In Experiment 1, classroom teachers were taught to delay their offers of help in naturally occurring situations, and thereby to provide additional opportunities for language use by six moderately retarded language-delayed children. The teachers introduced this delay technique in a multiple-baseline design across the six children. As delays were used, child verbal initiations increased. Follow-up assessment showed that teachers were maintaining greater than baseline levels of the delay technique after 10 weeks. Experiment 2 replicated the findings of Experiment 1, and included a more thorough maintenance assessment, while focusing on teachers' generalization of the delay technique. Teachers were found to generalize their use of delay to 56% of their monitored untaught opportunities. The two experiments show that (a) the delay technique is quick to teach and simple to implement, (b) delays do provide opportunities for children to initiate, (c) teachers can generalize their use of delay to novel self-selected situations, and (d) teachers can maintain their use of delays over time.