The Cost of Modularity 265
modularization than that found in actual biological organisms, reveal that modularity may
have disadvantages under a broad range of conditions. Consequently the adaptive explana-
tions of modularity as given by now are far from being satisfying.
In the case of modularization by duplication (Raff 1996; Calabretta et al. 2000), it is
clear that modularity or evolvability are not the traits being selected for. Instead selection
acts on the particular modules, which arose from the duplication process, or on individuals
possessing these modules. So again not modularity but the secondary adaptation of specifi c
modules requires evolutionary explanation (or better, as becomes clear in the next section,
the specifi c degree of modularity). But in this case, too, the adaptive argument does not
show why the process results to some degree in integration rather than in the highest pos-
sible independence of the modules.
15.4 The Desideratum of Explaining Nonmodularity
Integration, parcellation, or specialization of duplications, whichever way may have led
in any particular case to a modular organization: according to arguments of the kind dis-
cussed, the modular structure results in a selective advantage over a fully integrated
system. So one might tend to conclude that after ample time for evolution all structures
of biological organisms should be modular. But this is not the case, and it was not expected.
Even after biologists had become knowledgeable about modularity of metabolic networks,
it was an important and unexpected fi nding that genetic networks are modularized as well
(see Callebaut 2005). Moreover, modules are often much less clearly separated than the
paradigmatic examples in the debate might suggest, and there seems to be no borderline
between modular and integrated (sub)systems, but rather a continuity spanning the whole
range that lies between the extremes. Thus the citric acid cycle is a functional unit that is
not at all a structural module, as shown in section 15.6. What about the urea cycle? There
the internal interactions may be larger than the external ones, but is the difference large
enough for a structural delineation of a module? In β-oxidation of fatty acids this differ-
ence is larger, so in this case one might tend to talk about a structural module proper. So
the urea cycle seems to be an intermediate case. The conclusion needs to be as follows:
modularization comes in degrees—as does decomposability, its conceptually related ana-
lytic counterpart—and many networks show only intermediate, or even low degrees of
modularization. Metabolic networks, for instance, although unanimously classifi ed as
modular, are found to be much less nicely decomposable according to structural criteria,
and integrated much higher than straightforward mathematical methods can deal with (cf.
Davidson et al. 2002; Palsson 2006).
The situation thus is the following: while modularity in biological systems comes in
degrees, the present explanations of the evolution of modularity account for strong modu-
larity only. They do not state any reason that could explain why the evolution of modular