The problem we address is sentence planning in 3-8 year-olds and adults. We measure planning via patterns of non-fluency in studies with and without modeling. Participants in the study without modeling observed stories and then directed a blindfolded experimenter to pick up one of two identical toys in each story. Participants in the imitation study repeated a puppet?s request for a toy after each story. Both studies tested the same four types of relative clauses (varying gap position and depth of embedding). We analyzed time to utterance onset; frequency, duration, and distribution of filled and unfilled pauses; and use of optional functional elements. There were reliable effects of structural complexity on non-fluency patterns in both experiments, with some informative shifts across methods. For example, unfilled pauses distributed similarly across age groups, structures, and methods. But filled pauses (primarily, um) differed. In the elicited production study, adults preferred filled pauses before utterance onset; children also used them in the locations preferred for unfilled pauses. In the imitation study, the incidence of filled pauses sharply declined: Adults and older children produced almost none; young children?s pattern was more similar to that of the elicited production study.