Biology & Philosophy

, Volume 29, Issue 5, pp 747–759 | Cite as

Revisiting recent etiological theories of functions



Arguably, the most widely endorsed account of normative functions in philosophy of biology is an etiological theory that holds that the function of current traits is fixed by the past selection history of other traits of that type. The earlier formulations of this “selected-effects” theory had trouble accommodating vestigial traits. In order to remedy these difficulties, the influential recent selection or modern history selected-effects theory was introduced. This paper expands upon and strengthens the argument that this theory has trouble stemming from recent “no variation” cases. In addition, several influential arguments for the necessity of including a selection requirement in a theory of normative biological functions are contested. It is suggested that accounting for biological functions in certain areas of biology (such as physiology and the neurosciences) does not require adverting to selection.


Functions Malfunctions Normative functions Selected-effects theory of functions 

Normative biological functions are the sort of thing of which all of us are familiar. For instance, someone with 20/20 vision has eyes that are functioning normally. Someone whose heart has stopped beating generally has a malfunctioning heart. These are examples of normative functions in biological systems: the sorts of functions that can function properly or malfunction/dysfunction. In addition to the intuitive sense that many biological functions are normative, we find normative functional language throughout the biological sciences.

Arguably, the most widely endorsed account of normative functions in philosophy of biology is an etiological theory that holds that the function of current traits is fixed by the past selection history of other traits of that type (see e.g. Millikan 1989a; Neander 1991; Wright 1973). This is known as the selected-effects (SE) theory of functions.

However, the SE theory in its most basic form has trouble accommodating vestigial traits (see the section entitled “Introducing the (recent) selected-effects theory of functions”). This led to the introduction of the “modern history” or “recent selection” varieties of SE theories of functions by Griffiths (1993) and Godfrey-Smith (1994) that look specifically at the recent maintenance of traits under selection. This version of the theory is specifically designed to accommodate vestigial traits. The recent SE approach has been influential among SE theorists as we even find endorsements of it in later writings of some of the earliest proponents of the SE theory (see Millikan 1989b; Neander 2002).

This paper argues that the recent selection theory has significant difficulties stemming from cases where there hasn’t been any recent variation in a trait. This “no variation” issue has been discussed previously by Schwartz (1999, 2002). However, he mainly deals with this problem in the abstract while I develop a biologically realistic example that exemplifies the problem. Schwartz also too readily concedes a point that negates much of the force of this argument, namely, that variation will inevitably arise from mutation (I discuss this further in the section entitled “The “no variation” problem”). The present paper provides a stronger, more developed version of the argument related to “no variation” cases. In addition, I contest several influential arguments that attempt to establish the necessity of including a selection requirement in a theory of normative biological functions. I will suggest that we may be able to account for normative biological functions in certain domains of biology such as physiology and the neurosciences without adverting to selection.

Introducing the (recent) selected-effects theory of functions

The selected effects (SE) theory of normative functions1 is an etiological theory that holds that the functions of items, traits etc. are determined by the past history of other items, traits etc. of that type. Specifically, the relevant historical feature is the past selection history of items/traits of that type. Roughly, SE theorists claim that the function of a biological trait is to do that which previous tokens of that trait did in the past that contributed to the survival and reproductive success of past individuals with that trait and which thereby caused the trait to be selected by natural selection.

SE theorists were pushed towards a “recent selection” theory of functions by the need to be able to handle vestigial traits. For instance, it is thought that the human appendix has become vestigial with respect to the ability to digest cellulose but in some of our remote ancestors the appendix had the function of digesting cellulose. It would also appear that the appendix was selected for digesting cellulose in those ancestors. Thus, if an SE theory stipulated that the function of a trait was what past tokens of the type were selected for doing at any point in the past, the human appendix would mistakenly count as having the function to digest cellulose.

To remedy this problem, a number of SE theorists have stipulated that the function of a trait is what it was selected for doing in the recent past.2 On this view, the fact that appendices were selected for digesting cellulose in the distant past does not make this the function of human appendices because they were not selected for this recently. As is common in the literature, I will refer to this position as the recent selection (or recent SE) theory of functions. While this appears to be the preferred way for SE theorists to deal with vestigial traits, it leads to a significant problem as I discuss in the next section.

