1. Introduction

According to the selected effects theory of function, a biological trait’s function is, very roughly, whatever it was selected for by natural selection (or some comparable selection process). The function of the butterfly’s eyespots is to deflect attack away from vital organs because that is what they were selected for. A chief virtue of the selected effects theory is that it makes sense of how function statements can work as teleological explanations—which is implicit in at least some strands of biological usage. If functions are selected effects, then when we attribute a function to a trait (say, deflection of attack to eyespots) we are quite literally offering an explanation for why that trait exists. No other account of function—perhaps with the exception of the “organizational theory”—even purports to make sense of this feature of biological usage.

A consensus among philosophers of biology is that the selected effects theory was first formulated independently by Neander (Reference Neander1983) and Millikan (Reference Millikan1984), although perhaps earlier work such as Ruse (Reference Ruse1971), Wimsatt (Reference Wimsatt1972), and Wright (Reference Wright1973) gestured in that direction. One goal of this article is to challenge that consensus. The forgotten French philosopher of science Edmond Goblot (1858–1935) should be credited with formulating the theory, or at least an early incarnation of it. In a series of papers, Goblot (Reference Goblot1899, Reference Goblot1900, Reference Goblot1903) argued, quite rigorously and explicitly, first, that function statements are teleological explanations and, second, that function statements can be teleological explanations only if functions are selected effects. My goal is not in any way to undermine the originality and insight of Neander, Millikan, and their followers. It is rather to ensure that Goblot receives long-overdue credit for his prescient discovery.

But this article does not simply have the goal of insisting that Goblot receive some intellectual credit. That fact alone would be worthy of an extended footnote in a philosophy of biology textbook and not a whole paper. Goblot, however, did much more than that. He articulated a very distinctive (even by today’s standards) version of the selected effects theory. For Goblot, “selection” was much more inclusive than evolutionary natural selection. It was even more inclusive than the abstract notion of “differential reproduction,” or “differential retention,” as some would have it. For Goblot, “selection” refers to a very general process wherein one possibility is realized, to the exclusion of another, by virtue of an apparent advantage. For Goblot, evolutionary natural selection, and intelligent design, are two subtypes of this abstractly specified process.

Moreover, Goblot seemed to think that this claim—that functions are selected effects (when “selection” is broadly construed)—is a conceptual analysis of both lay and scientific use of “function.” If Goblot were right, that would be game-changing even by today’s standards. For it would imply that the selected effects theory, properly grasped, embraces both biological and artifact functions, scientific and lay usage, modern and ancient usage. This expansive construal of the selected effects theory deserves serious consideration.

To be fair, some philosophers of biology have flirted with expanding the selected effects theory to be more inclusive, that is, to allow processes other than evolutionary natural selection to produce new functions. These thinkers include Millikan (Reference Millikan1984) herself, Papineau (Reference Papineau1984), Godfrey-Smith (Reference Godfrey-Smith1992), Griffiths (Reference Griffiths1993), Kitcher (Reference Kitcher1993), and Garson (Reference Garson2011). Dennett (Reference Dennett1969), Wimsatt (Reference Wimsatt1972), and Wright (Reference Wright1973) also gestured toward the possibility of such an all-encompassing theory. Goblot, however, is unique in that he joined two ideas that nobody else joined: First, he attempted to specify, rigorously and precisely, the nature of this general process of which natural selection and intelligent design are subtypes. Second, he posited that this fact, that functions are selected effects—when selection is understood in this expansive way—is part of a correct conceptual analysis of “function.” This is a new thing.

