There is a curious habit in the way we talk about evolution.
We know that evolution has no foresight.
We know that natural selection is not a conscious selector.
We know that genes do not formulate intentions, populations do not deliberate, and evolution does not sit somewhere deciding what organisms should become.
And yet, when we describe evolutionary processes, agents keep appearing.
Genes want to replicate.
Organisms try to survive.
Selection chooses.
Evolution finds solutions.
Nature favours one form over another.
Adaptations solve problems.
The language is so familiar that we scarcely notice what has happened.
A process has occurred.
We have described it as though someone did it.
This is not necessarily a mistake. Much of the language is useful shorthand, and some of it captures real asymmetries in biological processes. But there is a conceptual danger in allowing grammatical convenience to become an ontology.
The problem is therefore not that evolutionary biology uses verbs of action.
The problem is that we can forget that an Actor in a clause is not necessarily an agent in the world.
That distinction will turn out to matter rather more than it first appears.
The gene that does things
Consider the simple sentence:
The gene replicates.
It sounds wonderfully straightforward.
There is a noun, the gene, followed by a verb, replicates. The gene appears to be the one doing the replicating.
And, within the SFL transitivity model, that is indeed a perfectly legitimate analysis: the gene is construed as Actor in the process.
But something very different happens if we ask a philosophical question:
Is the gene therefore an agent?
Not necessarily.
The transitivity category Actor describes how a participant is construed within a clause. It does not establish intentional agency.
And the distinction becomes even clearer if we look at the ergative model. The same process can be construed in a way in which the gene is Medium.
That is not a trivial technicality.
It reminds us that a process can be represented from different perspectives, and that the grammatical participant which appears to "do" something need not correspond to a metaphysical doer.
The sentence:
The gene replicates
can therefore be perfectly good biology without committing us to the proposition:
The gene is an agent that seeks its own reproduction.
The grammar does not entail the philosophy.
Yet the grammatical construal can make the philosophical inference feel natural.
That is precisely why it deserves attention.
Actor is not agent
This distinction gives us a useful piece of conceptual hygiene.
An Actor is a grammatical participant in a transitivity configuration.
An agent, in the philosophical sense relevant here, is something capable of initiating or directing action in a way that involves some form of agency.
The two notions overlap in ordinary discourse often enough to be confused.
But they are not equivalent.
A falling stone can be Actor in an appropriately constructed clause.
A chemical substance can be Actor.
A gene can be Actor.
A river can be Actor.
None of these facts, by themselves, establish that the participant possesses intentions, purposes or representations of possible futures.
This is why the distinction matters particularly in evolutionary biology.
Evolutionary processes are full of causal relations, but causal participation is not identical to agency.
A gene can participate causally in replication without wanting to replicate.
Natural selection can produce differential persistence without selecting anything consciously.
An organism can exhibit behaviour that contributes to its survival without having formulated survival as a goal.
Once this distinction is made, we can retain the explanatory force of action language without allowing it to smuggle in a little homunculus.
The grammatical metaphor of the replicator
The problem becomes even more interesting with the word replicator.
At first glance, replicator looks like a perfectly innocent noun for something that replicates.
But the suffix matters.
English forms an agentive noun by adding -or:
act → actorgenerate → generatorregulate → regulatorreplicate → replicator.
The resulting noun naturally suggests something that performs the action named by the verb.
The linguistic transformation is subtle.
A process has been construed through a noun that appears to designate an entity capable of carrying out that process.
And in the evolutionary context, the term can therefore make the gene look like a little biological operator whose essential activity is replication.
This is where our earlier discussion of grammatical metaphor becomes relevant.
We should be precise.
The term replicator is not simply "a metaphor" in the loose sense. Nor is "conceptual metaphor" an SFL category.
Rather, we can examine the grammatical resources through which a process is reconstrued as a participant-like thing, and then ask what conceptual construal that linguistic choice affords.
The suffix -or is especially suggestive because it packages an action together with an apparent doer.
And once the doer has been named, it becomes very easy to tell a story about what that doer is trying to achieve.
The linguistic sequence can therefore become:
replicate → replicator → something that replicates → something whose interest is replication → something that "wants" to replicate.
Each step may seem harmless.
Together, they can produce a remarkably strong image of agency.
Dawkins and the deliberate provocation
This brings us to Richard Dawkins.
Dawkins did not accidentally stumble into agentive language. He used it deliberately and brilliantly.
The Selfish Gene was designed in part to shift the level at which evolutionary processes were understood.
The gene became the focal point of evolutionary competition.
The language of selfishness was provocative because it captured something real about the differential persistence of hereditary variants while deliberately refusing to equate that persistence with conscious selfishness.
