In the apparent absence of suitable measures of idiom comprehension in normally developing and clinical populations, this study examined the relationship between a newly developed play task and a more conventional definition task. On the play task, children listened to a 1.5-minute, tape-recorded story into which were embedded 12 common idioms drawn from recordings of classroom teaching and children's television. As the story was then played again, sentence by sentence, the children were required to act it out using a play set and props. For each idiom, it was possible to act out either the idiomatic or literal meaning, but only the idiomatic meaning made sense in the context. The children's actions were video-taped and then played back to the child during the definition task. For this task, the video was stopped after each idiom occurred and the children were asked what they thought each idiom meant. Four groups of children were included. Twenty-six children (aged between 6-11 years), considered to have semantic-pragmatic difficulties, were compared with two groups of mainstream children (aged 6;6-7;6 and 10;6-11;6, respectively) and with a group of children (aged between 8-11 years) diagnosed with (other) language disorders not primarily of a semantic or pragmatic nature. The results indicate that the definition task underestimated common-idiom comprehension in normally developing children and, in particular, in children diagnosed with semantic-pragmatic difficulties or (other) language disorders. Furthermore, a significant difference in idiom comprehension between the two clinical groups evidenced on the play task was entirely masked on the definition task. It appeared that the expressive and metalinguistic demands of the definition task had a greater negative effect on the group of children with language disorders than on the children with semantic-pragmatic difficulties. Possible mechanisms through which the play task might have overestimated or underestimated idiom comprehension in these groups are examined and discussed. Although the nature of the play task probably facilitated idiom comprehension in these children, analysis supports the results of the play task being a reflection of true ability rather than an overestimation of idiom comprehension. A considerable incidence of false positive results on the definition task, coupled with its masking of significant differences in idiom comprehension across clinical groups militates against the use of definition tasks to assess idiom comprehension in research and clinical settings. In comparison with definition and multiple choice methodologies, the play task emerged as a more valid measure of common- and concrete-idiom comprehension in normally developing and clinical child populations.