PhysicsCounterintuitive resultExplained★ Famous
The Butterfly Effect
Also known as Sensitive Dependence on Initial Conditions, Lorenz Sensitivity
Can a butterfly in Brazil cause a tornado in Texas? Not by force — but by changing everything that follows.
- Difficulty
- Beginner
- Reading time
- 7 min
Counterintuitive result. Nothing is wrong with the reasoning. The result is simply correct and human intuition is simply unreliable here. Not a contradiction at all.
Discover
December 1961. Edward Lorenz is running a weather simulation on a Royal McBee LGP-30 computer at MIT. He wants to examine a particular sequence of results more closely, so he restarts the simulation from the middle — but to save time he types in the numbers from a printout rather than starting from scratch.
The printout shows 0.506. He types 0.506. The computer stores 0.506127. The difference is one part in a thousand — smaller than a rounding error, smaller than the variation from a puff of wind.
He goes to get coffee.
When he comes back, the simulation has diverged so completely from the original that it might as well be modelling a different planet. The two runs start in lockstep and separate gradually, then rapidly, until after a simulated month there is no similarity at all.
Lorenz understands immediately what he is looking at. The atmosphere is so sensitive to its initial state that an unmeasurably small difference propagates and amplifies until it changes everything. This means the future of the atmosphere is, in principle, unpredictable past some horizon — not because the physics is unknown, not because the equations are wrong, but because no measurement can ever be precise enough.
Understand
The butterfly effect is a property called sensitive dependence on initial conditions. Two states of a system that start arbitrarily close together diverge, exponentially, until they are as different as two random states. The butterfly is not the cause of the tornado — the tornado would have formed anyway, or something like it, from a different perturbation. The butterfly is a way of saying that the two futures — one where the butterfly flaps and one where it doesn't — are eventually completely unrelated.
This is counterintuitive because most systems we interact with are stable. Push a pendulum slightly off angle and it swings back to the same rhythm. Perturb a thrown ball and it lands a little further away — not in a completely different direction. The atmosphere is different. It is a dissipative system driven far from equilibrium, and its dynamics amplify small differences rather than damping them.
Determinism and predictability
The butterfly effect does not make the universe random. Lorenz's equations are perfectly deterministic — given exactly the same starting state they produce exactly the same result. The catch is exactly. In the real world, you cannot measure initial conditions to infinite precision. Every measurement has an error, however small. In a chaotic system, that error grows, exponentially, until it swamps the signal.
This means the butterfly effect draws a sharp line between two things that are often conflated:
- Determinism: whether the future is fixed given the present
- Predictability: whether we can know what that future is
A chaotic system can be perfectly deterministic and completely unpredictable. The future is fixed. We just can't get there from here, because we can never know the present precisely enough to compute it.
The Lyapunov exponent
The rate at which nearby trajectories diverge is measured by the Lyapunov exponent λ. If two states start a distance $\delta_0$ apart, after time $t$ they are approximately $\delta_0 e^{\lambda t}$ apart. A positive Lyapunov exponent means the system is chaotic. For the Earth's atmosphere, λ corresponds to a doubling time of about two to five days — which is why weather forecasts degrade rapidly beyond about ten days, not because the models are bad but because the physics is irreducibly unpredictable.
Examples
| System | Chaos? | Predictability horizon |
|---|---|---|
| Earth's atmosphere | Yes | ~10 days for regional weather |
| Solar system (short term) | Marginally | Millions of years stable |
| Solar system (long term) | Yes | ~5 million years for Mercury's orbit |
| Double pendulum | Yes | Seconds |
| Heart rhythm (healthy) | Slightly | N/A — variability is healthy |
| Heart rhythm (fibrillation) | High | Immediate danger |
Explore
The strange attractor. Lorenz's system, when plotted in three dimensions, traces out a shape that never repeats — a fractal…
The strange attractor. Lorenz's system, when plotted in three dimensions, traces out a shape that never repeats — a fractal object now called the Lorenz attractor. It looks like two wings of a butterfly (the coincidence is eerie and genuine). The system visits both wings unpredictably but never leaves the attractor entirely. This is what chaos looks like geometrically: bounded globally, unpredictable locally.
Is everything chaotic? No. Many systems are stable — perturbations damp out. A ball in a bowl returns to the bottom. A simple pendulum returns to its rhythm. Chaos requires a specific combination of nonlinearity and positive feedback. The atmosphere has it. Most engineered systems are deliberately designed not to have it.
