The Trouble With Error Patterns

Many attempts have been made to develop computer programs that diagnose student errors in subtraction (e.g., Brown & Burton, 1978; Ohlsson & Langley, 1985; Young & O'Shea, 1981). Because these systems have been unable to diagnose approximately 50% of those children they analyze, researchers have sought better computer algorithms to identify more of the errors children make. Furthermore, each of these systems uses only answer (product) data for analysis. Drawing from the literature in clinical mathematics (e.g., Underhill, Uprichard, & Heddens, 1980) and psychology (Gagné, 1985; Resnick, 1982), an attempt was made to develop other measures to determine more fully a child's knowledge of mathematics. Also, a subtraction test (Van Lehn, 1982) was used to determine the child's error pattern. These two approaches were then compared to determine how clinical analysis might aid an error-pattern diagnostic system. The results indicate that error-pattern diagnosis is unrelated to a richer clinical approach, and that the clinical approach seems to be a more appropriate method for diagnosis.