How does nature make decisions?

and what that says about how we treat patients

A leaf on a Tuesday

A tree drops a leaf. The leaf has been attached all summer, photosynthesising, in place. Then on a Tuesday in late October, with the abscission layer finally complete, the leaf falls. Nobody told it to. The tree has no nervous system. The leaf has no nervous system. Something committed.

A flower opens the same way. The bud is held closed by pressures and tensions that have been maintained for days. Cells on one surface press harder than cells on the other. Walls loosen in specific places. Then on some morning, when light and temperature cross a threshold, the petals separate. A bird lays an egg. A wound starts closing. A seedling breaks the soil. None of these systems has a brain. Each one decides.

How does that work. What is the mechanism by which biology decides things, without anyone deciding.

The answer turns out to be the same wherever you look. The same architecture appears in a T cell, in a falling leaf, in a synaptic vesicle, in a fly laying an egg. Different subfields have given it different names. The shape is one shape.

What follows is a description of that shape, why it matters for medicine, and what it asks of clinicians who work with bodies that decide.

Integration to threshold

The cleanest place to look is inside a lymph node.

A dendritic cell is an immune sentinel that lives in tissue. When it encounters something that looks dangerous, it picks the material up, processes it, and migrates through lymphatic vessels to the nearest lymph node. By the time it arrives, it is carrying fragments of the dangerous thing on its surface, displayed on receptors called MHC. It has become a delivery system for a report.

T cells in the lymph node sample. Each T cell carries a receptor that recognises one specific shape. They wander, touch dendritic cells, check what is being displayed, move on. Most touches mean nothing.

When the receptor does fit, something happens that is not yes-or-no on contact. The T cell starts a slow process inside itself. Each successful engagement triggers a small downstream signal. The signals accumulate. After enough contacts, sustained over enough time, the accumulated signal crosses a threshold. The cell commits to clonal expansion. It will spend the next two weeks producing thousands of copies of itself. The commitment is irreversible. The cell's epigenetics change. It cannot return to its naive state.

Byron Au-Yeung and colleagues at UCSF measured this carefully in 2014. The threshold is sharp and invariant. Below it, the cell does nothing. Above it, the cell enters a transcriptional programme it cannot leave. What matters is the total accumulated signal crossing the line.

This is the moment of decision. It is the licensing of a T cell to fight.

The same architecture appears in many other places. A neuron integrates postsynaptic potentials in its soma until the membrane voltage crosses a threshold, at which point an action potential fires. A cell in apoptosis accumulates cleaved caspase-3 until it crosses a threshold set by anti-apoptotic proteins, after which it proceeds inexorably to death. A fly preparing to lay an egg accumulates a calcium signal in a small set of descending neurons, with the signal modulated by the quality of available substrates, until it hits a threshold and the fly commits. The leaf does it through cell-wall changes in the abscission layer. The flower does it through turgor pressure.

Integration to threshold. Information accumulation. Temporal summation. Rise-to-threshold decision. The biology has many names. They are describing one machine.

The machine has four parts and they are not all the same kind of thing.

The first is a capacitor. A chamber that loads with substrate against a threshold. In the T cell, the accumulating signal. In the leaf, the cell-wall reorganisation. In the neuron, the rising membrane potential. The capacitor's job is to hold reserve and fire when the threshold is met. Pool dynamics: inflow, outflow, discharge events. Failures look like sub-threshold loading or threshold drift.

The second is a factory. Once the capacitor fires, the substrate enters a build phase. Parallel production lines with quality control. The T cell that committed to clonal expansion now divides into specialised effector subtypes, each one selected for survival against tumour conditions. The leaf detaching is being shaped by the wind into its fall trajectory. The factory's distinctive mechanism is competitive selection, a survival race rather than transformation. Failures look like throughput drop, skewed differentiation, low yield.

