learning environment

Comparison with generative-learning theory:


claims a very different role for the learner, as passive recipient rather than active generator

explains teaming as external neural induction instead of being neurally controlled internally; externally imposed not necessarily personally relevant.

explains the basic knowledge unit used in generative teaming theory

explains teaming as a transformation of information rather than generation

Provides a philosophical basis rather than a neuro logical explanation. of teaming

Comparative analogy.-task: purchasing clothing for an outing

approach: active and conscious selectionof items

approach: passive and reinforced by salesperson

approach: passive, externally driven

approach: not specified, but rather the units of clothing that can be chosen are more representative here

approach: active but not necessarily internally controlled- represents the process one goes through from attending to those items in the store, to trying various combinations, and then taking them home to store in the closet using a variety of grouping strategies

approach: active-- represents a philosophy that what is fashionable is constructed by individual tastes rather than predefined combinations proffered in society


item purchased: the intent is creating a new fashion statement

item purchased: outfit selected and laid out by another

item purchased: selection of combination of items for outfit determined by those most commonly seen on peers or in Magazines

item purchased: something to add to an existing outfit, tailoring of an old item, or creating of a new combination of items

item purchased: something to add to an existing outfit, tailoring of an old item, or creating of a new combination of items

item purchased: that which has been created by the individual rather than the salesperson

This connection is made by adding information to schema, restructuring, or tuning it. While connections are made by links, those linkages are not defined or labeled, as in creating a pattern note without labeling the lines. Generative-leaming theory, on the other hand, is similar in concept to creating a pattern note with all the links labeled. Activities designed by schema theorists would include those that reminded learners of prior knowledge and related the information to what the learner already knows. It is less relevant who selects those connection points over the fact that they are made.

Information-processing theory (Bell-Gredler, 1986) explains the process of thinking and memory storage (see 5.4. I)-in other words, the stages and levels of processing. What we get out of information-processing theory is an emphasis on how we think, rather than on what we think, or that we think. Its focus is on that process of transforming external stimuli into some recallable form to be stored in memory. The emphasis of generative-learning theory is on the generation of new conceptual understandings, not just on transforming information.

Finally, constructivism. (Jonassen, 1990) is a philosophy that underlies learning (see 7.3). It parallels generative theory in considering the learner to be an active processor of information; however, it is extreme in its position about the nonexistence of an objective reality. Wittrock has not addressed this notion in any of his writings.

To explain some of these subtle differences, a comparative analogy of an individual tasked with purchasing clothing for an outing was generated showing how the approach to the task (the purchase process) and the ultimate outcome (what item would ultimately be purchased) differ between schools of thought. The approach in generative buying would be exemplified as a buyer-controlled activity, with intention, motivation, and prior conceptions and beliefs about the outing and the people invited driving what types and styles of clothing are perceived as needed. The generative buyer would purchase cloth and create a new style based on those internally stimulated factors. The salesperson will query the buyer on how each article fits with those other influences, and how the articles would go with other items the buyer had at home. This buyer seeks a totally new fashion statement, rather than one already prepared.

A behavioristic example is simple: In a salespersondriven environment, the approach is very passive with an outfit having been preselected, and the salesperson giving much praise and lauding for purchasing the item. In a connectionism. scenario, the approach is also passive. Choices are driven by society-defined fashion that have been repeatedly seen (connected) in fashion magazines, television, and on peers. A buyer would purchase an outfit based on the frequency of seeing the outfit in fashion magazines. Intention or personal conception would already have been programmed.

Following schema theory, the buyer and seller play equal roles. Though the approach is not specified by this theory, the articles of available clothing in the store. and at home will play the key role. How these articles are combined is the important aspect of this theory. Accessories could be added, items rearranged, the outfit tailored to reach different desired effects. The coordination factors, however, would remain undefined. Intention, personal conceptions, or the generation of some new fashion item are irrelevant.

For information processing, both the buyer and sales environment are essential players, meaning that the approach by the buyer is active, but not necessarily intemally controlled. The information-processing buyer would attend to featured items that catch his or her attention, select a few, try on various combinations, and purchase a standard outfit, embellish it with accessories, or ask for it to be tailored. The key difference here is. that what is purchased is transformed from a rack in a store into an appropriate outfit following established rules of fashion, rather than generating totally new fashion statements or a totally new garment from different pieces of cloth.

A constructivistic buyer would hold a philosophy that what is fashionable is constructed by individual tastes and needs rather than predefined combinations proffered by society. The approach would be very active, independent, and individually driven. The ultimate item or combination of items would make an individual statement.

In each case, the notable difference is in the role of the buyer (learner) as he or she is related to the salesperson (instructor) or store items (instruction). This is exemplified through the approach (learning process) taken to the task, and the final selection of the item (learning).

*Recognition and appreciation are extended to Eileen Schroeder for her insightful comments on this manuscript, and to Christopher P. Rynd for his editorial assistance.




Updated August 3, 2001
Copyright © 2001
The Association for Educational Communications and Technology

1800 North Stonelake Drive, Suite 2
Bloomington, IN 47404

877.677.AECT (toll-free)

AECT Home Membership Information Conferences & Events AECT Publications Post and Search Job Listings