Unfortunately, it seems to me that, while there is something right about Goblot’s expansive way of thinking about functions, his particular construal of function cannot be right. That is because there is no single kind of process in the world of which natural selection and intelligent design are subtypes. The illusion that there is a single kind of process in nature arises from a hidden equivocation in the very idea of “selection for an advantage.” As I will show, one sense of the phrase points to human choice; the other, to evolutionary natural selection. These—as Darwin himself recognized—cannot be fused in any nonmetaphorical way. Although Goblot’s attempt fails, it is a quite noble sort of failure, one that still demands a serious philosophical reckoning

2. Goblot’s Basic Account of Function and Teleology

Goblot wrote two major papers on the topic of function and teleology: “Fonction et finalité” (Reference Goblot1899) and “La finalité sans intelligence” (Reference Goblot1900).Footnote ¹ Crucially, Goblot intended the two papers to be read as a continuous whole. This can be seen from the fact that the purpose of the first paper is to raise a general problem about biological functions, and the purpose of the second paper is to solve that problem. Together, the two papers form one long meditation on functions.

Although the two papers are meant to be read as one, each pursues a distinct question and offers a distinct thesis. Goblot (Reference Goblot1899) argues that function statements are teleological explanations. When we say, for example, “the function of the eyespots on butterfly wings is to deter attacks away from vital organs,” we are, in ordinary biological discourse, trying to explain why butterfly wings have eyespots. Goblot (Reference Goblot1900) argues that teleological explanations are grounded (in a way to be determined) by evolutionary natural selection. Hence, on the surface, his position seems nearly identical to that which Wright (Reference Wright1973) developed: function statements are teleological explanations (a trait has a function if the trait “is there because” it serves the function), and one way that a trait can have a function is if it was shaped by natural selection for the effect in question.

As we will see, however, Goblot goes much further by arguing that functions are just selected effects, where selection must be construed broadly enough to include both evolutionary natural selection and intelligent design.Footnote ² Before diving deeply into Goblot’s analysis, I turn to the text to draw out Goblot’s presentation of these two theses. Any further analysis we conduct must be based squarely on Goblot’s own words.

His first paper argues that the function of a trait is not just any useful effect it happens to have. Rather, a trait’s function is the effect that the trait (in some sense) was made for. It is an effect that plays into an explanation of the trait itself. Goblot begins his analysis by pointing out that functions, in the ordinary biological sense of the term, are peculiar and worthy of serious philosophical reflection: “Of the properties of cells, tissues, and organs, some are, and others are not, functions. Sometimes scientists intentionally use this word function; sometimes on the contrary they take care to avoid it; the definition is difficult, but the use is not at all arbitrary” (Reference Goblot1899, 495).Footnote ³

He then argues that, in ordinary biology, we only call something a “function” when we think that the effect in question is somehow part of an explanation for the trait’s existence: “The blood cell fixes atmospheric oxygen; it also fixes carbon monoxide and nitrogen dioxide. … Of these three chemical properties, only the first one is a function; and the only reason that one calls it that, is that the cell is made to draw, in its passage in the lungs, atmospheric oxygen. … If the cell also fixes other gases, these properties are not functions, for it is not made for that” (Goblot Reference Goblot1899, 497–98). The problem, of course, is that it is very difficult to see how an effect of a trait can be part of the explanation for that very trait, unless we are invoking some sort of supernatural principle, such as divine intervention or a mysterious vital force: “Certain physiologists seem to have a sort of distrust for this idea of finality, which, despite them, can be found in all parts of their science. They dare not look at it directly; finality seems unknowable; for them, it is an anti-scientific, and almost mystical, idea” (499).

Nonetheless, teleology is such a critical part of physiology itself that if we eliminate teleology, we eliminate physiology, too: “Does there exist, in the facts, a teleological order? Put differently, is physiology possible? … The existence of a teleological order is the postulate of the science of life. … The physiologist must therefore assume the reality of a teleological order, as the physicist assumes the reality of a constant and necessary order” (Goblot Reference Goblot1899, 504–5). Teleology seems both impossible and necessary.

