The size of an individual's phonological similarity effect for visually presented material is assumed to reflect his or her ability to recode, and by implication rehearse, information in verbal short-term memory. Many studies have shown that under these conditions, the size of this effect interacts with age, tending to be nonsignificant in children younger than 7 years and leading to the conclusion that children of this age do not rehearse. In the present study, the size of the phonological similarity effect was assessed in a total of 116 children aged between 5 and 9 years, manipulating the modality of both encoding and retrieval of the memoranda. Although the interaction between age and the size of the phonological similarity effect was replicated with visual presentation and verbal recall of material, this interaction was also present in other conditions that do not require recoding. In addition, the data from this "classic" condition were simulated by a model that assumed that the size of the similarity effect is (a) proportional to an individual's recall of dissimilar items and (b) constrained by a functional floor to recall of similar items. These findings undermine the evidence for a qualitative change in recoding and rehearsal at 7 years and question the extent to which rehearsal is necessary to explain the development of verbal short-term memory performance. (Contains 3 tables, 2 figures and 5 footnotes.)