The "selfish" gene does not have desires.
It does not sit in a tiny biological office worrying about its reproductive prospects.
The language was a metaphorical device for making a particular pattern of evolutionary explanation vivid.
And it was extraordinarily successful.
But precisely because it was successful, we need to examine what happened next.
A metaphor that helps us think can eventually become the environment in which we think.
Once selfish gene becomes familiar, it can cease to feel metaphorical.
The gene begins to seem as though it really is the kind of thing that has interests.
The distinction between:
a gene's differential persistence
and
a gene's pursuit of its own persistence
can become blurred.
The first is an evolutionary description.
The second is an attribution of agency.
And the two should not be allowed to collapse into one another.
The selector that isn't selecting
The same problem occurs with natural selection.
Consider:
Natural selection favours longer beaks.
This is a perfectly ordinary evolutionary statement.
But grammatically, natural selection is once again doing something.
It is the Actor.
The verb favours is a transitive process.
The beaks are its apparent target.
We have constructed a little scene:
Selection looks at beaks → prefers some → rejects others.
Yet that is not how the underlying process works.
There is no selector standing outside the population.
There is no criterion being consciously applied.
There is no comparison of alternative designs.
Instead, organisms with certain heritable characteristics have different reproductive consequences under particular conditions.
The population changes accordingly.
"Selection favours" is therefore a useful construal of a statistical and historical process.
But it is not a literal description of a choosing mind.
Again, the point is not to ban the sentence.
Indeed, doing so would be silly.
The point is to be able to move fluently between the construal and the process it construes.
Selection favours X
should remain translatable into something like:
Under these conditions, variants associated with X tend to leave more descendants than relevant alternatives.
The second formulation is less elegant.
It is also much harder to anthropomorphise.
And that is precisely its virtue.
Why the agent keeps returning
Why are we so tempted to restore agency?
Perhaps because agency is one of our most powerful explanatory resources.
When something changes in a coordinated way, we naturally ask:
Who did it?
When something appears to solve a problem, we ask:
Who solved it?
When something is organised for a function, we ask:
Who organised it?
In ordinary human affairs, these questions are often extremely effective.
Someone built the bridge.
Someone wrote the program.
Someone designed the engine.
Someone planted the garden.
Human culture trains us to interpret organised outcomes through the intentions of agents.
Evolution confronts us with a remarkable exception.
Here we have organisation without the organiser.
Function without foresight.
Accumulated structure without a plan.
The explanatory machinery we normally use for artefacts therefore becomes simultaneously useful and dangerous when transferred to biology.
It is useful because agency language can make complex causal relations intelligible.
It is dangerous because the very intelligibility it provides can conceal the absence of an agent.
But organisms really do act
There is an important complication.
We must not respond to this problem by swinging too far in the opposite direction.
Organisms are not rocks.
They move.
They regulate their internal conditions.
They respond to stimuli.
They seek resources.
They avoid threats.
Many organisms learn.
Some anticipate future conditions.
Some construct environments.
So there really are biological agents.
And that makes our conceptual problem more interesting, not less.
The question is not:
Does biology contain agency?
Of course it does, at least in some meaningful senses.
The question is:
Where is agency doing explanatory work, and where has agency been imported into our description of a process that does not require it?
An organism's behaviour may genuinely be agentive in a way that the replication of a DNA sequence is not.
A bacterium can move toward a chemical gradient without requiring us to imagine that the gene causing some component of the chemotactic machinery is itself pursuing a goal.
Different levels of organisation can therefore support different kinds of agency.
The danger is not anthropomorphism alone.
It is level confusion.
The gene is not the organism
This helps explain why the gene has sometimes become such a slippery object in evolutionary thought.
At one level, genes are indispensable causal participants.
At another, organisms are the entities that develop, behave and reproduce.
At another, populations evolve.
At another, ecosystems change.
There is no reason to assume that the participant that is most useful for explaining one process must therefore be the agent of every process.
A gene can be causally important to a phenotype without being an intentional actor.
An organism can behave purposively without every gene within it possessing purposes.
A population can evolve without being a mind.
Natural selection can describe population-level change without becoming a population-level chooser.
The biological world contains nested processes, and our language often compresses those levels.
That compression is convenient.
But it can make the grammar of one level look like the ontology of another.
From grammar to conceptual naturalism
This is where our second conceptual tool enters.
SFL can help us identify the linguistic construal.
But identifying the construal is not yet enough.
We then want to ask a further question:
What does this way of construing the process make easier for us to think?
This is the question we have been calling Conceptual Naturalism.
