The nature of a schema

Some of the limitations of the early work on schemas become strikingly evident when one faces the task of creating an explicit computer model of a particular schema. The characteristics of a schema, such as those illustrated in chapter 1, have not been determined at the detailed level required for computer modeling. They 

My purpose here is to join these two lines of thought in a single comprehensive theory. The theory is not a formal one of the type most frequently encountered in textbooks, with attendant definitions, axioms, postulates, and laws. Rather it is an informal theory of the sort that most often guides psychological research. At its heart is the premise that a schema consists of several different kinds of knowledge, as described later. A second and equally essential premise is that schema functioning involves both parallel and sequential processing. Both of these ideas have been noticeably absent from previous theories. 

Also key to the theory are a number of general schema characteristics, some of which have to do with form and some with content. Many of these characteristics have been articulated before by myself and other researchers, several have been generally accepted without having been put into words, and a few are new to the study of schemas. This chapter pulls them all together. 

To study the acquisition and use of schema knowledge, we must have a working definition that allows us to observe when such knowledge is present and when it is not. I propose the following  

A schema is a vehicle of memory, allowing organization of an individual s similar experiences in such a way that the individual 

---

Together, Bartlett s and Piaget s ideas provide the skeleton of schema theory. It begins to be understandable but does not yet have sufficient substance. Several recent researchers have sought that substance by performing more-detailed empirical investigations about the nature of a schema and by developing sophisticated computer models that simulate schema instantiation. 

Smolensky is more concerned with the mathematical development of harmony theory than with an elaboration of schema structure, although schemas play an essential role in harmony theory. He gives little attention to the nature of a schema, and his examples range from having a schema for a word to having a schema for a birthday party. By the definitions outlined here, only the latter qualifies as a schema. Thus, like Rumelhart et al.,... 

To begin, however, we must first have an idea of how space could be used in abstract cognition. The first part of this introduction seeks to illustrate what is meant by spatial schemas, what is meant by abstract cognition, and how the latter may benefit from the former. In the second part, the aim is to identify two important points in formulating accounts of spatial schemas. First, our account of spatial schemas explores the idea that space can be a mechanism for cognition, not merely a metaphor for cognition. Second, our account of spatial schemas aims to identify the nature of the correspondences between space and other domains, and how those correspondences, or mappings, are established. The third part of this introduction proceeds to provide an overview of the contents of the book. 

It is worthwhile noting that Aristotle, Plato, and Kant were all concerned primarily with concepts (e.g., both Aristotle and Kant devoted many pages to the nature of number and to describing what makes a triangle a triangle).18 Modern psychological usage moves beyond this scope and looks at events, experiences, and situations as well. However, much of what these philosophers observed about schemas for concepts is equally valid for our topics as well. 

Of the three researchers influencing schema theory in 1975 -Minsky, Schank, and Rumelhart - David Rumelhart has probably contributed the most to our current understanding of schemas. He has written a number of important articles about the nature of schemas, beginning with the one in 1975 and culminating with his work on parallel distributed processing (PDP) in 1986. Whereas Minsky introduced the frame and Schank the script, Rumelhart retained the word schema in his research. However, it takes on a substantially different meaning from that in earlier schema studies. 

Rumelhart s later work turned explicitly to the nature of the schema. In 1977, he and Andrew Ortony published a speculative study about what constitutes a schema. Rumelhart, as noted, approached the study of schemas through his work on story grammars (1975). Ortony arrived by way of studying the metaphor (1975). In their joint endeavor, they outlined four essential characteristics of the schema, and most of their chapter addresses these issues. However, the essence of their position can be found in a brief paragraph that precedes the discussion of these four characteristics. In that small paragraph (Rumelhart Ortony, 1977, p. 101), they lay out a number of important points ... 

The first point of interest is that a schema is a structure for organizing data. The implication is that we should pay attention to its form. This is a key issue. Unless we can create exact hypotheses about the nature of the structure and how it serves memory, we will be unable to create viable models of the schema. Together with Minsky and Schank, Rumelhart and Ortony pushed the notion of schema away from a purely conceptual or interpreta-tional construct toward one whose form could be specified. This approach moves well beyond Kant s speculations and foreshadows the importance of formal models. 

