Most developmental models of reading assert that knowledge of the alphabetic principle is necessary for satisfactory progress in the acquisition of word reading. Byrne and Fielding-Barnsley (1989, 1990) have provided operational procedures for determining a child's initial knowledge of the alphabetic principle. Their claims that both explicit letter-sound knowledge and phoneme awareness are necessary for initial learning of the alphabetic principle have been very influential. Much of their evidence is from two series of training studies (1989, 1990). None of these include pretest information on the phoneme awareness that was trained, or their tasks for testing knowledge of the alphabetic principle. Neither was there any control group or comparison with alternative training. Our purpose was to replicate one of these experiments but with an improved training design. In our study 34 children with an average age of 54 months from preschools in mainly low SES communities in New Zealand were pretested on knowledge of letter names, letter sounds, phoneme identity (for initial position of spoken words), receptive vocabulary, and word reading. Children were not included if they could read words, if they knew any letter sounds among those to be trained or tested, or knew many letter names. The children were randomly allocated to one of two groups. One group received training in phoneme identity (PI), as in the Byrne and Fielding-Barnsley study (1990, Experiment 1). The other group received alternative training on the initial letters of print words, instead of PI. There were no significant group differences on the pretests. The results showed no pre- to post-test gains in knowledge of the alphabetic principle in either group, with training taking much greater time for the PI instruction group. There were no pre- to post-test gains in PI performance on the two consonant phonemes, /m/ and /s/, claimed to be the easiest to learn in the Byrne and Fielding-Barnsley results. The failure to replicate with an improved design, calls into question the evidence claimed of this landmark study.