Emergence describes a situation in which a system’s large-scale behavior or properties cannot be straightforwardly predicted from the behavior of its individual parts. A flock of starlings wheels through the sky in coordinated patterns that no single bird is directing. A traffic jam materializes on a highway even though every driver is trying to move forward. Consciousness arises from neurons that, individually, do nothing resembling thinking. The concept cuts across nearly every scientific discipline, from physics and biology to economics and artificial intelligence, and carries a surprisingly rich set of debates about what it really means and whether it can be measured.
Where the Idea Came From
The formal study of emergence traces to a group of British philosophers working around the turn of the nineteenth and twentieth centuries. Figures like John Stuart Mill, George Henry Lewes, Samuel Alexander, and C. D. Broad argued that certain properties of complex wholes are genuinely novel and cannot be deduced from knowledge of their components alone. Lewes is often credited with coining the term “emergent” in this philosophical sense, distinguishing it from a mere “resultant” that you could calculate in advance. These thinkers were grappling with chemistry and biology: water’s properties seem qualitatively different from those of hydrogen and oxygen considered separately, and a living organism behaves in ways that a pile of its constituent molecules does not.
That early framework set up a tension that persists today. On one side is the intuition that the whole really is “more than the sum of its parts,” that new causal powers appear at higher levels. On the other is the reductionist objection that if you had complete information about the parts and their interactions, you could, in principle, predict everything. Modern debates about emergence are essentially sophisticated versions of this same argument.
Strong Versus Weak Emergence
The single most important distinction in discussions of emergence is between “strong” and “weak” versions. Weak emergence means that a higher-level pattern arises from lower-level rules in a way that is surprising or difficult to predict in practice, but not in principle. If you had an impossibly powerful computer and knew every detail of the lower level, you could simulate the system and watch the pattern appear. Most scientists who use the word “emergence” in day-to-day work mean this weak version. Ant colonies, weather systems, and the Game of Life all qualify.
Strong emergence makes a bolder claim: the higher-level property is not deducible from the lower level even in principle. It implies that something genuinely new enters the picture at the macro scale, something that cannot be captured by any amount of micro-level information. Consciousness is the most frequently cited candidate. A detailed review of modeling approaches in neuroscience found that models relying on strong emergence risk what the authors called “metaphysical implausibility,” because their loose mechanistic links to underlying biology leave them as just one of many possible explanations for observed behavior. Weakly emergent models built from biologically plausible components, by contrast, avoid that problem and scale more naturally.1PubMed Central. Conflicting emergences. Weak vs. strong emergence for the modelling of brain function
The practical upshot is that most working scientists treat strong emergence with caution. It is philosophically interesting but hard to test. Weak emergence, meanwhile, is everywhere and is the version you will encounter in most scientific papers, textbooks, and popular science writing.
Emergence in Physics
Physics provides some of the cleanest examples. When water freezes into ice, the crystalline lattice structure is not a property of any individual water molecule. It emerges from the collective interactions of trillions of molecules under specific temperature and pressure conditions. The technical term for many such transitions is spontaneous symmetry breaking: a system that looks the same in all directions (symmetric) suddenly “chooses” a preferred direction or arrangement. A magnet cooling below its critical temperature is a classic case. Above that temperature, the atomic magnetic moments point in random directions and cancel out. Below it, they spontaneously align, and the material becomes magnetized.
Researchers have argued that spontaneous symmetry breaking is itself a paradigmatic emergent property, arising when there is a broken symmetry in the relationship between microscopic and macroscopic scales.2arXiv. On emergence from the perspective of physical science A related mathematical treatment has shown how symmetry breaking emerges specifically in the transition from quantum to classical descriptions of physical systems.3International Journal of Geometric Methods in Modern Physics. The classical limit of Schrödinger operators in the framework of Berezin quantization and spontaneous symmetry breaking as an emergent phenomenon These are not just abstract curiosities. Superconductivity, superfluidity, and the mass-giving mechanism behind the Higgs boson all involve forms of emergent symmetry breaking.
