300 Speciesany other electron. Physics space is, so to speak, fully populated (or the states are
highly restricted in a physical NK to only a few possible occupiable coordinates).
Chemistry, being fully determined by physics, likewise has all kinds fully occupied,
as the periodic table indicates (although we may not yet have found some of the
higher atomic weight elements, and they may not even exist in the universe—but if
they do or will, they will occupy the location in the periodic table fully). But there
are several sciences that do not have indiscernible units, and biology is one of them,
and these sciences also do not have fully occupied state spaces. Historical processes
will not typically explore all possible states allowed by the properties of the sub-
strates; hence it is feasible that a planet might never have metamorphic rocks, or a
star might never go nova, although these states are permitted by the underlying phys-
ics, and are required if the initial and boundary conditions are satisfied. Geology,
astronomy, and biology, universalized to cover all such instances in the knowable
universe, are the historical sciences. In these cases, NGs are to be expected. In some
cases such as stellar evolution or ecology, there may be a generalized classification
based on NK physics, but to the extent that the objects are the special objects of these
disciplines—organisms or astronomical bodies—they form NGs.
In such disciplines classes must still be differentiated, and it is a mistake to think
that these classes must be differentiable in terms of definitions or essences. But the
classes will turn out to be those formed by common causal chains, and they will be
historical individuals as well. So NGs are not restricted to biology. Historical sci-
ences study and classify differentiable particulars.
I propose that we should conceive of biological groups as NGs rather than as NKs
unless they happen to coincide with each other. NKs are model-dependent, and they
are timeless and abstract classes, while real biological groups are time-indexed and
spatially located. This applies equally to other disciplines. It follows that while there
may be NKs in a historical discipline, and hence we may say they have laws, these
laws occur as generalizations within restricted or occasional models. The physical
systems covered by these models or theoretical structures will not always exemplify
them completely, and rarely in the same way as another instantiation. Thus we may
find that the metamorphic rocks of one region contain more or less of a trace element
than those of another, or that selection acts on the gross morphology of one species,
but on the metabolic processes of another, though the theoretical structure or expla-
nation is identical—the same selection “pressures,” the same “fitnesses,” and so forth
(Sober’s point about fitness being supervenient^48 ).
NGs can be generalized beyond biology, to cover any collective particular which
is caused by a unique set of causal processes at a time and place, and which has
similar, but not necessarily identical in any respect, members. For example, any
aggregate, such as a sand dune, in which the properties depend on the nature of the
members (the shape and size of the grains, for example) is a natural group. Waves,
hills, continents, and clouds are all NGs, although not every cloudlike formal entity
is—phenetic groups are only NGs if they are formed by unique causal processes.
It is no accident that Wittgenstein’s family resemblance predicates, so widely dis-
cussed, describe natural groups to the extent that the things predicated have the same
(^48) Sober 1984.