During the 1973-74 school year, 230 trainable mentally retarded (TMR) children (ages 7 to 14 years) were exposed to one of two language training conditions: Distar or Peabody. A population of 116 continuees from the first year of the project and 114 new entries were assigned in as random a fashion as possible to either Distar or Peabody. Ss were divided into low IQ (21-43) and high IQ (44-53). Sex was built into the design, as was pretest-posttest and new entries versus continuees. Thus, a five-factor, 2 x 2 x 2 x 2 x 2 repeated-measures design was subjected to analysis of variance for each of three basic criteria: Peabody Picture Vocabulary Test, Illinois Test of Psycholinguistic Abilities, and Mecham Verbal Language Development Scale. Seven children were selected randomly from each of the 16 between-factor cells to yield a total of 112 children. Longitudinal analyses were also conducted on just the continuees with pre- and posttest data from the three basic measures from both years of the project to yield a treatments-by-IQ-by-Sex-by-Measures (2 x 2 x 2 x 4) design. While no significant differences emerged for the high-IQ children, the low-IQ children were aided more by Distar than by Peabody. In the 5-way designs, gain in the total sample was not marked. However, when one considers only the continuees (in the second set of analyses), significant gain in language functioning did occur. (DB)