The third is a reactor. The built substrate meets its target across many parallel sites. Each meeting runs an A + B reaction with some rate, yield, and balance. The trained T cell engaging the tumour cell at the synapse. The fertilising sperm reaching the egg. The reactor's observables are reaction rate, yield, and the net economic balance — does the meeting produce more substrate than it consumes. Failures look like low rate (a brake on the synapse), low yield (binding insufficient), negative balance (the reactor net-consumes substrate it should net-produce).

The fourth is a homeostat. The work of the meeting consolidates into a setpoint held against drift. Memory T cells form. The leaf decomposes and returns substrate to the soil where new abscission layers will form next year. The downstream neuron accumulates its own potential. The homeostat's job is to detect drift and apply corrections, and critically, the correcting activity itself generates the next cycle's perturbation. Failures look like setpoint loss, patrolling gaps, over-correction, non-generative handoff.

These four engines are the four regimes of a dissipative cycle. We call them Potentiality, Construction, Encounter, Conservation. They run in that order. Each one outputs what the next one needs as input. The capacitor's discharge is what the factory builds with. The factory's product is what the reactor reacts. The reactor's yield is what the homeostat conserves. The homeostat's drift is what loads the next capacitor. The cycle closes by the last regime seeding the first.

The whole cycle is the decision. Not one of the engines. All four, in sequence, running once.

A dendritic cell does not arrive at the lymph node carrying generic antigen. It arrives carrying antigen captured at a specific tissue site, displayed through specific MHC alleles, in the context of specific danger signals from that tissue. The substrate the T cell binds to is all of that at once. The cell is not reading a representation of tissue events. It is participating, physically, in the same molecular material that captured them. The volume of substrate is the running judgment about whether to commit. The crossing is the verdict.

There is no measurement step. The substrate is the signal. When it meets the threshold, the commitment fires.

Two times in one decision

Decisions of this kind look instantaneous from outside. The leaf prepares through September and October, then on a Tuesday in late October it falls in three seconds. A T cell deliberates for hours and then commits in a moment. Days or weeks of nothing, then a binary event. But the slow part and the fast part are not the same kind of time.

Before the crossing, duration is whatever the substrate takes. A leaf might be ready in mid-October or late November. A T cell might commit in three hours or in twelve. There is no schedule. The arrival is when the substrate says it has arrived.

After the crossing, the cycle runs on clock time. Clonal expansion follows a schedule of hours per division and days per generation. The detached leaf falls at terminal velocity. Substrate that took an unknown amount of time to load is now being processed at a known rate.

The waiting was the work. From outside it looked like nothing was happening because clock time was not yet running. The crossing is the starter gun. Before it, silence; after it, the clock runs and everything downstream runs with it.

This asymmetry has a cost in clinical practice. Modern medicine measures in clock time — endpoints at six weeks, twelve weeks, six months. Drugs are scored by what happens between two timepoints on a calendar. When the intervention is operating in the loading phase, where clock time is not yet running, the measurement and the work are in different times. The cost shows up in trials. Patients whose loading was slower than the trial allowed get counted as non-responders. The drug worked. The clock ran out.

Two decisions in one cancer

Take cancer, where the temporal mismatch above has been most consequential, and where the cycle requires not one decision of this kind but two. The body's defence against cancer requires two decisions, days apart, in the same cycle.

The first happens in the lymph node that drains the tumour. The same scene as earlier in this piece: a dendritic cell arrives, this time carrying fragments of the tumour. A T cell that recognises the fragments begins the slow integration — touch, signal, accumulate. When the signal crosses threshold, the T cell commits. Over the next two weeks it divides into a small army of T cells trained to recognise this tumour.

The second decision happens days later, at the tumour itself. One of those trained T cells meets a tumour cell displaying the same fragment. The T cell does its own integration — engagement, accumulation, threshold. When it crosses, the cell commits to the kill. The tumour cell dies.

Two decisions, on the same substrate, days apart. The first loads the second. Without the first, no trained T cells exist; without the second, the trained T cells have nothing to act on. Both have to fire, in order, for the body to clear the cancer by itself.

