The Once and Future Function (Commentary on Pigliucci and Kaplan)[1]

Paul Griffiths

1. Introduction

Pigliucci and Kaplan subscribe to a 'modern history' (Griffiths 1993; Godfrey-Smith 1994) version of the etiological theory of function (Kaplan and Pigliucci 2001; Pigliucci and Kaplan 2006). Their particular concern is with the 'gene for' locution. They suggest that a gene is for that phenotypic effect which it is its proper function to produce, in the sense given by that theory of functions.

In earlier work (Griffiths 1992, 1993) I subscribed to the consensus view that there are two senses of function, selected (etiological) and causal:

·  Selected function: a sequence of nucleotides GAU has the selected function of coding for aspartic acid if that sequence evolved by natural selection because it had the effect of inserting that amino acid into some polypeptide in ancestral organisms

·  Causal function: a sequence of nucleotides GAU has the causal function of coding for aspartic acid if that sequence has the effect of inserting that amino acid into some polypeptide in the organism in which it occurs

The idea of causal function (Cummins 1975) is sometimes presented as a rival to selected function (Davies 2001; Lewens 2004). However, both notions are needed to capture the conventional idea of evolution by natural selection. Darwinism distinguishes between adaptations, which have evolved by natural selection, and adaptive traits, which increase the fitness of organisms that possess them relative to other types. By definition, every adaptation was once an adaptive trait, but not all adaptations are still adaptive and not every adaptive trait is yet an adaptation. If we use the language of functions, a trait is adaptive in virtue of some of its causal functions. The selected functions of a trait are those causal functions of the same trait in ancestors which caused it to be selected.

In this earlier work I also subscribed to the consensus that only the selected sense of 'function' has genuine overtones of teleology and allows us to discriminate between proper functions and the mere effects of a trait. In recent work I have come to doubt this (Griffiths 2006; Griffiths 2007). I still agree that there are some contexts in which the function/effect distinction is drawn in the manner suggested by the etiological theory. These are contexts in which scientists are focusing on evolutionary questions. However, in many of the 'proximal' , 'causal' or 'experimental' (Weber 2004) areas of biology – those concerned with the experimental elucidation of form and function (e.g. anatomy, physiology, developmental and molecular biology) - I believe that the use of functional language is richer than is captured by the bare causal account of functions. But I also believe that these uses of 'function' do not correspond to the etiological theory. One of my main reasons for believing this is that if it were true assignments of function in those sciences would be epistemically demanding, as PK make clearer that I could myself. Since ascriptions of function seem to be both ubiquitous and solidly-based in those sciences, I infer that they are not ascriptions of function in the etiological sense.

So I have come to tentatively believe that these function ascriptions correspond to a restricted set of causal functions – roughly, those that contribute to the causal capacities of the organism that are relevant to understanding its evolutionary fate So I have come to advocate a 'forward-looking', evolutionary theory of function as a supplement to the backwards-looking selected (etiological) theory. I am not the first to propose a future-looking, evolutionary theory, of course (Bigelow and Pargetter 1987; Canfield 1964).

2. Functions and Character Individuation

PKs theory of the 'gene for' locution amounts to a proposal to individuate genes (for some purposes) by their selected (etiological) functions. Since it is in the role of a principle of trait individuation that I am most sceptical of the value of the etiological theory, it is the potential clash between PKs proposal and my recent views that I want to explore here. Do we really need to know that it was selection for adding a phosphate group to some molecule or other that led to the proliferation of a particular gene in order to know that it is a protein kinase gene? Is the hypothesis that UCGis the codon for Serine really hostage to discoveries about whether the genetic code is the result of selection or only of chemistry and history (Knight, Freeland, and Landweber 1999)? PK are probably not committed to such strong views, but I want to make clear what the general area of disagreement is.

