Mathematics Education Research Journal, VolA, No.1, 1992. SOME COGNITIVE FAcTORS RELEVANT TO MATHEMATICS INSTRUCTION John Sweller and Renae Low, University of NSW Our understanding of cognitive processes has progressed sujficientLy in the last few years to enable us to generate novel instructional techniques that can enhance substantially learning of subjects such as mathematics. This paper will review briefly some research intended to contribute to this process. There are two relevant aspects. Firstly, recent work has thrown light on schema acquisition while learning mathematics, and on techniques for detecting schemas in mathematics learners. Secondly, other research has assessed the distribution of cognitive resources while learning mathemarics and other related subjects leading to the design of instructional techniques to facilitate schema acquisition. The Relevance of Schemas to Mathematical Expertise Cognitive research findings have revealed the importance of schema acquisition in successful learning and problem solving in mathematics (see Sweller, 1988). A problem schema can be defined as a cognitive construct that allows problem solvers to recognise a problem as belonging to a specific category that requires particular moves for solution. This definition includes problem states, problem-solving operators, and their relations. Consider a problem such as, "If a boat travelled downstream in 120 minutes with a current of 5 kilometres per hour and the return trip against the same current took 3 hours, what was the speed of the boat in still water?" Irrespective of the person's skill in computation and their basic linguistic and factual knowledge (e.g., what "still water" means, that rivers have currents that run only downstream), this problem is beyond the capability of someone who does not know the specific relations between speed of boat, rate of current, and time involved in river current contrast problems. In the absence of such schematic knowledge, there is no context for computation and calculation. The person needs to be able to identify the problem as belonging to a given type (e.g., river current contrast) in order to determine what information from the text should be used, in what sequence, and through what operations (see Mayer, 1987, pp. 345-373). 84 Sweller and Low Novices and experts have been shown to differ not only in solution rates, but in schematic knowledge. When asked to remember problem states, experts who have acquired appropriate schemas can use them to encode information with reference to problem structure. A novice, not possessing such schemas, will find problem configurations more difficult to reproduce (Chase & Simon, 1973). Similarly, because a schema classifies problems according to solution steps, it is likely to be used as a mode for classification. Experts can categorise problems as similar or different on the basis of underlying principles, but novices are readily deceived by context (Chi, Glaser & Rees, 1982). Definitions of a Problem Readers familiar with the area will recognise that we are using a particular conceptualisation of a problem. This conceptualisation distinguishes between experts and novices but not problems and exercises. It is assumed that we cannot define a particular task as constituting a "real" problem in isolation, without reference to the person who is attempting to solve it. A problem for one person may be an exercise for another and only problems unsolved by anyone (such as unresolved research questions) could universally be classified as "real" problems for everyone. Once a solution to such problems is discovered and learned by sections of the population, these people can be classed as experts with respect to the relevant problems while the rest of us are novices. From this discussion it can be seen that by distinguishing between experts and novices we have defined all tasks as problems. We know how to solve some problems and so are experts; other problems we do not know how to solve and so are novices. No task can be classed as either an exercise or problem simply by referring to its structure and components. Alone, the structUre of a task cannot reveal its "problem" status. Its status is revealed fully by the novice-expert distinction. From an educational perspective, obviously we should be concerned with helping novices become experts and this paper is concerned primarily with that issue. We know also that some people are much better than others at solving novel problems and it is equally legitimate to conduct research on techniques for improving people's ability to solve such problems. Nevertheless, we do not believe that these techniques have been discovered as yet. They may not exist. We do believe that techniques for facilitating the transition from novice to expert in domain-specific areas are available and some of these procedures are discussed below. Measures of Schematic Knowledge Schematic knowledge has been studied mainly through three kinds of tasks: problem classification, problem recall, and problem construction. Hinsley, Hayes and Simon (1977) assessed schematic knowledge by