Structural Learning Theory

Developed in the late 1960s by University of Pennsylvania professor Joseph Scandura, structural learning theory came onto the scene just as the fields of psychology and education were transitioning from a behaviorist to a cognitive orientation. This article outlines the ways in which structural learning theory adopts a cognitive orientation to learning, as well as some of the elements it shares with behaviorist theory. The basic tenets of the theory are outlined, with a discussion of the ways in which it has been applied. Although structural learning theory hasn't generated a significant amount of discussion in the academic literature, it has had a substantial impact in the fields of instructional design and software development.

Keywords Behaviorism; Deterministic; Inputs; Operations; Outputs; Probabilistic; Response; Rules; Scandura, Joseph; Stimulus

Educational Theory > Structural Learning Theory

Overview

In the late 1960s, Joseph Scandura, a professor at the University of Pennsylvania, invited his colleagues to a conference to discuss the nature of human learning. As a result of that conference, Scandura developed what has become known as structural learning theory. At that time, he later wrote, "we were not interested in reporting finished research, but, rather, we were unsatisfied with the kind of research being done on complex human learning and wanted to see what we could do about improving it" (Scandura, 1973, p. vii). Whether or not Scandura and his colleagues improved research on human learning is up for debate; however, it's important to understand the ways in which they believed they were extending the conversation.

Scandura marketed his theory as the first to approach the understanding of human learning from a deterministic point of view as opposed to a probabilistic one. In other words, Scandura believed it possible to identify the conditions under which human behavior could be predicted and controlled with near certainty. He opens his book, Structural Learning: Theory and Research, with the following statement:

In spite of the diversity which presently exists in behavioral theorizing, reference to probabilistic notions is all-pervasive. Even support at the .05 significance level is often enough to elicit whoops of glee from most cognitive theorists. Given this milieu, it is not too surprising that…no one seem to have seriously pursued the possibility that deterministic theorizing about complex human learning may actually be easier than stochastic theorizing. And yet, this is precisely what in my own work I have found to be the case (Scandura, 1973, p. 1).

What allowed Scandura to suggest that human behavior could be predicted and controlled was his belief that theory building should occur under idealized conditions. That is, Scandura eliminated the variables that might reduce certainty and predictability in human learning – mainly, our limited capacities for information processing and memory – and devised a theory to explain how learning might occur were our resources unlimited. "One of the major reasons why we have been relatively unsuccessful in devising adequate theories of complex learning and behavior, I believe, is because we have tended to tackle the problem as a whole. With a few exceptions, the possible value of ignoring the effect of memory in theorizing about…behavior has not been taken seriously by most psychologists" (Scandura, 1973, p. 171). Scandura was not suggesting, however, that memory should be excluded from theorizing altogether – although he strongly believed that memory was of minimal consequence in many real-life learning tasks – but rather that it should be considered only after theorizing at an idealized level.

Move from Behaviorism to Cognitivism

At the time Scandura was developing his theory, the fields of education and psychology were experiencing a transition. Prior to the 1970s, behaviorism reigned supreme; researchers equated learning with changes in behavior brought about by the environment, and ignored almost entirely the internal experience of individuals. Some denied the existence of the mind altogether, often referring to it as 'the black box.' Critics began realizing, however, that behaviorist theories – with their emphasis on the stimulus and response – were limited in their ability to explain more complex learning such as language acquisition. As a result, the cognitive revolution began, and attention was turned toward the very construct behaviorists had ignored – the human mind. Scandura's theory reflects the tension of this transition, simultaneously rejecting and upholding certain behaviorist tenets.

Like many of his cognitivist peers, Scandura rejected the notion that all learning could be explained by simple associations between stimulus and response. Clearly positioning himself in the cognitive camp, Scandura argued that human learning is the result of rules, not associations. "The to-be-proposed theory," he wrote, "is clearly of the information processing variety; learning is the result of internal operations and not of contingencies between overt stimuli and responses. Information processing theorists generally view learning as a problem solving process and mine is no exception" (Scandura, 1973, p. 7). The notion that all human behavior is rule-governed is central to Scandura's theory.

