Two young children who were intellectually disabled and whose expressive language skills consisted of single signs were taught two-word semantic combinations. The study utilized combined multiple baseline and alternating treatment designs to compare the effectiveness of unimodal (speech and signs) and multimodal (speech, signs, and an electronic communication aid) augmentative communication conditions. The results indicated that, overall, treatment resulted in increases in the number of spontaneous productions of the target two-word combinations. However, the subjects showed differences in (a) their rates of learning, (b) the comparative effectiveness of the two augmentative communication conditions, and (c) their tendencies to use signs versus the communication aid. The need to specifically target two-word combinations and the relative effectiveness of the two conditions are discussed in relation to the nature of the experimental task and learner characteristics.