Abstract
Kinds - also known as 'natural sets' or 'universals' - are a very intuitive assumption about the way the world is put together. As a piece of metaphysical theory, however, they give rise to the Identification Problem: which of all sets are the ones that in fact qualify as kinds? In this thesis an answer is given starting out from the assumption that kindhood always coincides with similarity. From this it follows that similarity must be similarity 'with respect to', and p properties - kinds - must be arranged in 'similarity systems'. To turn this insight into a credible answer to the Identification Problem, however, a wide variety of (physical) objects must be considered, whose common ground is that they are all 'mereologically complex'. Therefore in the second part of the thesis the focus will be on the derivation of 'composite kinds', thus allowing the classification of larger objects in terms of kinds. It will be concluded that classical (Boolean) mereology is sufficient for this purpose. I shall argue that this approach is therefore preferable to that whereby kinds are reified to be (structural) universals