Scandura's theory also reflected a second new trajectory in his profession. Behaviorists, and other theorists before them, had been in search of a 'holy grail' of sorts – a single theory or law that could explain all human learning. B.F. Skinner, one of the most well-known behaviorists, for example, suggested all learning could be explained according to the principles of operant conditioning. Cognitivists, however, were increasingly recognizing the futility of such a search, emphasizing the different processes – memory, perception, attention, and motivation, to name just a few – that confound human learning. Scandura too made this discovery, even suggesting his theory surpassed other cognitive theories in this respect. He believed that in most present-day information processing theories the questions asked were: how do subjects remember x? How do subjects learn y? Or, solve problem z? It is assumed that there is a unique answer to such questions. Scandura's theory renders such a question meaningless, positing that there are many different possible ways in which any particular subject might perform (Scandura, 1973, p. 7).

In an important way, structural learning theory maintains an alliance of sorts with behaviorists. Scandura believes, for example, that educators and researchers must focus on that which can be observed – in other words, behavior. "Behavior is the only thing that instructional scientists can observe. It is impossible to know all or exactly what any person does or does not know that causes him to behave as he does. Instructional scientists do not have license or means to look inside" (Scandura, 2001, p. 2). Thus, structural learning theory combines what many view as diametrically opposed elements – behaviorism and cognition. Scandura (2001) himself describes it as a cognitive theory, but one with "methods for operationally defining human knowledge in terms of behavior" (p. 3).

Description of the Theory

Scandura (2001), like many of his colleagues, was attempting to answer four basic questions:

• What does it mean to know something, and how can that be represented behaviorally?

• How do learners use and acquire knowledge?

• How does one determine what an individual does and does not know?

• How does knowledge change as learners interact with the environment?

What Does It Mean to Know Something?

In structural learning theory, knowledge must be demonstrated through performance, and all performance (or human behavior) is rule-governed. As he writes in 2001, "the competence required to successfully perform tasks…is represented in structural learning theory in terms of a finite set of higher and lower order rules" (Scandura, 2001, p. 2). What then, is a rule? Rules are conceptualized as ordered triplets (D, O, R), where D refers to the determining properties of a stimuli, and O to the operation which occurs to derive a particular response, referred to as R. Scandura uses the addition algorithm as a concrete example, where a set of whole numbers serves as the stimulus (D), the addition operation represents O, and a different set of whole numbers (sums) serves as the response. Because rules can be devised to account for all behavior, there are many different kinds; Scandura (1973) identifies encoding, decoding, decision/branching, operation, retrieval, and storage rules as examples of six different types.

How Do Learners Use & Acquire Knowledge?

It is easy to recognize that there are some similarities between Scandura's terminology and the terminology used in behaviorism, leading to questions of whether or not the underlying processes described are the same. Is Scandura suggesting, for example, that a stimulus leads to a response in the same way Skinner proposed? Scandura anticipated such questions, and although he used S-R terminology, he also referred to D, O, R as inputs, operations, and outputs with equal if not greater frequency. More importantly, however, he placed great emphasis on underlying distinctions between structural learning theory and behaviorism.

This distinction between overt stimuli and responses, on the one hand, and input and output properties on the other, may seem to be a relatively minor point, but it is not. It represents a major departure from the kind of thinking associated with S-R…psychology. Specifically, the present view rejects the idea that overt (and/or potentially overt) stimuli cause behavior. We assume instead that behavior is caused by rules, an underlying construct - in effect, that subjects actually do use rules (Scandura, 1973, p. 15).