Ruth Millikan and Karen Neander are the main advocates of the view that proximal biology individuates parts and processes by their selected functions. Neander writes:

“[function and structure] are both ‘normative’ in the sense that they are both notions of the normal, in the teleological as opposed to the statistical sense of the term, if we assume an etiological account of each of them, Abnormality inclusive categories involve a notion of structure and function that is…with, not without, a purpose.” (Neander 2002, 414)

Millikan offers a very general argument for supposing that biological characters must be defined by selected function. She argues that a purely descriptive biology, unaided by evolutionary teleology, has no principled way to determine what counts as the system in need of analysis. Biologists need to identify whether something is part of an organism's biological functioning, as opposed to an irrelevant causal processes like that a kangaroo undergoes when it is burnt in a bushfire. They need to distinguish variation in the population from pathology, and to delimit the boundaries of a single organism, so as, for example, not to confound the kangaroo and its fleas. However,

"Living chunks of matter do not come, just as such, with instructions about what are allowable conditions of operation and what is to count as allowable input. Similarly, they do not come with instructions telling [what is] damage, breakdowns or weardowns. Nor do they come with instructions about which processes…are to count as occurring within and which are irrelevant or accidental to the system.” (Millikan 2002, 121)

We can draw all these distinctions, Millikan argues, by studying only those features of an organism which are part of its evolutionary design, and by classifying the parts of the organism in terms of the purposes for which they were designed – their proper functions. Interestingly, this same point was made by the founders of ethology, Lorenz and Tinbergen, who criticised American comparative psychology for analysing capacities of animals that they believed were not ecologically valid. It was as if, to use Millikan’s metaphor, one were to study washing machines by examining how they behave in space, or at great depths in the ocean.

These are genuine problems, and they require a solution. But this cannot come from adaptive teleology. It can't be the case that biologists need to know the selective history of an organism before they describe the contributions of its parts to its biological functioning. If this were true, biology would be trapped in a vicious regress, as follows:

1.  Ascriptions of selected function are generated by (hypothetical) causal analysis of the capacities of ancestral organisms to survive and reproduce in ancestral environments (Griffiths 1993)

2.  Hence, if we cannot identify which capacities of ancestral organisms to subject to causal analysis without knowing what the parts of those organism were selected for in their ancestors, then we face a vicious regress

3.  Therefore, a purely causal analysis of the adaptive role played by parts of ancestral organisms must be possible without knowing what those parts were adaptations for

4.  Furthermore, ancestral organisms cannot be easier to causally analyze than living organisms on which we can actually experiment (Stotz and Griffiths 2002)

3. Can we do without an evolutionary perspective in proximal biology?

The leading advocate of this view is probably Arno Wouters (1995; 2003; 2005; 2005). He writes that:

“Functional biology without evolution is incomplete in the sense that it ignores many important questions about life, but not in the sense that no aspect of life can be understood without invoking evolution” (Arno Wouters 2005, 55)

Wouters distinguishes between ‘biological role’ – "the manner in which that item/activity contributes to the activity of a complex system" (my ‘causal function’) - and 'biological advantage’ – "the way in which that trait influences the life chances of an organism as compared to other traits that might replace it" (2005a, 41-2). He argues that biologists study biological role not biological advantage. The specific biological role that is causally analysed by biologists is ‘viability’ – the ability to stay alive. Other theorists who offer similar accounts of function have identified the privileged causal capacity as autopoesis, or self-reproducibility (e.g. Schlosser 1998).

Marcel Weber (2004) has endorsed a similar position to Wouters, but influenced by the function theory of Peter McLaughlin (2000) rather than Wouters himself, so the idea that proximal biology can do without an evolutionary perspective is quite popular. In my view, however, it is deeply mistaken. The ideas of viability and self-reproducibility are impoverished accounts of biological functioning and cannot draw the necessary distinctions identified by Millikan. Capacities we would not identify without a genuinely evolutionary perspective include mechanisms controlling variance in offspring number, which no biologist had realised were part of the organism's functioning until post-Hamiltonian population genetics revealed that they affected the course of evolution. Another example is way that somatic stem cells are set up so as to deal with the problem of cancer. This was invisible until the relevant evolutionary models were put forward. The need for an evolutionary perspective is overlooked by Wouters, Weber and others because a great deal of experimental biology is still documenting basic mechanisms and is simply not yet sensitive to the details of the causal capacity to which these causal mechanism contribute.