The purpose of “La finalité sans intelligence” (Goblot Reference Goblot1900) is, as the title indicates, to point the way to a solution. If, in the past, a trait was shaped by evolutionary natural selection for a certain effect, then that trait exists now precisely because of that effect. If the flower’s nectar glands were selected for attracting insects, then we can rightfully say, now, that the nectar glands exist (i.e., one reason flowers have nectar glands) because they attract insects. When selection is present, a trait’s effect can be cited as part of an explanation for its existence, without appealing to theism or vitalism.

And later: “[After selection,] the final term [i.e., the trait with the function that is now a “fixed” species character] no longer has an accidental character, since it is this very advantage, which has become a species character. Utility is the origin of finality; utility characterizes the initial term, it serves a certain end, but it is not made for this end; finality characterizes the final term; it is well made for this usage, since it is because of its utility that it became fixed as a species character” (404).

In sum, Goblot holds that function statements are teleological explanations and that natural selection can vindicate such explanations, since natural selection shows how a trait’s effect can play a role in an explanation of that trait’s existence. As we will see, however, Goblot thinks that the relation between function and selection is, in fact, even more intimate than this.

3. Rethinking Teleology as Selection

What is really innovative about Goblot’s thought, even by today’s standards, is what I take to be his core thesis about biological functions, one that is never stated explicitly but implied throughout the text.Footnote ⁴ I will try to articulate the view as follows: As a matter of conceptual analysis, an object has a function when it is the result of an abstract kind of selection process. In this selection process, one possibility is realized to the exclusion of another on account of something like “the appearance of an advantage” (l’apparition d’un avantage; Goblot Reference Goblot1900, 405–6). The function of the object is just this advantage. Evolutionary natural selection, and intelligent design, are two different subtypes of this abstract process.

My main textual evidence that this was, in fact, Goblot’s view of function stems from the extraordinary closing section of Goblot (Reference Goblot1900). There, he states that all teleology, intelligent or not, somehow involves a selection between possibilities and the preferential realization of one over another:

In this passage, Goblot leads us through three major areas where teleological statements loom large and shows us that, in each of the three domains, teleology exhibits the same fundamental pattern. First, he asks us to consider intelligent design in the creationist worldview. Suppose we believe that some feature of the world is designed by God for certain ends. We can then ask ourselves: What exactly is God up to when he designs something? It consists in none other than this: God somehow surveys a vast array of possibilities and chooses to realize one possibility over another because of an apparent advantage.

In the lengthy passage cited above, Goblot is quick to point out that the same pattern is exemplified in human decision-making. When a person designs something, something takes place in her mind that is like a competition between imagined possibilities, and one possibility is ultimately realized, over another, because of an apparent advantage.Footnote ⁵

Finally, and most importantly, Goblot sees evolutionary natural selection as conforming to this basic pattern. Natural selection, he thinks, involves a competition between possibilities, wherein one possibility is realized, over another, because of the appearance of an advantage. At this juncture, one might suspect that Goblot is playing a semantic game with us or that he is abusing the natural contours of ordinary language. Surely, natural selection is not a competition between possibilities. To the extent that natural selection is a “competition,” it is a competition between actual organisms (cells, groups), not possible ones.

Although natural selection must be seen as a competition between actual, rather than possible, beings, for Goblot, natural selection is also, and at the same time, a competition between possible species characters. When a new variant arises in a population through a genetic mutation—say, the first butterfly with eyespots on its wings—that variant represents a possible species character. It is not yet an actual species character; it must compete with other variants to earn that title. One thing that natural selection does is that it takes a possible species character and transforms it into an actual species character because of an advantage it possesses: “But if it happens that an individual character is an advantage, natural selection will make of it a species character, and that because it is an advantage. Hence again there is finality, but finality without intelligence” (Goblot Reference Goblot1900, 402). It is because of this somewhat unconventional perspective on natural selection that Goblot can see it as conforming to the basic pattern of teleology in other domains.

