Skip to content

Tautology 5b: The issues, continued

In this post I will discuss these issues:

3. What is a function? Is it in the mind/theory, or in the world?

4. Is natural selection a mechanism? If so, what kind?

5. Is the principle of natural selection a law?

Again, this will be pretty short.

3. What is a function?

One way we might evade the tautology is if we can establish that there are some functional properties that are not defined by the PNS, and are independent of it. There are fundamentally three approaches to what a function is: the Cummins (1975) approach – a function is some trait, organ or process maintains the organism homeostatically; the Wright (1973) approach – a function is whatever task a trait, organ or process was selected to do; and the Bigelow-Pargetter (1987) version: a function is whatever the part has a propensity to do.

We cannot appeal to the Wright, or etiological, function, because that would be viciously circular: NS would be defined by whatever it is that NS selects. The Cummins, or “causal role” function isn’t much better, because it relies on the organism surviving, which would define NS as differential rates of survival of things that make organisms survive. And the propensity version fails to give us independent access to which properties are functional and which properties are merely side effects: my arm has a propensity to drop when I relax – is that its function, or one of them?

There is, again, an extensive literature of this (ably covered by Allen, Bekoff and Lauder 1998), but it seems at a quick look that none of these are able to help us. One other view, mentioned by Bigelow and Pargetter, is eliminativism. This is the view that in the end, functions are simply mind-relative or theory-dependent,, and have no independent existence in the world. They are, in short, the aspects of biological systems that we deem to be interesting enough to model. This is my opinion: functions are literally mathematical functions; transformations within the model of the form y = fx, where f is a mathematical transformation. To say, then, that hearts have the function of pumping the blood in a circulatory system is to give hearts a particular transformative role with respect to blood; to say that hearts were “selected to pump blood” is to give the task “pumping blood” a particular role in a selection model. To say that hearts have a propensity to pump blood is to say that “pumping blood” is one of the infinite number of intrinsic and relational properties that hearts have that interests us in that model of the organism.

Functions therefore do not help us. But I will argue next post that they give us a clue how to resolve this problem.

Is natural selection a mechanism?

Another possible solution is to treat NS as a mechanism. Mechanisms in science have a causal role in explanation. If I explain the behaviour of a system in virtue of, say, being a lever, the simplest form of a mechanism, then I have no further physical explanation necessary for that behaviour (lifting the weight in a g-field). I may then have to explain the strength of the level and fulcrum, and so on, but so far as the dynamics goes, I have given a complete explanation.

So if we could make out that natural selection is a mechanism, then we may evade tautology. Recently, several attempts have been made to make natural selection out to be a mechanism of this kind. Assuming that the strategy works if the case can be made, let us look briefly at them.

A well-known book is entitled Selection: The mechanism of evolution (Bell 1996). What sort of mechanism could NS be? Elliot Sober, in his famous The nature of selection (Sober 1984, see also his 1993, 2008) and since has argued that selection is a force, which explains (some of) the variation from the “default state” of existing population structure and features in the case of divergent or directional selection. In the case of stasis it explains it by invoking stablising selection.

But what kind of force is it? In mechanics, the forces are energetic; in other cases the forces are equally physical. But there appears not to be a physical force that is all and only selection (unless some kind of thermodynamic account of selection could be made, which cannot be ruled out yet). In a series of papers (Machamer 2004, Machamer, Darden and Craver 2000, Craver 2001) a view of mechanisms as active causal mechanisms has been advanced, and Roberta Millstein (2006) and Rob Skipper (Skipper and Millstein 2005) have applied this to NS. This seems, however, to be rather different to the traditional “push-pull” sense of mechanism, as Skipper and Millstein point out.

What sense of mechanism here could be useful? This is a complicated matter, but I will propose that a mechanism in science is anything that you could in principle build a machine to cause the outcomes. By this, I do not mean that it would simulate the outcomes, but cause them in the same regular manner as the putative mechanism. An example of what I am talking about is a story told by John Maynard Smith in his Theory of Evolution, I think, recounted how as an aeronautical engineer he built a spring and pulley model of the equations for drag and lift coefficients to find the optimal cross section for wings. That is a mechanism that merely simulates the outcome, and could be done these days as easily on a computer. But abstract models like this are not mechanisms – they do not fly. If NS is a mechanism in this sense, then it is merely an abstract model not a physical, active, causal mechanism.

But if there is a physical process going on, then we might think NS is a mechanism, and there’s the rub. Since Sober noted it in Nature of Selection, it’s been clear that fitness and selection are not realised in a single kind of physical process. A virus and a behaviour may have the same fitness, and yet they share no unique physical substrate. Plants, symbiotic organisms in lichens, mosquitoes and, if one allows a NS process in society, an idea, may all have the “same” fitness and yet be physically very different.