The “no variation” problem

The two most developed recent selection theories are those of Paul Griffiths (1993) and Peter Godfrey-Smith (1994). In this section, I will focus primarily on Godfrey-Smith’s theory though as I will point out the issue that I expound for his theory also vitiates Griffiths’ account. As I mentioned in the previous section, Godfrey-Smith’s recent SE theory holds that the function of a biological trait is what it was selected for doing in the recent past. However, he leaves the notion of “recent” fairly vague. The extent of Godfrey-Smith’s elaboration on what “recent” amounts to is the following:

Some might wonder how recent the selective episodes relevant to functional status have to be. The answer is not in terms of a fixed time- a week, or a thousand years. Relevance fades. Episodes of selection become increasingly irrelevant to an assignment of functions at some time, the further away we get. (Godfrey-Smith (1994), p 356)

Since Godfrey-Smith does not define “recent” in terms of an exact length of time, my challenge to his theory will be in the form of a dilemma. The dilemma suggests that he is pressed to count a certain length of time into the past as recent to accommodate a particular “no variation” case. While this allows his theory to handle the “no variation” case, it leads to another significant problem, as I will argue.

The “no variation” case that I will present involves a genetically and phenotypically uniform population of phenylthiocarbamide (PTC) tasters. In order for the case to make sense to a general audience, I need to provide some background on PTC tasting and its genetic basis.

Since the 1930’s, it has been known that some people are capable of tasting a substance, phenylthiocarbamide (PTC) whereas others cannot. So called “tasters” taste small to moderate amounts of PTC as being rather bitter whereas “non-tasters” do not taste the compound at these amounts and only taste it, if at all, at very high concentrations. Although PTC does not occur naturally, it is structurally similar to compounds (isothiocyanate and glucosinolate) found in cruciferous vegetables such as broccoli and cabbage (Tepper 2008). Isothiocyanates impair the uptake of iodine by the thyroid and can effectively be thyroid toxins if an individual has a low consumption of iodine in his/her diet (ibid.).

It was long suspected that there is a significant genetic component to the ability to taste PTC but in 2003, it was determined that a single gene, TAS2R38, appears to be responsible for the ability to taste the substance in the vast majority of cases. For ease of exposition, I will refer to this as the “PTC gene” in what follows. Haplotype3 differences in this gene accounted for up to 85 % of the variance in the trait (i.e. ability to taste PTC) in one sample (Kim et al. 2003) and 75 % of the variance in the trait in a different sample (Prodi et al. 2004). In building my “no variation” case, I will assume that the remaining variance in PTC tasting ability that is not explained by differences in the PTC gene can be accounted for entirely by non-genetic factors such as environmental and developmental differences. This assumption may not be correct.

The PTC gene consists of 1,002 base pairs in a single exon (Kim et al. 2003). Earlier studies found seven different allelic forms of this gene but only two of them existed at high frequencies outside of sub-saharan Africa (Kim and Drayna 2004). The two common alleles differ in three SNP’s that result in three amino acid changes to the protein encoded at amino acid position 49 (proline or alanine), 262 (alanine or valine) and 296 (valine or isoleucine) (Kim et al. 2003). These two common alleles are named according to the amino acids that they code for at those positions: PAV (proline, alanine, valine) and AVI (alanine, valine, isoleucine).

The PAV allele is associated with sensitivity to PTC while AVI is not. Most homozygotes with AVI/AVI can’t taste PTC when presented with moderate amounts of the substance while most homozygotes with PAV/PAV can taste it at those amounts (Kim et al. 2003). Individuals with only a single copy of the PAV allele usually display an intermediate sensitivity to PTC (ibid.).

Now that we have seen some empirical information about PTC tasting and its genetic basis, I will turn to the evolutionary benefit of this trait. It has been suggested that the ability to taste substances with a similar structure as PTC as bitter helps prevent the ingestion of goitrogens (i.e. substances that suppress thyroid function) particularly in areas with little natural iodine and thereby provides an adaptive advantage compared to less sensitive tasters and non-tasters (Tepper 2008; Greene 1974).4 In line with this, I will assume in what follows that the PTC tasting trait and the PAV allele conferred a reproductive advantage on those who possessed them and were selected in the past for preventing consumption of goitrogenous vegetables in conditions of iodine deficiency.

Now, there are several entities with functions related to the ability to taste PTC-like substances under iodine deficient conditions. First, there is the taster trait itself, which seems to have the function of preventing individuals from consuming thyroid toxins (Greene 1974, p 139). Similarly, the taster allele (PAV) is thought to have the function to code for a receptor that detects thyroid toxins (Mennella et al. 2010, p 5). Finally, there is the receptor coded by the PAV allele, which appears to have the function of detecting PTC-like, anti-thyroid substances on the tongue (Bufe et al. 2005; Drayna 2005, p 223).

