This study compared the error patterns of third- and fourth-grade children on addition, subtraction, and multiplication exercises in horizontal and vertical formats. The research design used parallel eight-item tests for each operation; the only difference between items on parallel forms was the formal (horizontal or vertical) of the computational problems. Data were summarized for the entire group and for two subgroups defined as low and average socioeconomic status (SES). In general, the error rate was higher on horizontal format problems. Over all problems, the average SES group performed better than the low SES group. Appended to the report are data summary tables, a testing schedule, and copies of the tests. (SD)