How exactly do subjects use rules? Scandura identifies two situations requiring different uses of rules. In the first, the subject has the necessary rules within his or her repertoire to solve the given problem. In this case, solving a problem is simply a matter of searching the set of rules an individual has already acquired and applying it appropriately. In structural learning theory, such a process is governed by a universal control mechanism, assumed to vary across individuals according to the speed at which it occurs. In other words, some people find and apply the necessary rule faster than others. The second scenario presented by Scandura is the one in which an individual does not have the necessary rule to solve a problem. After searching his or her repertoire, and recognizing that none of his or her given rules will suffice, the individual then begins creating a new rule. As such, Scandura characterizes his theory as constructivist in nature, recognizing that learning is an active process in which people create new knowledge.

How Does One Determine What an Individual Does & Does Not Know?

Scandura's thesis that all behavior is rule-governed is central to assessing the abilities of any individual student. Determining what an individual knows is dependent upon determining what class of rules is necessary in any problem-solving situation. Because rules are ordered hierarchically, with lower order rules subsumed by higher order rules, testing a student on all rules identified in each problem set is unnecessary. As Scandura (1973) writes, "A path which contains all of the atomic rules of another path plus some of its own would occupy, relatively speaking, a higher position in the ordering… From this it follows (on largely logical grounds) that if a subject is successful on an instance associated with a higher level path, then the subject should also be successful on items associated with all (relatively) lower level paths" (p. 185).

Scandura emphasizes the singular nature of this exercise; that is, determining what an individual does and does not know happens one student at a time, not with groups of students or even entire classrooms of students. Working with individual students in the manner described above is what allows his theory to be deterministic when put into practice, as opposed to probabilistic. "It is not a question of running groups of subjects or of averaging over different tasks. We experiment with a number of different subjects on a number of different tasks under a number of different conditions, to see if we can predict exactly what the subject can and cannot do" (Scandura, 1973, p. 188).

How Does Knowledge Change as Learners Interact with the Environment?

The last question addresses the interaction of the learner with his or her environment; how does the teacher, for example, or the design of instruction, impact what new knowledge a learner might acquire? Scandura has devoted a significant portion of his career addressing this question by designing software and technology to optimize teaching and learning. As Scandura (2001) writes, "methods used to teach various kinds of content derive from a common set of SLT principles. Optimal instructional techniques depend as much on the learner's state (of knowledge) as on the kind of content" (p. 13). More specifically, Scandura believes a competent tutor – whether real or computerized – is able to choose instructional and evaluative exercises that maximize the amount of information gained about the learner's knowledge, and therefore, also about what a learner needs to know to overcome her deficiencies. By doing so, teachers become increasingly efficient, by providing just the information needed to build on what is already known.

Further Insights

The value of a theory can be evaluated two different ways – by how often it is discussed and/or by how easily it can be applied. If structural learning theory were evaluated according to the former, it might not fare so well. It has generated much less discussion in the academic literature than other theories of learning, such as, for example, either behaviorism or Piaget's cognitive developmental theory. If evaluated according to the latter, however, its impact becomes much more substantial.

SLT in the Academic Literature

To say that structural learning theory has received less attention is not to say that it hasn't received any. Indeed, an international interdisciplinary journal titled "Technology, Instruction, Cognition, and Learning" began in 2003 and has published over 13 volumes. The purpose of the publication is to "improve interdisciplinary communication and to promote scientific dialogue on both fundamental issues and their real world application" (Scandura, n.d., TICL section, p 9). Scandura himself has arguably been the most prolific contributor to the discussion surrounding structural learning theory, having published over 200 books, articles, and software systems. In addition, the American Education Research Association sponsors a special interest group in technology, instruction, cognition, and learning.

SLT in Software Development

Scandura has contributed significantly to the development of software applications founded upon the principles of structural learning theory. His website, www.scandura.com, describes his software development company as follows: "Scandura products build on over four decades of advanced R&D by Scandura principles in structural learning and the cognitive sciences. Millions of dollars in public and private funds have gone into educational software and software engineering development. Major advances have led to products with numerous benefits and unique features, many protected by patent" (Scandura, n.d., p 1).

