What life is — biology between physics and intentionality
Why the question won't take a single answer — and what the partial answers tell us about the kind of thing life is.
The Question the Textbooks Duck
Open any introductory biology textbook and look for the definition of life. You will not find one. You will find a list. Metabolism. Replication. Response to stimulus. Homeostasis. Growth and development. Evolution. The list varies in length — some sources add cellular organization, some add energy use as separate from metabolism, some collapse two items into one — but the structure is always the same. A set of criteria, sometimes presented as necessary and sufficient, sometimes presented as a typology of features that "living things" share. What is conspicuously missing is the move from criteria to what the criteria pick out. The book tells you what living things do; it does not tell you what makes them living. The criteria are advertised as a definition while functioning as a description.
This is not a pedagogical failure. The textbooks duck the definitional question because the working scientific literature ducks it too. Open Molecular Biology of the Cell and search for the phrase "definition of life" [Alberts et al. 2014, pending+pointer]; the book is two thousand pages of life and never quite says what it is. Open Campbell Biology and the same — the introductory chapter lists features, the rest of the book describes the features in detail, the question of what they collectively are never gets adjudicated [Urry et al. 2020 Campbell, pending+pointer]. NASA's working definition for the search for extraterrestrial life — "a self-sustaining chemical system capable of Darwinian evolution" — is a compromise hammered out for instrumental reasons rather than a metaphysical claim about the nature of the living [Joyce 1994 (NASA exobiology working definition), pending+pointer]. Even Schrödinger's What Is Life? — the book that gave the question its modern title — does not, in the end, answer it [ref: Schrödinger 1944, pending+pointer]. The question is in the title because it is the question the lectures are organized around, not because the lectures answer it.
Why does the working literature handle the question this way? Three possibilities. First, the question is empty — there is no fact of the matter, "life" is a folk-biological category that happens to track a useful cluster of features but does not pick out a natural kind. Second, the question has an answer but it is so directly the conjunction of the criteria that stating it adds nothing. Third, the question is real and has resisted answering because life is the kind of object whose ontology is not exhausted by any single discipline's apparatus, and the criteria-list is what working biology has produced as a coordination strategy in the absence of an answer it does not need in order to predict.
This essay is about the third possibility. The criteria-list, on this reading, is not a substitute for a definition that biology has been too lazy to write. It is the trace of a discipline that has discovered that its central object is genuinely multi-aspectual — that life is the kind of thing for which several different sciences each produce a true and partial description, none of which is the whole story, and that biology operates by holding the partial descriptions in working alignment without claiming to have unified them. This is structurally identical to the situation chemistry finds itself in with the chemical bond [14.essays.01]. The bond is the place where chemistry's pictures coordinate; life is the place where biology's lenses coordinate. Neither object is exhausted by any single picture or lens. Each is real in the sense that the coordination is robust, and not real in the sense that no underlying substrate corresponds, in the bare physical Hamiltonian, to the integrative concept.
What follows is a tour of the lenses through which the question has been asked. Each lens picks out a real feature of life. None alone captures it. The relationship between them — the way they coordinate when they coordinate, and diverge when they diverge — is what the discipline of biology is.
The Physical Lens: Life as Thermodynamic Disequilibrium
Schrödinger's 1944 What Is Life? — a small book based on lectures given at Trinity College Dublin — asked the question from the side of physics. Schrödinger's question was not "what feature distinguishes a cell from a crystal" but "how does life persist as an ordered structure in a universe whose statistical tendency is toward disorder?" The second law of thermodynamics says that closed systems evolve toward maximum entropy. Living systems are massively ordered. They are not closed; they are open, exchanging matter and energy with their environment. The Schrödinger move was to argue that life maintains its low-entropy order by exporting entropy to the environment faster than its internal entropy production accumulates. He coined the term "negative entropy" — later sharpened by physicists into the more precise statement that life consumes free energy (or, equivalently, that it dissipates entropy into a sink at higher temperature than its own).
This is, in the canonical-ensemble language of statistical mechanics [11.04.01], the statement that a living organism is a system maintained far from thermal equilibrium by continuous coupling to an environment that can absorb its entropy production. The internal degrees of freedom of a cell — the configuration of its proteins, the concentration gradient across its membranes, the sequence-specific arrangement of bases in its DNA — are all low-probability configurations relative to the thermodynamic equilibrium they would settle to if the cell were sealed off. The cell holds itself away from that equilibrium by continuously processing high-free-energy inputs (glucose, photons, reduced inorganic compounds) into low-free-energy waste (carbon dioxide, water, heat). The waste carries away the entropy. The cell stays ordered.
