Rethinking Functions: The Role of Selection

By Hanzhe Dong and Gualtiero Piccinini

Introduction

What are biological functions, and how should we think about them? This question has been controversial in the philosophy of biology (and technology).

The Selected Effects (SE) account ties (teleological) functions to selection history, asserting that a trait’s teleofunction F is an activity it was selected for—typically, the idea is that a trait was selected for because its performing F enhanced fitness in (recent) ancestors. By contrast, the Goal-Contribution Account (GCA) defines teleofunctions as traits’ regular contributions to an organisms’ biological goals, such as survival, reproduction, development, and helping (Piccinini 2020, Ch. 3).

In recent Brains Blog posts, Hundertmark et al. offer a thoughtful defense of SE and critique of GCA. They argue that selection is inescapable for understanding teleofunctions. While we acknowledge that selection plays an important role in explaining how teleofunctions stabilize and evolve, we will argue that GCA (along with systemic-organismic accounts, which are compatible with GCA) accounts for the core notion of function in biology and other disciplines. Therefore, GCA provides a more foundational framework for understanding functions. Below, we explore this debate by addressing three points: malfunction and normativity, the boundaries of functional attribution, and novel traits and artifacts.

Malfunction and Normativity

The Critique: Hundertmark et al. argue that GCA struggles with systemic dysfunctions and normativity. They present two cases:

1. Pandemic Dysfunction: When a systemic issue (e.g., pneumonia) prevents most lungs in a population from oxygenating blood, GCA supposedly implies that lungs lose their teleofunction entirely because they are no longer contributing regularly to survival.
2. Pandemic Futile Functioning: Traits often depend on other systemic factors to contribute to biological goals (e.g., eyes require a functioning body). If systemic failure prevents these contributions, Hundertmark et al. argue that GCA undermines the normative aspect of functions, potentially failing to maintain the distinction between what a trait does and what it ought to do.

According to them, SE avoids these problems because it grounds normativity in selection history. A trait retains its teleofunction under SE regardless of current performance or systemic disruptions.

Our Response: Addressing pandemic dysfunction and futile functioning requires the more careful formulation provided Garson and Piccinini (2014). Here we will paraphrase Garson and Piccinini (2014) using language from Piccinini (2020). Strictly speaking, teleofunctions are those regular contributions to goals of organisms (within a reference class) that traits make when they do contribute to those organisms’ goals. To see what a trait’s teleofunctions FF are, we must observe well-functioning tokens of those traits when they do contribute regularly to organisms’ goals—i.e., under appropriate conditions (which do not include pandemics, of course). Since even during pandemics it remains the case that, were those traits to contribute to organisms’ goals, they would do so by performing FF, FF remain their teleofunctions. In other words, traits retain their teleofunctions even when they generally fail to perform them, as during cases of pandemic dysfunction.

Take the lungs during a pneumonia pandemic. Even if widespread dysfunction prevents most lungs from oxygenating blood, their teleofunction persists because the normativity of functions is tied to the biological goal of oxygenation. Similarly, eyes retain their function to convert light into electrical signals, even if systemic failures render this futile for survival.

Hundertmark et al. appear to assume that GCA ties normativity directly to performance under all circumstances, but this is not the case. Teleofunctions persist as long as the underlying capacity exists. As Garson and Piccinini (2014) explain, GCA capture this dispositional aspect of functions​.

What Counts as a Function?

The Critique: Hundertmark et al. argue that GCA lacks clear criteria for distinguishing functions from incidental effects:

1. Frequency and Reliability: They claim that GCA’s reliance on “regular contributions” is ambiguous, particularly for rare but essential functions like sperm fertilization.
2. Incidental Effects: They imply that GCA might be unable to exclude incidental benefits (e.g., body heat produced as a byproduct of neural activity contributing to warmth in cold environments) because these effects sometimes contribute to survival.

SE resolves these issues, they argue, by tying teleofunctions to past selection, which provides a clear boundary between functions and incidental effects.

Our Response: GCA distinguishes functions from incidental effects based on regular contributions to biological goals, not selection.

1. Frequency and Reliability: GCA accommodates teleofunctions that are rarely performed, like sperm fertilization, in the same way that it accommodates functions during pandemic diseases. As Garson and Piccinini put it, “even though the unconditional probability of a sperm’s fertilizing an ovum is low, the conditional probability that a sperm fertilizes an ovum, given that it is contributing to survival or inclusive fitness, is extremely high. That is, on those occasions in which the sperm contribute to survival or inclusive fitness, it is very likely that they do so by fertilizing the ova” (Garson and Piccinini 2014, 6).
2. Incidental Effects: GCA excludes incidental effects by focusing on contributions that are regular enough. Hair does not have the function of slowing bullets because this effect does not reliably or systematically contribute to survival or reproduction. What is regular enough? Previous formulations leave this vague, and that may well be ok. But one possible way to define what counts as a contribution to organisms’ goals that is “regular enough” to count as a teleofunction is based on whether natural selection could act on it. If an effect is regular and impactful enough that natural selection could act on it, then it’s a teleofunction. This way of drawing the boundary does appeal to selection, but the role selection plays is the reverse of the role it plays in SE accounts. SE accounts claim that an effect is a function because it was selected for. GCA claims that an effect can be selected for because it’s a function.