requiring relatively experienced students to group algebraic story problems Cognitive Factors 85 into clusters, categorise a given problem after hearing only part of the text, solve problems in which content words were replaced by nonsense words, and solve problems when irrelevant information in the text had produced ambiguity. They found extensive agreement between subjects in the categorisation of problems. Further, after hearing the words "A river steamer. .. ", many subjects judged that the problem was going to be one involving the river current type. From a number of converging demonstrations, Hinsley et a1. (1977) concluded that the encoding and retrieval of information in the process of solving algebraic word problems is governed by a person's schematic knowledge. In further research emphasising the role of problem schemas, Mayer (1982) required adults to recall problems to which they had been exposed earlier, and to construct problems within a theme (for example, "trains leaving stations"). Recall was best for the types of problems that occur frequently in textbooks, and schema-relevant material was more likely to be recalled accurately than schema-irrelevant information. In addition, the problems constructed by students tended to approximate standard rather than nonstandard formats. These results, Mayer argued, suggest that while students possess schematiC knowledge for simple versions of a problem category, their schematic knowledge for the more complex versions are less accessible. Text Editing as a Measure of Schematic Knowledge Schematic knowledge also can be assessed by requiring students to edit problem texts to specify what information is necessary and sufficient for solution. It has been claimed that schemas provide the basis for filJing gaps and for selecting relevant information (Chi & Glaser, 1985). To be able to supply information that is missing from a problem text but is necessary for the problem to be soluble, or to tease out irrelevant information contained in the problem text, one has to know the minimal elements that have to be present in the text in order for the problem to be open to solution. Since the content of a problem schema includes principles, formulae, concepts, and characteristics of the problem situation (De long & Ferguson-Hessler, 1986), it can be assumed that successful text editing of algebraic word problems requires schematic knowledge. To illustrate, when presented with a problem that is incomplete in the sense that further information is needed before a solution is possible (such as "A rectangular lawn is 72 square metres. What is the width of the lawn?"), the student needs to know that the area of a rectangle is the length of one side multiplied by the length of the adjacent side in order to identify the essential component missing from the problem. Similarly, when presented with a problem such as "A truck leaves Melbourne for Sydney at 1.00 a.m. The distance between Melbourne and Sydney is 1200 kIn. If the truck travels at 80 km/h, how long will it take to reach Sydney?" a student is able to specify which information in the text is irrelevant for solution only through knowing about the structure of this class of problem. If solution requires schematic knowledge, a student who can solve a given problem should be able to identify the minimal text that will allow the problem to be 86 Sweller and Low solved. In a series of three studies in which tenth grade students were required to classify algebraic word problems in terms of whether a problem contained sufficient, insufficient, or irrelevant information for solution, relationships were demonstrated between text editing and solution, as well as between text editing and several indices of schematic knowledge (Low & Over, 1990). Text editing performance correlated with judgments of similarities and differences between problems and with memory for problems. Further, text editing as a skill was predictive of solution rate, and of mathematical achievement in general. Students with higher scores on tests of general mathematical ability were more accurate in cClassifying algebraic word problems in terms of whether they contained sufficient, insufficient, or irrelevant information. This relationship applied even after allowance was made for contributions from verbal skill. As noted earlier, measures that have been used in past research to index schematic knowledge include identification of similarities and differences between problems and recall for problems that were previously presented. Low and Over (1990) have shown that scores on the text editing task correlated not only with the probability that problems were solved, but with recall memory for problems and discrimination of whether problems are similar or different. These relationships are to be expected if text editing, in common with problem solution, problem classification, and recall memory, assesses the extent to which a student has processed information based on understanding of the structure of the problem. Hierarchical Ordering of Schematic Knowledge Low and Over (1990) noted that the nature of errors