Emergent properties are characteristics that appear in a system but do not exist in any of its individual parts. A single water molecule is not wet, not liquid, and has no surface tension, yet a vast collection of water molecules together produces all of these familiar traits. The physicist Philip Anderson captured this idea in a landmark 1972 paper arguing that at each level of complexity, entirely new properties appear that cannot be understood simply by extrapolating from the behavior of a few particles.1Science. More is different Emergence runs through nearly every scientific discipline, from chemistry and developmental biology to ecology, neuroscience, and artificial intelligence, and it raises some genuinely thorny questions about how the universe organizes itself.
Why Water Is the Textbook Example
Water is probably the most intuitive illustration of emergence because the gap between a single molecule and the substance we drink is so dramatic. One water molecule is a bent arrangement of two hydrogen atoms bonded to an oxygen atom. It has no viscosity, no boiling point, and no ability to dissolve salt. Bring trillions of these molecules together, though, and their orientation-dependent hydrogen bonds create open, cage-like structures that give rise to water’s remarkable volumetric and thermal properties.2PubMed Central. How Water’s Properties Are Encoded in Its Molecular Structure and Energies The liquid expands as it freezes, floats as ice, absorbs enormous amounts of heat before its temperature rises, and climbs up narrow tubes against gravity. None of those behaviors belong to a single molecule; they emerge from how the molecules interact collectively.
Even something as familiar as a water droplet sitting on a countertop turns out to be an emergent phenomenon. Wetting, the way water spreads on or beads up on a surface, can be traced to the interplay between the surface chemistry of the material and the energetic cost of disrupting hydrogen bonds in the liquid. Researchers have shown that across a wide spectrum of surface types, contact angles collapse onto a single universal curve when expressed through a molecular wetting coefficient, meaning wetting behavior arises from the liquid itself imposing a threshold that separates spreading from beading up.3PubMed. Wetting as an Emergent Property of Water: Reformulating Young’s Equation on Molecular Grounds You could know everything about one water molecule and one surface atom and still not predict whether the droplet would spread or bead. That prediction only becomes possible when you consider how an enormous number of molecules behave together.
Patterns That Build Themselves in Living Things
Biology is full of patterns that no single cell “knows” it is building. The stripes on a zebrafish, the regular spacing of hair follicles on a mouse, the arrangement of fingers on a developing hand: these forms emerge from local chemical interactions between cells rather than from some master blueprint that dictates where every feature should go. The theoretical groundwork for this was laid by the mathematician Alan Turing, who showed that a system of chemicals diffusing and reacting with each other could generate spatial patterns out of an initially uniform mixture.4PubMed Central. Towards an integrated experimental-theoretical approach for assessing the mechanistic basis of hair and feather morphogenesis His idea was elegant but remained largely theoretical for decades because finding the actual molecules responsible proved difficult.
That picture has changed considerably. In zebrafish, researchers used lasers to erase part of the stripe pattern on the skin and watched as surrounding stripes slid sideways to fill the gap, restoring normal spacing. The dynamic matched what Turing’s reaction-diffusion model predicted, providing strong evidence that a Turing-type process is genuinely at work in stripe formation.5PubMed Central. Studies of Turing pattern formation in zebrafish skin In mice, proteins called Nodal and Lefty appear to act as an activator-inhibitor pair during early embryonic development, and the quasi-regular spacing of mouse hair follicles involves a similar interaction between activating and inhibiting proteins. When mutant mice produce abnormally high levels of the inhibiting protein, the resulting follicle patterns match theoretical predictions from Turing-style models.6PubMed Central. Forging patterns and making waves from biology to geology: a commentary on Turing 1952 ‘The chemical basis of morphogenesis’
The point here is not that cells follow some choreographed plan. Instead, cells produce chemicals, those chemicals diffuse outward and interact, and the result is a stable repeating pattern. No individual cell contains information about the stripe or the follicle spacing. The pattern is a property of the system, not of any part within it.