One might wonder whether the functions referred to here are normative functions or whether they are perhaps non-normative causal role functions of the sort described by Cummins (1975). There is a compelling reason to think that biologists are referring to normative functions in these instances. For, there is evidence that in cases where an individual who has the genotype associated with PTC tasting loses this tasting ability, biologists describe him/her as dysfunctional. For instance, Sharma et al. (2008), p 1070 describe a (possibly hypothetical) case of an individual who tastes PTC as sweet (instead of bitter) as having a taste dysfunction. This seems to be a fairly clear indication that if some factor were to prevent an individual with the genetic predisposition to taste PTC from tasting it as bitter, that this would be viewed as a dysfunction. In addition, given that it is thought that these traits have been selected in certain environments for preventing the ingestion of thyroid toxins, the SE theory would also hold that the functions here are normative.

We are now in a position to consider the “no variation” case that I will be presenting as part of the dilemma against Godfrey-Smith’s recent SE theory. It is worth emphasizing that the case is somewhat hypothetical and involves numerous assumptions that I will point out along the way. However, I think the case is biologically realistic and that it is reasonable to think that such a case may exist or have existed at some point in the past though as far as I know, no such case has been documented. In fact, whether it existed or not is somewhat immaterial. The more important question is whether it is realistic; because if it is, it will help demonstrate how Godfrey-Smith’s theory must treat cases like it. Having a realistic “no variation” case on the table will be immensely useful to expose the difficulties that Godfrey-Smith’s theory has when it attempts to accommodate it.

The case is the following: Suppose that there is a small, relatively isolated population where the “founder” members of the population were mostly PAV homozygotes. Furthermore, the PAV allele has been driven by local selection pressure to fixation in this population due to a lack of iodine and an abundance of goitrogenous vegetables. Hence, everyone in the population is also a PTC taster. There is also little to no immigration into the population for a considerable length of time of several thousand years or longer. As I will show below, mutation is not expected to introduce much (if any) genetic variation in the PTC gene into the population during that stretch of time. The population is hence phenotypically and genotypically uniform with respect to this trait/gene for an extended period of time of several thousand years or longer.

Now, by definition, for selection to take place there must be variation in the population with respect to the trait (or gene) selected. If a trait (or gene) is uniformly present in a population then selection for it cannot occur. In that case, it follows that there won’t be selection for PTC tasting or the PAV allele during the stretch of time of uniformity. But the trait still has a normative function during that time, i.e. to prevent the ingestion of thyroid toxins, which does not simply disappear. I see no reason to think that biologists (or anyone else) would withdraw their normative function attributions from a trait simply because it became uniformly present in a population.

To accommodate this case, Godfrey-Smith’s theory is forced to hold that “recent” can extend backwards potentially for several thousand years or longer. That would enable it to hold that the ability to taste PTC was “recently” selected in this population and it still retains a normative function even when the population has been uniform for some time. I will explain why this is problematic shortly.

An important question that first needs to be addressed about this case is that even if a population is composed of all PAV homozygotes, could it realistically remain genetically uniform with respect to this allele for a considerable stretch of time? I will argue that it can. The three major sources of new genetic variation in a population are recombination, migration and mutation and I will consider each of them in turn.

In this case, since I am supposing that all members of the population are homozygous for the PAV allele that means that each person has two identical copies of it. As a result, there will be no new genetic variation arising from recombination (i.e. crossing-over) as would occur, if say, there were heterozygotes in the population. Any crossing-over would swap identical portions of DNA with each other. Hence, recombination will not introduce new genetic variation into the population.

Next, there is the issue of migration. Given that the population that I’m considering is a relatively isolated one, I will be assuming that the amount of immigration into the population is extremely minimal for a long stretch of time and non-existent for large portions of that time span. Can I assume this? Geographic isolation of human populations has occurred throughout our history. For instance, the Sentinelese people of the Andaman islands are thought to have lived there in isolation for between 30,000 and 60,000 years (Jobling 2012). Furthermore, given that the population at issue lives in a low iodine region, it is possible that neighboring populations also have iodine deficiency issues. That could drive the PAV allele to high frequencies in those populations as well. Thus, even if individuals migrated into this population from neighboring ones, there is a good chance that the migrant will also be a PAV homozygote and therefore, won’t introduce genetic variation into the population.