What we call cancer is, in many cases, one of these two decisions failing to fire. The tumour has learned to interfere. Sometimes it suppresses the dendritic cell before it can reach the lymph node — the chamber never loads. Sometimes it installs a brake on the T cell so the accumulated signal cannot cross — the chamber loads but does not fire. Sometimes both decisions complete and the trained T cells cannot find their target because the tumour has hidden the fragment they were trained against — the second decision has nothing to meet. In each case the machinery is intact. What has failed is the body's capacity to decide.

This changes what the clinician is doing. The clinician is not killing the tumour. The clinician is removing whatever the tumour installed to stop the body from clearing it. The body does the killing. The drug is the lockpick on the brake.

The clearest example is the family of drugs called PD-1 blockers. PD-1 is the brake the tumour exploits: it sits on T cells, and when the tumour engages it, the activation threshold rises so high that the accumulating signal can no longer cross. The chamber loads correctly, the report is assembled, but the threshold has been artificially lifted. The drug binds to PD-1 and removes the engagement. The threshold drops back. The signal crosses. The first decision fires. The trained T cells set out. Days later, at the tumour, the second decision fires. The body clears the disease itself.

The timing of this drug matters in a way that is easy to miss. The clinician can give it before the tumour is removed by surgery, or after. Given before, both decisions have what they need: the lymph node still sees the tumour, and the trained T cells still have a tumour to meet at the second decision. Given after surgery, the first decision still fires. But the trained T cells set out into a body whose tumour has already been removed. They have nothing to encounter. The cycle stalls for lack of substrate. Same drug, different time, different outcome.

Four ways a cycle fails

What cancer shows is one shape of failure: a brake installed on one of the engines so the cycle cannot complete. There are four such shapes, because there are four engines, and the diagnostic question for any patient is not what disease they have. It is which engine has failed, and what kind of help that engine needs.

The capacitor undercharges. The body wants to commit, but the substrate is not there. The decision is structurally unavailable. The lymph node has no loaded T cell response. The marrow has no neutrophils to release. The clinical signature is a body that should be acting but is not, where the absence is a vacancy. This is a Potentiality failure. The orientation is to load the chamber. Restore substrate. Protect the source that feeds the chamber. Wait. Trilaciclib protecting marrow stem cells before chemotherapy is this kind of work. Prehabilitation before surgery is this kind of work. The clinician does not push. The chamber takes the time it takes.

The factory cannot build. The capacitor fired, but what came out of the chamber is not being shaped into deployment-ready substrate. Throughput is low, differentiation is skewed, yield drops. T cells committed to clonal expansion but the expansion is producing exhausted or misdifferentiated effectors. The clinical signature is loaded readiness that does not convert into action. This is a Construction failure. The orientation is to find what is jamming the build line and remove only that. Anti-PD-1 in a patient whose factory is being held back by tumour-installed brakes is this kind of work. CTLA-4 blockade removing a different brake on the same build line is this kind of work. IL-2 muteins providing expansion substrate that the patient's factory is starved of is this kind of work. The intervention is precise. Identify the block. Remove it.

The reactor cannot run. The factory built the effectors, but the meeting at the target is not running. Reaction rate is low, yield is low, the balance is negative. The trained T cells circulate but cannot infiltrate the tumour. The antibody is in the bloodstream but the antigen has been internalised. The clinical signature is action without contact. This is an Encounter failure. The orientation is to stage the meeting. Bring substrate and target into reach. Restore the reaction conditions. Neoadjuvant timing keeps the tumour in place so trained effectors have something to meet. Engineered T cell receptors in adoptive cell therapy build a substrate that can find a target the patient's own immune system was missing. A surgeon repositioning a fracture is the same orientation at a different scale. The intervention is structural. Where are the substrates, where is the target, what is keeping them apart.

