What it has been selected for

As usually happens when philosophers off er defi nitions, Wright’s defi nition of function has been attacked on the grounds that things that fi t the defi ni- tion are clearly not functions. For example, Christopher Boorse (1976) off ers the following scenario. A gas tap is left on in a laboratory. Th e gas knocks out the only scientist in the laboratory, so he cannot turn it off or open a window, so aft er a while there is more gas in the laboratory. Knocking out the scientist is why the gas is there – that is, in the air in the laboratory – which seems an acceptable use of “is there” on Wright’s construal. Knocking out the scientist is

also a consequence of the gas being there. So we would have to say, on Wright’s defi nition, that knocking out the scientist is the function of the gas.

Millikan (1984; 1989) proposes getting rid of such problems by making the function of X what X was selected for. What she is interested in, she says, is

biological function. Accordingly, she is not concerned about satisfying Wright’s

criterion (i), that the concept of function that we apply to parts of living things is the same concept that we apply to man-made artefacts. Her defi nition of “proper function” is, A has function F as a proper function if either:

(1) A originated as a “reproduction” (to give one example, a copy, or a copy of a copy) of some prior item or items that due in part to pos- session of the properties reproduced, have actually performed F in the past, and A exists because (causally historically because) of this or these performances. [Or] (2) A originated as the product of some prior device that, given its circumstances, had, performance of F as a proper function and that, under those circumstances, normally causes F to be performed by means of producing an item like A. (Millikan 1989: 288) Th is formulation does well in terms of the rest of Wright’s requirements. It recognizes that a thing’s function has a role in explaining why it is there. It allows that something can have a function even if it never performs that function. It does not make usefulness a suffi cient condition for something being a function; that is, it distinguishes between genuine functions and mere accidental eff ects. It does not say that the thing has to be necessary for the system as a whole of which it is a part, and it allows that something may be unable to perform its function. It does all this by making the fact that something has done F the reason for saying its function is F. So eff ects of biological functions that were not selected for do not count. Moreover, the gas tap example does not count either, as the gas in the room is not there in virtue of being a reproduction of something that had that eff ect.

Millikan’s defi nition seems to do a good job of accounting for biological func- tion. But we still might be worried about the fact that it no longer works for the functions of man-made artefacts or their parts. Wright wanted a defi nition of function that would apply equally well to parts of living things and to man-made artefacts. And this does seem to be an attractive feature of his account. Millikan’s formulation does not actually mention natural selection, let alone make it a requirement that the entity in question has to be biological. So, in principle, it can apply to man-made artefacts. If we think of mass-produced items (such as cars), or items that human beings have been making for some time (such as ships), we might plausibly say that the reason they have the various functional parts they have is that those parts have worked and hence have been reproduced. On this construal of things, the functional parts of such items can be said to

fi t Millikan’s defi nition. Th e steering wheel of your car really did originate as a copy of some prior item (the steering wheels of earlier cars), and has its function in virtue of those predecessors performing that function. Millikan is perfectly happy to say that any man-made items that fi t her defi nition count as having functions. But what about a man-made item that is unique, or the fi rst of its kind? Th e Wright brothers’ fi rst aeroplane clearly had parts that were functional, but they cannot be considered reproductions of parts of anything else that per- formed the same function. Even if there were earlier prototypes, they did not work. We would probably say that it was something to do with the intentions of the Wright brothers that gave those parts the functions they had. But Millikan’s account does not allow intention, even intention intelligently carried out and successfully achieved, to count as a reason for calling something a function.

Her justifi cation for taking this stance is that it is biological function that she is concerned with. Th is means, incidentally, that even if the gas tap example had not been dealt with, it would not matter. So whether or not it turns out that some non-biological items wrongly fi t or fail to fi t the defi nition, is of no consequence. More generally, she is interested in producing a defi nition of func- tion that refl ects scientifi c practice, not one that conforms with whatever our everyday intuitions or ordinary language tells us. Th is tough-minded approach allows her to brush aside another class of counter-examples, those that are based on science-fi ction type counter-examples. For example, suppose an atom-by- atom replica of a lion appeared suddenly by a chance confi guration of atoms. Th is would not have the right kind of history for any of its parts to fi t Millikan’s defi nition, but we would still want to say that its “heart”, “lungs” and so on have functions. For that matter, they would not fi t Wright’s defi nition either. However, Millikan argues that there is no need for any defi nition of function to be required to fi t such cases, as this is not a situation that would ever arise in the real world. She would say: none of the instant lion’s parts have any func- tions, but so what? Science is none the worse for working with a defi nition that confl icts with what ordinary language would say in cases that never arise in the real world. So for Millikan the only requirements that we should require a defi nition of “function” to satisfy are that it apply to things that could actually exist in the real world, and that it conforms to how scientists actually use the term in the contexts in which they actually use it.

