The prevailing view of phonological development is that changes in pronunciation are driven by phonological changes. This view (it is argued here) derives from the particular form of the data that has most often been used in studies of phonological development, namely broad phonetic transcriptions. Transcribing an earlier pronunciation with one phoneme symbol and a later pronunciation with a different symbol encourages the interpretation that the child has made a flip from one category to another. However, broad transcriptions may have misrepresented the facts of speech development. We review some auditory-based studies which have used a more fine-grained phonetic transcription and discuss the significance of findings on the development of long-lag plosives. We argue that gradient change is the typical fashion in which children's speech output development progresses; that it is therefore not appropriate to use rules of the sort that are employed for morphophonemic alternations in adult phonology to explain revisions over time in children's pronunciations; and that a child's speech output is not the best guide to their phonology.