那是另一道题：论”若当初不这样”恰如”若是3+3”

关于过去的反事实设想，恰如设想发生的不是1+1而是3+3——结果就不是2而是6。 “若是3+3，则得6”——这句话完全正确，谁也驳不倒。 而它对”实际发生的是1+1”这件事，一个字也没说。 你不是在讲那道算式，你在讲另一道算式。 反事实的处境一模一样。“若钥匙没丢，我现在就进屋了”——作为一句蕴含，它也许没错；可它讲的是另一道题。 由此说清了它为什么听起来那么有力：因为算得对。 那份笃定不是关于世界的，是关于运算的：3+3确实得6，这一点毫无疑问。而我们把”这一步推得对”的确定感，误当成了”这件事本来会那样”的确定感。 两种确定感的分别，可以一句话说清：一句蕴含的正确，只保证从那个前提到那个结论这一步没走错；它不保证那个前提是真的。 而这里不必说”更不保证它曾经差一点是真的”——“差一点是真的”就是”不是真的”，加上一个语气。 那个词没有刻度：多近算差一点？ 只有两档：是真的，不是真的。中间没有第三档。 而更要紧的一步是：即使那个反事实宇宙是可行的，它所对应的也不是1+1的世界。 3+3不需要任何人去证明它”可行”——它可行得毫无疑问：算得出来，结果确定，内部无矛盾。而它照样不对应1+1。 于是”可行”和”对应”是两件事，而我们一向把前者当成后者的证据。可行是那一侧内部的性质；对应是两侧之间的关系。而两侧之间，没有任何关系。 这也把”请拿出计算来”那一路问法摆到了它该在的位置：那不是主战场。就算那份计算交出来了，交上来的也是另一道题的答案。 这个类比还带出两件事。 其一：式子不会自己变，是有人换了一个数。 没有人会无缘无故去设想3+3——“若是3+3”这句话之所以被说出来，是有人先把3写在了那儿。 其二：结果不回流。 算3+3不是白算，你确实得到了一个正确的结果；可那个结果属于3+3那道题，它不流回1+1那儿去。 同样，一个反事实推得再周密，周密全部留在那一侧。 而最见分晓的是一处对照：没有人为”没能是3+3”而懊恼。 同样的形式，同样正确的一步推理——一个不带任何东西，一个能让人半夜坐起来。 所以那份不甘，不是从那一步推理里来的：推理两边一模一样。它从别处来。而本文不追那个别处——本文只指出：它不在这一步里。 Abstract A counterfactual about the past is just like supposing that what occurred was not 1+1 but 3+3—whereupon the result is not 2 but 6. “If 3+3, then 6”—entirely correct; no one can refute it. And about the fact that what occurred was 1+1, it says not a word. You are not discussing that expression; you are discussing another one. Counterfactuals are in exactly this position. “If the keys had not been lost, I would be inside by now”—as an implication it may well be correct; but it is about another problem. This explains why such sentences sound so forceful: because the arithmetic is right. That confidence is not about the world but about the operation: 3+3 does make 6, beyond doubt. And we mistake the certainty of “this step is correctly taken” for the certainty of “this is how it would have gone.” The difference between the two certainties can be put in a line: the correctness of an implication guarantees only that the step from that premise to that conclusion was not misdrawn; it does not guarantee that the premise is true. And one need not add “nor that it once nearly was”—“nearly true” just is “not true,” with a tone attached. The word has no scale: how near counts as nearly? There are two settings only: true, and not true. There is no third. More important still: even if that counterfactual universe is feasible, what it corresponds to is not the world of 1+1. No one needs to prove 3+3 “feasible”—it is feasible beyond doubt: computable, determinate, internally consistent. And it still does not correspond to 1+1. So feasibility and correspondence are two different things, and we have always taken the former as evidence for the latter. Feasibility is a property internal to that side; correspondence is a relation between the two sides. And between the two sides there is no relation at all. This also puts the demand “produce the computation” in its place: that is not the main field. Even were the computation delivered, what is delivered is the answer to another problem. The analogy brings out two further things. First: the expression does not change itself; someone changed a number. No one supposes 3+3 for no reason—“if 3+3” gets said because someone first wrote a 3 there. Second: the result does not flow back. Computing 3+3 is not wasted—you do get a correct result; but that result belongs to 3+3, and does not flow back to 1+1. Likewise, however carefully a counterfactual is worked out, the care stays entirely on that side. Most telling is one comparison: no one regrets that it was not 3+3. The same form, the same correctly drawn step—one carries nothing, the other can wake a person at night. So that sting does not come from the step of reasoning