Forty-seven educable mentally handicapped adolescents between the ages of 11 and 16, living in a residential training facility, were given programmed instruction in phonics. Error measures on the two phonics programs and achievement scores on two administrations of the Wide Range Achievement Test (WRAT) were intercorrelated to determine the extent to which programmed instruction error measures could predict the criterion. Two error measures were used for instructional program performance: (1) error score, the total number of errors made on a program; and (2) error-frame score, the number of frames with errors in them. The instructional programs were selected because they elicited a wide range of error measures from different subjects, with substantial inter-subject variability. Correlations between the error measures and WRAT scores were consistent with, but somewhat higher than, those reported in previous studies. There was relatively high correlation, despite the use of comparatively short instructional programs. Two practical implications were noted. First, instructional programs may be used for mentally handicapped subjects instead of traditional tests when there are appreciable individual differences in the program error rates. Secondly, instructional programs used as tests may have educational value.