4. Convergence and Divergence

Goblot is not the only person to suggest a deep similarity between natural selection and intelligent design. Many philosophers have hinted at a deep connection between the two sorts of things. It is impossible to do justice here, in a rather short article, to the rich similarities and differences between these theorists. Here it will have to suffice to say this: nobody, with the exception of Goblot, has ever held this convergence of ideas:

1. 1. Functions are selected effects.

2. 2. “Selection” in 1 must be understood very generally to encompass natural selection and intelligent design.

3. 3. Claim 1, understood in terms of 2, is a conceptual analysis of both ordinary and scientific language.

An all-too-brief perusal of the literature will show exactly how and where Goblot departs from others. To begin with, Wright (Reference Wright1973) did not accept 1, at least not as a conceptual analysis. He thought that, as a matter of conceptual analysis, a function of a trait is just an effect that explains the trait’s existence. He does discuss the similarity between natural selection and intelligent design, and even the idea that they both exemplify a kind of “selection process” (see 163–64), but he never identifies function, as a matter of conceptual analysis, with this abstract “selection process.” He identifies it with what he calls a “consequence-etiology.”

Wimsatt (Reference Wimsatt1972, 13) seemed to accept that, empirically speaking, functions probably always involve selection, where “selection” is understood broadly to encompass both natural selection and intelligent design: “the operation of selection processes is not only not special to biology, but appears to be at the core of teleology and purposeful activity wherever they occur.” But he adamantly rejected that this should be understood as a conceptual analysis (15).

Dennett (Reference Dennett1969), too, describes a deep analogy between natural selection and learning, and even suggests that selection is at the root of teleology itself (64), although he does not develop this insight into a theory of “function” per se. In fact, he has pursued this analogy throughout much of his work, particularly in his classification of “Darwinian,” “Skinnerian,” and “Popperian” creatures (Dennett Reference Dennett1995), each of which involves the operation of different sorts of selection processes.

Millikan (Reference Millikan1984) defines functions in terms of a general process involving the differential reproduction of one type of entity over another. Her view of reproduction is expansive enough to include trial-and-error learning and learning by imitation (28). But in her view, differential reproduction does not include the process wherein a person creates an artifact for the first time. (The first hammer did not have a history of differential reproduction on account of its past success.) Goblot’s view is, therefore, more inclusive than hers. Similarly, Papineau (Reference Papineau1984, 557–58) states that mental states can undergo natural selection within the lifetime of an individual and thereby acquire (selected effects) functions—but not, presumably, the artifacts that are produced by said mental states.

Kitcher (Reference Kitcher1993), to whom I return in the next section, says that function is design, as a matter of conceptual analysis, and that natural selection and intelligent design are two subtypes of this design. Unlike Goblot, however, he does not articulate what design is supposed to be such that it encompasses both. That is, he never articulates what this nonmetaphorical form of design is supposed to amount to, other than alluding to Darwin’s view that natural selection can be seen as a kind of “design without a designer.”

Neander (Reference Neander1991) holds that functions are selected effects but only in the sense of evolutionary natural selection. Hence, her theory of function is only supposed to apply to the biological sort of function, and it is only intended as a conceptual analysis of modern biological usage. She notes, in passing, the possibility of a theory like Goblot’s but chooses not to develop it in any detail (175).

Griffiths (Reference Griffiths1993) sketches a theory of artifact function that rests on the idea that artifacts come from a kind of “competition” between ideas. But he says explicitly that natural selection and artifact design are quite different things and that there is no single concept of function that applies to both (421).

5. A Critique

If Goblot were right, that would be game-changing for contemporary philosophical discussion of function and teleology. That is because it would give us a version of the selected effects theory that effortlessly captures teleology in every domain in which it arises, both natural and conscious, human and divine. It would also, as a conceptual analysis, unify both modern and ancient usage, as well as scientific and lay usage. For Goblot, by “function,” everybody has always meant selected effect.

Unfortunately, Goblot’s expansive analysis of function does not work. The reason is that there is no single kind of process in the world, loosely called “selection for an advantage,” of which both natural selection and intelligent design are two subtypes. There is only a strained analogy. Goblot’s view relies, ultimately, on an unacceptable anthropomorphism.

