Over the past two decades, a growing number of randomised controlled trials have assessed the impact on children's language skills of interventions encouraging parents to read books to their children. We present the results of a meta-analysis of the impact of 30 such interventions. Results indicate that they are often ineffective, and that only one specific methodology (dialogic reading), displays systematically positive impacts. Moreover, effective interventions display weaker impacts on low-socioeconomic groups, thus raising equity issues, and on younger children. Our systematic analysis of the research designs of these studies points at three major weaknesses. First, only short-term outcomes are measured, most often within six months after the conclusion of the intervention, and even within such a narrow time window, we find indications that treatment impacts fade out. A second limitation concerns the limited range of outcomes measured (receptive or expressive vocabulary). Finally these studies display low external validity (ad hoc sampling, small sample sizes, lack of multi-site experiments, scant evidence outside Anglo-saxon countries).