What Does Function Mean in Science and Biology?

“Function” in biology carries two distinct meanings that often get blurred together, and the difference matters more than most people realize. One meaning is mechanical: what does a structure or molecule actually do right now? The heart pumps blood, hemoglobin carries oxygen, stomata let gases in and out of a leaf. The other meaning is evolutionary: what was this trait selected to do over generations? These two senses usually overlap, but when they don’t, real scientific disputes erupt, sometimes over billions of dollars in research funding and the basic question of how much of your genome actually matters.

What It Does Versus What It’s For

Philosophers of biology have spent decades trying to pin down what scientists actually mean when they say something has a function. Two frameworks dominate the conversation. The first, and currently the most widely accepted among philosophers, is the “selected effects” account. Under this view, a trait’s function is whatever effect contributed to that trait being favored by natural selection over time. The function of the heart is to pump blood because pumping blood is why hearts were selected for across evolutionary history.

1PubMed. Biological functions are causes, not effects: A critique of selected effects theories

The second framework is the “causal role” account, which takes a more engineering-style approach. Here, a part’s function is whatever capacity it contributes to some larger system’s ability to do something. If you want to explain how the circulatory system delivers oxygen, the heart’s function is to pump blood, simply because that’s the role it plays in making the system work. This definition doesn’t require any evolutionary backstory. You could apply it just as easily to a machine.

2Philosophy of Science. Role functions, mechanisms, and hierarchy

For most practical biology, the two definitions point to the same answer. But they part ways in important edge cases, and those edge cases are where the interesting science lives.

The ENCODE Controversy and Why Definitions Have Consequences

The gap between these two definitions blew up publicly in 2012 when the ENCODE project, a massive international effort to catalog the activity of every stretch of the human genome, announced that roughly 80 percent of human DNA was “functional.” For decades, the conventional understanding had been that most of the genome was non-coding and had no clear purpose, sometimes loosely called “junk DNA.” The ENCODE claim seemed to upend that.

The backlash was fierce, and the core objection was about what “function” meant. Critics argued that ENCODE had used the causal role definition in an extremely loose way: if a piece of DNA showed any detectable biochemical activity at all, like being transcribed into RNA even briefly, it was labeled functional. Under an evolutionary definition, much of that activity might be meaningless noise, the molecular equivalent of a machine vibrating without doing useful work. One prominent critique called the 80 percent figure an “absurd conclusion” reached by applying the causal role definition of function inconsistently across different biochemical properties.

3PubMed Central. On the immortality of television sets: “function” in the human genome according to the evolution-free gospel of ENCODE

This wasn’t just a philosophical squabble. If 80 percent of the genome is functional, then mutations almost anywhere could potentially cause disease, and the medical research agenda shifts accordingly. If only a small fraction is functional in the evolutionary sense, the research priorities look very different. The definition of “function” was steering real-world decisions about where to aim scientific resources.

When Functions Shift Over Time

One of the reasons the selected effects definition gets complicated is that natural selection doesn’t plan ahead. A trait can be shaped by selection for one purpose and then get co-opted for something entirely different. The classic concept here is exaptation, a term coined by Stephen Jay Gould and Elisabeth Vrba to describe features that now serve a role different from the one they originally evolved for. Feathers, for example, likely evolved first for insulation or display and were later co-opted for flight. Gould and Vrba argued that calling flight the “function” of feathers obscures this history, so they proposed restricting “adaptation” to features built by selection for their current role, and labeling everything else an exaptation.

4Paleobiology. Exaptation—a Missing Term in the Science of Form

This idea connects to a broader critique of what Gould and the geneticist Richard Lewontin called the “adaptationist programme,” the tendency to assume that every biological feature must have been sculpted by selection for some purpose. Their famous 1979 paper pointed out that many traits exist as side effects of development or physical constraints, not because they were selected for. A trait that looks functional could be a structural byproduct, what they called a “spandrel,” borrowing the term from architecture.

5PubMed. The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme

Vestigial organs sit in a related gray zone. Structures like the human appendix or wisdom teeth are often described as functionless leftovers, but research consistently finds that many vestigial structures still do something biologically useful, just not what they originally evolved for. The appendix, for instance, appears to play a role in immune function and may serve as a reservoir for beneficial gut bacteria. A recent review of the scientific literature on vestigiality concluded that vestigial organs should not be interpreted as useless, since many still perform important biological roles.

6INTERNATIONAL JOURNAL OF MULTIDISCIPLINARY RESEARCH AND ANALYSIS. Study of Vestigial Organs: Disuse and Evolution of the Human Body

So a vestigial organ has lost its original function, may have gained a new one, and still exists in a reduced form. Under the selected effects view, its “proper function” might be the ancestral one. Under the causal role view, its function is whatever it’s doing right now. Depending on which definition you use, the same structure can be labeled functional or non-functional.

