Three studies examined the sources of learning by which children, very earl
y in learning to read, formed correspondences between letters and phonemes
when these were not explicitly taught in the whole language instruction the
y received. There were three classes of predicted knowledge sources: (a) in
duced sublexical relations (i.e., induction of orthographic-phonological re
lations from the experience of print words), (b) acrophones from letter nam
es, and (c) transfer from spelling experience. The results of Study 1 indic
ated that children used both sources (a) and (b). Study 2 results showed th
at source (a) dominated when the letters were initial components of pseudow
ords rather than isolated items. The transfer from phoneme-to-grapheme corr
espondences of the children's spelling was examined in Study 3. The results
were not consistent with the use of source (c). The findings of these stud
ies have implications for the question of how early in learning to read chi
ldren are able to use knowledge from their experience of print words as a s
ource for phonological recoding.