Even if a migrant did introduce an allelic variant of the PTC gene into the population, it could be removed by genetic drift or be selected against for reasons unrelated to the PTC sensitivity that it confers (I explain this more below). Hence, it seems wholly possible for there to be minimal migration of PTC tasting variants into the population for a time as I have stipulated. I will assume that for at least several thousand years in this population there is either no new genetic variation at this locus introduced by migration or that the amount that is introduced is not sufficient to overcome drift or is selected against for “other reasons”.

Finally, there is the important issue of mutation. Some might argue that my PTC case is unrealistic because mutation is bound to introduce new genetic variants of the PTC gene in any given population. Neander (2006), p 594 comes very close to making this argument:

For one thing, it needs to be stressed that cases where there is absolutely no variation over an evolutionarily significant length of time will almost never-perhaps never-occur… To say that a trait has “gone to fixation” is NOT to say that mutations cease to occur.

In fact, Peter Schwartz, who has most vocally pressed the “no variation” concern for recent SE theories, appears to concede this point (see Schwartz 1999 p S216, 2002 p 248). The problem posed by “no variation” cases dissipates dramatically if we admit that there is bound to be inevitable variation introduced by mutation. I will now show that variation from mutation is not, in fact, inevitable in the PTC case by calculating how often we would expect mutation to produce genetic variation at this locus in the population at issue, given the best information that is available.

As mentioned previously, the PTC gene contains 1,002 base pairs in a single exon. Different estimates have been offered for the mutation rate in the human genome. It should be noted that these are average mutation rates for single nucleotides across the entire human genome. I know of no mutation rate estimates that are specific to the PTC gene, so I will assume that this section of the genome is “average”. I will utilize two particular human mutation rate estimates in my calculation that appear to be among the most inclusive estimates available.

Roach et al. (2010) extrapolated a human intergeneration mutation rate of 1.1 × 10−8 per position/per haploid genome/per generation which corresponds to a rate of 2.2 × 10−8 per position/per diploid genome/per generation. Their estimate was only for single nucleotide mutational events. Another estimate of the mutation rate in humans is provided by Lynch (2010), who found a rate of single nucleotide mutations of 2.7 × 10−9 per site/per generation. As with Roach et al. multiple nucleotide substitutions were not included in Lynch’s estimate.

In performing my calculation, it is important to differentiate synonymous from non-synonymous mutations. Synonymous mutations change a codon for an amino acid into another codon for that same amino acid whereas non-synonymous mutations change the codon from coding for one amino acid to one that codes for a different amino acid. It is usually assumed that synonymous mutations are generally neutral and I will follow suit. 1/3 of the potential mutations in the coding exon of the PTC gene are expected to result in synonymous substitutions (Campbell et al. 2011). That means that 2/3 of all mutations at this locus are expected to be non-synonymous. It is the latter that pertains to my calculation since it is only genetic variation that might be selected against that is relevant.

Not all non-synonymous mutations are detrimental to the fitness of the organism. Some few are beneficial while others change the codon to code for a similar amino acid with small (if any) effects on protein function. Boyko et al. (2008) estimate that 27–29 % of non-synonymous mutations are neutral or nearly neutral. Taking the intermediate value of 28 %, it follows that 72 % of non-synonymous mutations are non-neutral. It is unknown exactly what effects on PTC tasting would result from mutations at most locations along the PTC gene. Though Boyko et al.’s estimate does not guarantee that 72 % of non-synonymous mutations over any given stretch of DNA are non-neutral, given the absence of this data for the PTC gene, I will assume for the sake of the calculation that the 72 % rate applies here.

Using these estimates of mutation rate, the percentage of mutations that are non-synonymous and the percentage of those that are detrimental to fitness, we can use some basic math to calculate a rough estimate of how frequently fitness-decreasing variants in the PTC gene would arise from genetic mutation. I will calculate the expected number of fitness decreasing mutations per generation (and per year) at the TAS2R38 site in our small uniform population. This involves multiplying the mutation rate for this site, the number of base pairs in the PTC gene, the percent of mutations that are non-synonymous, the percent of non-synonymous mutations that are fitness decreasing and the population size. Given that I am hypothesizing a small, isolated population, a population size of 1,000 individuals seems appropriate.