As Scandura described in a 2012 article, “the use of computers for instructional purposes is much more diverse now than even a few years ago when Macromedia’s (now Adobe’s) Authorware was dominant in computer-based instruction (CBI).” Computers are used as a medium to support various forms of eLearning, Scandura stated, ranging from “communicating with human tutors over the Web to sophisticated simulations.” Lacking major advances, however, CBI has “taken a back seat. With its focus on ‘one size fits all,’ traditional CBI does not lend itself to truly adaptive instruction” (Scandura, 2012c).

Scandura has focused on the development of three different software applications. The first, titled TutorIt, is the only instructional tool, Scandura argues, to separate content from instructional logic. In essence, TutorIT can be used with any content, provided it is given a detailed analysis of the knowledge to be acquired. TutorIT can be used as a diagnostic tool, a tutoring tool, an evaluation tool, a simulation tool, and a tool that provides progressively difficult instruction.

Scandura wrote in 2012 that a “broad base of TutorlT tutoring systems has been developed in the last year with minimal resources. These math tutorials range from basic facts though alge- bra 2. A comprehensive tutorial on algebra word problems is in process. Included as well is a series of TutorlT tutorials based on a critical reading workbook series developed back in the 1970s” (Scandura, 2012a).

AuthorIt is a second software application designed to help teachers define the domain of knowledge to be learned; it is used in combination with TutorIt, since the latter is dependent on a clear understanding of desired outcomes. AuthorIT results in "a rigorously defined Abstract Syntax Tree (AST) representation of the to-be-learned behavior and cognitive processes" (Scandura, n.d., AuthorIT section, p 3). Finally, Scandura has also developed a series of software development programs, designed with the intention of being user-friendly and accessible, so that people other than software specialists can program with them.

Viewpoints

In 1996, Ikegulu wrote "It is quite appalling that Scandura's theory did not make it past 1988. Literature about this theory actually stopped in 1985 after Scandura's second phase of SLT. It seems somewhat that the theory has been forgotten, not refined, and totally discarded. Therefore, the structural learning theory has no influence and applications in school settings…" (p. ii). Structural learning theory continues to be a topic of interest in the literature and Scandura himself has developed numerous software applications founded upon SLT principles. Ikegulu's (1996) assessment, therefore, is exaggerated, if not inaccurate. And yet, Ikegulu (1996) might have been correct to suggest that structural learning theory has contributed more significantly to the merging of technology and instruction, than to the discussion of learning theory in general.

As such, much of the discussion about structural learning theory isn't argumentative in nature, with scholars firmly planted on two sides of an issue. Rather, the discussion centers on how the theory can be applied, and the ways in which structural learning theory informs the use of technology in the classroom. A brief look at the topics covered in the international journal of Technology, Instruction, Cognition and Learning (TICL) demonstrates the point. The topics covered include, but are not limited to, learning in artificial environments, model-centered learning, emerging theory in educational technology, the use of computerized tutors in assessing reading skills, fostering self-regulation through the use of technology, and the role of multi-media in problem-solving tasks. An editorial, written by Scandura himself, published in the January 2012 issue of TICL discusses the individual’s “invariant” capacity for information processing and its effect on cognitive load. “Perhaps it will take several more decades for the implications of individual processing capacity to be realized in advanced instructional systems,” Scandura writes. “A major issue will be how to adapt instruction for students with different processing capacifies. It would seem that automated tutoring systems should be up to the task” (Scandura, 2012b).

To the extent that technology will continue to serve educational needs, structural learning theory will continue to contribute to the discussion and practice of teaching and learning.

Terms & Concepts

Behaviorism: The predominant theory of learning in the early-mid twentieth century. Behaviorists equated learning with changes in behavior, and said all learning could be explained via environmental factors. Many denied the existence of the mind altogether, referring to it as 'the black box.' Those who didn't deny its existence argued that it couldn't and shouldn't be studied.

Deterministic: Scandura characterizes structural learning theory as a deterministic one, as opposed to a probabilistic one. He argues that the behavior of individuals can be determined, on the basis of what one knows about what rules that individual does or does not have in his repertoire.