Ilya Prigogine, working in Brussels through the 1960s and 1970s, gave this picture its mature physical form [Prigogine 1977 Nobel lecture and *Order Out of Chaos*, pending+pointer; Prigogine and Stengers 1984, pending+pointer]. Prigogine showed that certain classes of open systems driven far from equilibrium can spontaneously organize into spatially or temporally structured patterns — dissipative structures. The Bénard cell, the Belousov-Zhabotinsky reaction, the chemical clocks that oscillate between colors with steady frequency — these are not life, but they are systems that demonstrate that the formation of macroscopic order from the dissipation of high-free-energy inputs is a generic physical phenomenon. Life, on this view, is a dissipative structure with extraordinary refinement: a system that has evolved an enormous and specific machinery for capturing free energy and using it to maintain itself in a particular non-equilibrium configuration.
Eric Smith and Harold Morowitz have developed this view into a full-scale theory of biology as planetary-scale thermodynamics [Smith and Morowitz 2016, pending+pointer]. On their account, the Earth's biosphere is a heat engine driven by the temperature difference between the sun (high temperature) and deep space (low temperature). Life is what the planet does to dissipate the resulting free-energy gradient. The molecular details — the specific catalysts, the specific genetic code, the specific architectures of cells — are contingent solutions to the thermodynamic imperative, refined over four billion years. The imperative itself is generic.
This lens is powerful. It explains why life requires energy sources and waste sinks. It explains why metabolism is universal. It explains the energetic foundation of biochemistry — why ATP, with its high-free-energy phosphoanhydride bond, is the universal currency of cellular energy; why the citric-acid cycle takes the form it does as a stepwise extraction of free energy; why photosynthesis is structured as a series of redox reactions descending a free-energy ladder. It also, less visibly, explains why life spreads: a thermodynamic system that has solved the problem of self-maintenance under a given free-energy gradient will tend to expand its access to that gradient, because expansion is favored by the same thermodynamics that favored the original structure.
But the physical lens leaves something out. A candle flame is a dissipative structure. It captures free energy (the chemical potential of the wax), dissipates entropy (heat, carbon dioxide, water), maintains a spatially organized macroscopic pattern, and even propagates itself (lighting another candle). It is not alive. Whatever the additional feature is that distinguishes a cell from a candle, thermodynamics alone does not name it. The physical lens shows us that life is consistent with the second law, and indeed that it is a particular elaboration of the way the second law plays out on a planet exposed to a free-energy gradient. It does not show us what kind of elaboration.
The Chemical Lens: Autocatalytic Networks
The chemical lens picks up where the physical lens leaves off. If life is a dissipative structure with extraordinary refinement, the refinement consists in the chemical machinery that captures and channels the free-energy flow. The question becomes: what kind of chemical organization counts as that machinery?
Manfred Eigen, working at the Max Planck Institute in Göttingen through the 1960s and 1970s, proposed that the chemistry-becomes-biology transition is the emergence of hypercycles — networks of mutually catalytic molecules in which each member catalyzes the production of the next, and the network as a whole closes on itself [Eigen 1971, pending+pointer; Eigen and Schuster 1979, pending+pointer]. The hypercycle is a chemical structure that catalyzes its own existence. Any single member, alone, would decay; the network sustains its members because each member sustains its neighbors.
Stuart Kauffman generalized this in The Origins of Order [Kauffman 1993, pending+pointer]. Kauffman's argument was that in any sufficiently rich chemical soup, the formation of an autocatalytic set is generic — once the number of reactions in the soup exceeds a critical threshold, the probability that some subset of molecules forms a closed catalytic cycle approaches one. The emergence of life, on this picture, is not a low-probability accident requiring rare conditions; it is a phase transition in chemical-network connectivity. Cross a threshold of complexity, get an autocatalytic set. The autocatalytic set is the seed of biology.
The hypercycle and autocatalytic-set theories sit alongside competing accounts of the chemistry-biology threshold. The RNA world hypothesis, articulated by Walter Gilbert in 1986 [Gilbert 1986, pending+pointer], holds that the first self-sustaining biological system was a self-replicating RNA molecule — a single polymer that could both store information and catalyze its own copying. The lipid world hypothesis, defended by Doron Lancet and collaborators [ref: Segré, Ben-Eli, Deamer, and Lancet 2001, pending+pointer], holds that the first biology was a self-organizing assembly of amphiphilic lipids forming compositional structures that could grow, divide, and inherit compositional state without nucleic acids. The metabolism-first hypothesis, associated with Günter Wächtershäuser and later with Mike Russell's hydrothermal-vent scenarios [ref: Wächtershäuser 1988, pending+pointer; Russell and Martin 2004, pending+pointer], holds that biology began as a network of redox reactions on mineral surfaces — that the metabolism came before the genes, and the genes were a refinement of an already-functional chemical network.
These accounts disagree about the order of events. They agree on the structural claim that the chemistry-biology threshold is the emergence of a self-maintaining chemical network — a set of reactions that produces, from environmental inputs, the very catalysts that sustain the set. The §15 organic chemistry units describe the molecules; the §19.15 origin-of-life unit (when produced) will describe the scenarios. The question this essay asks is structural: at what point does a network of chemical reactions become a biological one?
