266 Ulrich Krohs
structures did not bring about modules that are separated more nicely than actually found.
So the arguments about the evolution of modularity are incomplete.
There are several possibilities for supplementing the arguments. Some of these could be
assumed to be tacitly accepted—that the time available for evolution was not long enough
to bring about full modularization; that, by chance, the additional mutations required for
further modularization did not occur; that the actual degree of modularity makes the organ-
isms suffi ciently adapted; or that, in the case of extreme separation, morphologically dis-
tinct modules would fall apart without being viable anymore. None of these explanations
of the limitations of modularity is satisfying as long as no support by adequate data is pro-
vided, because all can be brought up as ad hoc arguments whenever needed. They have no
specifi c explanatory power. (An exception may be the argument about the separation of
morphological modules, but this is hardly applicable to modular networks.)
Though this critique is focused on parcellation arguments, it covers assembly arguments
as well. It even gets additional strength from data collected on assembled systems. After
an assembly has taken place, be it on the endosymbiotic pathway or by duplication, an
evolutionary integration process takes place that lowers the degree of decomposability of
the assembled system. So we have a process—usually described as an adaptive process
under natural selection—by which additional interaction and interconnection of the
modules is established. This integration seems to proceed much further than is physiologi-
cally required. Mitochondria may serve as an example. Genes are transferred from the
bacteria that became mitochondria into the nucleus, which increases the integration of the
modules. But the modularization argument does not explain why this integration (i.e., this
lowering of the degree of modularity) goes much further in the case of mitochondria than
in cases of endosymbiotic bacteria as found in certain fl agellates. Even the mere existence
of the adaptive part of the integrative path to modularity—the adaptive part being the
modifi cation that follows the integrative step—shows that both modularity and integration
bring selective advantages. Schank and Wimsatt (2001) have pointed to such advantages
of integration. However, the adaptive explanations of the evolution of modularity men-
tioned do not recognize this issue.^10
Let me return to the argument that explains modularity by adaptive parcellation of an
integrated system. Since it refers to selective advantages, it is quite clear that the incom-
pleteness of modularization has to be explained by reference to selective disadvantages of
a modular organization or to the advantages of integration as mentioned with respect to
the integrative pathway. The consideration of the benefi ts of modularity ought to be supple-
mented by the cost side. Modularity has fi tness-decreasing effects as well.
While the development of the demanded argument has to be left to biologists, philoso-
phy can give, in addition to criticisms of the argument, some hints in which direction to
proceed. In the following I therefore single out some kinds of possible costs of modularity
by comparing the biological with the technological case, and ask for their consideration
in biological explanations of the evolution of modularity.