Phoneme awareness (PA) is undoubtably the most important and well-studied predictor of reading development. Yet, 20 years ago, Castles and Coltheart made the provocative claim that there was no convincing evidence for the causal role of phoneme awareness (PA) in learning-toread because previous studies had failed to control for pre-reading skills, such as letter knowledge. In the present study, we first wanted to know whether studies published after Castles and Coltheart's seminal article had adequately controlled for pre-reading skills. The results of our systematic review clearly show that this is not the case. Thus, the evidence in favor of a direct causal role of PA in learning-to-read is not stronger today than it was 20 years ago. Given the absence of any conclusive data, we thus analyzed the results of a large-scale longitudinal investigation on predictors of reading development that has been conducted nation-wide on all first graders in France (i.e., N = 810,328 children). We estimated not only the effect of PA on reading fluency measured one year later but also the interaction effects with letter-knowledge (LK), knowledge of the alphabetic principle (KAP) and oral comprehension.The results clearly showed that when interaction effects are integrated into the statistical models, PA is no longer a good predictor of reading development, while LK and KAP remain strong individual predictors. These results have important consequences for our understanding of early predictors of reading development, educational policy recommendations, and early intervention strategies.