that students make in text editing suggest that individuals who lack knowledge of the higherorder template appropriate for the problem in hand will conceptualise the problem as one entailing a lower-order template. Consider, for example, the problem, "The length of a rectangular window pane is twice its width. If the area of the pane is 98 cm2 , what are the dimensions of the pane?" In order to classify this problem as one containing information that is necessary and sufficient for solution, a student needs knowledge of what Mayer (1981) termed the "area relative" template. Most students who classified this problem as providing insufficient information indicated that either the width or the length of the pane must be provided in the text in order for solution to be possible. Since these students specified the length of one side as the information that was missing, they were clearly employing the "simple area" template. Using four templates for area-of-rectangle (adjacent sides, ratio of sides and one side, perimeter and side, and diagonal and side) in three different task formats (text editing, knowledge of formula, problem solution) Low and Over (l992b) demonstrated hierarchical organization of the schematic knowledge underlying the solution of the four problems. Solution, text editing, and . knowledge of formula task were most accurate for the adjacent sides Cognitive Factors 87 templates, followed by the ratio and side template, the perimeter and side template, and then the diagonal and side template. For all three tasks, students rarely were successful on a problem requiring a template above one on which they had failed. Performance on problems involving a lower-order template could thus be predicted by performance involving a higher-order template. Although the basis for the hierarchical ordering of templates found in the Low and Over (l992b) study is unclear, this investigation, together with Mayer's (1981) taxonomy offer a starting point for exploring the manner in which schematic knowledge is organised and integrated. The methodology used by Low and Over (1992b) can be used to establish relations between families, categories, and templates, particularly whether there is hierarchical ordering of templates within families or categories. Such work may have important implications for classroom instruction. For example, if the acquisition of competence in the use of one template proves to be dependent on the mastery of another, instruction should be programmed to ensure that students are not exposed to superordinate templates until they have demonstrated understanding of subordinate templates. Gender Differences in Schematic Knowledge Following Marshall and Smith (1987) who characterised gender differences in mathematical performance in terms of differences in information processing, Low and Over (1992a) compared boys and girls in terms of schematic knowledge. In one experiment, boys and girls in tenth grade classified algebraic word problems in terms of whether the problems contained sufficient, irrelevant, or insufficient infonnation for solution. Among students with similar levels of general mathematical ability, girls were less likely than boys to identify missing information or irrelevant information within problems. More girls than boys perceived irrelevant information within the text of a problem as being necessary for solution. It was found in a subsequent experiment that girls who were as able as boys to solve algebraic word problems containing sufficient information had lower solution rates than boys on problems containing irrelevant infonnation. On these latter problems the girls more often than the boys incorporated the irrelevant information into their attempted solution. Low and Over (1992a) suggested that the results point to differences between boys and girls in their schematic knowledge. Marshall and Smith (1987) suggested that traditional instructional procedures initially produce, and later maintain gender differences in the cognitive processes that are involved in learning mathematics. There may be scope for modification of educational practices so that neither boys nor girls are disadvantaged in acquiring skills that are basic to competent performance in mathematics. The results obtained by Low and Over (1992a) point to the need for instruction designed to ensure that all students (giris as well as boys) acquire schemas pertinent to the solution of mathematical problems, and that students adopt a schema-oriented approach to problem solution. 