Text integration and mathematical connections: A computer model of arithmetic word problem solving

Understanding arithmetic word problems involves a complex interaction of text comprehension and mathematical precesses. This article presents a computer simulation designed to capture the working memory demands required in ''bottom-up'' comprehension of arithmetic word problems. The simulation's sentence-level parser and text integration component reflect the importance of processing the problem from its original natural language presentatation. Children's probability of solution was analyzed in exploratory regression analyses as a function of the simulation's sentence-level and text integration processes. Working memory variables measuring the combined effects of concepts to remember and text integration inferences account for a significant proportion of variance in children's solution probabilities across the first four grade levels (K-3). Consistent with previous results from others, which highlighted the significance of small changes in problem wording, the simulation offers a process-oriented perspective as to why natural language presentation constrains the comprehension of mathematical relationships.