Cognitive research has demonstrated that variations in a task can cause changes in strategy use, yet information about strategy use by problem solvers is not usually available to test designers or teachers. For example, some of the word problems presented in algebra courses may be solved instead through a non-algebraic strategy, "plug-in", which involves testing the validity of hypothesized answers.; The model of strategy choice developed in the current research asserts that for a given problem, students will choose a strategy requiring less cognitive "effort" (less of a burden on working memory) than alternative strategies. Working memory load was operationally defined as the maximum amount of information considered simultaneously during problem solving.; The memory load theory was instantiated and tested by varying four problem features in mathematics word problems. These problem features were expected to affect the amount of cognitive effort required to employ one of the strategies (either algebra or plug-in), but to have no effect on the other strategy.; The problem features manipulated were: (1) the number of different types of quantities (units of measurement) in the problem; (2) the number of semantic relationships in the problem; (3) whether or not the problem statement mapped directly onto algebraic equations; and (4) the difficulty of calculations required to employ the plug-in strategy. Strategies were determined from students' written solutions. Verbal protocols were also collected.; The strongest effect was found for the number of different types of quantities in the problem: as the number of different types of quantities increased from one to two, a greater proportion of students used plug-in. The number of semantic relationships and type of mapping did not affect strategy selection. Students avoided using plug-in when calculations were extremely difficult. The results applied to students at all levels of mathematics ability.; The theoretical understanding of students' problem-solving processes gained from this research can aid in design of algebra instruction. Tailoring test problems to elicit particular strategies may help to increase the construct validity of traditional tests in the difficult domain of mathematics word problems.
展开▼