Fourteen mother–child dyads were videotaped four times over 16 days as they repeatedly read an initially unfamiliar storybook. Videotapes were transcribed and a variety of discourse measures were coded for both mother and child. Principal components analysis was used to combine three maternal and three child variables into a single composite variable, engagement, which described levels and quality of maternal and child participation in extratextual interactions during the first reading session. Dyads with the highest levels of engagement during the first session showed the greatest rates of change for a number of discourse measures. Results are discussed in terms of maternal scaffolding approaches, child participation, the importance of context and the nature of the task, and implications for intervention.