Data modelling plays a pivotal role in Computer Integrated Manufacturi
ng (CIM) as a vehicle for integration among the various component syst
ems and users. Standardization in the representation of product inform
ation is receiving growing attention in recent years. In data modellin
g, the concepts of generalization, aggregation and classification are
well known. They form the foundation of many, if not most, semantic an
d object-oriented models. Class hierarchy and the attendant inheritanc
e semantics provide powerful modelling mechanisms whereby a more effec
tive schema can be constructed in which common structural and behaviou
ral specifications are shared. Aggregation has mainly been treated as
an information structuring concept in most models, with focus on how c
omplex-valued properties can be accommodated. In the application of co
mputers in such areas as CAD/CAM and Geographic Information Systems wh
ere the structure of objects is of primary importance, the current agg
regation semantics are inadequate. In this paper, we examine the aggre
gation concept from a number of perspectives. We discuss a proposed mo
del of attribute semantics which can more effectively support the defi
nition of constraints and derived data in an aggregation hierarchy. In
the model, object attributes are categorized according to their meron
ymic type. The attribute categories serve as expansion scope for gener
ating attribute references on a composite object. In the latter part o
f the paper we illustrate how part-of semantics can be incorporated in
to the current international standard for manufacturing data exchange
of STEP/EXPRESS.