Example: The Australian brush Turkey Alectura lathami incubates its eggs in mounds of rotting compost tended by the male. It has recently been discovered that seasonal variations in temperature have a small but significant effect on sex ratio (via differential survivorship) (Göth and Booth 2004). This raises the question of whether the mechanisms that underpin this dependence on temperature are part of the biological functioning of the bird, or whether they are like the 'mechanisms' that cause kangaroo joeys to perish in bushfires that their mothers cannot outrun. Is the difference in survivorship merely the failure of the Brush Turkey to buffer environmental insults to the embryo, or is it a mechanism linking sex ratio to seasonal resource fluctuations, as seen in other birds, such as the New Zealand Kakapo Strigops habroptilus. To draw this distinction it is not enough to ask if they are part of how the bird remains viable or how it is capable of self-reproduction. We must ask how they contribute to the birds ability to evolve. Biological function is survival and reproduction and, properly understood, both of these mean 'activity relevant to future evolution'.

4. Millikan's paradox

I have argued that an evolutionary perspective on biological functioning is necessary, but that the usual version of what this would mean leads to a vicious regress. To avoid this paradox we must distinguish two kinds of functioning which are privileged from an evolutionary viewpoint. The first is selected functioning (the operation of adaptations). The second is causal functioning which contributes to survival and reproduction (adaptive functioning). The importance of the second notion for causal analysis is clear in Niko Tinbergen’s ‘On the Aims and Methods of Ethology’ (1963). To understand an organism from a biological point of view we need to answer four questions:

  1. Causation
  2. Survival value
  3. Ontogeny
  4. Evolution

Questions of causation ask about the mechanisms by which organisms do what they do, and questions of ontogeny ask how those mechanisms are built (‘causal biology’).

Questions of survival value ask: “whether any effect of the observed process contributes to survival if so how survival is promoted and whether it is promoted better by the observed process than by slightly different processes.” This approach is simultaneously ‘evolutionary’ (it is guided in by our best current models of selection), and ‘methodologically creationist’: “To those who argue that the only function of studies of survival value is to strengthen the theory of natural selection I should like to say: even if the present-day animals were created the way they are now, the fact that they manage to survive would pose the problem of how they do this.” (1963, 423 my emphasis). Even creationists must study survival value if they are to do biology properly.

Questions of evolution have “two major aims: the elucidation of the course evolution must be assumed to have taken, and the unraveling of its dynamics.” (1963, 428) The course of evolution is revealed by inferring phylogenies and homologies. The dynamics of evolution are revealed by the study of 1) population genetics and 2) survival value (428), which correspond to Sober’s (1984) evolutionary ‘consequence laws’ and ‘source laws’.

With Tinbergen’s analysis in hand we see that Millikan is right - a biologically meaningful causal analysis must be carried out from an evolutionary perspective. But rather than focus on those causal capacities that featured in past episodes of selection, we should focus on causal capacities that contribute to survival and reproduction (survival value). This was how Tinbergen and his successors tackled Millikan’s ‘washing machine problem’. How we define the ‘system’ will reflect our theories about evolution. If there is more than one level of selection, there will be more than one ‘system’ to analyse. More than one ‘system’ may be needed to capture all the dynamics at one level of selection if not all the measures of ‘fitness’ needed to model evolutionary dynamics can be grounded in a single physical propensity (Rosenberg and Bouchard 2002). Turning to boundary conditions for an organism’s functioning, a principled choice, at least for evolutionary analysis, would be the parameter ranges of the ‘fundamental niche’ – the ecological hyperspace within which a population could maintain itself indefinitely (Sterelny and Griffiths 1999, 270). Finally, the distinction between pathology and polymorphism is sometimes obvious (most pathology is not heritable) and sometimes problematic. It has been argued that alleles for haemochromatosis (excessive iron accumulation in tissues) are advantageous to women living under scarcity, whilst under abundance they are neutral in women and harmful in men. It is unclear that a geneticist or physiologist needs to definitively answer whether carriers of these alleles are deformed or merely different.