Another physical example comes from systems far from equilibrium. When energy flows through a system continuously, it can spontaneously organize into ordered structures called dissipative structures. Research on friction surfaces has shown that once a system loses thermodynamic stability, self-organization can kick in and produce structures that dramatically change the system’s behavior, in that case reducing wear rates.4PubMed Central. The Conditions Necessary for the Formation of Dissipative Structures in Tribo-Films on Friction Surfaces That Decrease the Wear Rate The general principle applies well beyond engineering: hurricanes, convection cells in heated fluid, and certain chemical oscillations are all dissipative structures that emerge when energy gradients push a system past a tipping point.
Patterns in Living Things
Biology is arguably where emergence feels most intuitive. A leopard’s spots, the branching of blood vessels, the spiral arrangement of seeds in a sunflower head: none of these patterns is painted on by some master blueprint. They arise from local chemical and cellular interactions.
The foundational insight came from Alan Turing in 1952, in a paper that had nothing to do with computers. Turing proposed that chemical substances he called morphogens, reacting with each other and diffusing through tissue, could generate spatial patterns from an initially uniform state. Small random disturbances could trigger an instability in the homogeneous equilibrium, and the system would spontaneously develop structure.5Philosophical Transactions of the Royal Society B. The chemical basis of morphogenesis This reaction-diffusion framework has since been validated as a working model for a wide variety of biological patterns, from the stripes on zebrafish to the spacing of hair follicles.6PubMed. Reaction-diffusion model as a framework for understanding biological pattern formation
What makes this emergent rather than simply “complicated” is that the pattern is not encoded anywhere. No gene says “put a spot here.” Instead, genes set the parameters of the reaction-diffusion system, and the pattern self-organizes. Change the parameters slightly and you get stripes instead of spots, or a different spacing between features. The macro-level pattern is a genuine product of collective dynamics.
Flocks, Swarms, and Collective Behavior
Animal groups offer some of the most visually striking emergence. A school of fish evading a predator moves as a single fluid entity, splitting and reforming in ways that look choreographed. But no fish is in charge. Each individual follows simple local rules about spacing, alignment, and speed relative to its nearest neighbors. The coordinated group behavior emerges from those rules applied simultaneously across thousands of individuals.
Research on collective animal behavior has shown that phenomena as varied as ant pheromone trail networks, cockroach aggregation, the synchronized applause of opera audiences, and the migration of fish schools can all be accurately described in terms of individuals following simple sets of local rules.7PubMed Central. The principles of collective animal behaviour The key insight is that the complexity lives at the group level, not the individual level. You could study a single ant for years and never predict the colony’s highway-like trail networks.
A formal study of flocking models highlighted the interplay between consensus (individuals trying to align with neighbors) and frustration (conflicting signals from different neighbors). The tension between these two tendencies produces highly complex, unpredictable, coherent behavior at the group scale.8Complexity. Defining emergence: Learning from flock behavior This is a useful template for emergence generally: simple competing rules at the micro level can generate rich, surprising order at the macro level.
Evolutionary Transitions
Emergence also shows up on the grandest timescales in biology. The history of life includes several “major transitions” in which previously independent individuals began cooperating so tightly that they became a new, higher-level organism. Single cells teamed up to form multicellular life. Individual organisms formed eusocial colonies where most members gave up reproduction. At each step, a new level of biological organization emerged.
These transitions have been broken down into two stages: first, a cooperative group forms; then that group transforms into an integrated entity with division of labor, communication, mutual dependence, and minimal internal conflict.9PubMed Central. Major evolutionary transitions in individuality Each transition also involved the evolution of a new way of using, transmitting, or storing information, from genetic codes to language.10PubMed. The major evolutionary transitions and codes of life You could view the entire arc of life on Earth as a series of emergent leaps, each one producing a new kind of individual from the interactions of older, simpler ones.