In sum, GCA’s focus on regular contributions to biological goals provides clear enough boundaries for functional attribution without resorting to selection history.

Novel Traits and Artifacts

The Critique: Hundertmark et al. claim that SE can account for the functions of novel traits or artifacts:

1. Novel Traits: They argue that typically, novel traits arise gradually within an existing lineage. Thus, SE’s stepwise selection process, which links novel traits to ancestral functions, ascribes functions to (gradual) mutations. If a trait is so novel that it has not been selected for, they say, it lacks teleofunctions altogether.
2. Artifacts: They also suggest that artifacts, such as beaver dams or chimpanzee tools, may require a different account of teleofunctions from biological traits. Either way, they also argue that SE can account for artifact functions in terms of selection.

Our Response: These are good points, but GCA remains better positioned to account for the core notion of teleofunction of novel traits and artifacts.

1. Novel Traits: Mutations are often gradual, but they don’t have to be. In addition, novel and often beneficial traits can be acquired by an organism not via selection but via other processes. Consider horizontal gene transfer. There are three kinds of horizontal gene transfer: transformation (acquiring DNA from the environment), conjugation (acquiring DNA from another organism), and transduction (acquiring DNA from a virus that, in turn, acquired it from another organism). While the novel gene that an organism acquires may have its own selection history, the organism that receives the gene acquires a beneficial trait by means other than selection. This shows that selection is not necessary for teleofunction. Novel traits have teleofunctions insofar as they contribute reliably to biological goals. For instance, if a bacterium receives a gene from another bacterium that allows it to resist antibiotics, this antibiotic-resistance function exists immediately because it supports survival—regardless of how the bacterium acquired the gene or whether the gene was selected for. SE’s stepwise explanation is an important part of the story but it’s not the whole story. In the most basic sense, teleofunctions arise from traits’ contributions to goals, whether or not they were selected for​. After they come into being, teleofunctions are often selected for, precisely because they are beneficial to organisms (Piccinini 2020, Ch. 3).
2. Artifacts: Many artifacts are selected for in ways that resemble natural selection. For instance, consumers within a relatively free marketplace select products of a given type based at least in part on how well the products perform and, as a result, some companies continue to make copies of their products while other companies go out of business and their products stop being made. So, artifact selection can account for the existence and proliferation of some artifacts at the expense of others in a way that parallels the explanation of evolution by natural selection. But novel artifacts can be created from scratch, and novel features can be added to existing artifacts. Those novel artifacts and features often have desirable effects, which fulfill their users’ goals. GCA assigns teleofunctions to such artifacts and features regardless of how they come about. They might have come about because their inventors “selected them” in their own conscious mind but they might also have come about because their inventors designed them without any conscious “selection” process, copied some other artifact (analogously to horizontal gene transfer), created them by trial and error, or even created them by accident. In the end, the story of how a novel artifact or artifact feature came about does not change whether it provides a regular contribution to its users’ goals. If it does, that regular contribution is its teleofunction.

In short, both biological traits and artifacts can be selected for precisely because their teleofunctions are there in the first place, and GCA can account for that. Selection is a very important part of the story of how functions become stabilized, refined, and evolve, but it’s not the whole story.

Conclusion: A Pluralistic Perspective

While selection is not necessary to define and account for teleofunctions in their core sense, we agree with Hudertmark et al. that there are multiple useful notions of function, at least one of which involves selected effects. Teleofunctions exist insofar as traits contribute regularly to organisms’ goals. Selection can act on these traits, ensuring their persistence, refinement, and evolution across generations. Thus, selection is an important part of the explanation of why many traits (and artifacts) come to be where they are. It is legitimate to define a notion of “function” (aka adaptation) stronger than the goal-contribution notion, a notion that entails that an effect was selected for.

For example, the heart’s teleofunction to pump blood exists because it supports survival and development. Selection stabilizes and refines this function by favoring traits that perform it well. As Garson and Piccinini (2014) argue, teleofunctions are primary, and selection explains their evolutionary persistence​. That said, pumping blood is surely an adaptation, and it’s ok to imply that it is when claiming that it’s the function of the heart.

In the end, GCA and SE can coexist as accounts of different notions of function. GCA provides a robust framework for understanding teleofunctions in the core sense, emphasizing regular contributions to biological goals. SE enriches this picture by explaining how traits stabilize and evolve over time, becoming adaptations. By integrating both perspectives, we achieve a richer understanding of the complexity of biological functions.

Reference

Piccinini, Gualtiero & Garson, Justin (2014). Functions Must Be Performed at Appropriate Rates in Appropriate Situations. British Journal for the Philosophy of Science 65 (1):1-20.

Piccinini, G. (2020). Neurocognitive mechanisms: Explaining biological cognition. Oxford University Press.