The crux of the matter is this: in the standard selected effects theory, an object’s effect becomes that object’s function by virtue of the fact that that sort of object has an actual historical track record of producing that effect. Having an actual historical track record of producing a given effect is necessary for having a function, in the ordinary selected effects sense. Artifacts, however, are not subject to this constraint. As far as artifacts go, it is possible for the effect of some artifact to be its function even if that artifact has no actual historical track record of producing that effect. Neander (Reference Neander1991, 174–75) makes precisely this point in enumerating the differences between natural and artifact functions. The very first twist corkscrew that was ever invented, back in 1795 by the Reverend Samuel Henshall, no doubt had the function of opening wine bottles, even though it had no actual historical track record of doing so. True, there may have been a kind of “virtual selection process” involved in the production of the first spiral corkscrew, a selection process that took place in the reverend’s mind. But the physical corkscrew, that is, that physical type of thing, was not selected because it actually ever opened a wine bottle, since it had never done so. It was selected because someone (i.e., Henshall) thought or surmised or believed or reckoned or figured that it would have that advantage. But anticipated advantages are not real advantages, any more than imaginary ponies are real ponies. To say that all function involves something like “selection for an advantage” obliterates that distinction.

Let me put the point somewhat differently: Goblot’s argument involves a fallacy of equivocation. In the fallacy of equivocation, two or more premises only seem to support a conclusion because of a critical ambiguity in a word or phrase that appears in the premises. Goblot, I maintain, is guilty of such an equivocation. We can reconstruct his argument as follows: natural functions involve selection for an advantage; artifact functions involve selection for an advantage; so, natural and artifact functions both involve selection for an advantage. The ambiguity is this: in the first premise, the “advantages” in question are real, actual advantages; in the second, the “advantages” in question are merely imagined or hoped for. But imagined advantages are not real advantages—any more than imagined ponies are real ponies.

The problem is analogous to the problem in Kitcher’s (Reference Kitcher1993) theory of function. Kitcher attempts to define “function” simply and solely in terms of “design.” He then says that human invention, and Darwinian natural selection, are two subtypes of this design. The problem is that the apparent unity of the concept of function is purchased at the cost of an equivocation: there is no single kind of thing called design, of which human choice and natural selection are subtypes. The latter is “design” in name only; it might be a useful anthropomorphism to speak of natural selection as a form of design, but this analogy cannot bear any real theoretical weight.Footnote ⁶

6. Conclusion

If Goblot is wrong, then what is the right way to think about function and selection? First, if we are to maintain that functions are selected effects, we should continue to understand “selection” in a relatively narrow sense that requires (not as a sufficient condition but as a necessary one) something like an actual history of differential reproduction, or differential retention, on account of the effect in question (Garson Reference Garson2019). It is not enough that some agent hopes or anticipates or surmises that the object will have the relevant effect. In contrast, functions in the realm of artifacts come about because the artifact bears the right kind of relationship to an agent’s mental states—although the precise nature of that relationship remains highly contentious. But I think we should give up the search for a unified theory of biological functions and artifact functions. Too many smart people have tried and failed for that to be a fruitful endeavor.

Now, it may very well be true that, as a rule, when someone produces an artifact, that event of production is preceded by something like a virtual selection process in the agent’s mind. Dennett (Reference Dennett1995) calls us “Popperian creatures” for our ability to carry out a hypothetical trial and error in our minds before implementing our schemes in the real world. But this, in my view, is incidental to an artifact’s having a function. It is not by virtue of the fact that a selection process takes place “in the designer’s mind” that the artifact acquires a function. As Wimsatt (Reference Wimsatt1972, 15–16) argued some time ago, if God is real, and if God had a creative hand in designing the universe or some of the things in it, he would not have had to go through anything like a virtual form of trial and error in order for his creations to have functions. He would have just known what to do.