The National Early Literacy Panel (2008) report identified early predictors of reading achievement as good targets for instruction, and many of those skills are related to decoding. In this article, the authors suggest that the developmental trajectories of rapidly developing skills pose problems for traditional statistical analyses. Rapidly developing skills yield correlations with later reading success that change with learning, so the predictive strengths are temporary and unstable. The correlations are strong only briefly, when children demonstrate partial learning of a skill that they will master completely later. Thus correlations with rapidly developing skills exaggerate the strength of the relation to later achievement, ignore the transient developmental window, and inflate effect sizes of interventions.