Halliday & Matthiessen (1999: 90-1):
And this is in fact how we shall model them — as local meronymic taxonomies, applicable to some particular region of the hyponymic taxonomy, and not as one global meronymy superimposed on our taxonomy. There would seem to be far too many discontinuities to create a global meronymy; meronymies tend to occur only where there are contiguous parts of an independent whole. Thus while concrete objects are regularly construed meronymically (with the human body as both a representative example and a model for other meronymies), substances are not; substances are extended through measure ('unit of') rather than through part ('part of'). Similarly, taxonomies tend to be more extended meronymically for concrete regions than for abstract ones (although even things in an abstract region can have parts, e.g. aspect of an idea). Thus there is a generalised set of categories such as part, element, component, aspect; and also a generalised set 'facets' of spatial or temporal orientation, top, bottom, side, front, back, middle, centre; beginning, middle, end. There are more specific variants for parts of particular concrete things, such as facade, roof, wall [of a building]; ceiling, floor, wall [of a room]; slope, peak [of a mountain]; limb, trunk, root, bark [of a tree]; skin, core, pips [of an apple]; crust, crumb [of bread]; preface, epilogue [of a book].