So if NS is a mechanism, I do not think it is a mechanism of a physical kind. So our last question, and one raised by this discussion, is:

5. Is selection a law?

Laws are abstractions, but they do not need to have the same physical substrates (unless they are laws of physics). As we have seen a great many of the critics of PNS hold that it isn’t a law, because (i) it doesn’t hold universally over its domain, and (ii) is doesn’t predict outcomes. I think we’ve eliminated the prediction question, but the exception question still holds. Much recent work on laws in science generally hold that they are idealisations (Cartwright 1983, see her 1999). Indeed, Cartwright holds that given the “dappled” nature of the world, at best laws of science are merely models that apply in restricted cases (for example, how many frictionless smooth infinite planes have you encountered?).

So, if this is correct, and I think it is, if selection is a law, it is a model, and if it is a model, then we are left with our initial problem of how to interpret it in physical terms without merely stipulating that it applies when it does. Which is our next post.


Allen, Colin, Marc Bekoff, and George Lauder, eds. 1998. Nature’s purposes: analyses of function and design in biology. Cambridge, MA: MIT Press: A Bradford Book.

Bell, Graham. 1996. Selection: The mechanism of evolution. New York: Chapman and Hall.

Bigelow, John, and Robert Pargetter. 1987. Functions. Journal of Philosophy 84:181-196.

Cartwright, Nancy. 1983. How the laws of physics lie. Oxford: Clarendon Press.

Cartwright, Nancy. 1999. The dappled world: a study of the boundaries of science. Cambridge: Cambridge University Press.

Craver, Carl F. 2001. Role Functions, Mechanisms, and Hierarchy. Philosophy of Science 68 (1):53-74.

Cummins, Robert. 1975. Functional analysis. The Journal of Philosophy 72:741-765.

Machamer, Peter. 2004. Activities and Causation: The Metaphysics and Epistemology of Mechanisms. International Studies in the Philosophy of Science 18 (1):27 – 39.

Machamer, Peter, Lindley Darden, and Carl F. Craver. 2000. Thinking about Mechanisms. Philosophy of Science 67 (1):1-25.

Maynard Smith, John. 1975. The theory of evolution. 3rd ed. Harmondsworth; Baltimore: Penguin.

Millstein, Roberta L. 2006. Natural Selection As a Population-Level Causal Process. British Journal for the Philosophy of Science 57 (4):627-653.

Sober, Elliott. 1984. The nature of selection: evolutionary theory in philosophical focus. Cambridge, Mass.: MIT Press.

Sober, Elliott. 1993. Philosophy of biology, Dimensions of philosophy series. Boulder, Colo.: Westview Press.

Sober, Elliot. 2008. Evidence and evolution: the logic behind the science. Cambridge, UK; New York: Cambridge University Press.

Wright, Larry. 1973. Functions. Philosophical Review 82:139-168.


  1. Eh? Where have functions sprung from? I’m confused!

    The PNS doesn’t have to hold universally: if it doesn’t, you just call it a ceteris paribus law, i.e. one that holds except when it doesn’t.

  2. bob koepp bob koepp

    Seems to me that the above discussion of functions ends on a notably ahistorical note; i.e., seems to me that historically, the notion of a biological function is not assimilable to the notion of a mathematical function.
    But this might be simply beside the point, since I don’t understand how the notion of function is thought to relate to the problem of providing a non-circular definition of ‘fitness’.

  3. ckc (not kc) ckc (not kc)

    …the trouble with “mechanism” is that it is often tied up with “why” (function, purpose, etc.) rather than with “how”. I think there are really no “why” components in natural selection, and that it all amounts to “how” (energetics, ability to compete for resources, including reproductive resources, etc.) Some may view that as a “why”, but I think that is a mistake, and that we should focus on the “how”.

  4. ckc (not kc) ckc (not kc)

    …and the trouble with exception (…”it doesn’t hold universally over its domain”…) as an argument about tautology, is that NS is no different from other mechanisms of evolution (drift, etc.) in that there is no expectation that it be the exclusive mechanism. One could equally argue that random drift is a tautology (“the survival of the – randomly selected– survivors”), but it doesn’t make it any less real as a process in evolution, wherein the survivors differ from (are evolved from) their ancestors.

  5. jeff jeff

    The law approach definitely has some things going for it. A law can be any abstract description observed to be true. They don’t necessarily have to assume anything causal, and in the long run, it doesn’t even matter how they were arrived at, as long as they are observed to be true. Laws don’t have to be universal – Newton’s laws don’t hold at relativistic speeds. And laws can be superseded by other laws, as new observations are made.

  6. ckc (not kc) ckc (not kc)

    …observed to be true.

    …a sticky point, there. The “true” thing, that is.

  7. jeff jeff

    Yeah I know, the whole falsification/verification thing. You push down one wrinkle in the carpet, and three others pop up somewhere else. At least it keeps philosophers busy (if not necessarily employed).

  8. heleen heleen

    The basic model of selection is often held to be the following:
    the change in phenotypic mean is equal to the genetic covariance between phenotype and fitness.
    That would not be a mechanism, as a covariance can scarcely count as a mechanism. The idea function would surface in the question: what biology of the phenotype leads to a covariance with an expected value other than zero to appear – discounting the situation that even if the expected value of the covariance equals zero, the covariance need not be zero in any one generation or environment, and much selection might therefore being stochastic rather than functional. Moreover, the function part of the phenotype should be something that can be analysed by separately without reference to fitness.

Comments are closed.