When Does Something Begin to Think?
On cells, intelligence, and the words we use when nature surprises us
I have long thought of the brain as the place where thinking takes place.
It seems almost self-evident. The brain receives sensory impressions, processes information, remembers, evaluates, and enables us to act. When the brain is damaged, memory, language, and our ability to orient ourselves in the world may also be impaired. We have therefore learned to associate thinking with the nervous system, and above all with the brain.
Then I read about research that unsettles this picture.
Not because researchers have discovered a new part of the brain, but because they are studying organisms and cells that have no brain at all.
And yet they do things that resemble what we normally associate with thinking.
They learn. They adapt. They solve problems. They preserve traces of what has happened before.
And so the question arises:
When does something begin to think?
Life Without a Brain
In February 2024, science journalist Rowan Jacobsen wrote in Scientific American about research into what is often called basal cognition. In the print edition, the article was titled Minds Everywhere. The online version carried the more provocative headline Brains Are Not Required When It Comes to Thinking and Solving Problems—Simple Cells Can Do It.
The starting point is simple, but the implications are not.
Researchers are asking whether processes we associate with cognition—learning, memory, information processing, problem-solving, and goal-directed behavior—may also exist in biological systems far simpler than organisms with advanced nervous systems.
One of the central researchers in this field is Michael Levin at Tufts University. His laboratory investigates, among other things, how cells communicate and coordinate their activity during development and regeneration. Bioelectrical signaling between cells plays an important role in this work.
Levin draws on concepts from cognitive science when describing such processes. Groups of cells can be understood as systems that process information and respond to change in ways that maintain or restore biological form.
At this point, a philosophical question already appears.
What does a problem mean to a cell?
A Flatworm Loses Its Head
The most striking example in Jacobsen’s article concerns planarians—small flatworms with a remarkable capacity for regeneration.
In 2013, Tal Shomrat and Michael Levin published a study in the Journal of Experimental Biology. They investigated learning and memory in the flatworm Dugesia japonica.
The experiment was more interesting—and more nuanced—than the popular formulation of a worm that “remembers after losing its head.”
The worms were familiarized with a particular experimental environment, including a textured surface and regular access to food. The researchers were later able to show that this familiarity was retained for at least fourteen days.
They then removed the heads of some of the worms in a way that ensured the brains were gone.
The flatworms regenerated new heads and new brains.
When the regenerated animals were later tested directly, the difference between previously trained and untrained worms was not statistically significant.
But the experiment had an intriguing continuation.
The worms were given a brief re-exposure to the environment before being tested again. Those descended from previously trained animals then showed a shorter latency before beginning to feed than the control group. The researchers described this as a savings paradigm: previous learning appeared to make relearning easier.
This is an important distinction.
The experiment does not simply show that a complete memory remained stored somewhere in the body while a new brain grew. It shows something more puzzling: the results are consistent with the possibility that a trace of earlier experience survived the removal of the brain and later influenced learning in the regenerated organism.
The researchers discussed several possibilities. Information may have been preserved outside the brain, perhaps in the peripheral nervous system or through other cellular mechanisms. The new brain may also have been influenced by tissue that survived the procedure.
We do not know for certain.
And that is precisely what makes the experiment interesting.
What, exactly, survives?
A memory?
A trace?
An altered biological readiness?
Or something for which we do not yet have a good word?
The Problem Lies in the Words
It would be easy to draw a dramatic conclusion:
Cells think.
But the moment I write the sentence, the problem appears.
What do I actually mean by think?
If an organism changes its behavior on the basis of previous experience, we tend to call it learning. If that experience affects later behavior, we speak of memory. If the organism finds a way around an obstacle, we may describe it as problem-solving.
The words come from our own experience.
I know what it means to remember a summer day from childhood. I know the experience of struggling with a problem, putting it aside, returning to it, and suddenly seeing something I had not seen before. I also know that there is a difference between repeating a correct answer and understanding what the question means.
But when I use the word memory about a flatworm, am I speaking about the same thing?
Probably not.
That does not mean the word is wrong.
It means that we need to examine what the word is doing.
An Old Hermeneutic Question
Here biology suddenly becomes hermeneutics.
We encounter something we do not fully understand. In order to approach it at all, we must use concepts we already have.
Gadamer called these our prejudices—not primarily in the sense of unreasonable or morally objectionable opinions, but as the preliminary understandings we always bring with us when we encounter something new.
We never begin from nothing.
When a researcher says that an organism “solves a problem,” the expression already has a history. Problems are something human beings know. We perceive a situation, encounter an obstacle, consider possibilities, and try to reach a goal.
Then we observe an organism encountering an obstacle, altering its activity, and reaching the outcome nonetheless.
We say:
It solved the problem.
That may be a precise and useful description. But the description is not the phenomenon itself.
Between the phenomenon and the words we use about it, there is always an interpretation.
That is true of science as well.
