In a previous article, Moore, Rosenberg and Coleman (Brain and Language, 2005, 94, 72-85) reported evidence for significant improvements in phonological awareness in mainstream children following 6 h of exposure to a commercially available phoneme discrimination training programme, but not in a control group. In a follow-up study, we failed to replicate this finding, despite using an almost identical training programme (Halliday, Taylor, Millward, & Moore, Journal of Speech, Language, and Hearing Research, 2012, 55, 168-181). This paper directly compares the methods and the results of the two studies, in an effort to explain the discrepant findings. It reports that the trained group in Moore et al. (2005) showed significantly greater improvements in phonological awareness following training than the trained group in Halliday et al. (2012). However, the control group in Halliday et al. (2012) showed significantly greater improvements in phonological awareness than the control group in Moore et al. (2005). The paper concludes that differences in the randomization, blinding, experimenter familiarity and treatment of trained and control groups contributed to the different outcomes of the two studies. The results indicate that a plethora of factors can contribute to training effects and highlight the importance of well-designed randomized controlled trials in assessing the efficacy of a given intervention. © 2014 The Authors. Dyslexia published by John Wiley & Sons, Ltd.