blood that this heart exists at all. The having of a function is constituted in having the right kind of etiology, the right kind of evolutionary history.
I will not focus on details of these models, but wish to simply point out that such a model entails that nothing can have a function if it does not have the right history, whether
or not it contributes to self-maintenance. Millikan accepts the consequence that if a lion were to magically pop into existence that was molecule for molecule identical to a lion in the zoo, the heart of the science fiction lion would not have a function because it would not have the proper history (Millikan, 1984). Accepting such examples that will never occur may be a price well worth it if the etiological approach satisfied all other desiderata, but, unfortunately, what the science fiction example demonstrates is that the etiological approach
cannot model function in terms of current state of a system. But only current state can be causally efficacious. The etiological approach, therefore, yields a model of function that is inherently epiphenomenal — nothing causal in the world depends on the presence or absence of having a function, that is, on the presence or absence of having the proper
history, per se. What does make causal difference is current state — Does the heart pump
blood, or not? — but current state is not sufficient for function in this view.
Clearly, the model of function as contribution to self-maintenance is definable in terms of current state — that is how it has been defined — and is therefore not epiphenomenal.
Representation. Function emerges in self-maintenant systems, and
representation in recursively self-maintenant systems. The basic idea is that the selection of some process of interaction between the system and its environment has the function of contributing to the self-maintenance of the system, but it could be in error. That selection involves a kind of anticipation about the environment — it anticipates that the current environment is of the kind for which the selected processes are appropriate for helping to maintain relevant far-from-equilibrium conditions. The paramecium “anticipates” that swimming is the right thing to do, and it could be wrong — it could be swimming up a
saccharin gradient, for example, not a sugar gradient.
This, I claim, is the most primitive level of emergence of representational truth value. Recursively self-maintenant selections involve anticipations that can be false; they
involve implicit predications — e.g., predicating of this environment that it is of a type
appropriate for swimming — that can be false. Representational truth value, in this manner, emerges quite naturally in the solution to the evolutionary problem of action selection.