When Individuals Become a Swarm
Collective animal behavior is one of the most visually striking forms of emergence in nature. A murmuration of starlings wheeling through the sky looks coordinated to the point of seeming choreographed, but no bird is directing traffic. Each bird follows simple local rules, primarily reacting to the speed and direction of its nearest neighbors, and the large-scale formations arise spontaneously from those interactions.7PubMed Central. Schools of fish and flocks of birds: their shape and internal structure by self-organization
Fish schools show similar self-organization. Studies of golden shiners in groups of 30 to 300 found that schools exhibit three distinct collective behaviors: a disorganized swarm with low speeds and little order, a strongly aligned state where the fish all move in the same direction at higher speeds, and a milling state where individuals circle around the group’s center. The transitions between these states resemble phase transitions in physics, like water switching between ice and liquid, and they emerge from the local alignment rules each fish follows.8PLOS Computational Biology. Collective States, Multistability and Transitional Behavior in Schooling Fish
Fireflies provide a particularly clean example. In the species Photinus carolinus, individual fireflies flash with no regular rhythm when they are alone. Put thousands of them together in a mating swarm, though, and something remarkable happens: they begin flashing in synchrony with a stable, rhythmic periodicity. The periodicity itself is a collective property that no single firefly possesses. Researchers found that a simple model of the transition, with no fitting parameters, agrees strikingly well with observed data.9eLife. Emergent periodicity in the collective synchronous flashing of fireflies Similar math has been applied to populations of identical biological oscillators more generally, showing that pulsatile coupling between oscillators, where firing by one unit pulls others slightly closer to firing, leads almost inevitably to full synchrony.10SIAM Journal on Applied Mathematics. Synchronization of Pulse-Coupled Biological Oscillators Your heartbeat relies on exactly this principle: thousands of pacemaker cells synchronize through local electrical coupling to produce a single coordinated contraction.
Life Itself as an Emergent Property
One of the deepest questions about emergence is whether life is itself an emergent property of chemistry. No individual molecule is alive, yet certain combinations of molecules, enclosed in a membrane and capable of replicating information, cross a threshold into something we recognize as living. Research on the origin of life increasingly frames this transition as the emergence of self-sustaining chemical networks from simpler precursors.
One line of work models these early chemical networks as autocatalytic sets, groups of reactions in which the products of some reactions catalyze others, creating a self-reinforcing loop. Starting with a minimal set of inorganic molecules like water, hydrogen, carbon dioxide, and ammonia, researchers found that only a tiny autocatalytic network of about eight reactions could form. But the sequential addition of organic cofactors expanded this network dramatically, from 16 reactions up to over 1,300, spanning a quarter of the known anaerobic metabolic network.11PubMed Central. Autocatalytic chemical networks at the origin of metabolism The implication is that metabolism did not need to appear all at once. It could have grown incrementally, with each new molecule opening up new catalytic possibilities, a kind of snowball effect in chemical complexity.
A complementary approach focuses on the compartment side: fatty acid vesicles that can grow, divide, and encapsulate informational polymers. The integration of a dynamic fatty-acid membrane with replicating genetic material would yield a system capable of Darwinian evolution, the defining feature of life.12PubMed Central. The origins of cellular life Other researchers have modeled this with lipid-based catalytic networks, showing that non-equilibrium assemblies with cell-like reproduction can emerge from catalysis-based growth and occasional fission of lipid clusters.13PubMed Central. Systems protobiology: origin of life in lipid catalytic networks In both cases, “aliveness” is not a property any single molecule possesses. It emerges when molecules interact in particular, self-sustaining ways.
Consciousness and the Brain
If life emerging from chemistry feels conceptually challenging, consciousness emerging from neurons is even more so. Your experience of seeing the color red, feeling anxious, or understanding a sentence cannot be found in any single neuron. Modern theories of consciousness generally propose that subjective experience arises from interactions between large-scale neuronal networks, a classic emergent phenomenon.14PubMed Central. Consciousness as an Emergent Phenomenon: A Tale of Different Levels of Description But the details remain deeply contested. Is consciousness just what large-scale neural computation feels like from the inside, or is something fundamentally new happening at the system level that you could never predict even with perfect knowledge of every synapse?
Some modeling work suggests that brain dynamics sit at a critical point between order and disorder, a state called self-organized criticality. Simulations of evolving neural networks show that spike avalanches, cascading bursts of neural activity, follow a power-law distribution, a signature of criticality that has also been observed in real human brain data.15Physica A: Statistical Mechanics and its Applications. Self-organized criticality and structural dynamics in evolving neuronal networks: A modified sandpile model This kind of critical-state behavior is thought to maximize the brain’s ability to process information and respond flexibly, and it is itself an emergent property of how neurons wire and rewire themselves. Whether this gets us closer to explaining consciousness itself, rather than just the brain’s computational properties, remains an open and fiercely debated question.
Emergence in AI and Financial Markets
The concept has found unexpected relevance in artificial intelligence. Large language models trained on text have displayed what researchers describe as emergent abilities: capabilities that are absent in smaller versions of the same model but appear abruptly once the model crosses a certain size threshold. Tasks like multi-step arithmetic or complex analogical reasoning seem to switch on rather than gradually improve as the model scales up. This has been characterized as genuinely unpredictable from the performance of smaller models.16Transactions on Machine Learning Research. Emergent Abilities of Large Language Models Whether this represents “true” emergence or is an artifact of how we measure performance has been debated, but the phenomenon has influenced how AI labs think about scaling: the possibility that crossing a threshold could unlock qualitatively new behaviors adds both excitement and unpredictability to the field.