The honest answer is that there is no sharp threshold. An autocatalytic set in a beaker, with no membrane, no genetic code, no replication-with-variation, is closer to chemistry than to biology — but it is not pure chemistry either, because pure chemistry does not have catalytic closure. Add a membrane and the set begins to look like a protocell. Add inheritance of compositional state and it looks more so. Add a polymer that encodes the catalyst identities and you have something nobody would deny is alive. The progression is continuous. The question "at what step did chemistry become biology" admits no honest single answer.
This is the first hint of the multi-aspectual character we will keep encountering. The chemical lens does not give us a definition of life so much as a threshold-region — a range of organizational complexity in which a chemical network starts to satisfy the structural conditions for the higher-order phenomena (Darwinian evolution, autopoiesis, agency) that the other lenses will pick out. Where exactly the line falls is, on the chemical lens alone, indeterminate.
The Cybernetic Lens: Autopoiesis and Self-Reference
In 1972, Humberto Maturana and Francisco Varela — two Chilean biologists working in the wake of the cybernetics tradition of Norbert Wiener, Ross Ashby, and Heinz von Foerster — published a slim book in Spanish that introduced the concept of autopoiesis [Maturana and Varela 1972 (*De Máquinas y Seres Vivos*), pending+pointer]. The English-language elaboration came in 1980 [Maturana and Varela 1980, pending+pointer] and again in the more accessible The Tree of Knowledge [Maturana and Varela 1987, pending+pointer]. The concept is the most rigorous attempt at a definition of life produced in the twentieth century, and it has the kind of structural elegance that means it is either profound or wrong — possibly both.
An autopoietic system, in Maturana and Varela's definition, is a network of processes that (1) continuously produces the components that constitute it, (2) maintains a boundary that distinguishes it from its environment, and (3) does both of these through the operations of the network itself. The cell is the paradigm. The metabolic network of a cell produces, from environmental inputs, the enzymes that catalyze the metabolic network. The membrane that bounds the cell is produced by the metabolic network it bounds. The DNA that encodes the enzymes is replicated by enzymes the DNA encodes. The whole thing is recursive: the system produces itself, including the boundary that says where it ends and the environment begins.
This is more than the autocatalytic network of the chemical lens. The autocatalytic network is a necessary feature of an autopoietic system, but it is not sufficient. What autopoiesis adds is the boundary condition: the network must produce its own enclosure, and the enclosure must be the boundary that defines the system as a unity. A hypercycle in a beaker is autocatalytic but not autopoietic — the beaker, not the network, is providing the boundary. Add a membrane that the network maintains by its own operations, and the system becomes autopoietic: a chemical individual.
Maturana and Varela's claim was strong. They argued that autopoiesis is both necessary and sufficient for the living. Anything that is autopoietic is alive. Anything that is alive is autopoietic. The criteria-list of textbook biology, on their view, is a description of the consequences of autopoiesis: a system that produces itself will, by structural necessity, also metabolize, respond to stimuli, maintain homeostasis, and so on. Strip the consequences and look at what they descend from; you find autopoiesis. Definitions accomplished.
The definition has its critics. Some philosophers of biology — Pier Luigi Luisi notably [Luisi 2006, pending+pointer] — have defended autopoiesis as the best candidate for a structural definition of life. Others have argued that autopoiesis is too restrictive (it appears to exclude obligate parasites that cannot produce themselves without a host) or too permissive (some kinds of self-maintaining chemical systems satisfy the criteria without being intuitively alive). The literature has produced refinements — autonomy, operational closure, self-individuation — that try to preserve the structural insight while addressing the edge cases.
For our purposes the structural insight is the point. The cybernetic lens reveals that life is self-referential in a way no other physical system is. The cell is not just a system that does things; it is a system whose doings constitute the self that does them. The boundary is not given; the boundary is produced by what the boundary contains. This is a kind of organization unfamiliar to physics — not because it violates any physical law, but because it is a closure condition that physics qua physics has no vocabulary for. Physical systems can be open or closed; they cannot, in the physical-systems sense, be operationally closed in the autopoietic sense, because operational closure is a relation between the productions and the producers, not a relation between the system and its surroundings.
Information theory enters here. César Hidalgo, in Why Information Grows [Hidalgo 2015, pending+pointer], argues that life — and complex systems more generally — are systems that have evolved the capacity to embed information in physical structure, where embedded information is the difference between random matter and matter arranged to encode a function. Luciano Floridi, in The Philosophy of Information [Floridi 2011, pending+pointer], pushes further: information is the right level of description for systems whose causal structure runs through symbolic encoding, and biology is the paradigmatic case. The autopoietic system has DNA; the DNA encodes the catalysts; the catalysts produce the cellular structure; the cellular structure maintains the conditions under which the DNA is read. Information flows through the system in a way that is not reducible to the matter-and-energy flow, because the meaning of the information — what the DNA codes for — depends on the cellular machinery that interprets it. There is no DNA-in-itself that has biological meaning; there is only DNA-as-read-by-the-cell-it-is-part-of.
