Functions 541towards the bottom of a cup may not reach that state simply because, let’s say,
Armageddon occurs before such equilibrium is attained. Now we could say that
in some weak sense systems like these have equilibria as their goals. One might
want to rule these out as full-blooded goals, especially if one wishes to use goal-
directedness as definitional of life. The distinction that helps to mark out weak
and strong goal-directedness is between systems that merely tend to some end
state, and those that tend to some end state in virtue of internal compensating
changes that offset the perturbing effects of environmental alterations [Sommerhof,
1950]. The ball bearing is not directed towards reaching the bottom of the cup
in this sense, because the bearing is not disposed to make adjustments to its own
internal structure that increase the chances of this end state being attained if the
cup itself changes structure. A maturing organism, on the other hand, will make
internal compensating changes in the face of perturbations to the developmental
environment, which increase the chances of reproductive capacity being attained.
Thus far, we have suggested that a system directed at goal G is one such that,
were environmental circumstances to change such that the attainment of some
state G becomes less likely, the system would make compensating alterations in
its internal structure with the result that the attainment of G becomes more likely.
One should not quibble over the fact that some environmental perturbations will
yield no such compensating adjustments. Obviously if one smashes a fertilised
hen egg into a frying pan, nothing will happen that raises the chances of hen
development occurring. The important point is merely that some more mild per-
turbations will yield compensating adjustments in the developing embryo. Perhaps
more worrying is that we have left open in this analysis any fine specification of
which end state some system is directed towards. A torpedo’s trajectory is af-
fected by a strong current, and it makes compensating changes to the position
of its fins. This series of events may initially reduce, and then raise once again,
not merely its chances of hitting a particular boat, but of its passing through an
area of water drawn around the boat, of its producing a loud noise on impact
with the boat, and of its creating a wave after impact with the boat. So all of
these will count as goals on the analysis I just sketched. It is not clear to me that
this is a great problem for the analysis, at least not if we are trying to demarcate
systems that are goal-directed from those that are not. It does suggest, however,
that if we are tempted to attribute a very specific goal to a torpedo (its goal is to
destroy the boat, not to make a loud noise), we do so because we make a guess
about the intention behind the torpedo’s launch. If that is right, then there is
no fact of the matter about whether plants are directed towards the attainment
of adult reproductive states, or towards the release into the soil of a decent-sized
quantity of organic fertiliser that only the death of an adult organism can yield.
The compensating adjustments that occur during early development tend to make
this second state likely, as well as the first.
Goal-directed accounts of function were especially popular in the middle of the
twentieth century (e.g. [Rosenbluethet al., 1943; von Bertalanffy, 1950; Braith-
waite, 1953]), particularly in the wake of developments in the science of cybernet-