The Cost of Modularity 263
most widely applied class of explanations and show that current arguments belonging to
this class are incomplete and therefore not yet satisfying. In the following sections of the
chapter I propose a way of completing the arguments. Arguments of the kind in question
are adaptive explanations that refer to the evolutionary options that a modular structure
opens up. In short, since modularity allows for evolutionary plasticity, it is regarded as
favored by natural selection (e.g., Altenberg 1995; Galis 1999; Wagner and Altenberg
1996). Some authors even claim that the modular organization of metazoa is the result of
selection for evolvability (e.g., Gerhart and Kirschner 1997).
Before discussing the argument behind this claim, and reframing it in a way that makes
it more plausible than it appears now, I proceed to investigate its equivalent within the
realm of technological evolution, where a sound argument in favor of selection for evolv-
ability can be made when a latent premise is made explicit. Technical artifacts with a
modular design can be modifi ed by substituting or reassembling modules without much
effort, as it is familiar from construction kits. Modularity thus allows to cover a huge
design space easily. Not only substitution and reassembly but other modifi cations are pos-
sible as well. Baldwin and Clark (2003) give a long, but still incomplete, list of operations
that can be performed on modules: splitting, substituting, augmenting, excluding, invert-
ing, porting as well as replicating, combining and extending. Given this fl exibility in
modifying a modular design, it is obvious that new kinds of systems can evolve more
easily from modular systems than from fully integrated ones. The latter requires complete
redesigning to end up with another functional system, while the former can be modifi ed
stepwise, module-by-module, with a high probability that the intermediate forms are still
working (stable).^7 Industry takes advantage of this: “Through widespread adoption of
modular designs, the computer industry has dramatically increased its rate of innovation”
(Baldwin and Clark 1997; see also Langlois and Robertson 1992). Design methodology
relies heavily on a modular approach, exactly for the reason that this allows for quick
evolution of products (e.g., Ulrich and Eppinger 2003; Pahl et al. 2007). Therefore, in the
fi eld of technology, modularity is present and favored because of the high evolvability it
enables. It may well be that in some cases there is selection not only of evolvability but
also selection for evolvability.
Accepting that one reason why modularity drives technical innovation is that modular
design is often chosen because it allows further evolvability, it is tempting to draw the paral-
lel between technological and biological evolution, resulting in the aforementioned claim
that modularity has evolved because of the evolvability of modular systems. Modularity in
fact opens up possibility spaces also for the evolution of biological organisms, and high
evolvability is found with respect to many modular traits of organisms. So modularity
allows for quick biological evolution and high evolvability (Schlosser and Wagner 2004;
Callebaut and Rasskin-Gutman 2005). However, it is diffi cult to see how high evolvability
could be a character that natural selection can act upon. Mutations, that is, undirected heri-
table modifi cations occurring in the offspring, do not usually increase the fi tness of an indi-