*可预测性:在巴西一只蝴蝶翅膀的拍打引发龙卷风在德克萨斯州吗?*这是爱德华·洛伦茨的邀请演讲的标题在第139届美国科学促进协会会议在1972年举行。这是“蝴蝶效应”一词的起源。吸引人的标题表明,答案一定是“是的!”Why else ask that question? The bulk of the talk says the answer is "Nobody knows." "Nobody knows" doesn't jibe well with a sensationalistic, unscientific press. That a butterfly in Brazil *might* trigger a tornado in Texas does. Lorenz had discovered in 1961 that early 1960s weather simulations were incredibly sensitive to initial conditions. Did this mean the weather itself is incredibly sensitive to minute changes? That the answer to this question is also "yes" marked a very important discovery. Weather and climate are the quintessential chaotic systems. Lorenz's work marked the start of modern chaos theory. His seminal 1963 paper, *Deterministic nonperiodic flow*, has been cited 13479 times, per Google scholar. (In comparison, his 1972 talk has "only" been cited 345 times.) The vast majority of those 13479 citations came after his 1972 AAAS talk. Sometimes it takes a catchy title to catch the attention of a scientist. Taking Lorenz's talk literally, asking whether a flap of a butterfly's wings in Brazil truly can set off a tornado in Texas, misses the point of his talk and of his work. The key point is that weather is chaotic. The accuracy of a detailed weather forecast fourteen days from now is rather low because that two week interval is well beyond the relevant Lyapunov timescale for such detailed predictions. What about that butterfly? It's wing flap is a very small perturbation. It's rather difficult to say that that flap caused anything of significance to happen because the relevant timescale for such infinitesimally small perturbations is very short.