This study examined whether the effects on cognitive and language outcomes of a recently developed home-based educational intervention program, Opstap Opnieuw, for 4–6-years-old disadvantaged children could be explained by improved mother–child interaction. The present sample (n=30) was drawn from a larger sample of Turkish–Dutch families (n=181) for which in a previous study significant effects of Opstap Opnieuw were found on children's (first) language and cognitive pre-math skill, 5 months after the program ended. The present study focused on two facets of interaction quality as possible mediators of these program effects: the mean cognitive distancing level of mothers' communication and instruction behaviour as an indicator of the cognitive and verbal stimulation provided, and the degree of cooperation as an indicator of mothers' social-emotional support to their children. Both measures were based on systematic observation of mother–child interaction during sorting tasks. Participation in the program appeared to improve mothers' social-emotional support behaviour substantially, but not their cognitive distancing behaviour. For Turkish (first language) vocabulary, about half of the program effect appeared to be mediated by the improved social-emotional support. For cognitive pre-mathematical skills, two-thirds of the program effect appeared to be mediated by improved social-emotional support. Mothers' cognitive distancing was moderately-strongly related to children's vocabulary development, but did not mediate program effects. Some implications of the results are discussed. Copyright © 2004 John Wiley & Sons, Ltd.