Population biology. Simple equations modelling predator-prey populations, or the growth of a bacterial colony, can exhibit chaos. Robert May showed in 1976 that even a one-parameter equation — the logistic map — produces chaos for certain parameter values. The butterfly effect is not confined to weather.
The popular misconception. In films and popular writing, the butterfly effect is often used to mean "small causes have large effects" — the idea that a tiny decision changes everything. This is not quite what Lorenz meant. His discovery is more specific: in chaotic systems, any uncertainty in initial conditions, however small, renders long-term prediction impossible in principle, not just in practice. It is a fundamental bound, not an engineering limitation.
Historical context
1963. Lorenz publishes "Deterministic Nonperiodic Flow" in the Journal of the Atmospheric Sciences. It is initially ignored by…
1963. Lorenz publishes "Deterministic Nonperiodic Flow" in the Journal of the Atmospheric Sciences. It is initially ignored by meteorologists and physicised by physicists. It is not recognised as the founding paper of chaos theory until the 1970s.
1972. Lorenz presents "Predictability: Does the Flap of a Butterfly's Wings in Brazil Set Off a Tornado in Texas?" at an AAAS symposium in Washington, DC. Philip Merilees, the session chair, invented the title when Lorenz failed to supply one. Lorenz used it anyway. The butterfly is born.
1975. James Yorke and Tien-Yien Li publish "Period Three Implies Chaos" — the paper that introduces the word chaos in its mathematical sense into English.
1987. James Gleick's Chaos: Making a New Science brings the field to a general audience. The butterfly effect becomes a popular cultural concept.
1993. The film Jurassic Park (based on Michael Crichton's 1990 novel) popularises chaos theory further, via Jeff Goldblum's repeated invocation of it.
Why it matters
The limits of prediction. The butterfly effect gives a precise, physical reason why long-range weather forecasting is impossible…
The limits of prediction. The butterfly effect gives a precise, physical reason why long-range weather forecasting is impossible — not "we don't have good enough computers" but "the physics makes it impossible, full stop." This matters for planning, for policy, for any human activity that depends on knowing what next month's weather will be.
Determinism without predictability. Before chaos theory, there was a broad assumption that a deterministic universe was a predictable one — that if the laws were known, the future could in principle be computed. Lorenz demolished this. Determinism and predictability come apart completely in chaotic systems, and the atmosphere we live in is chaotic.
Nonlinearity is everywhere. The butterfly effect forced scientists across many fields to take nonlinear dynamics seriously. Before 1963, most mathematical models of real systems were linearised — simplified to a form that was tractable. Lorenz showed what could be hiding in the parts that were being thrown away.
Modern relevance
Ensemble forecasting. Since individual forecasts degrade rapidly, modern weather prediction runs hundreds of simulations with…
Ensemble forecasting. Since individual forecasts degrade rapidly, modern weather prediction runs hundreds of simulations with slightly varied starting conditions and reports the distribution of outcomes — a direct operational response to the butterfly effect. The forecast you see is a summary of that ensemble.
Climate vs. weather. The butterfly effect limits weather prediction but not climate projection. Climate is the statistics of weather — the average temperature in July, the frequency of hurricanes — and these averages can be stable even when individual trajectories are chaotic. Climate models are not trying to predict next Tuesday; they are modelling the long-run statistics of a chaotic system.
Financial markets. Many economists apply chaotic dynamics to financial markets: prices are path-dependent, small perturbations can trigger large movements, and long-range prediction is fundamentally limited. The models are disputed, but the butterfly effect is part of the serious intellectual case against believing that markets are reliably foreseeable.
Complex systems generally. Ecology, epidemiology, neuroscience, and economics all contain systems that are nonlinear and potentially chaotic. The butterfly effect is the reason "predict and control" is often the wrong ambition — "understand and adapt" is sometimes the only one available.
Sources & further reading
Primary source
(1963)Deterministic Nonperiodic FlowJournal of the Atmospheric Sciences, 20, pp. 130–141.
The founding paper of chaos theory, introducing the strange attractor and sensitive dependence on initial conditions.
How this connects
- Related to
- Moore's LawBoth describe systems where small early changes compound dramatically. The butterfly effect concerns nonlinear sensitivity to initial conditions; Moore's Law concerns systematic doubling that makes a 50-year-old prediction feel like a law of nature.