This is the cybernetic lens's contribution. Life is not just a thermodynamic structure, not just an autocatalytic network: it is a self-referential system whose own operations produce both its physical components and the informational structure by which those components coordinate. The cell is its own producer and its own interpreter. This is closer to a definition than anything the physical or chemical lenses gave us, and it is still not the whole story.
The Darwinian Lens: Life as What Evolves
Daniel Dennett, in Darwin's Dangerous Idea [Dennett 1995, pending+pointer], proposed that the cleanest empirical handle on the question "what is life" is to ask "what does Darwinian natural selection act on?" The answer — populations of imperfectly replicating entities whose variation in heritable traits correlates with variation in reproductive success — is, in his framing, the working definition of life. NASA's exobiology definition is a refinement of this: a self-sustaining chemical system capable of Darwinian evolution. Whatever else life is, it is the kind of thing that evolves.
This lens has the advantage of being operational. We can check, for any candidate system, whether it has heritable variation and differential reproductive success — and if it does, by hypothesis, it is alive. The Hardy-Weinberg framework [19.02.01] gives us the formal language for what evolution looks like at the population level: allele frequencies that change over generations under the pressures of selection, mutation, migration, and drift, against a baseline of what frequencies would be in the absence of these forces. A system that exhibits Hardy-Weinberg-violating allele-frequency dynamics in response to environmental pressure is, by that token, an evolving system. By the Darwinian definition, it is alive.
The lens is powerful in part because it is substrate-neutral. The Darwinian definition does not specify carbon, DNA, water, or membranes. It specifies a structural condition — heritable variation plus differential reproduction — that could in principle be satisfied by systems made of very different stuff. This is what makes it useful for astrobiology: we can ask whether a candidate Martian system, or a hypothetical silicon-based system, satisfies the Darwinian criterion without prejudging its chemistry. And it has the further advantage that the criterion is cleanly violated by the obvious non-life cases. A candle flame does not evolve. A crystal does not evolve. A computer simulation does not evolve in any sense not parasitic on the evolution of its programmers. The Darwinian criterion picks out what we want it to pick out, in most clear cases.
But it has costs. The first is at the origin-of-life boundary. The very first replicator, by hypothesis, replicated imperfectly and was subject to selection — but it was a single molecule, or a single small chemical network, not yet differentiated from its abiotic precursors by anything except this one capacity. Was it alive? The Darwinian definition says yes, by the moment that capacity emerged. The autopoiesis definition says probably no, because a single molecule is not yet operationally closed. The thermodynamic definition is silent. The intuitive answer most biologists would give is "barely," or "the question is not well-posed."
The second cost is at the individual-versus-lineage boundary. Natural selection acts on populations, not individuals. An individual organism does not, strictly speaking, evolve — it lives, reproduces, dies. The evolution is the statistical fact about the population over generations. So the Darwinian definition, strictly read, makes "life" a property of lineages rather than of organisms. An individual sterile worker ant, by this strict reading, is not alive — only the lineage is — which is not what we want the definition to say. Refinements exist (the inclusive fitness framework of W. D. Hamilton [Hamilton 1964, pending+pointer] makes the sterile worker alive as part of a colony-level reproductive strategy), but they reveal that the Darwinian lens, by itself, does not give us a clean answer at the individual level. It gives us an answer at the population level and treats the individual derivatively.
The third cost is the case of obligate non-replicators that are the products of evolution but cannot themselves evolve. A mule is the offspring of two evolved lineages but is sterile. By a strict Darwinian definition, the mule is not alive, because it cannot replicate. The intuitive judgment is that the mule is alive — it metabolizes, responds, maintains itself, suffers — and that the strict Darwinian criterion has failed in this case. Refinements (the mule is alive as a member of a hybrid class whose parental species evolve) again rescue the intuition but reveal that the Darwinian criterion is, on its own, a description of life at a level (lineages) different from the level (organisms) at which we usually apply the concept.
What this lens contributes, despite its limits, is the historical dimension that the others lack. Life is not just a kind of physical organization — it is the kind of physical organization that has been shaped, over billions of years, by cumulative selection on heritable variation. The complexity of the eye, of the protein-synthesis machinery, of metabolic regulation — none of these is explicable from physical or chemical principles alone. They are explicable as the products of a long history of differential reproduction. The Darwinian lens is what gives biology its temporal depth: every cellular feature is also a fossil of the lineage that produced it. To understand a feature is, in part, to understand its history. No other discipline thinks this way about its central object.