For example, utilizing Roach et al.’s mutation rate estimate leads to an expectation of 1 deleterious mutation in the exon of this gene (in this population) every 96 generations. If the generation length (i.e. average time until reproduction) is 25 years then this is an expectation of 1 deleterious mutation every 2,400 years. If we instead employ the mutation rate estimate of Lynch we end up with an expectation of 1 deleterious mutation every 770 generations. With a generation length of 25 years, that is 1 expected deleterious mutation every 19,250 years.

Given this low rate of expected deleterious mutations and if migration is extremely minimal as I’m assuming, then in order for Godfrey-Smith to be able to hold that PTC tasting (and the PAV allele) were selected “recently” in this population, he would have to maintain that “recent” extends back potentially for thousands of years since the relevant variation might not arise from mutation for that length of time. Even if a relevant mutation did occur, there is no guarantee that it would be selected against for there is a good chance that it will be removed by genetic drift before it has a chance to be selected against. For instance, the individual with the mutation might fail to reproduce for reasons unrelated to his PTC tasting ability as he may simply decide not to have children. Even if he does reproduce, the mutation might not be passed on to his children due to the independent assortment of alleles (i.e. the sperm cells that produce his children may not carry the mutated allele). Supposing that the mutation was passed on to the next generation, there will be new opportunities for it to be eliminated by drift in each generation.

The mutation might also be eliminated by selection but for reasons unrelated to its affect on the PTC tasting phenotype. It is possible that the PTC gene has other pleiotropic effects and a mutation in it could for instance be lethal if it caused a deficiency in an enzyme that catalyzes an essential chemical reaction.5 This would involve selection against the mutant allele but it would be for the “wrong reasons”.

These considerations suggest that “recent” has to extend back even further in time to ensure that there was a variant that was “recently” selected against for the “right reasons”. Otherwise, the recent SE theory could not say (as it should) that PTC tasting has a normative function in this population after thousands of years of uniformity. It is hard to say precisely how far “recent” needs to extend backwards in this case. With Lynch’s mutation rate estimate, it might be on the order of tens of thousands of years. Even using the more conservative estimate of Roach et al. once we take into account the possibility of drift and selection for the wrong reasons, recent might very well have to extend back over three thousand years.6

The dilemma that the PTC case poses for Godfrey-Smith’s theory is that in order to accommodate that “no variation” case, he must maintain that “recent” extends backwards in time potentially for many thousands of years. This stipulation is problematic because if many thousands of years into the past can count as “recent”, then selection for a trait may have occurred in the very distant reaches of the “recent” past but over the many (thousands) of years hence, the trait could have made no contribution to fitness and been subject to regressive evolution due to an accumulation of degenerative mutations. It also may have been subject to extensive regressive evolution because there was selection for some other trait where reduction of the trait at issue was an inevitable side effect of that selection. In either case, Godfrey-Smith’s theory would hold that the trait still has a normative function today, which seems wrong and inconsistent with biological usage.

Perhaps Godfrey-Smith would respond by maintaining that how far into the past counts as “recent” varies depending on the context of the case.7 As we saw, he claimed that “recent” is not defined in terms of a fixed period of time. He also stated that his general aim in introducing his recent SE theory was to provide an account of biological functions for a particular “core biological sense of the term” (Godfrey-Smith 1994, p 344). Perhaps he believes that biologists do not interpret “recent” in terms of a fixed time when they are thinking about recent selection as it relates to this “core” sense of biological functions. In that case, he might draw on contextual differences between distinct cases to define “recent” differently in each of them in a way that allows him to sidestep the problem that I raised. In comparing the PTC case to one where a trait was selected thousands of years ago but no longer contributes to fitness, Godfrey-Smith might maintain that the definition of “recent” could change dependent upon whether a trait is currently contributing to fitness or not.

The basic problem with this response is that it seems ad hoc and Godfrey-Smith has not provided any evidence that biologists define “recent” differently based on this contextual difference. Furthermore, it isn’t clear what reason there is to recommend that “recent selection” should extend farther into the past when a trait is still contributing to fitness as opposed to when it’s not. This looks like an arbitrary move whose sole purpose is to save the theory from a counter-example. In other words, there is no conceptual connection between the length of time that counts as “recent selection” and whether a trait is currently contributing to fitness or not.

