Tests are made of an aspect of the "knowledge sources" theoretical account of acquisition of reading in which, contrary to the developmental bypass hypothesis, it is postulated that sublexical relations between orthographic and phonological components are formed very early in learning by spontaneous induction from stored print word experience. Experiments 1 and 2, conducted with 5- and 6-year-old children, indicated as predicted that positional frequency of orthographic components in experienced print words influenced reading responses to unfamiliar pseudoword items. In Experiment 3 positional frequency of an orthographic component was manipulated in a training-transfer paradigm. Transfer to pseudoword reading was as predicted. The results could not be given alternative explanations by the developmental bypass hypothesis nor by accounts which predict exclusive use of onset and rime units at this early reading level.