The Agency Lens: Goal-Directedness Without Telos
Pre-Darwinian biology was teleological in a metaphysical sense: living things were taken to have purposes given to them by some external designer or by an internal principle of organization that aimed them at their natural ends. Aristotle's telos. Aquinas's final cause. Paley's watchmaker. The trouble with this kind of teleology was not that it was unscientific; it was that it explained nothing. To say that the heart pumps blood in order to circulate it is to redescribe the function as a purpose without saying what makes it a purpose rather than a side-effect.
Darwin's contribution, on the agency side, was to replace metaphysical teleology with what Colin Pittendrigh in 1958 called teleonomy [Pittendrigh 1958, pending+pointer]. Teleonomy is functional purposiveness without metaphysical telos. The heart pumps blood because cardiovascular pumping was selected for in the lineage of vertebrates; the pumping is a function, in the technical sense of "a property maintained by selection because of what it does", not a purpose in the sense of being aimed at an external end. The distinction matters. Post-Darwinian biology can use functional language without commitment to a designer or to an Aristotelian principle of organization, because the functional language tracks the selective history that produced the structure. Functions are real; they are explained naturalistically; they require no extra-physical commitments.
The teleonomy concept does serious work in modern biology. Ernst Mayr made it the cornerstone of his philosophy of biology [Mayr 1961, pending+pointer; Mayr 1988, pending+pointer]. Functional explanation in molecular biology — "this protein functions as a chaperone," "this gene functions to regulate cell-cycle entry" — is teleonomic explanation. The function is a feature of what the structure does as selected for, not of what it is in itself. The same physical machinery, in a context where it was not selected for that function, would not have that function.
But the agency question, in modern philosophy of biology, has pushed past teleonomy. There is a real phenomenon — call it minimal agency — that is not exhausted by teleonomic functional ascription. Consider a bacterial cell undergoing chemotaxis. The cell detects a chemical gradient (using receptors that signal through a kinase cascade), processes the signal (through a feedback-coupled network of phosphorylation states), and modulates its motor activity (flagellar rotation) so that, on average, it moves up the gradient toward higher concentrations of the attractant. The cell, in the working description biologists give, is seeking the chemoattractant. It is responding to information about its environment. It is deciding — in the limited but real sense that the same kind of cell, given different sensory input, will behave differently in a way that is appropriate to the input.
This is more than teleonomy. Teleonomy is a property of structures: the chemotaxis machinery was selected for navigating gradients. Minimal agency is a property of the cell's current operation: the cell, right now, is using its machinery to do something. The chemotaxis behavior is a present-tense activity directed at a future state (being near more food), mediated by information (the gradient measurement) that the system is currently producing and currently using. This kind of present-tense future-directed information-mediated behavior is what we usually call goal-directed action. To call the bacterium an agent — even a minimal one — is not to overcommit to consciousness or to deliberation; it is to acknowledge that the system has the functional structure of goal-pursuit.
The agency lens is therefore not the teleonomic lens; it is something more. Daniel Dennett's intentional stance [Dennett 1987, pending+pointer] is one well-known articulation: a system is best described as having goals, beliefs, and information when treating it that way is predictive in a way that no lower-level description matches. The bacterial cell predicts well under the intentional stance ("it wants to find food"); the rock does not. Whether the cell really has goals in some deeper sense is, on Dennett's view, the wrong question — the intentional stance is a stance, not a metaphysical claim, and what it tracks is a real pattern.
More recent work in the philosophy of biology has tried to go further. Terrence Deacon, in Incomplete Nature [Deacon 2012, pending+pointer], argues that biological agency is a real ontological category — that systems that have evolved to maintain themselves against entropy by acting on their environment have a kind of causal structure that no non-living system has, and that this causal structure underlies the appearance of goal-directedness without reducing to it. Whether Deacon's view is correct is a live question in philosophy of biology. The point for our purposes is that the agency lens is doing work that the other lenses do not: it is asking what kind of thing a living system is as an actor in its environment, and answering that it is something for which goal-language is not metaphorical but tracking.
This lens connects directly to the unit on action potentials [17.09.02], which is the cellular substrate of information processing in nervous systems, and to the unit on muscle contraction [18.04.02], which is the cellular substrate of action in the strict sense — molecular motors converting chemical free energy into directed mechanical work. The agency lens says: the cell that fires the action potential is informing itself about its environment; the muscle that contracts is acting on its environment. The system that does both is a behaving system, in a way no purely physical or chemical system is.
The Phenomenological Lens: Life as What Experiences
The hardest lens to apply, and the one we have most trouble defending in scientific register, is the one that asks what it is like to be a particular living thing. Thomas Nagel's famous paper "What Is It Like to Be a Bat?" [Nagel 1974, pending+pointer] framed the question. Some — perhaps all — living systems have interiority. They are not merely objects in a world; they are subjects to whom a world appears. The question of subjective experience is the question of what makes the difference between a system that processes information and a system that has experiences.
