Explanations at Multiple Levels1

The preference for `reductive explanations', i.e., explanations of the behaviour of a system at one `basic' level of sub-systems, seems to be related, at least in the physical sciences, to the success of a formal technique –- perturbation theory –- for extracting insight into the workings of a system from a supposedly exact but intractable mathematical description of the system. This preference for a style of explanation, however, can be justified only in the case of `regular' perturbation problems in which the zeroth-order term in the perturbation expansion (characterizing the `basic' level) is the uniform limit of the exact solution as the perturbation parameter goes to zero. For the much more frequent case of `singular' perturbation problems, various techniques have been developed which all introduce a hierarchy of levels or scales into the solutions. These levels describe processes or sub-systems operating simultaneously at different time or spatial scales. No single level, no reductive explanation in the above sense will provide an adequate explanation of the system behaviour. Explanations involving multiple levels should be recognized as far more common even in supposedly reductionist disciplines like physics.