88 Sweller and Low Cognitive Load Theory and Techniques to Design Instruction to Facilitate Schema Acquisition If schema acquisition is a major component of mathematical problem solving expertise, clearly it is important to determine ways of facilitating the acquisition of schemas. Cognitive load theory (Sweller, 1988; 1989; see also Ayres & Sweller, 1990) has been developed for this purpose. The theory assumes that instructional designs must be structured in a manner that focuses attentional resources on problem states and their associated moves because it is familiarity with problem states and their associated moves that provides the basis of schemas. If instructional designs require students to devote cognitive resources to other aspects of a problem or other features of instruction, a heavy, extraneous cognitive load may be imposed that interferes with learning and problem solving. As it happens, many conventional instructional procedures are inappropriate from the perspective of schema theory and cognitive load theory. The remainder of this paper will be concerned with some ofthese difficulties and techniques for overcoming them. Using Problem Solving as a Learning Technique The solution of many problems or exercises as a means of gaining experience in a particular area probably is a universal procedure. Most teachers have their students solve problems, most texts have a series of problems to solve at the end of each section. The strategy most of us use to solve novel problems for which we do not have a schema is called meansends analysis. This strategy requires us to attempt to reduce differences between each problem state we encounter and the goal state. If we are presented with a geometry problem for which we must find a value for Angle X (as in Figure 1), the diagram provides the initial problem state, Angle X is the goal, and the theorems of geometry are the problem operators that we must use to transform the initial state into the goal state. Thus, using meansends analysis, we may work backward from the goal angle attempting to find theorems that will allow us to connect the goal angle to the givens of the diagram. Using means-ends search we can establish that since we have a value for Angle CDE, if we had a value for Angle DEC we could find X using the theorem, external angles of a triangle equal the sum of the opposite internal angles. Angle DEC becomes a subgoal. We can find a value for Angle DEC from Angle FEG using the theorem, vertically opposite angles are equal. Once we have succeeded, in constructing this chain we can then reverse direction and substitute values until we have a value for the goal, thus solving the problem. Searching for relevant angles and theorems is the major intellectual process of.the task. This means-ends strategy is an excellent technique for attaining a problem goal but except in tests, we do not present students with problems to solve in order that they will attain the goal. We give students problems so that they will learn and it is learning that is the real goal of problem solving in most educational ·contexts. From the perspective provided above, we know Cognitive Factors 89 that schema acquisition is a major aspect of learning. Means-ends analysis does not place an emphasis on learning to recognise and categorise problem states and their associated moves. Instead, it requires a complex search process involving relations between problem states and the operators that can be found to reduce differences between states. We can predict that solving problems by means-ends analysis imposes a heavy extraneous cognitive load that interferes with learning. In other words, under many commonly found conditions, learning and problem solving may be incompatible. (See Owen & Sweller, 1989, for a more detailed exposition of this position, Lawson, 1990 for a response, and Sweller, 1990, for a reply to Lawson). o _ _ _ _ _r#- ~r_._,:_-F A G II Solution: J. Angle DEC = Angle FEG (verlically opposile angles are equal) = 50' 2 Goal Angle X = Angle CDE 'Angle DEC (ExternaJ angles of a triangle equaJ the sum of the opposite interoal angles) = 60', 50· = 110· Figure 1. A conventionally structured worked example demonstrating a procedure for finding a value for Angle X. Two techniques have been used to simultaneously provide evidence for this suggestion and to test alternatives to conventional problem solving that are more in accord with the requirements of schema acquisition. The 90 Sweller and Low techniques are a heavy use of (a) goal-free problems and (b) worked examples. These will be discussed next. Goal-Free Problems If means-ends analysis imposes a heavy cognitive load that interferes with learning, then preventing students from using a means-ends strategy and having them attend to problem states and their associated moves should facilitate schema acquisition. Goal-free problems should have this effect. Consider the previous geometry problem again in which the goal is to find a value for Angle X. Assume this goal is replaced by the statement,"Find the value of as many angles as you can." With this statement, it is no longer possible to reduce differences between current problem states and the goal state because the goal state is insufficiently specified to be used in this manner. All students can do is attend to each problem state and attempt to find moves that can be made from that state; precisely what we might assume they need to do to acquire a schema. In many experiments comparing goalfree with conventional problems, this effect was obtained with learning substantially enhanced by goal-free problems (See Owen & Sweller, 1985; Sweller, Mawer & Ward, 1983; Tarmizi & Sweller, 1988. In addition, Silver, 1990, gives