Previous research has established that the degree of "wordlikeness" of nonwords affects young children's nonword repetition performance. Experiment 1 examined the possibility that output processes are responsible for the wordlikeness effect by using a probed recall procedure. Wordlikeness was defined in terms of phonological neighbourhood density, although this measure was found to be related to the traditional measure of wordlikeness involving adult ratings. A significant effect of number of phonological neighbours wordlikeness was observed in favour of nonwords with many neighbours. In Experiments 2 and 3, the wordlikeness effect was qualified by a significant interaction with nonword repetition ability. Children with poorer repetition ability were affected by number of neighbourswordlikeness, while children with better repetition ability were not. Children with poorer repetition ability were significantly poorer than the better repeaters with nonwords with few neighbours. The results were interpreted in terms of theories of phonological development that suggest progressive segmentation of lexical representations. In Experiment 4 the relationship of children's nonword repetition ability to phonemic discrimination ability was investigated. The results demonstrated that children with better nonword repetition ability had superior phonemic discrimination performance than children with poorer nonword repetition ability.