This lens cannot be applied without entering territory the natural sciences are not equipped for. The cell that responds to a chemoattractant exhibits behavior that, from outside, looks like seeking. Whether there is anything it is like to be that cell — whether the seeking is accompanied by some minimal phenomenal character — is a question that the cell's behavior cannot answer, because no behavior, in principle, settles the question of accompanying experience.
For most of the history of biology, the question was left to philosophy and theology. The mid-twentieth century revival of consciousness studies in cognitive science (David Chalmers, Patricia Churchland, Daniel Dennett, Thomas Metzinger) has begun to make the question respectable again, though it remains the area of biology most exposed to disciplinary doubt about whether the question is well-posed. The §20 reflexivity essay [20.essays.01] frames the structural side of the question: there is something that sees, in any seeing, that cannot be located as an object among the things seen. The biology of experience must, at minimum, address what biological organization produces the seer.
The honest contemporary position is that we do not know. We know that brains correlate with experience in humans, and we have good reasons to extend the inference to other vertebrates and probably to invertebrates with sufficiently complex nervous systems. We have less good reasons to extend it to plants, fungi, and unicellular organisms — but we also have no principled reason to deny it, because we have no theory that says what kind of physical organization is sufficient or necessary for experience. The integrated-information theory of Giulio Tononi [Tononi 2008, pending+pointer] gives a quantitative measure of integrated information and proposes that systems above some threshold of have experience proportional to it; the global-workspace theory of Bernard Baars and Stanislas Dehaene [Baars 1988, pending+pointer; Dehaene 2014, pending+pointer] proposes that experience requires a specific neural architecture for broadcasting information. Both theories are testable; neither is settled.
What the phenomenological lens contributes — even in its unsettled state — is the recognition that some living systems are experiencers, and that the fact of experience is a feature of the biological world that the other lenses do not address. A thermodynamic description of an animal can be complete and not mention that the animal is a subject. A Darwinian description can be complete and not mention that the animal experiences pain or pleasure. An autopoietic description can describe the self-production of the system without addressing whether there is a self to whom anything appears. Whatever lens is doing the work, experience is not in the picture unless we deliberately put it there. And yet, in the case of at least some living systems, the experience is a real feature of what is going on, and any account of life that ignores it is incomplete in a particular way: it leaves out what life is from the inside.
The reflexivity essay's structural claim — that the seer cannot be seen — maps onto biology with surprising precision. The biological organization that produces a seer cannot itself contain the seer as one of its inventory items. The brain that produces the experiencing subject is not, when described in neuroanatomical detail, a brain plus a subject; it is a brain, and the subject is somehow of the brain in a way that brain-vocabulary does not capture. This is structurally identical to the autopoietic point that the boundary is not given but produced by what it contains. The seer is not given but produced by the biological organization that has, somehow, become its own subject. How this happens is the consciousness problem, and biology has not solved it.
The Mereological Problem: When Does Life Begin and End?
A final lens, less often named as a lens, complicates all the others. When is something alive? The standard intuition is that there is a moment of beginning and a moment of ending, with life in between. The biological reality is that both edges are processes, not moments, and that the question "is this alive now?" admits no sharp answer at the edges.
Consider the boundary at conception. A fertilized human egg is alive in the sense that it metabolizes, maintains itself, divides. The sperm was alive in the same sense; the egg was alive in the same sense. The combination is a new individual in a genetic sense, but the "life" was never absent during the transition — it was redistributed. The same logic runs through cell division generally: a cell does not pause being alive in order to be replaced by two cells. Life is continuous across the discontinuity of individuation.
The boundary at death is even less sharp. Cardiac death (the cessation of heartbeat) and brain death (the cessation of measurable brain activity) and cellular death (the progressive failure of individual cells to maintain homeostasis) occur at different times, in different tissues, across hours to days. A heart-stopped body has tissues that are alive for some time after the heart stops; transplanted organs from a brain-dead donor can continue to live in a recipient for decades. There is no single moment of death. There is a process of dying, distributed across tissues, with the legal and medical definitions of death drawn for convenience at points within the process where the question of further reversibility becomes operationally answerable.
Cross-species cases compound the problem. Viruses are the classic edge case. A virion outside a host is essentially inert — a packet of nucleic acid in a protein capsid, doing nothing, exhibiting none of the textbook criteria for life. Inside a host cell, the virus uses the cellular machinery to replicate itself, with substantial heritable variation, and is subject to natural selection. Is the virus alive? Most biologists, asked, will say "not quite" or "it depends" or "by some criteria yes, by others no." By the Darwinian criterion, yes — viruses evolve, and evolve consequentially. By the autopoietic criterion, no — viruses do not produce their own components. By the thermodynamic criterion, maybe — they are not dissipative structures on their own, but they participate in the dissipative structure of their host. By the agency criterion, marginally — viruses do not deliberate in any sense, but they navigate their environment via specific molecular recognition. The virus is alive by some lenses, not alive by others, and there is no fact of the matter beyond the lens-dependence.