Consciousness and the Brain
If there is a “hard problem” of emergence, it is consciousness. How does subjective experience arise from the electrochemical activity of neurons? Most neuroscientific theories treat consciousness as emergent from large-scale interactions among brain networks, though proposals vary in how literally they mean this. Some researchers have argued that consciousness can be identified with specific patterns of connectivity, and that the framework could extend even to subcellular networks exhibiting quantum phenomena, suggesting the concept of emergent “conscious networks” is not exclusive to large brain areas.11PubMed Central. Consciousness as an Emergent Phenomenon: A Tale of Different Levels of Description
Whether consciousness is weakly or strongly emergent remains one of the deepest open questions in science. If it is weakly emergent, then a sufficiently detailed simulation of a brain would, in principle, be conscious. If it is strongly emergent, no simulation would suffice because something non-physical or non-computational enters the picture. Philosophers and neuroscientists continue to disagree, and the question has real implications for how we think about artificial intelligence.
Cities, Markets, and Traffic Jams
Human systems are rich with emergence. Cities grow and organize themselves without anyone designing their overall structure. Research on urban self-organization has described how complex regional-level patterns emerge from the local interactions of individual agents, often with consequences that nobody intended or predicted.12Environment and Planning B: Planning and Design. Indicators for self-organization potential in urban context Neighborhoods specialize, traffic corridors form, and land values stratify, all from millions of individual decisions about where to live, work, and shop.
Markets work the same way. Adam Smith’s “invisible hand” is an early description of emergence: individual actors pursuing their own interests can produce a coordinated economic order that nobody planned. Experimental economists have tested this idea directly, investigating whether groups of people can, without any central coordination, converge on outcomes that benefit the group.13Experimental Economics. Testing for the emergence of spontaneous order The results show that spontaneous order is real but fragile, and the conditions under which it arises matter a great deal.
Traffic jams are a particularly relatable example. You are cruising at highway speed when the flow suddenly collapses and you are at a standstill, yet when you finally clear the jam there is no accident, no construction, nothing visible that caused it. These “phantom jams” emerge from small perturbations in driver behavior that amplify through the chain of following vehicles. One driver brakes slightly harder than necessary, the next driver brakes a bit harder still, and the disturbance propagates backward through traffic as a standing wave. The jam is not caused by any single driver; it is a collective phenomenon.
Emergent Abilities in Artificial Intelligence
One of the most contested recent uses of the word “emergent” involves large language models like GPT-4. Researchers noticed that as these models were scaled up, adding more parameters and training data, certain abilities seemed to appear abruptly rather than improving gradually. A model that was completely unable to perform a task at one scale could suddenly perform it well at a larger scale. A widely cited paper defined an ability as “emergent” if it is not present in smaller models but is present in larger models, in a way that cannot be predicted by extrapolating from smaller models’ performance.14arXiv. Emergent Abilities of Large Language Models
This framing generated intense debate. Critics have argued that the apparent emergence may be an artifact of how performance is measured. If you use a metric that gives zero credit for partial success, abilities can look like they appear suddenly when they were actually improving gradually all along. A survey of the field noted that the debate hinges on whether these abilities are truly emergent or simply depend on external factors like training dynamics, problem type, and choice of metric.15arXiv. Emergent Abilities in Large Language Models: A Survey The argument is a live one and mirrors the broader philosophical tension between strong and weak emergence: is something genuinely new happening at scale, or are we just failing to see the gradual buildup?
Can Emergence Be Measured?