The homeostat cannot hold. The reactor produced its yield, but the residue is not consolidating into a setpoint that the next cycle can build on. The kill happened but no memory formed. The wound healed but function did not restore. The response will not return when the next demand arrives. The clinical signature is a successful cycle that produces no substrate for the next one. This is a Conservation failure. The orientation is to support consolidation. Memory T cell support after a complete response. Rehabilitation after surgery. Setpoint protection in chronic disease. The intervention is temporal. What just happened needs to become substrate for what comes next.

Four engines, four failures, four orientations. The disease tells the clinician what the substrate is. These four questions tell the clinician which engine has stalled. The orientation tells the clinician what kind of work meets the stall.

A patient with metastatic disease might need all four at different points in the same illness. Capacitor work to restore the marrow after chemotherapy. Factory work with anti-PD-1 to remove the block on the build line. Reactor work with neoadjuvant timing to stage the meeting. Homeostat work to support memory formation against recurrence. The disease is one. The cycle is one. The four engines are four.

Look closely at any of these four failures and the same thing is happening one level down. A T cell that has crossed its licensing threshold but cannot infiltrate the tumour looks like a reactor failure at the immune-cycle scale. Zoom in to the cell's own machinery and the chemokine-receptor programme that should drive tissue infiltration has not loaded enough to fire against the local gradient. The reactor failure at one scale is a capacitor failure at the next scale down. The memory that fails to form after a successful kill is a homeostat failure at the cycle scale, but at the cell scale it is the differentiation programme that did not load its substrate to threshold. Look one level deeper at any cycle stall and the failure has the same shape one level down. Four engines all the way down, with the bottom of each cascade always pointing back to a capacitor that did not fire.

This is the framework's structural claim. Every regime, at every scale, is itself a four-engine cycle. The failure in any one of them traces, at depth, to a capacitor somewhere whose substrate did not meet its threshold. The four diagnostic questions above are the surface diagnostic, the level at which a clinician can act. The depth diagnostic is that all four resolve, eventually, to a chamber that did not load. The claim is testable wherever a cycle stall can be resolved one level finer.

The bias in modern medicine

Of these four orientations, modern medicine knows mostly one. A clinician trained in event-based medicine is trained, narrowly, in the Construction orientation. The moment of action. The intervention that acts on the disease, with measurable kinetics, on a clock that can be billed. Insurance reimburses clock time. Trial endpoints measure clock time. Career advancement runs on clock time. The infrastructure of modern medicine evolved to handle problems where the work happens between two timepoints on a calendar, and within that constraint the field has done real work. Bleeding stopped now. Infection cleared in ten days. Fracture immobilised for six weeks.

What the constraint cannot see is the work that happens in the loading, in the staging of the meeting, in the carrying forward. The other three orientations exist in practice, scattered across specialties, but they are systematically undervalued because they do not fit the clock the system uses to measure work. Potentiality work that loads a chamber and waits has no event to point to. Encounter work that times an operation correctly relative to the body's immune cycle does not appear in any procedure code. Conservation work that consolidates residue happens after the headline result and is rarely measured.

The deeper problem is one of honesty. The body is the actual agent of healing. The clinician does not heal the wound. The body heals the wound and the clinician keeps the conditions right for it to happen. The drug does not kill the tumour in cases of effective immunotherapy. The body kills the tumour and the drug removes a brake the tumour had installed. This is a structural claim. Across the cases in which medicine succeeds at chronic and complex disease, the body did the work and the intervention restored the conditions for the body to do the work. The cases in which medicine acts as though the clinician were the agent and the body the object are the cases of acute and mechanical disease. There the framing is honest because the body has been overwhelmed and needs something done for it.

For everything else, the intervention class that names the body as the agent and the clinician as the restorer of conditions is the most honest class available. It is honest because it describes what is happening. The other orientations can work, and often do, but they work by accident from a stance that does not see what it is doing.

This is what the four orientations are for. Each one is a different way the clinician restores conditions. Potentiality work restores conditions for the chamber to load. Construction work restores conditions for the crossing to fire. Encounter work restores conditions for the meeting to happen. Conservation work restores conditions for the residue to carry forward. None of them perform the decision on the body's behalf. All of them make the decision possible. The cycle is the body's. The work is the body's. The intervention is the conditions.

