The Cost of Modularity 269
vant for the evolution of modularity, or they themselves may be in need of an evolutionary
explanation: 1) functional modules do not in general coincide with structural ones in bio-
logical systems, and 2) functionality in metabolic and gene regulatory networks is not
localized at particular components of the system but delocalized or distributed over entire
subnetworks (Boogerd et al. 2005). The relevance of functional modularity for evolution
needs to be investigated.
The concept of a biological function is notoriously problematic from the philosopher’s
point of view, so I need to explain which concept is to be applied. First, I should point
out that “function” must not be equated with “dynamics” (see Krohs 2004: 41). As any
complex physical entity, biological entities have a structure and display change in time,
which is described as its dynamics. (Even being static is a kind of dynamic in this sense:
a change of measure zero.)^15 “Function,” then, is ascribed to an entity that exhibits its
dynamics if these dynamics contribute to some capacity of the biological (or technical)
system of which the entity is a component.^16 While “dynamics” denotes the processes an
entity undergoes, “function” refers to the relation of this process to a more comprehensive
process within a system of a certain kind. A function can thus be considered the contribu-
tion of an entity to a capacity of the system the entity is embedded in—with the caveat
that the system and/or the capacity at all qualifi es as being functionally organized. The
dispute among philosophers about the concept of function is over how to specify these
further conditions. In the following paragraph I indicate an explication of the concept of
function that is adequate for use in systems biology.
In general, functions are ascribed to metabolic and other networks and to their substruc-
tures by physiological analysis, not by an analysis of adaptive processes in the evolution
of the network. Consequently systems biologists do not refer to etiological functions,
which are specifi ed with respect to evolutionary processes, but to systemic functions, to
be conceived as roles in a system (Boogerd et al. 2005). For the latter, merely the present
contribution to a capacity or disposition to such a contribution is relevant. However, it is
not satisfying to rely on Cummins’s (1975) approach as an explication of the concept of
function in question, since this approach does not allow to make any difference between
function and dysfunction. But it is possible to modify the Cummins approach suitably by
introducing reference to a norm for functionality and dysfunctionality. Such a norm already
is present in the way systems (and other) biologists refer to their subject of inquiry: they
do not describe the metabolic network of one individual of the species E. coli or Dro-
sophila melanogaster but the type of network present in individuals belonging to these
species. The type of the network is described as fi xed by genetic and epigenetic factors,
growth conditions (which exactly for the reason of sticking with one type need to be
standardized when networks are to be investigated), and so forth. Not only the networks
as wholes but their components are described as being of fi xed types. The fi xed types also
fi x the roles of the components within a network. Such fi xed roles are conceived as func-
tions. Any deviation from the function that corresponds to the fi xed type may therefore be