The effects of prompting and confirmation on automated presentation of materials in discrimination learning were studied. Eight pairs of words or pictures were presented to 48 mentally retarded subjects (mean IQ 63, mean chronological age 163.4 months, mean mental age 103.3 months). Each subject's correct responses advanced the program and, in the prompting condition, the correct response was always underlined. The number of errors in the original learning was less under prompting (p .001). Treatment by order interaction was also significant (p .01) with better performance for prompting from word-picture order and for confirmation from picture-word order. Seven days later, prompting subjects committed fewer errors in relearning (p .001) and were superior in recognition (p .01). Confirmation was further tested under two new conditions: in one, praise and candy reinforced all correct responses; in the other, they were given only upon correct first choices during practice trials. The first group had fewer errors (p .01) than the second or those in the analogous condition in the earlier experiment; and performance in the second experiment was not significantly different from that seen in prompting. (JD)