The same shift in leadership

The shift this article describes is not unique to medicine. Anyone who has worked in or near an organisation in the last thirty years has watched the same shift unfold in management. Command-and-control treats employees as objects on which the manager acts. Decide for them, instruct them, monitor them, correct them. The work happens when the manager makes it happen. Servant leadership, empowerment, self-managed teams, and most of what the agile movement and the modern HR literature has argued treats the team as the actual agent of work. The leader's job is to clear blockers, secure resources, provide context, hold conditions, and step back.

Most readers in business recognise this shift. The argument has been made for thirty years. The evidence is in. Organisations that empower the people closest to the work outperform organisations that try to direct the work from above. The cost of command-and-control is not invisible in management the way it has remained invisible in medicine. Teams that have been micromanaged do worse, retain people less, recover from setbacks more slowly. The data is there. The shift is broadly accepted, even where it is incompletely practised.

The shift in medicine is the same shift. Treat the body as the actual agent. Restore the conditions for the body to decide. Cross boundaries only when other options are off the table.

Daniel Goleman's six leadership styles plot onto the four orientations cleanly. Visionary leadership loads the chamber by giving the team a shared sense of where things are going. Coaching is the work of removing what blocks a team member's crossing into action they were already prepared for. Democratic leadership stages the meeting where the team's perspectives contact the problem they were brought together to address. Affiliative leadership consolidates the residue of past work into the relational substrate that loads the next cycle. Four styles, four orientations. Each one is a way the leader restores conditions without performing the work.

Goleman's other two styles, pacesetting and commanding, are the override styles. The leader performs the work themselves, on a clock, sometimes by example and sometimes by direct control. Goleman's own data shows these styles have the worst long-term effects on climate and performance when used as default. Both styles have their place. Commanding is the right move when the building is literally on fire. Pacesetting is the right move when the team is loaded and needs to see what crossing looks like. The mistake is reaching for them by default.

The structural rule is the same in both domains. Override intervention is honest when the system has been overwhelmed and needs something done for it. The override stance is corrosive when used as default in conditions where the system is the only agent that can do the work. Acute disease and emergencies are honest territory for the override. Chronic and complex disease, and any organisational work that depends on people choosing to commit, is not.

Cross boundaries only when other options are off the table.

Back to the leaf

This piece opened with a question: how does biology decide things, without anyone deciding. The answer is that nothing does. A chamber loads, and when it is full enough the cycle that comes next fires by itself. Four engines run in sequence — load, build, meet, hold — each one outputting what the next one needs. There is no decider standing outside the cycle, telling it when to commit. The substrate is the signal. The crossing is the decision.

The leaf works this way. The flower works this way. The T cell, the fly, the neuron, the dividing cell, the closing wound. They are not metaphors for each other; they are running the same shape of process at different scales and in different materials. The leaf-falls and flower-opens you can see, and the cell-divisions and wound-closures and immune commitments you cannot, are the same machinery being expressed at the scales we happen to look at.

The body has been doing this for the patient's entire life. The clinician's job, when it works, is to read which part of which cycle needs help and to support the part that has stopped. The body does the deciding. The body does the discharging. The body does the healing.

The intervention's job is to restore the conditions under which all of that can happen.

---

References

Au-Yeung BB et al. A sharp T-cell antigen receptor signalling threshold for T-cell proliferation. PNAS 2014; 111(35): E3679-E3688.

Groschner LN et al. Dendritic integration of sensory evidence in perceptual decision-making. Cell 2018; 173(4): 894-905.

Mitra A et al. Quantifying information accumulation encoded in the dynamics of biochemical signalling. Nature Communications 2021; 12: 1264.

Roux J et al. Fractional killing arises from cell-to-cell variability in overcoming a caspase activity threshold. Molecular Systems Biology 2015; 11(5): 803.

Vijayan V et al. A rise-to-threshold process for a relative-value decision. Nature 2023; 619: 563-571.