Abstract

In the classification literature, there has been much discussion of the no-overlap principle, which allows a categorical distinction between natural kinds from a realist perspective. However, cases of crosscutting natural kinds in scientific practice provide a serious challenge to the no-overlap principle (Khalidi, 1998, Dupre 1993, Tobin 2010). The HPC view of natural kinds has emerged in order to accommodate the fuzzy boundaries we associate with clusters of properties, by the introduction of a homeostatic regulating mechanism. Craver (2009) claims that there is a difficulty in deciding where a particular mechanism begins and another ends. The strategy of lumping and splitting is designed in order to accommodate the no overlap principle; namely if you find that a single cluster of properties is regulated by more than one mechanism, then because there can be no overlap between mechanisms, then we must split the natural clusters. However, a closer analysis of dynamic mechanisms in scientific practice reveal that the overlap problem re-emerges with the strategy of lumping and splitting in that there is an assumption that once a mechanism is found to be responsible for a property cluster, that this is sufficient for delineating the boundaries of that cluster. Scientific practice reveals that there might be multiple causal routes that could result in a similar functional output. Moreover, there might be different kinds of mechanisms, which produce the same property cluster and depending on which one we are using, the decision to lump or split may be different. The paper concludes with some observations about the implications for the classification of mechanisms and to the additional question as to whether mechanisms are themselves natural kinds.