Though I have been focusing on Godfrey-Smith’s theory, “no variation” cases also present a problem for Griffiths’ recent selection theory. Griffiths ((1993), pp 417–418) frames his theory in terms of selection during the last “evolutionarily significant time period”. This is defined as, “a period such that, given the mutation rate at the loci controlling T and the population size, we would expect sufficient variants for T to have occurred to allow significant regressive evolution if the trait was making no contribution to fitness.” On his view, so long as a trait was selected to x during the last evolutionarily significant time period defined in this manner, then it has the function to x. If it was not selected to x during that period, then it does not have the function to x.

It is difficult to say precisely how long an evolutionarily significant time period would extend as it will vary with the population size and mutation rate. Its length also depends on what “significant” regressive evolution amounts to, which isn’t entirely clear. In any case, it seems likely that there will be instances where the evolutionarily significant time period extends for several thousand years or more and as I’ve shown, there might not be variation (and hence, no selection) for that long. As a result, Griffiths’ theory would objectionably deny a function to a beneficial trait if that trait failed to be selected due to a lack of variation for an extended period of time.

Do we need selection in a theory of functions?

Based in part on concerns about “no variation” cases, Peter Schwartz (1999, 2002) proposed his continuing usefulness (CU) theory as an alternative to the recent SE theory. The CU theory contains both a selection component and a separate “fitness” component. However, his theory does not require that the selection be recent so a lack of recent variation is not a problem for his account. More specifically, the CU theory states that,

A trait-type X has the proper function F (at time t) if and only if

C1) X has arisen, been modified, or been maintained by natural selection at some point (prior to t) because its doing F contributed to the fitness of individuals with X, and

C2) X’s doing F has recently and importantly (before t) causally contributed to the survival and reproduction of organisms in this species with this trait (Schwartz 2002, p 253).

Elsewhere, I have defended a biostatistical theory of functions that builds off of something like Schwartz’s C2 but unlike his CU theory, it does not require past selection to confer a normative function on a trait (see Kraemer 2013). There is not space in the present paper to provide a defense of the theory. However, what I would like to do in the remaining space is rebut Schwartz’s arguments that an account of normative functions must include a selection requirement.

One of Schwartz’s primary reasons for including a selection requirement is that he believes that a purely “fitness” account based upon C2 would confer too many proper (or normative) functions to traits. For instance, zebra eyes can contribute to survival and reproduction by transporting blood from the optic artery to the optic vein and by blocking parasites but he resists calling either of these the normative function of zebra eyes. C1 could prevent these effects from being normative functions of zebra eyes if it turned out that they were never selected effects. Without a selection requirement, Schwartz’s concern is that the theory would assign normative functions “too liberally”.

There are several important responses to this argument. First, any plausible “fitness” theory of functions must hold that for a trait to possess the normative function to x, it must typically or frequently contribute to survival and reproduction by x-ing, perhaps only under certain circumstances (see e.g. Boorse 1977). Otherwise, the theory would assign functions too liberally such as the function of escaping from cars to the squirrel’s tail when it prevents the animal from being run over by getting caught in a crevice. If zebra eyes do not typically contribute to survival and reproduction by transporting blood or blocking parasites, “fitness” theories would not count those effects as their normative function.

Secondly, if it turns out that these effects are typical ways for zebra eyes to contribute to survival and reproduction (under certain circumstances), I want to suggest that there is nothing objectionable to regarding them as their normative function. Schwartz’s intuition seems to be that the proper function of eyes is to enable sight and that they should only be said to malfunction if they lose the capacity to see. However, we should recognize that many traits have multiple functions at different levels and that a trait can malfunction at one level while functioning properly at another level. For instance, an eye that enabled sight but failed to produce adequate amounts of lysozomes to kill bacteria entering the eye would both function properly in one sense and malfunction in another sense. It would function properly in so far as it enabled sight but malfunction in so far as it failed to protect the eyes from bacterial invaders. In sum, if it is typical in certain circumstances for zebra eyes to contribute to survival and reproduction by protecting the animal against parasites, it does not seem problematic to regard this as one of their normative functions.