useful examples of goal-free problems.) Worked Examples 'While worked examples are vastly different in structure from goal-free problems, from a cognitive load perspective they should have similar effects. Consider a worked example in elementary algebra: Solve for a (a + b)c = d a + b = dlc a = (dlc) - b To process this worked example, students must do no more than concentrate on each problem state (line) and the operator (rule of algebra) needed to transform the problem state into the next state. In contrast, if the problem is presented without the solution (first line only), students must engage in the usual means-ends search to fmd a solution. They must consider the difference between (a + b)c and a and try to find operators to reduce the differences. Cooper and Sweller (1987) demonstrated that this can be a very difficult task for novice algebra students. The cognitive load associated with a means-ends strategy interfered with learning. Both in that paper and in Sweller and Cooper (1985), results of many experiments demonstrated that giving students many algebra worked examples to study was substantially superior to having them solve the equivalent problems. The Split-attention Effect It might be considered that the appropriate conclusion from the above studies on worked-examples is that they should universally replace Cognitive Factors 91 conventional problems. In fact, the only legitimate conclusion is that when teaching mathematics, extraneous cognitive load should be reduced. Conventionally structured worked examples in algebra focus attention almost solely on problem states and their associated moves, and this is important if schema acquisition is the goal of the exercise. Conventionally structured worked examples in other areas have quite different properties. Consider a geometry worked example as in Figure 1. It consists of a normal diagram and associated statements. Neither the diagram nor the statements are intelligible until they have been mentally integrated. Once they have been integrated, the worked example can be processed. The act of mental integration requires cognitive effort. The sole reason for the cognitive effort is to mentally restructure the example so that it can be understood. With a slightly different structure as in Figure 2, the cognitive resources required to mentally integrate the diagram and statements might not be required for this purpose and so could be used entirely to better understand the mathematics associated with the example. D G) Go3J Angle X CDE • Angle DEC (External angles of a triangle equal the sum of the = Angle opposite internal angles) 60'· 50' 110' ____ = = A CD Angle DEC Angle FEG (verticaJJyopposite angles are equ3J) = 50' = ~<#_------------------ i"_~-F B G Figure 2. An integrated worked example demonstrating a procedure for finding a value for Angle X. Several studies have demonstrated that if worked examples are presented in a form that includes multiple sources of information that are unintelligible until they have been mentally integrated, then such worked examples are no more effective as a student learning device than solving the equivalent problems. Mentally integrating disparate sources of information imposes an approximately equal extraneous cognitive load as solving the 92 Sweller and Low equivalent problems by means-ends analysis. In contrast, such examples can almost always be restructured to elimin.ate split-attention between disparate sources of information. For example, in geometry, the geometric statements are placed at appropriate locations on the diagram as in Figure 2 rather than physically separated from it as .in Figure 1. When physically restructured to reduce the need for mental restructuring and thus reduce extraneous cognitive load, worked examples are massively more effective than either conventional examples or conventional problems. It should be noted at this point that the split-attention effect applies to all instructional material, not just worked examples. Any instructional material that unnecessarily requires students to split their attention between disparate sources of information should be physically integrated if possible. Results suggest that the seemingly minor alterations involved in physical integration can have enormous benefits unless we are dealing with redundant rather than essential sources of information. (Experiments on the split-attention effect may be found in Sweller, Chandler, Tierney & Cooper, 1990; Tarmizi & Sweller, 1988: Ward & Sweller, 1990.) Conclusions Work on cognition and instruction in mathematics has provided us with a new conception of what is involved in mathematical skill. A skilful mathematician has a huge knowledge base consisting of tens of thousands of schemas acquired over many years. For any individual, mathematical problem solving skill is restricted to those areas of mathematics for which he or she has access to a large, readily accessed knowledge base. The major task of someone wishing to acquire mathematical problem solving expertise is not to learn the rules of mathematics (although the rules of mathematics obviously