Although some school based smoking prevention programmes have shown positive impacts, others have had only short term effects, no effects, and in some cases reverse effects. The St-Louis du Parc Heart Health Project was a five year heart health promotion programme targeting children in eight elementary schools aged 9-12 years in disadvantaged multiethnic neighbourhoods in Montreal. In a controlled, longitudinal evaluation, the programme produced reverse effects on smoking--children exposed to the programme were more likely to initiate and to continue smoking than control children. This article explores hypotheses to explain the reverse effects. Following work by an in-house committee, a consensus workshop with international experts was conducted to develop hypotheses to explain the reverse effects. This was complemented by an analysis of the programme's concordance with the standard guidelines on the school based prevention of tobacco use, and discussions with experts at the Centers for Disease Control and Prevention. The programme respected most standard guidelines for smoking prevention programmes with respect to content and mode of delivery. Hypotheses to explain the reverse effects include: an unfavourable environment characterised by strong pro-smoking models and resistance to environmental interventions; heightened sensitivity to smoking among children most exposed to the programme; defence mechanisms among children stimulated by cognitive dissonance or anxiety; unanticipated effects associated with the health educator who delivered the programme; inadequate attention in programme development to the diverse cultural origins of the population targeted; and intervention content inappropriately targeted to children's stages of cognitive development. Elementary school based interventions should aim to develop a clear and coherent social norm about the non-use of tobacco, as a precursor to or in close conjunction with having children as their primary target. Programme design should take key student characteristics into consideration and ensure that the modes of communication are adapted to the targeted group's characteristics. Neighbourhood level interventions should be orchestrated to complement regional, provincial, and national programmes.