Recent work on raising attainment in education has raised theoretical issues about organizational responses to innovation and qualitative aspects of programme implementation. In particular, the "depth" of an implementation (the extent to which a programme actually changes the educational interactions between pupils and their teachers) is now thought to be as important as its "spread" (the number of schools in which it is replicated). Such issues, together with the existing literature on Reading Recovery implementation, suggest that it is time to review the models of implementation held within the Reading Recovery community. We used a range of methods at four levels of data collection to analyse variation in "depth" of implementation in a countrywide Reading Recovery programme. From our findings we conclude that Reading Recovery has achieved resounding success as an individual programme, but that it would be still more efficient if the models of implementation were to be reviewed. (Contains 4 tables.)