are essential) but rather, to acquire the huge range of schemas that the subject demands. Frequently, there is a discrepancy between what is taught in mathematics classes and what must be learned. While mathematics instructors may place a heavy emphasis on the rules of mathematics, students, to be competent, must both learn the rules and acquire schemas. For many students, learning the rules is a relatively simple task. Unfortunately, simply knowing the rules needed to solve a particular problem frequently is insufficient to allow us to find a solution. We have difficulty applying previously learned rules in unfamiliar contexts and our students are no different. We require the relevant schemas and acquiring a large body of schemas may be overwhelming for many students. Successful schema acquisition may require teaching techniques that differ from those traditionally used. Our knowledge of the processes of cognition may have generated new procedures to facilitate schema acquisition. Hopefully, these procedures may make mathematics classes more productive and enjoyable for both students and teachers. Cognitive Factors 93 References Ayres,P., & Sweller,J. (1990). Locus of difficulty in multi-stage mathematics problems. American Journal ofPsychology, 103,167-193. Chase,W.G., & Simon,H.A. (1973). Perception in chess. Cognitive Psychology, 4, 55-81. Chi,M., Glaser,R., & Rees,E. (1982). Expertise in problem solving. In R.Sternberg (Ed.), Advances in the psychology of human intelligence (pp. 7-75). Hillsdale, NJ: Erlbaum. Chi,M.T., & Glaser,R. (1985). Problem solving ability. In R.J. Sternberg (Ed.), Human abilities: An information-processing approach (pp. 227250). NY: Freeman. Cooper,G., & Sweller,1. (1987). The effects of schema acquisition and rule automation on mathematical problem-solving transfer. Journal of Educational Psychology, 79, 347-362. De Jong,T., & Ferguson-Hessler,M.G.M. (1986). Cognitive structures of good and poor novice problem solvers in physics. Journal of Educational Psychology, 78, 279-288. Hinsley,D.A., Hayes,J.R., & Simon,H.A. (1977). From words to equations. In M.A.Just & P.A.Carpenter (Eds.), Cognitive processes in comprehension (pp. 89-106). Hillsdale: Erlbaum. Lawson,M. (1990). The case for instruction in use of general problemsolving strategies in mathematics: A comment on Owen and Sweller (1989). Journalfor Research in Mathematics Education, 21,403-410. Low,R., & Over,R. (1990). Text editing of algebraic word problems. Australian Journal ofPsychology, 42,63-73. Low, R., & Over, R. (1992a). Sex differences in solution of algebraic word problems containing irrelevant information. Unpublished manuscript, University of New South Wales. Low,R., & Over,R. (l992b). Hierarchical ordering of schematic knowledge relating to the area-of-rectangle problems. Journal of Educational Psychology, 84, 62-69. Marshall,S.P., & Smith,J.D. (1987). Sex differences in learning mathematics: A longitudinal study with item and error analyses. Journal of Educational Psychology, 79, 372-383. Mayer,R. (1981). Frequency norms and structural analysis of algebra story problems into families, categories and templates. Instructional Science, 10, 135-175. Mayer,R. (1982). Memory for algebra story problems, Journal of Educational Psychology, 74, 199-216. Mayer,R.E. (1987). Educational psychology: A cognitive approach. Boston: Little, Brown. Owen,E., & Sweller,J. (1985). What do students learn while solving mathematics problems? Journal ofEducational Psychology, 77, 272-284. Owen,E., & Sweller,J. (1989). Should problem solving be used as a learning device in mathematics? Journal for Research in Mathematics 94 Sweller and Low Education, 20, 322-328. Silver,E. (1990). Contributions of research to practice: Applying findings, methods and perspectives. In T.Cooney (Ed.), Teaching and Learning Mathematics in the 1990s: 1990 Yearbook (pp. 1-11). Reston: National Council of Teachers of Mathematics. Sweller,J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12,257-285. Sweller,J. (1989). Cognitive technology: Some procedures for facilitating learning and problem solving in mathematics and science. Journal of Educational Psychology, 81, 457-466. Sweller,!. (1990). On the limited evidence for the effectiveness of teaching general problem solving strategies. Journalfor Research in Mathematics Education, 21, 411-415. Sweller,J., Chandler,P., Tiemey,P., & Cooper,M. (1990). Cognitive load and selective attention as factors in the structuring of technical material. Journal of Experimental Psychology: General, 119, 176-192. Sweller,!., & Cooper,G.A. (1985). The use of worked examples as a substitute for problem solving. Journal of Experimental Psychology; General, 112, 634-656. Sweller,J., Mawer,R., & Ward,M. (1983). Development of expertise in mathematical problem solving. Journal of Experimental Psychology: General, 112, 634-656. Tarmizi,R., & Sweller,!. (1988). Guidance during mathematical problem solving. Journal of Educational Psychology, 80,424-436. Ward,M., & Sweller,!. (1990). Structuring effective worked examples. Cognition and Instruction, 7, 1-39.
© Copyright 2026 Paperzz