Rumelhart and Ortony s third point is that there are a number of interrelations in a schema. A schema is not simply a list of features but rather is a collection whose parts are linked together. The nature of these linkages remains unspecified, but this notion is an... 

Minsky, Schank, Rumelhart and Ortony, and Winograd made important hypotheses about the nature of the schema. However, a limitation in their presentations is that they were confined to hypothesis or speculation. Brief hints appear of how models might be constructed or how individuals might use schemas, but they are not elaborated. We are left at this point with several good ideas but no evidence of whether schemas provide a viable means of describing human cognition. We must look elsewhere for such evidence. 

Stories of these kinds are used to demonstrate that individuals do use prior experience to comprehend stories. However, as evidence of schemas, they are somewhat weak. What we have is confirmation that prior experience is stored in memory and that it influences recall. Studies such as these cannot provide evidence about the structure of the memory storage. Why not We cannot make any such claims because the nature of the storage has not been specified nor have the specific contents. All that we can determine from these studies is that the prior knowledge does influence the subjects as they read and encode the story for later recall. However, these studies have made an important contribution to the overall understanding of schemas by expanding the focus of the research beyond the formal symbolism of the structural rule set to a more general conceptual framework. 

There are potentially three big areas in schema research the investigation of how schemas influence memory recall, the specification of models to simulate schema instantiation, and the study of how schemas develop and change. All require substantial theoretical and empirical study, but the nature of the theories differ. A considerable body of research exists for the first two areas, as is evident from the preceding discussion, but the third is largely untapped. For the most part, researchers have taken schemas as given, and few have dealt with issues having to do with schema creation or modification. 

Adopting schema theory as the basis of instruction almost certainly necessitates a complete restructuring of the curriculum. It also raises questions about the nature of learning. In this part and the next, I address these two issues. 

A point to be highlighted is that students will develop schemas about the subject matter whether instruction takes a schema-based approach or not. Learners will search for structure and relationships. Thus, the questions to be asked about schema-based instruction do not center on whether or not students create schemas. They do. The questions focus instead on the nature of the schemas that are developed. We ask whether the instruction itself can promote more cohesive and better structured schemas than would instruction having another foundation. It is here that the basis set of schemas becomes important, because these lay the groundwork for the instructional design. 

Over the past 25 years, visual representations have received considerable attention under the various designations of graphics, diagrams, visual aids, or simply pictures. No matter which name is used, the issue is whether their inclusion in instruction makes a significant difference in how and what students learn. The question of their importance extends naturally to schema theory. The most interesting issues have to do with how visual representations fit into the knowledge organization of a schema and what role they have in its access and use. 

The most obvious consequences of schema-based assessment show up in the content of test items. Innovative items are needed for the simple reason that traditional items will not suffice. Different questions are to be posed different answers are to be sought. The innovations may derive from a modification in the nature of instruction, or they may result from a shift in test objectives. Both of these reasons are operative when we consider schema-based tests. 

The most noticeable difference in test content between a schema-based test and a traditional test is the emphasis in the former on intermediate decisions. A schema-based test will attempt to tease apart the four knowledge components of one or more schemas, and these knowledge components may be utilized at different points in the formulation of a student s test response. It is very difficult to construct a test item in which all four knowledge components for one schema are assessable. Almost certainly some will be invisible. Thus, to evaluate a single schema, one probably will need several items, each devoted to the evaluation of one or perhaps two of the components. The questions that tap into the various knowledge components are unlike those that typically appear on tests because they target the intermediate steps at which these knowledge components are needed. Consider, for example, the nature of the questions asked in the exercises of SPS (see Chapter 5). 

Test construction and item generation under schema theory require new approaches. It is worth considering how test items can be constructed to reflect the linkages expected in students knowledge of a subject. The nature of the linkages points to two possible representations a hierarchy and a network. 

---