Does a Cell Have a Goal?
Biology is full of the language of purpose.
A wound “heals.” An organism “tries” to maintain balance. Cells “cooperate” to build an organ. The immune system “recognizes” a threat.
At the same time, we know that such formulations do not mean that cells gather for deliberation or imagine a future outcome.
And yet something happens that can reasonably be described as coordinated and goal-directed activity.
It is tempting to choose between two extremes.
One says that everything is merely mechanics, and that talk of goals, choices, and problem-solving is human decoration.
The other lets goal-directed behavior become evidence that cognition exists almost everywhere life exists.
But we do not need to choose so quickly.
A biological process can be goal-directed without the organism representing the goal to itself. Information can be processed without our thereby having shown that anyone understands the information. A biological system can preserve traces of past events without this being memory in the same sense as when I remember something from my own life.
The differences do not become less important.
They become clearer.
From the Cell to Artificial Intelligence
It is difficult to read this without thinking about artificial intelligence.
Here too we use words that were once closely associated with human beings.
The machine learns.
It remembers.
It reasons.
It solves problems.
It answers.
At times, we even say that it understands.
But what have we actually established when we use these words?
That the system performs certain operations?
That it produces results resembling those human beings can produce?
Or that there is also something corresponding to the human experience of understanding?
These are different questions.
A calculator can arrive at an answer that I cannot work out in my head. I would not therefore say that the calculator understands mathematics.
With a language model, the distinction becomes harder to hold on to, because the activity resembles our own more closely. It uses language, answers questions, corrects itself, and participates in something that has the form of a conversation.
The question then easily shifts from:
Does it think?
to:
Does it understand me?
And from there to another question:
Can it meet me?
Here Martin Buber’s distinction between I–It and I–Thou becomes relevant.
Buber’s Thou is not simply a name for something that produces an appropriate response. I–Thou refers to a relation in which the other does not appear merely as an object for my use, analysis, or control.
The question of artificial intelligence therefore cannot be settled simply by asking how intelligent an action appears.
A language model can perform something that resembles a conversation.
Whether it can thereby enter into what Buber calls an I–Thou relation is another question.
The flatworm, the cell, and artificial intelligence each pose, in different ways, the same problem for us:
What do we mean by the words we use?
Are We Asking the Wrong Question?
“Can a cell think?” is a crude question.
It invites a yes or a no.
But nature is not obliged to organize itself according to our binary distinctions.
We can ask more concrete questions.
What can the organism register?
What can it preserve from previous experience?
How does its behavior change when the environment changes?
Can it reach the same outcome through different pathways?
Can it correct itself when something unexpected happens?
And at what point does it become meaningful to use words such as learning, memory, problem-solving, or cognition?
We then move away from the question of whether the organism “really” thinks and closer to what can actually be investigated.
The philosophical question does not disappear.
It becomes clearer.
The Boundary Moves
The history of science is also a history of boundaries being moved.
Human beings have repeatedly had to surrender the idea that qualities we value in ourselves exist only in us.
Animals feel pain, learn, and enter into complex relationships. Birds can solve problems. Octopuses display remarkable behavioral flexibility. And organisms with very simple nervous systems—or none at all—show forms of learning, information processing, and adaptation that make old boundaries less secure.
That does not mean that all differences disappear.
Quite the opposite.
The more we learn about different forms of information processing, learning, and adaptation, the more important it becomes to distinguish carefully.
A cell is not a human being.
A biological trace in a flatworm is not my memory of a summer day.
A language model that answers me is not therefore a Thou in Buber’s sense.
But differences do not become clearer by denying similarities.
They become clearer when we examine what the similarities consist of—and where they end.
Letting the Question Remain Open
I began with an apparently simple question:
When does something begin to think?
After reading the research, I am less certain what the question means than I was before.
I do not regard that as a defeat.
Quite the opposite.
It may be a sign that something has happened to understanding.
Before, I thought I knew roughly where the boundary lay. Thinking belonged in the brain. Without a brain, no thinking.
Now the boundary has become less clear.
Not because the cell has suddenly become a little human being, but because life turns out to be capable of registering, preserving, coordinating, and adapting in ways that make our old categories less certain.
It may be that what we call thinking has deeper biological roots than we previously assumed.
It may also be that one word—thinking—conceals several different phenomena that we now need to learn to distinguish.
For sometimes philosophy does not begin when we find the answer.
It begins when a word we thought we understood is no longer quite so self-evident.
References
Jacobsen, R. (2024). Minds everywhere. Scientific American, 330(2), 44–51. https://doi.org/10.1038/scientificamerican0224-44
Shomrat, T., & Levin, M. (2013). An automated training paradigm reveals long-term memory in planarians and its persistence through head regeneration. Journal of Experimental Biology, 216(20), 3799–3810. https://doi.org/10.1242/jeb.087809
The history of science is also a history of boundaries being moved.