Correctly construed, the concept of intervention requires that active consideration be given to the mechanisms by which any behaviour change produced by an intervention procedure may subsequently be generalized—across settings, across behaviours and, particularly, across time. Broad strategies that have been adopted with the aim of fostering generalization are reviewed. These may be seen to define a continuum along which scope for engineering behaviour change is “traded” with the probability of the contingencies required to maintain this change being assimilated within the “post-intervention” environment. Viewing strategies in this manner prompts consideration of interventions representing the extreme of this continuum, where scope for engineering change is minimal, but the probability of maintaining any change produced is very high. The implications of adopting such “minimal interventions” are examined, and these strategies shown to have some empirical support. A broad structure for implementing “minimal interventions” is proposed.