Schwartz also appears to believe that we need a selection requirement to allow assignments of functions to explain why traits are present. This explanatory role of function assignments was insisted on by Wright (1973) and it may be one of the central roles of function attributions in evolutionary biology. However, not all areas of biology use functions to play this explanatory role. As numerous authors have noted, normative functions in physiology and in the neurosciences are primarily related to the current effects of traits and not to their past evolutionary history (see e.g. Walsh and Ariew 1996; Roe and Murphy 2011). Although textbooks and scholarly papers in these fields make many claims about the functions of traits, they rarely mention evolution in this context. In these disciplines, normative function categories (i.e. functioning properly, dysfunctioning, etc.) are used to distinguish different groups in populations based upon their differing current functional abilities. They are not directly concerned with explaining how a trait came to occupy the current role that it has. Evolution by natural selection may be lurking in the background in all areas of biology, but that doesn’t mean that all biological concepts are defined directly in terms of it and that all explanatory enterprises in biology revolve around it. Thus, a theory of normative functions that is adequate for physiology and the neurosciences need not include a selection requirement to capture the explanatory role that this concept plays in those disciplines.

I have argued here that Schwartz does not make a compelling case for the necessity of adding a selection requirement to a “fitness” theory of normative functions. To fully establish that a biostatistical/fitness theory could stand on its own would require showing that the theory can handle numerous other objections that have been raised in the literature such as the alleged problem of epidemic and universal diseases. I believe that it can handle these objections if it carefully defines the normal functional range for the performance of a function and is especially careful about what the relevant reference classes are. I don’t have the space here to discuss the details of the theory nor its defense from these objections, though I do so elsewhere (See Kraemer 2013). 


  1. 1.

    For ease of exposition, I will henceforward just say “functions” to refer to normativefunctions.

  2. 2.

    This comes closest to Godfrey-Smith (1994)’s formulation of the recent selection theory. Griffiths (1993)’s formulation instead refers to selection during the “last evolutionarily significant time period”.

  3. 3.

    Haplotypes in this context appear to refer to combinations of single nucleotide polymorphisms (SNP’s), which are relatively common single nucleotide differences between different alleles of a gene.

  4. 4.

    Kraemer (2012) discusses some of the evidence in favor of this suggestion.

  5. 5.

    Schwartz (1999) makes a similar point.

  6. 6.

    My calculation should be taken with a grain of salt because it does not include multiple nucleotide mutations, which though they are much less common than single nucleotide mutations, still occur. I also did not factor in the possibility of mutations along the regulatory regions of the PTC gene, which might also potentially impact PTC tasting.

  7. 7.

    Robert Brandon and an anonymous reviewer brought this potential response to my attention.



I am grateful to Robert Brandon, the late Fred Dretske, and especially Karen Neander for helpful comments on previous versions of this paper. I also benefitted from suggestions made by an anonymous reviewer and by the editor of this journal, Kim Sterelny.


