Butterfly Effect, explained.
The butterfly effect refers to sensitive dependence on initial conditions: small starting differences can grow substantially in a deterministic nonlinear system.
Why it happens
Deterministic means the update rule determines the next state. It does not guarantee that limited-precision measurements permit useful long-range prediction. Repeated amplification can turn a tiny input difference into a large later divergence.
Lorenz's work brought attention to sensitive dependence on initial conditions. A deterministic rule can still have a limited practical prediction horizon.
Read the result
Compare nearby starting states under the same rule, then change the system parameter. Not every parameter produces chaotic behavior. A short overlap followed by divergence is more informative than a single final difference.
A worked example
Two almost identical forecasts
Two model runs start from values that differ by a tiny amount and use the same update rule.
At a sensitive parameter setting, repeated updates amplify the discrepancy until the trajectories separate.
The uncertainty comes from initial precision within the model, rather than an extra random event added at every step.
OPTIONAL DEEPER DETAILGo deeper: inside the model
Inside this model
The logistic map is x(t+1) = r × x(t) × (1 − x(t)). Starting values are 0.4 and 0.4 + epsilon/1,000,000. Both use the same parameter. The chart shows normalized population for each generation, not Lorenz's atmospheric equations.
Where this idea is useful
A practical use
Small measurement errors can matter greatly when forecasting a nonlinear system. Comparing regimes shows why that sensitivity is not equally strong everywhere.
A common misconception
“Chaos means the system has no rules.”
A chaotic system can follow a completely specified deterministic rule. The difficulty is prediction under uncertainty about starting conditions.
What this explanation leaves out
- This one-dimensional toy map is not a weather forecast. Some parameter values have stable cycles; deterministic chaos is different from independent random noise.
Does every small change cause a huge outcome?
No. Sensitivity depends on the system, parameter regime, and time horizon. Stable dynamics can dampen differences, while some nonlinear regimes amplify them.
Is your prediction limited by randomness in the rule or by uncertainty about the initial state?
Associated thinkers
Further reading
Explore the original research or the teaching reference behind this experiment.