The study of computational errors among sixth grade students included identification and classification of errors, investigation of the effects of two feedback treatments and of classwork and homework on error patterns, and investigation of the relationships of error patterns with intelligence, mathematics achievement, attitudes toward mathematics, teacher appraisal, and combinations of these variables. The two feedback treatments consisted of eight weekly homework or eight weekly classwork assignments involving the use of the four fundamental operations on the digits 6, 7, 8, 9, and occasionally 5 and 0. One group received feedback only during the first four weeks, while the second group received feedback only during the last four weeks. Scores on independent variables were determined from cumulative records and administration of the Suydam-Trueblood Attitude Toward Mathematics scale and an experimenter constructed teacher appraisal scale. Data were submitted to analysis of variances, multiple linear regression, and regression analyses. No significant differences were uncovered for the feedback treatments, nor for the homework vs. classroom treatments. Feedback appeared to increase the number of errors when it was first provided. Different types of errors were best predicted by different independent variables, and five of the 24 pairs of dependent and independent variables showed significant interactions under the feedback treatment conditions. Several other effects were observed. (Author/SD)