  1. Boorse C (1977) Health as a theoretical concept. Philos Sci 44(4):542–573CrossRefGoogle Scholar
  2. Boyko A, Williamson S, Indap A, Degenhardt J, Hernandez R, Lohmueller K, Adams M, Schmidt S, Sninsky J, Sunyaev S, White T, Nielsen R, Clark A, Bustamante C (2008) Assessing the evolutionary impact of amino acid mutations in the human genome. PLoS Genet 4(5):1–13CrossRefGoogle Scholar
  3. Bufe B, Breslin P, Kuhn C, Reed D, Tharp C, Slack J, Kim U-K, Drayna D, Meyerhof W (2005) The molecular basis of individual differences in phenylthiocarbamide and propylthiouracil bitterness perception. Curr Biol 15(4):322–327CrossRefGoogle Scholar
  4. Campbell M, Ranciaro A, Froment A, Hirbo J, Omar S, Bodo J-M, Nyambo T, Lema G, Zinshteyn D, Drayna D, Breslin P, Tishkoff S (2011) Evolution of functionally diverse alleles associated with PTC bitter taste sensitivity in Africa. Molecular Biology and Evolution, Advance Access, pp. 1–37Google Scholar
  5. Cummins R (1975) Functional analysis. J Philos 72(20):741–765CrossRefGoogle Scholar
  6. Drayna D (2005) Human taste genetics. Annu Rev Genom Human Genet 6:217–235CrossRefGoogle Scholar
  7. Godfrey-Smith P (1994) A modern history theory of functions. Nous 28(3):344–362CrossRefGoogle Scholar
  8. Greene L (1974) Physical growth and development, neurological maturation, and behavioral functioning in two ecuadorian andean communities in which goiter is endemic II. PTC taste sensitivity and neurological maturation. Am J Phys Anthropol 41:139–152CrossRefGoogle Scholar
  9. Griffiths P (1993) Functional analysis and proper functions. Br J Philos Sci 44:409–422CrossRefGoogle Scholar
  10. Jobling M (2012) Significant others. Investig Genet 3:21CrossRefGoogle Scholar
  11. Kim U-K, Drayna D (2004) Genetics of individual differences in bitter taste perception: lessons from the PTC gene. Clin Genet 67:275–280CrossRefGoogle Scholar
  12. Kim U-K, Jorgenson E, Coon H, Leppert M, Risch N, Drayna D (2003) Positional cloning of the human quantitative trait locus underlying taste sensitivity to phenylthiocarbamide. Science 299:1221–1225CrossRefGoogle Scholar
  13. Kraemer D (2012) Making sense of normative functions and information in neurobiological systems. PhD Dissertation, Duke University, Durham, USAGoogle Scholar
  14. Kraemer D (2013) Statistical theories of functions and the problem of epidemic disease. Biol Philos 28:423–438Google Scholar
  15. Lynch M (2010) Rate, molecular spectrum, and consequences of human mutation. Proc Natl Acad Sci 107(3):961–968CrossRefGoogle Scholar
  16. Mennella J, Pepino YM, Duke F, Reed D (2010) Age modifies the genotype-phenotype relationship for the bitter receptor TAS2R38. BMC Genet 11(60):1–9Google Scholar
  17. Millikan R (1989a) In defense of proper functions. Philos Sci 56(2):288–302CrossRefGoogle Scholar
  18. Millikan R (1989b) An ambiguity in the notion “function”. Biol Philos 4:172–176CrossRefGoogle Scholar
  19. Neander K (1991) Functions as selected effects: the conceptual analyst’s defense. Philos Sci 58(2):168–184CrossRefGoogle Scholar
  20. Neander K (2002) Why history matters: four theories of functions, published as “Warum Geschichte Zahlt: Vier Theorien von Funktionen” in M Weingarten, G Schlosser (eds) Formen der Erklaerung in der Biologie (Verlag fuer Wissenchaft und Bildung)Google Scholar
  21. Neander K (2006) Moths and metaphors. Review essay on organisms and artifacts: design in nature and elsewhere by Tim Lewens. Biol Philos 21:591–602CrossRefGoogle Scholar
  22. Prodi DA, Drayna D, Forabosco P, Palmas MA, Maestrale GB, Piras D, Pirastu M, Angius A (2004) Bitter taste study in a sardinian genetic isolate supports the association of phenylthiocarbamide sensitivity to the TAS2R38 bitter receptor gene. Chem Senses 29:697–702CrossRefGoogle Scholar
  23. Roach J, Glusman G, Smit A, Huff C, Hubley R, Shannon P, Rowen L, Pant K, Goodman N, Bamshad M, Shendure J, Drmanac R, Jorde L, Hood L, Galas D (2010) Analysis of genetic inheritance in a family quartet by whole-genome sequencing. Science 328:636–639CrossRefGoogle Scholar
  24. Roe K, Murphy D (2011) Function, dysfunction and adaptation? In: Adriaens P, De Block A (eds) Maladapting minds: philosophy, psychiatry, and evolutionary theory. Oxford University Press, USA, pp 216–237CrossRefGoogle Scholar
  25. Schwartz P (1999) Proper function and recent selection. Philosophy of Science (66) Supplement, p. s210–s222Google Scholar
  26. Schwartz P (2002) The continuing usefulness account of proper function. In: Ariew A, Cummins R, Perlman M (eds) Functions: new essays in the philosophy of psychology and biology. Oxford University Press, New YorkGoogle Scholar
  27. Sharma K, Sharma P, Sharma A, Singh G (2008) Phenylthiocarbamide taste perception and susceptibility to motion sickness: linking higher susceptibility with higher phenylthiocarbamide taste acuity. J Laryngol Otol 122:1064–1073Google Scholar
  28. Tepper B (2008) Nutritional implications of genetic taste variation: the role of PROP sensitivity and other taste phenotypes. Annu Rev Nutr 28:367–388CrossRefGoogle Scholar
  29. Walsh D, Ariew A (1996) A taxonomy of functions. Can J Philos 26(4):493–514Google Scholar
  30. Wright L (1973) Functions. Philos Rev 82(2):139–168CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media Dordrecht 2014

Authors and Affiliations

  1. 1.Florida Atlantic UniversityBoca RatonUSA

Personalised recommendations