One Protein, Multiple Jobs

The problem of assigning a single function to a biological part gets even messier at the molecular level. Many proteins moonlight, meaning they perform completely different jobs depending on where they are in the cell, what other molecules are nearby, or what developmental stage the organism is in. A protein might act as an enzyme in one context and a structural component in another, or it might regulate gene expression inside the nucleus and serve as a signaling molecule on the cell surface.

7PubMed. Moonlighting Proteins

These aren’t proteins that do the same basic thing in slightly different ways. Moonlighting proteins perform multiple independent cellular functions within a single chain, switching roles based on factors like cell type, location, binding partners, and whether they’re working as a single unit or clustered together.

8PubMed Central. Computational characterization of moonlighting proteins

Moonlighting is common enough that it complicates any attempt to assign “the” function of a gene or protein. If you knock out a gene to study what it does and observe a defect, you might be seeing the loss of one function while remaining unaware of two or three others that only manifest in different tissues or at different times during development. The practical upshot is that gene-function databases are mapping one-to-many relationships rather than simple one-to-one pairs.

Different Parts Doing the Same Job

Moonlighting is about one part with many functions. The flip side, called degeneracy, is about structurally different parts that can all perform the same function. Your immune system is a prime example: many different combinations of cells and molecules can mount a successful defense against the same pathogen. If one pathway is knocked out, others can partially compensate.

9PubMed. Measures of degeneracy and redundancy in biological networks

Degeneracy is not the same as redundancy. Redundancy means identical backup copies doing the same thing. Degeneracy means structurally different components that overlap in what they can do without being identical. This partial overlap is thought to be a fundamental source of robustness in biological systems, because it means the system can tolerate damage or mutation without collapsing. It also appears to be tied to how complex systems evolve in the first place, because having components that partially overlap in capability creates more pathways for natural selection to explore.

10PubMed Central. Degeneracy: a link between evolvability, robustness and complexity in biological systems

For anyone trying to assign functions to biological parts, degeneracy is a headache. If three different sets of neural circuits can all produce the same behavior, what is “the” function of each individual circuit? The answer might be that assigning a single clean function is the wrong goal, that the circuits are better understood as overlapping contributors to a flexible system.

Function in Medicine and the Definition of Disease

The concept of function also underpins how medicine defines disease. One influential framework, the biostatistical theory developed by philosopher Christopher Boorse, holds that disease is essentially a failure of some internal mechanism to perform its biological function. Under this theory, “health” is statistically typical functioning relative to a reference population, and disease is atypically diminished functioning.

11PubMed Central. Reference-Class Problems Are Real: Health-Adjusted Reference Classes and Low Bone Mineral Density

This sounds straightforward, but it ties the entire concept of disease to whatever definition of “function” you’re using. If function means what a part was selected to do, then a condition is a disease only if it impairs a trait’s evolved purpose. If function means what a part typically does in a healthy population, disease is a statistical departure from the norm. The two approaches can disagree. Consider aging: many of the changes that come with getting older involve declines in biological function. Under a strict reading, aging is a disease. Under a more contextual reading, it’s a normal part of the lifecycle, and the “reference class” problem of deciding what counts as typical functioning for a given age and sex becomes genuinely tricky.

12PubMed. The biostatistical theory versus the harmful dysfunction analysis, part 1: is part-dysfunction a sufficient condition for medical disorder?

These aren’t abstract puzzles. Insurance coverage, disability determinations, and drug approval decisions all rely on some working notion of normal function. Whether low bone mineral density in a 75-year-old woman is “disease” or “aging” depends on whether you define the expected function of bones by the 75-year-old reference class or the 30-year-old reference class.

How Gene Databases Handle the Problem

Bioinformatics needed a practical solution to the ambiguity of “function,” and the Gene Ontology (GO) project is the most widely used one. Rather than trying to settle the philosophical debate, GO sidesteps it by splitting function into three separate categories: molecular function (what does the gene product do at a biochemical level, like binding DNA or catalyzing a reaction), biological process (what larger-scale activity does it contribute to, like cell division or immune response), and cellular component (where in the cell does it operate).

13PubMed. Linking molecular function and biological process terms in the ontology for gene expression data analysis

The Gene Ontology provides a structured, species-neutral vocabulary so that researchers working on yeast, mice, and humans can annotate gene products in the same standardized way. The project’s explicit goal is to unify how function is described across organisms.

14PubMed Central. The Gene Ontology’s Reference Genome Project: a unified framework for functional annotation across species

This three-way split is a pragmatic workaround, not a theoretical resolution. A protein annotated with one molecular function might participate in a dozen biological processes. And the GO categories don’t capture whether the function is an evolved adaptation, an exaptation, or a byproduct. The system is useful precisely because it doesn’t try to resolve the deeper debate; it just organizes observable data in a consistent format.

