Basic HTML Version

of that level than from appealing to other levels, though there are well-defined circumstances when
it is appropriate to do so (Wimsatt 2007: Chapter 11). Levels of organization are the joints of
nature, as it were, and in giving explanations we should cut nature at its joints. Theories will be
simplest and most powerful if we respect these joints in theorizing and explaining. They are where
the regularities are to be found, and objects at those levels will tend to engage in processes
interacting at roughly the same timescale.
If levels of organization did not exist, our evolution could not have happened. Evolution
requires that there be stabilities in nature that one can utilize in constructing adaptations that
respond to these regularities and in building species that can evolve and utilize these regularities,
what one might call unique stable building blocks. Furthermore, it applies all the way up: stable
systems made out of building blocks, stable supersystems made out of subsystems and so on. There
are multiple connections between robustness in not only what exists but also in how we ought to
structure our explanations. I reiterate: We should try to explain things that are not robust in terms
of things that are robust.
The relationship between robustness and function
Consider any adaptive system or machine. The machine has differentiated parts and its
performance will depend in different ways on what those parts do, which means for most
interesting machines that there are parts that do not cause much damage if they break. You can
just about use it without them; maybe you are not able to do everything you could before, but you
can get by. In contrast, breaking some other parts will mean that you are dead in the water, you
just cannot do anything. So some parts have more downstream consequences in the operation of
that machine than anything else. In fact, that turns out to be an extremely general principle of
organization. It applies for artifacts. It applies for organisms. It even applies for ecosystems: so-
called keystone species are those that if you lose them many other things collapse. Indeed, it can
arguably be said that the principle applies to any functionally organized system, any roughly
adaptive system, any system that is a product of a selection process that differentiates it in order
to accomplish a complicated function. It will have different kinds of parts and they will have
differential impact. Suppose you ask whether there are any ways of making changes deep in the
functional architecture. That is just what sometimes happens: a scientific revolution changes the
basic principles and that forces you to make changes all the way up. However, because of the
multiple connectedness of the fundamental principles of nature, all hell breaks loose if you change
them, so when we do make changes, we try not to change things too much.
For example, the relation of Newtonian mechanics to special relativity involves Lorentz
transformations. You may say that classical kinetic energy is not quite what we thought it was, but
it is very interesting to see how it changes. Relativistic kinetic energy is obtained from classical
kinetic energy in this way. First of all, in relativistic kinetic energy there is a rest mass term, but it