This study examined short-term motor memory of 100 mildly retarded and 100 nonretarded children during the recall of a preselected, arm movement. The focus was to determine the effects of anatomical differences in arm lengths between subjects on the measurement of movement accuracy. Absolute Error (AE) in its linear form was analyzed in a 2 (groups) X 5 (retention conditions) X 5 (trials) factorial yielding significant main effects for both groups and retention but not for trials. The analysis performed on the same data transposed to angular form (AE') gave a significant main effect for retention only. These results generally support the premise that differences between groups using a linear representation of movement error resulted in part from inaccuracy in estimating the actual anatomical (angular) movement.