A total of 411 elementary school children seven to twelve years old read short prose passages and selected the most appropriate continuation sentence from four alternatives. The completion sentences were constructed so that the correct (target) response involved either an explicit (simile) or an implicit (metaphor) metaphorical comparison. It was found that the children selected significantly more explicit target sentences than implicit target sentences. The children also made more correct selections when the referent of the metaphorical comparison was specifically denoted than when the identity of the referent had to be inferred. Thus, the measurement of children's understanding of metaphorical language was all too easily confounded by general language variables (indirectness and referential specificity) having no particular connection to metaphorical language. (RL)