Background: "Synthetic phonics" is the widely accepted approach for teaching reading in English: Children are taught to sound out the letters in a word then blend these sounds together. Aims: We compared the impact of two "synthetic phonics" programmes on early reading. Sample: Children received "Letters" and "Sounds" (L&S; 7 schools) which teaches multiple letter-sound mappings or "Early Reading Research" (ERR; 10 schools) which teaches only the most consistent mappings plus frequent words by sight. Method: We measured phonological awareness (PA) and reading from school entry to the end of the second (all schools) or third school year (4 ERR, 3 L&S schools). Results: Phonological awareness was significantly related to all reading measures for the whole sample. However, there was a closer relationship between PA and exception word reading for children receiving the L&S programme. The programmes were equally effective overall, but their impact on reading significantly interacted with school-entry PA: Children with poor PA at school entry achieved higher reading attainments under ERR (significant group difference on exception word reading at the end of the first year), whereas children with good PA performed equally well under either programme. Conclusions: The more intensive phonics programme (L&S) heightened the association between PA and exception word reading. Although the programmes were equally effective for most children, results indicate potential benefits of ERR for children with poor PA. We suggest that phonics programmes could be simplified to teach only the most consistent mappings plus frequent words by sight.