Finally, our findings may not generalize beyond the treatmentteachers and children. It is possible, for example, that better teachers demonstrated more fidelity to the intervention, thus asking more association-level questions, subsequently related to higher gain among their students. Similarly, it is possible that some intervention groups of students had strengths, such as self-regulation(not measured in the present study), that both elicited richer teacher instruction and led to greater vocabulary gains. Neitherstudents nor teachers were randomly assigned to differing levelsof distribution of time or types of questions, so caution must beused in generalizing beyond our study. Conclusions can only bedrawn about teacher talk in the context of a shared book reading intervention but not about teachers in a business-as-usual context