Prions are sharper. A prion is a misfolded protein that catalyzes the misfolding of normally folded copies of itself. It has no nucleic acid. It replicates, in the broad sense that it produces more of itself; it is subject to selection in the sense that some prion variants are more transmissible than others. By a Darwinian criterion, prions might be alive. By any other criterion, they are not. Plasmids — small circular DNA molecules in bacteria that replicate semi-autonomously — sit somewhere similar. Mitochondria and chloroplasts, descended from once-free-living bacterial ancestors, are now organelles within other cells and are not separately alive; but the line between symbiont and organelle is a continuum that evolution has crossed many times.
What the mereological problem reveals is that "life" is not a property of a single instant but of an extended pattern over time, and the pattern admits boundary cases the criteria-list cannot adjudicate. The candle flame is not alive even when it is propagating; the prion is not alive even when it is replicating; the virus is alive only inside a host; the dying organism is alive in some tissues for hours after the integrated organism has ceased. None of this is a defect of biology; it is a feature of life as a category. The category covers a robust core and has fuzzy edges, and the edges are exactly where the lenses disagree with one another.
The Synthesis: What the Lenses Agree About
Six lenses. Each picks out a real feature. None alone gives a definition.
The thermodynamic lens picks out the energetic feature: life is a system held far from equilibrium by continuous coupling to a free-energy gradient. The chemical lens picks out the organizational feature: life is an autocatalytic network with closure properties. The cybernetic lens picks out the self-referential feature: life is a system that produces its own components and its own boundary. The Darwinian lens picks out the historical feature: life is the kind of system whose present organization is the cumulative product of selection on heritable variation. The agency lens picks out the active feature: life is goal-directed action mediated by information about the environment. The phenomenological lens picks out the subjective feature: at least some life is experience, accompanied by interiority of some kind.
What is striking is that these features are concomitant in the systems we uncontroversially call alive. A human cell is all six things at once. A bacterial cell is all six (with the sixth contested). A multicellular organism is all six, organized at multiple levels. Each feature implies the others in the canonical cases, not as a matter of logical necessity but as a matter of how biological systems have actually come to be. The thermodynamic gradient is necessary for the autocatalytic network to run; the autocatalytic network is necessary for autopoiesis; autopoiesis is necessary for Darwinian individuation; Darwinian selection produces the goal-directed structures of agency; sufficient complexity of agency-machinery produces, at least in some systems, the conditions for experience. The lenses pick out features that, in real biology, coordinate.
This is the key structural fact, and it is the answer the discipline of biology has settled into without quite stating. Life is the coincidence of features. Where the six features convene in a single self-maintaining system, what we have is alive in the full sense the word demands. Where some convene but others do not — a candle flame (thermodynamic but not autocatalytic-with-closure, not Darwinian, not agential), a crystal (organizational but not thermodynamic-far-from-equilibrium, not Darwinian), a prion (replicating but not autopoietic, not agential), a computer simulation of a cell (informational but not thermodynamic, not autopoietic in the material-realization sense) — we have something else. Something with one or more of life's features. Not life.
This is why the textbooks duck the definition. They are not being lazy. They are operating from the implicit recognition that what makes a system alive is not a property but a convening — a coordination of multiple features that no single feature is identical to. The criteria-list is a list of the convening features, given without the explicit claim that life is the convening. The claim is hidden in the list. Biology proceeds as if the list is a definition because, for predictive purposes, the list is good enough.
But the philosophical work of saying what life is requires naming the convening as the answer. Life is the configuration in which a thermodynamic dissipator is also an autocatalytic network is also an autopoietic system is also a Darwinian individual is also an agent is also (in the cases where the right organization is reached) a subject. Remove any feature and you get something else — a different kind of object, often a kind of object that has its own name (a flame, a crystal, a virus, an organelle, an algorithm). What we call life is what is left when none of the features is removed.
The Closing Move: The Cell Exceeds the Picture
There is a structural parallel to the chemical-bond essay [14.essays.01] that wants to be named here. Chemistry has its central object — the bond — and that object resists a single ontology. The Lewis picture, the molecular-orbital picture, the wavefunction, the AIM density: each is a partial disclosure of the bond, and the bond is what makes the disclosures coordinate when they coordinate. The chemist who has internalized this uses each picture fluently while knowing that none is the bond. The bond exceeds the picture.
The cell — and more broadly, the living organism — has the same structural status. The thermodynamic description, the chemical-network description, the autopoietic description, the Darwinian description, the agency description, the phenomenological description: each is a partial disclosure of the living system, and life is what makes the disclosures coordinate when they coordinate. The biologist who has internalized this uses each description fluently while knowing that none is life. Life exceeds the description.
