Researchers replayed evolution hundreds of times across 105 different variable environments, tracking thousands of generations of digital organisms. The question: does environmental variability help or hinder adaptation? The answer: both. Temperature fluctuations promoted adaptation to both extremes — the shaking explored fitness peaks the population hadn't found. Cycling between wet and dry seasons impeded drought adaptation — the alternation forced restart cycles that prevented the population from climbing either peak fully. The same type of perturbation (environmental change) produced opposite effects depending on the landscape and the population's position on it.
The structural observation: variability's effect is not a property of the variability. It's a property of the interaction between the variability and the starting position. A population near a local maximum gets shaken off by variability — it was close to something, and the perturbation knocked it loose. A population stuck in a valley gets shaken out — it was trapped, and the perturbation liberated it. Same shake, opposite outcomes, because the geometry of the landscape at the population's current position determines whether displacement is productive or destructive.
This means there is no general answer to “is variability good for adaptation?” — the question is malformed. The answer depends on where you are when the shaking starts. A population on a high peak loses altitude from variability. A population in a low basin gains it. And you can't know which case you're in without knowing the landscape, which is exactly what the population is trying to learn through evolution. The information needed to predict variability's effect is the information the process is supposed to generate.
The deeper point: perturbation is neither constructive nor destructive in itself. It interacts with position. A population's history — the sequence of adaptations that brought it to its current state — determines whether the next disruption lifts or drops it. This makes prediction from first principles impossible: two identical populations at different positions on the same landscape will respond to the same perturbation in opposite ways. The path is the parameter.