The issue as to whether an atomistic or holistic view of knowledge and meaning is
correct relies on the way part/whole relationships is analysed, exactly as the issue as to whether a constructive or realistic view of knowledge and meaning is correct relies on the way internal/external relationships is analysed. Both the principle of compositionality and the context principle depend on how finely the constituents, the nature and the size of the context are identified; both the notion of meaning and the notion of truth depend on the resources of internalisation/externalisation. Thus the spectrum of semantic and epistemological theories varies from (global) atomism to (global) holism, and from minimal to maximal internalisation. Are compositional theories necessarily extensional? Does formal semantics necessarily rely on set theory?Does the domain-specific character of the notions of element, part and whole prevent any general, non-trivial account? The aim of the present paper is to provide a negative answer to these questions by exploring some of the features a theory covering the phenomenology of part and whole should have. This phenomenology will only be sketched through a few paradigmatic examples, showing how the reference of notions of part and whole varies and which are the constraints inherent in such variation. Category theory provides the tools for fashioning this framework, since it allows describing any coherent collection of objects (with actions defined over them) and (action preserving) maps between the objects, as well as the variation of such collections in terms of suitable functors, coding the ways parts and wholes undergo co-variation. The main thesis is that there are interference patterns between the two pairs Local/Global and Internal/ External, only in terms of which the above phenomenology can be properly described.