Th e thought seems plausible if we think of cases like this: suppose scientists have a defi nition of “liquid” according to which glass is a liquid. (In actual fact, glass is not clearly either a solid or a liquid according to physicists.) A phi- losopher might then object: in ordinary language we do not call glass a liquid, so the defi nition must be wrong. But a scientist could respond that we need a defi nition of “liquid” for scientifi c purposes, and the defi nition that makes glass into a liquid is good enough for that. Moreover, if we wait around for a defi ni- tion that fi ts our ordinary-language usage perfectly, we may be waiting a long

time. Philosophers are very clever at thinking up counter-examples: examples of things that fi t a proff ered defi nition of an X but are clearly not X’s accord- ing to ordinary language. To get around this, Millikan proposes that we say, in eff ect, to hell with ordinary language. For scientifi c purposes, we do not need to deal with instantaneous lions, so whether they have functional parts or not according to this defi nition is irrelevant. Th e selected-eff ect defi nition of func- tion, then, is perfectly adequate for scientifi c purposes, and to demand more of it is unreasonable:

Now I fi rmly believe that “conceptual analysis”, taken as a search for necessary and suffi cient conditions for the application of terms, or as a search for criteria for application by reference to which a term has the meaning it has, is a confused program, a philosophical chimera, a squaring of the circle, the misconceived child of a mistaken view of the nature of language and thought. (Millikan 1989: 290) In my opinion, however, this is all too quick. Millikan may well be right that ordinary language should not be treated as the arbiter of whether a defi nition that is to be used in science is good enough for science. And she may be right that a defi nition that deals with all the counter-examples that philosophers can dream up will probably be a long time coming. But scientifi c purposes are not the only purposes we have in trying to defi ne something. Moreover, saying that man-made artefacts are for the jobs they are designed to do hardly seems like a crazy philosopher’s counter-example. It is not clear that, in deciding if a defi ni- tion is adequate, we are faced with the choice of either deciding on the basis of whether it fi ts what ordinary language would say about the counter-examples, or deciding on the basis of whether it is good enough for science. Millikan appears to think that if we are not willing to accept a defi nition that is “good enough for science” then we must be unreconstructed ordinary-language philosophers. But that presupposes that these are the only two possible grounds we might have for accepting or rejecting a defi nition.

Imagine that someone came up with a defi nition of “alive”, and it turned out that light bulbs fi t the defi nition. A philosopher produces the counter-example of light bulbs to show that there is something wrong with the defi nition. But in doing so, that philosopher is not necessarily arguing that there is something wrong with the defi nition because it confl icts with ordinary language. It is pos- sible that the philosopher is arguing that there is something wrong with the defi nition because light bulbs are not, in fact, alive.

Th is might seem a silly example, indeed, another case of the philosopher’s disease of thinking up highly unlikely scenarios to create problems that never arise in practice. Is it remotely plausible, it might be asked, that a defi nition of life might be produced that would fi t light bulbs? Is this not just as unlikely as

instantaneously appearing lions? Moreover, we know that science has some- times given us good reason to reclassify things. Who would now think it rea- sonable to doubt the classifi cation of earth as a planet, for example? Yet that classifi cation came about as a result of a revolution in science. So, this argument might go, if the purposes of science require us to move things into, or out of, categories contrary to everyday usage, so much the worse for everyday usage. It would have been no good for a seventeenth-century astronomer to argue that we cannot reclassify earth as a planet because it is not a planet. We would need, at least, to show that there was some reason for saying it is not a planet, other than ordinary usage, and that that reason is suffi ciently important to outweigh the scientists’ reasons for calling it a planet. By analogy, we would need some reason for accepting that some things that fulfi l the defi nition of function are not actually functions, or vice versa, other than in ordinary usage. Th at reason would have to be suffi ciently important to outweigh the consideration of sci- entifi c practice. Can we do this?

I think we can. Th e notion of function, even within science, is not confi ned to biological function. Engineering may be considered a part of science, but engineers will talk of the parts of things having functions. Moreover, the way engineers talk about function in machines is clearly the same way that biolo- gists talk about function in body parts. Th e outcome of a part being there, is in both cases to be accounted for by the same physical laws. Biologists may even borrow information from engineering to understand how an organ works. Th e same principles of aerodynamics apply to a bird fl ying as to an aeroplane fl ying, even though they fl y using diff erent mechanisms. And an engineer may study a previously unseen machine to discover its function – for example, in industrial espionage – in just the same way that an anatomist may study a pre- viously unseen organ to discover its function. Dennett (1995) calls both proc- esses “reverse engineering”. Th e machine–organism analogy, and specifi cally its “what-is-for-what” aspect has a role even in science. It may not be a matter for concern that the parts of instantaneous lions turn out not to have functions on this defi nition. But it is a matter for concern that the parts of the Wright brothers’ fi rst aeroplane turn out not to have functions either.

In any event, the appeal to scientifi c practice as the arbiter has led some philosophers to challenge accounts of function that focus on the purposeful or “for something” aspect.