Operations: Scandura defines a rule as an ordered triplet, composed of a stimulus or input, an operation, and a resulting response, or output. When students are asked to perform addition, for example, the operation is the act of summing whole numbers.

Probabilistic: Most theories of learning, Scandura argues, are probabilistic in nature. That is, they predict human behavior on the basis of probability statements; the .05 significance level, for example, suggests we can have confidence in an investigation's results with 95% certainty. Scandura believes human behavior can be predicted with 100% certainty; by identifying what rules are necessary to solve a particular problem, and identifying the rules available to a particular individual, researchers should be able to predict exactly what an individual can and cannot do.

Response: Scandura defines a rule as an ordered triplet, composed of a stimulus or input, an operation, and a resulting response, or output. When students are asked to perform addition, the resulting whole number sums are the output.

Rules: According to Scandura, human learning can be explained in two ways - as a result of simple stimulus response associations, or as the results of rules. Even though Scandura believes what a person knows must be demonstrated via performance, he rejects the behaviorist notion that all behavior reduces to S-R connections. Scandura believes, rather, that all behavior is rule governed, and defines rules as a tripartite set of inputs, operations, and outputs. Finally, rules are not just abstractions to explain what researchers observe in humans; rather, individuals use and create rules in their everyday lives.

Stimulus: Scandura defines a rule as an ordered triplet, composed of a stimulus or input, an operation, and a resulting response, or output. When students are asked to perform addition, for example, the set of whole numbers they are asked to sum serves as the stimulus, or input.

Bibliography

Ikegulu, N.T. (1996). Scandura's structural learning theory: A critique. (ERIC Document Reproduction Service No. ED410216). Retrieved August 21, 2007 from EBSCO Online Education Research Database. www.eric.ed.gov/ERICDocs/data/ericdocs2sql/content_storage_01/0000019b/80/16/cd/2e.pdf

Scandura, J.M. (2012a). Comments on Ansari & Sykes and Gogus, and suggestions for future research. Technology, Instruction, Cognition & Learning, 9(1/2), 51–56. Retrieved December 14, 2013, from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=85423612&site=ehost-live

Scandura, J.M. (2012b). Editorial comment on the expertise reversal effect. Technology, Instruction, Cognition & Learning, 9, 133–135. Retrieved December 14, 2013, from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=85423618&site=ehost-live

Scandura, J.M. (n.d.). Overview. Retrieved August 21, 2007, from http://www.scandura.com/

Scandura, J.M. (1973). Structural learning: Theory and research. New York, NY: Gordon and Breach Science Publishers, Inc.

Scandura, J.M. (2001). Structural learning theory: Current status and new perspectives. Instructional Science, 29, 311-336. Retrieved August 21, 2007 from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=16980119&site=ehost-live

Scandura, J.M. (2001). Structural learning theory in the 21st century. Journal of Structural Learning and Intelligent Systems, 14, 271-306. Retrieved August, 21, 2007 from EBSCO Online Database Academic Search Premier. http://search.ebscohost.com/login.aspx?direct=true&db=aph&AN=5887805&site=ehost-live ..F.T.-Scandura, J.M. (2012c). The role of automation in instruction: Recent advances in AuthorIT and TutorIT solve fundamental problems in developing intelligent tutoring systems. Technology, Instruction, Cognition & Learning, 9(1/2), 3–8. Retrieved December 14, 2013, from EBSCO Online Database Education Research Complete. http://search.ebscohost.com/login.aspx?direct=true&db=ehh&AN=85423609&site=ehost-live

Suggested Reading

Scandura, J.M. (1977). Problem solving: A structural/process approach with instructional implications. New York, NY: Academic Press.

Scandura, J.M., & Scandura, A.B. (1980). Structural learning and concrete operations. New York, NY: Praeger Publishers.

Scandura, J.M. (1976). Structural learning: Issues and approaches. New York, NY: Gordon and Breach Science Publishers, Inc.