This is not an evasion. It is the working ontology of biology as a mature discipline. A molecular biologist treating the cell as a chemical-network problem is doing real work, producing predictive results, and is right to do so. A physiologist treating the organism as an integrated regulatory system is also doing real work, and is also right. An evolutionary biologist treating populations as lineages-under-selection is doing real work. A neuroscientist treating the brain as an information-processing organ is doing real work. None of these workers, when they are doing the work well, makes the mistake of confusing their lens for the thing the lens discloses. They use the lens because it predicts. They know — explicitly or implicitly — that the cell exceeds the lens. The discipline coheres because the lenses coordinate; the discipline does not pretend to a unified ontology because the unity, if there is one, is not yet anyone's to articulate.
The reflexivity essay's structural claim returns here. The reflexivity essay said: the seer cannot be seen, and what we call the seer is what exceeds the seen. Biology, in the same structural register, says: the cell cannot be exhausted by any single description, and what we call life is what exceeds the descriptions. The cell is to its descriptions what the seer is to the seen. This is not a coincidence of language. It is the same point at two different scales of organization. Wherever there is a system whose self-production includes the production of the very things by which the system is described, the system exceeds the descriptions. The seer who produces the seeing exceeds what is seen. The cell that produces the conditions of its description exceeds what is described.
This is the deepest thing biology has discovered about its object, and it is the thing biology has been least willing to say out loud, because saying it threatens the picture of biology as a normal science with a settled ontology. Biology is not, in this sense, a normal science. It is the science that studies the kind of system that produces, among other things, the conditions of its own being-studied. The criteria-list in the textbook is the trace of this: a discipline writing around its object because the object is not exhausted by any single way of writing about it.
The honest closing is therefore not a definition but a recognition. Life is the convening of features that no single feature names, the coordination of disclosures that no single disclosure is, the operation of self-production that produces, among its products, the very pictures by which it is partially seen. The working biologist already knows this in practice. The textbook ducks it because the pedagogy has not figured out how to say it. The philosophy of biology is, slowly, learning to say it. This essay has tried to say it once more.
Cross-references. This essay synthesizes across the action-potential unit [17.09.02] (cellular information-processing), the muscle-contraction unit [18.04.02] (purposive cellular action), the Hardy-Weinberg unit [19.02.01] (the Darwinian lens at the population level), the canonical-ensemble unit [11.04.01] (the thermodynamic foundation), the chemical-bond essay [14.essays.01] (the structural parallel: a central object exceeded by its pictures), and the reflexivity essay [20.essays.01] (the seer-seen structure mapped onto biology's lens-multiplicity). The §15 organic-chemistry chapter and the §19.15 origin-of-life unit (when produced) are the chemistry-side and the deep-history-side companions of the chemical-lens discussion.
Citation status. All references are pending+pointer: — the reference/ archive carries no biology material as of Wave 1 bio production. The citations are author/year placeholders to be resolved during the biology-side sourcing pass (BIOLOGY_PLAN.md §5.4). Sources to obtain: Schrödinger 1944 (What Is Life?); Prigogine 1977 (Nobel lecture) and Prigogine and Stengers 1984 (Order Out of Chaos); Smith and Morowitz 2016 (The Origin and Nature of Life on Earth); Eigen 1971 ("Self-organization of matter and the evolution of biological macromolecules"); Eigen and Schuster 1979 (The Hypercycle); Kauffman 1993 (The Origins of Order); Gilbert 1986 ("The RNA world"); Segré et al. 2001 (compositional lipid genome); Wächtershäuser 1988 (pyrite-pulled prebiotic metabolism); Russell and Martin 2004 (alkaline hydrothermal vents); Maturana and Varela 1972, 1980, 1987 (the autopoiesis trilogy); Luisi 2006 (The Emergence of Life); Hidalgo 2015 (Why Information Grows); Floridi 2011 (The Philosophy of Information); Dennett 1987 (The Intentional Stance), 1995 (Darwin's Dangerous Idea); Pittendrigh 1958 ("Adaptation, natural selection, and behavior"); Mayr 1961, 1988 (philosophy-of-biology papers, esp. "Cause and effect in biology"); Hamilton 1964 (inclusive fitness papers); Deacon 2012 (Incomplete Nature); Nagel 1974 ("What Is It Like to Be a Bat?"); Tononi 2008 (integrated-information theory paper); Baars 1988 (A Cognitive Theory of Consciousness); Dehaene 2014 (Consciousness and the Brain); Ladyman and Wiesner 2020 (What Is a Complex System?); Alberts et al. 2014 (Molecular Biology of the Cell, 6th ed.); Urry et al. 2020 (Campbell Biology); Joyce 1994 (NASA exobiology working-group definition of life).