For a long time, emergence was treated as a qualitative concept, something you could point to but not quantify. That has started to change. One influential approach uses a quantity called effective information to compare how much causal work is being done at different levels of description. The key finding is that for certain system architectures, the effective information peaks at a macro level rather than the micro level. This happens when coarse-grained macro mechanisms are more deterministic and less degenerate than the underlying micro mechanisms, enough to overcome the smaller number of macro states. The authors describe this as genuine causal emergence: the macro level supervenes on the micro but causally supersedes it.16PubMed Central. Quantifying causal emergence shows that macro can beat micro
A related information-theoretic framework defines causal emergence as occurring when a higher-level feature of a system carries dynamically relevant information, meaning it helps predict the system’s future, and that information goes beyond what you get from looking at the parts separately.17PLOS Computational Biology. Reconciling emergences: An information-theoretic approach to identify causal emergence in multivariate data These tools are still young, but they represent a significant step: they allow researchers to test, with data, whether a macro-level description of a system captures causal structure that the micro-level description misses. This moves emergence from a philosophical claim to an empirical one.
Downward Causation and Why It Matters
One of the trickiest questions about emergence is whether higher-level properties can reach back down and influence their own components. Can the flock change the behavior of the bird? Can the mind affect the brain? This is known as downward causation, and it is both intuitive and philosophically explosive. If higher levels really do cause things at lower levels, emergence starts to look like a fundamental feature of reality rather than just a useful description.
A recent philosophical analysis argues that downward causation can avoid the logical problems traditionally attributed to it, as long as we think about “wholes” and their causal powers in relational terms. The proposal distinguishes between contextual causation, where the whole constrains the parts, and downward causation proper, where second-order structural relationships between levels drive the effect.18History, Philosophy and Theory of the Life Sciences. Emergence, Downward Causation, and Interlevel Integrative Explanations In plainer terms, this means the organization of the whole system can genuinely constrain and reshape how its parts behave, without violating any physical laws at the lower level. Your anxiety (a mental-level state) changing your heart rate (a cellular-level process) is the kind of everyday example this framework tries to make rigorous.
How Language Itself Emerges
Language is something most people take for granted, but its structure is itself an emergent phenomenon. No individual or committee designed English grammar. Instead, the systematic regularities of language appear to arise from the constraints of human cognition, especially memory limitations, passed through generations of learners. Experiments using iterated learning, where one person learns a set of sequences and then teaches them to the next person in a chain, show a cumulative increase in structure over generations. The sequences that survive this transmission process end up closely paralleling the statistical structure found in natural language corpora.19PLoS ONE. Sequence Memory Constraints Give Rise to Language-Like Structure through Iterated Learning
This line of research has been extended to artificial neural agents. When neural networks are placed in an iterated learning setup, communicating and then passing their language to the next “generation,” the language that develops becomes more compositional and structured over time.20ICLR. Compositional Languages Emerge in a Neural Iterated Learning Model The implication is striking: the basic structure of language is not a product of some uniquely human genetic endowment alone but an emergent consequence of any system with limited memory trying to transmit complex information across generations. The rules of grammar, in a sense, invent themselves.
Emergent Network Structures
Many real-world systems, from the internet to social networks to protein interaction maps, share a common structural feature: a few nodes have vastly more connections than most. This “scale-free” pattern was long thought to require a growth process in which new nodes preferentially attach to already well-connected ones. But recent work has shown that scale-free structure can emerge even in networks that are not growing. Starting with an arbitrary network of fixed size, if connections are allowed to detach and reattach under a mix of preferential and random rules, the degree distribution evolves toward a power law. Running the dynamics with equal amounts of preferential and random reattachment produces a realistic power-law exponent of about 3.21PNAS Nexus. Emergent scale-free networks The upshot is that the familiar “rich get richer” topology of real networks may not require network growth at all; it can emerge purely from rewiring.
Cellular automata offer another angle on emergent computation. These are grids of cells that update according to simple local rules, and some of them produce astonishingly complex global behavior. Researchers have shown that you can construct coarse-grained versions of these automata that capture the large-scale behavior without tracking every cell. The fact that a simpler macro-level description can emulate the system’s behavior is itself a signature of emergence: the meaningful dynamics live at a higher level than the individual cell.22PubMed. Coarse-graining of cellular automata, emergence, and the predictability of complex systems