Economics and finance offer another domain where emergence matters. Individual traders each making local decisions about buying and selling can collectively produce phenomena no one intended, such as market bubbles and recessions. Simulation work using AI-driven agents in a modeled stock market has demonstrated how individual actions trigger group behaviors that lead to these emergent outcomes.17Neural Information Processing Systems. TwinMarket: A Scalable Behavioral and Social Simulation for Financial Markets The emergent properties of animal collectives, including self-organization, robustness, and adaptability, have also inspired the design of autonomous swarm robotic systems, where engineers deliberately harness emergence to build systems that are more flexible than any centrally controlled robot could be.18National Science Review. From animal collective behaviors to swarm robotic cooperation
The Weak-Versus-Strong Debate
Not everyone agrees on what emergence really means, and the disagreements are more than semantic. The central divide is between what philosophers call weak and strong emergence. Weak emergence describes properties that are surprising or hard to predict in practice but that could, in principle, be deduced from complete knowledge of the parts and their interactions. The wetness of water is a common example: it is not obvious from looking at one water molecule, but there is nothing mysterious about it once you understand hydrogen bonding at scale. Strong emergence, by contrast, claims that some higher-level properties are genuinely irreducible, that they cannot even in principle be derived from lower-level descriptions no matter how complete.
This distinction matters practically. Computational models of the brain built on strong emergence tend to be loosely connected to the underlying biology, and a critique from the neuroscience literature argues that this loose mechanistic link makes such models metaphysically implausible, leaving them as just one of many possible explanations for the same observed phenomena.19PubMed Central. Conflicting emergences. Weak vs. strong emergence for the modelling of brain function Models based on weak emergence, built from biologically plausible units that interact according to known rules, do not face this problem. They can be scaled up and tested against real data. The practical upshot for researchers is that invoking emergence is not an explanation in itself. It is a description of a phenomenon that still needs a mechanistic account.
Downward Causation and Measuring Emergence
One of the most counterintuitive ideas associated with emergence is downward causation: the possibility that properties at a higher level can constrain or influence events at a lower level. This sounds paradoxical if you believe that all causation flows upward from physics to chemistry to biology, but several lines of argument suggest it is more straightforward than it seems.
One concrete way to think about it comes from multiscale modeling. In developmental biology and systems biology, researchers have shown that macroscale parameters act as boundary conditions for models at lower scales. The large-scale organization of a tissue, for instance, constrains which genes individual cells express, even though the tissue is made of those cells. This is a specific, mathematically demonstrable form of downward causation, not a vague philosophical claim.20Philosophy of Science. Scale Dependency and Downward Causation in Biology A related argument focuses on information control in microorganisms, where the genome’s operating system processes and controls information in ways that constitute top-down causation by feedback.21PubMed Central. Downward causation by information control in micro-organisms
Emergence has also started to be measured quantitatively rather than just debated philosophically. Using a measure called effective information, researchers have shown that in certain systems, coarse-grained macro-level descriptions carry more causal information than micro-level descriptions. In other words, the macro level is not just a convenient summary; it is genuinely more causally informative than the micro level.22PubMed Central. Quantifying causal emergence shows that macro can beat micro A complementary framework introduces formal criteria for identifying causal emergence in multivariate data, offering practical tools that can be applied to real systems in neuroscience, ecology, and elsewhere.23PubMed Central. Reconciling emergences: An information-theoretic approach to identify causal emergence in multivariate data This work is moving the field away from the somewhat frustrating position of “we know it when we see it” toward something more rigorous.
Evolutionary Novelty and the Creation of New Parts
Emergence also shows up in how evolution produces genuinely new structures. The origin of a novel body part, like feathers in dinosaurs or the turtle shell, is not simply the modification of something that already existed. It requires the evolution of a new gene regulatory network that integrates developmental signals into a gene expression pattern unique to that structure.24Cell Press (Current Biology). Evolutionary novelties The individual genes involved typically have other functions elsewhere in the body. What makes the novel structure possible is a new combination of regulatory interactions, a network-level property that no single gene possesses. This framing helps explain why evolution can produce genuinely surprising innovations rather than only incremental modifications. The raw genetic material may already be present; what emerges is a new way of organizing it, and with that new organization comes a structure with properties that did not exist before.
This connects to a broader theme across all the examples above. Whether the system is a puddle of water, a developing embryo, a swarm of fireflies, or a population of neurons, the key ingredient is the same: interactions among components generate properties that belong to the system as a whole. Understanding those interactions, rather than cataloguing parts, is increasingly where the interesting science lives.