The Language of Purpose in Biology

Talking about function in biology inevitably sounds like you’re attributing purpose to things that have no intentions. “The heart is for pumping blood” sounds a lot like “the hammer is for driving nails,” as if someone designed the heart with that goal in mind. Biologists are well aware of this linguistic trap. The technical term for this kind of goal-directed language is teleology, and biology has a complicated relationship with it.

One attempt to resolve the tension came from the concept of teleonomy, introduced in the mid-twentieth century to distinguish legitimate, program-driven goal-directedness (like a bird migrating south guided by inherited behavioral programs) from the illegitimate sense that nature has conscious goals. Ernst Mayr, one of the towering figures of evolutionary biology, championed the term. For Mayr, the key question was whether an apparently goal-directed process was governed by a program or code of information, like DNA, rather than by some mysterious inner purpose.

15PubMed Central. Teleonomy: Revisiting a Proposed Conceptual Replacement for Teleology

In practice, most working biologists use teleological language as a convenient shorthand. Saying “the function of hemoglobin is to transport oxygen” doesn’t commit anyone to believing hemoglobin has a purpose in the way a fire escape has a purpose. It’s an epistemological tool: biologists view a structure as a means to an end because doing so generates productive hypotheses and experiments, not because they think nature is planning anything. The distinction between biological function and artifact function matters here. An artifact like a fire escape has a function because someone designed it with that intention, whether or not it’s ever actually used. Biological function depends on actual use and evolutionary history, not on anyone’s intent.

16Design Studies. Function and forethought in design

Temporary Structures and Developmental Function

Not all biological functions are permanent. During embryonic development, structures appear, serve critical roles, and then disappear entirely before birth. In the developing human brain, for example, a layer called the subplate acts as a “waiting compartment” where incoming nerve fibers form temporary connections before the permanent wiring of the cortex is ready. The subplate guides growing nerve fibers to their correct destinations and helps establish transient neural circuits that are essential for the formation of adult brain connections.

17PubMed. Transient structures of the human fetal brain: subplate, thalamic reticular complex, ganglionic eminence

These transient structures have clear functions during a narrow developmental window, then are largely dismantled. Their function is scaffolding: they enable the construction of the permanent system and then become unnecessary. This is yet another layer of complexity in the question of what “function” means, because the function exists only in a specific time window. A structure can be critically functional for a few weeks of fetal development and entirely absent in the organism it helped build.

Engineered Function in Synthetic Biology

Synthetic biology introduces a context where biological function and artifact function collide. When researchers design synthetic gene circuits in plants or microbes, they think about function in engineering terms: sensors that detect signals, processors that transform information, and actuators that produce an output. A typical synthetic circuit is built from three modules, each defined by what it does within the engineered system.

18PubMed Central. The design of synthetic gene circuits in plants: new components, old challenges

In this context, function genuinely is assigned by a designer with forethought, much like an artifact. A synthetic gene circuit has a function because someone built it to do something specific. But the biological components used in that circuit, the promoters, the regulatory proteins, the signaling molecules, all have their own evolutionary histories and their own natural functions that may conflict with or complicate the engineered purpose. A protein co-opted for a synthetic circuit might moonlight, performing its natural function in ways the designer didn’t intend. The overlap between designed function and evolved function in engineered living systems is a practical challenge, not just a philosophical curiosity. Unintended interactions between a component’s engineered role and its native biological role can cause synthetic circuits to behave unpredictably.

Function Across Scales

One of the reasons “function” is so slippery in biology is that it operates at every scale simultaneously. A single molecule has a biochemical function. That molecule participates in a cell-level process. That cellular process contributes to a tissue-level function, which contributes to an organ-level function, which contributes to the organism’s survival and reproduction. At the level of ecosystems, the concept stretches further still: the processes by which ecosystems interact with human well-being are sometimes framed as “ecosystem services,” a functional language applied to entire ecological systems rather than individual organisms.

At each scale, the meaning of function shifts subtly. A molecule’s function might be fully captured by its chemistry. A heart’s function involves physics, anatomy, and physiology. An ecosystem’s function involves interactions among thousands of species and their environment. The word stays the same, but the kind of explanation it invokes changes with the level of organization you’re looking at. This is part of why the debate over what “function” means in biology has lasted for decades and shows no sign of resolution. The concept is doing too many different jobs across too many contexts to have one tidy definition, and biologists working at different scales routinely use the word to mean different things without realizing they’re talking past each other.

Cognitive science adds yet another wrinkle. The idea that the same mental function, such as recognizing a face or making a decision, might be carried out by different physical arrangements of neurons in different individuals or species has been a major argument for studying the mind at a functional level rather than purely at the level of brain wiring.

19Philosophy of Science. Multiple Realizability Revisited: Linking Cognitive and Neural States

If a psychological function can be “realized” in very different neural hardware, then understanding the function and understanding the physical parts are related but separate projects. This echoes the degeneracy problem seen at the molecular and systems levels: different structures, same outcome. “Function” in this sense is defined by what the system accomplishes, not by what it’s made of.