A conceptual form becomes natural when it fits so comfortably within our existing ways of thinking that we cease to experience it as a particular construal.
"Selection chooses."
"The gene wants."
"The organism solves."
"Evolution finds."
These formulations are cognitively economical.
They compress complicated causal histories into familiar structures of agency.
That is precisely why they are powerful.
But the same economy can hide relations that matter scientifically.
The gene appears independent of the developmental system.
Selection appears independent of the population.
The organism appears independent of its ecological relations.
Evolution appears to possess a direction of its own.
The conceptual environment has changed.
And because the change is so comfortable, we may not notice it.
The danger is not metaphor
It would be tempting to conclude that we should simply eliminate these metaphors.
But that would miss something important.
Scientific thought depends upon productive metaphorical and grammatical resources.
Dawkins' language was powerful precisely because it enabled people to see evolutionary processes from a different perspective.
"Selfishness" made differential replication vivid.
"Competition" captures real population-level asymmetries.
"Selection" captures a genuine pattern in the differential persistence of variants.
"Information" allows biologists to describe remarkably complex molecular processes.
The question is therefore not:
Is this metaphor true or false?
It is:
What does this construal reveal, and what does it conceal?
That is a more demanding question.
And it is one that allows us to preserve the intellectual achievements of evolutionary biology while examining the conceptual machinery through which those achievements become thinkable.
The agency of grammar
There is an irony here.
We began by asking how evolution can produce organised outcomes without an agent.
And we have now discovered that our language repeatedly supplies the missing agent.
Where nature gives us:
processes,
grammar gives us:
actors.
Where nature gives us:
differential consequences,
grammar gives us:
choices.
Where nature gives us:
historical persistence,
grammar gives us:
interests.
The linguistic transformation is not merely decorative.
It can shape the conceptual space in which explanation takes place.
But we should resist saying that language therefore creates the biology.
It doesn't.
Genes replicate whether or not we call them replicators.
Populations change whether or not we say selection "favours" something.
Organisms develop whether or not we describe development as the execution of a program.
The biological processes are there first.
Language gives us ways of construing them.
That distinction will remain important throughout our journey.
The process beneath the actor
Perhaps the most useful habit we can develop is therefore a simple one.
Whenever we encounter an evolutionary sentence containing an apparent agent, ask:
What process would remain if I removed the agentive construal?
Take:
The gene wants to replicate.
We might reconstruct it as:
A hereditary variant has consequences for its persistence in a population.
Take:
Selection favours camouflage.
We might reconstruct it as:
Under particular ecological conditions, organisms possessing camouflage leave more descendants.
Take:
Evolution solves the problem of thermoregulation.
We might reconstruct it as:
Heritable variation affecting thermoregulation can become differentially represented across generations under particular environmental conditions.
The reconstructed formulations are cumbersome.
But something important has happened.
The agent has disappeared.
The relations have become visible.
And the explanation has not disappeared with it.
Yet something is still missing
This is where we must resist congratulating ourselves too soon.
We have removed the imaginary agent.
Good.
But have we explained evolution?
Not yet.
Indeed, we have merely exposed a deeper question.
If nobody is directing evolutionary change, and if no gene is literally pursuing its own replication, then where does the apparent organisation come from?
We know part of the answer.
Variation.
Inheritance.
Differential reproduction.
Selection.
But we still have not examined the things those processes operate upon and through.
What exactly is inherited?
What exactly is being replicated?
What makes one variation developmentally accessible and another not?
What is the relationship between a gene and the organism in which it participates?
What does it mean for something to be a "unit" of evolution?
These questions become increasingly difficult once we stop treating the gene as a little agent.
And that is precisely why the next step matters.
For if the gene is not an agent, perhaps we should look more carefully at what kind of participant it is.
From agency to participation
This gives us the first major turn in the series.
We began with the apparent designer.
We have now found that the designer can disappear from evolutionary explanation.
But we have also found a subtler problem.
When we remove the designer, we are tempted to put agency back into the system in smaller pieces:
selection becomes the chooser;
the gene becomes the strategist;
the replicator becomes the competitor;
evolution becomes the problem-solver.
The vocabulary changes.
The conceptual structure remains.
So perhaps the more fruitful question is not:
Who is the agent?
but:
What participates in the process, and how?
That question is less dramatic.
It is also much more promising.
A participant need not have intentions.
A Medium need not be an agent.
A causal contributor need not be a chooser.
A hereditary structure need not want anything.
And once we begin thinking in those terms, the gene starts to look rather different.
Not as a tiny organism inside the organism.
Not as a little strategist pursuing its interests.
But as something embedded in a much larger organisation of processes.
The gene is not alone.
And that is where we must go next.