In our recent paper (Muter, Hulme, Snowling, & Taylor, 1997) we argued that measures of segmentation were better predictors of early progress in learning to read than were measures of rhyme. Bryant (1998, this issue), in his comment on our paper, has argued that this conclusion is flawed because the instructions used in our rhyme detection measure included the phrase "rhymes with or sounds like." We present new data showing that the instructions used do not have the effect Bryant claims: asking children which word "rhymes with" or which word "rhymes with or sounds like" a target word produces identical patterns of responses. We argue that Bryant's new measure derived from our data simply reflects children's global sensitivity to the similarity in sound between different words and that this measure provides no convincing support for his conclusion that sensitivity to onset and rime is a predictor of children's success in learning to read. We conclude that the data in our paper, as well as other recent evidence, support the view that measures of phonemic segmentation are better predictors of early reading skills than are measures of onset-rime sensitivity. Copyright 1998 Academic Press.