Xiaoyang Yu

**时间在流动，当前的时刻在时间轴上移动，我们之所以有这种感觉，是因为我们感觉当前时刻具有特殊的属性，它分割了之前与之后：之前的历史是一条时间线，而之后的未来则似乎可以有多条不同的时间线。** **如果我们摒弃这种多条时间线的认知，而承认总共就只有一条时间线，那么我们可以把时间认知为一个新的空间维度，那么时间作为时间维度就不存在了，或者说时间不再流动，而是静止的。** **须先把"此刻"这个说法说精确：它不是线上的一个点，是一道边界。** 边界之前，那些原本"活着的"分岔都已经关闭、收缩成了一条；边界之后，分岔还"开着"。**"此刻在流动"，说的其实是——这道"收缩已完成"的边界，在往前推进。** **按这个定义看，"过去只有一条线"和"未来有很多条线"，不是时间自带的两种性质，是同一件事在边界两侧的两种说法：边界之前＝分岔已关闭，边界之后＝分岔未关闭。"此刻"就是这道关闭线本身，不是线上的一个特殊点，是这道关闭线所在的位置。** **而这道边界，从来没有真正存在过。** 反事实从没有第二条线可以比对，分岔从没有被真正算出来过，每一个"未来的可能性"，都只是被情感从无限个同等适用的候选里选出来的一个构造。**——说的正是：那道"未来仍有活分岔"的边界，从来就不曾真的存在过。分岔那半从来没有开过，只是被感觉成开着。** **若是这样，那句"如果我们摒弃"里的"如果"，可以拿掉。** 不是"如果我们摒弃多条时间线的认知"，是**我们已经有大量理由该摒弃它**——它一直是被凑出来的，不是被发现的。 **而摒弃之后，"此刻"这道边界找不到立足点了。** 若这条线上任何一点，都不曾真的有过"活的分岔"（不论在它之前还是之后），**那么这条线上就没有一个点，比别的点更特殊**——每一点的处境，跟线上别的每一点，完全相同。**"此刻在流动"，要求有一道边界可以移动，而边界没了，移动这件事，连主语都没有了。于是时间不再流动，而是静止的。** **而"之前""之后"不会因此消失，它们会变得跟空间里的"左边""右边"一样：作为一种排序关系，照样成立**——这一点在那一点的前面，是一件跟"那栋楼在这栋楼的左边"同类的事实。**消失的，只是"此刻"作为其中一个点被特殊标记出来的那份地位。** ## Abstract **Time flows; the present moment moves along the timeline. We have this sense because we feel the present moment carries a special property, one that divides before from after: the history before it is one timeline, while the future after it seems capable of having several different timelines.** **If we set aside this cognition of multiple timelines and accept that there is, in total, only one timeline, then we may cognize time as a new spatial dimension—and time as a time dimension ceases to exist, or, put otherwise, time no longer flows but stands still.** **First, "the present" must be made precise: it is not a point on the line but a boundary.** Before the boundary, the branches once "alive" have all closed, collapsed into one; after it, the branches remain "open." **"The present is flowing" really says: this boundary of "collapse complete" is advancing.** **On this definition, "the past has only one line" and "the future has many" are not two properties time itself carries, but one fact stated from either side of the boundary: before it, branches closed; after it, branches not yet closed. "The present" just is this closing-line itself, not a special point on it, but the location where the closing-line sits.** **And this boundary has never truly existed.** Counterfactuals have never had a second line to compare against; branching has never actually been computed; every "future possibility" is merely a construct selected by emotion out of infinitely many equally valid candidates. **This says precisely: the boundary of "the future still has live branches" has never once existed. The branching half was never open; it was only felt to be.** **If so, the "if" in "if we set this aside" can be dropped.** Not "if we set aside the cognition of multiple timelines," but **we already have ample grounds to set it aside**—it has always been manufactured, never discovered. **And once set aside, "the present," as a boundary, has nowhere left to stand.** If no point on this line ever truly had a "live branch"—whether before or after it—**then no point on the line is more special than any other**—every point is in exactly the same situation as every other point. **"The present is flowing" requires a boundary that can move, and with the boundary gone, "moving" no longer even has a subject. So time no longer flows, but stands still.** **And "before"/"after" do not disappear on this account; they become like "left"/"right" in space—still holding as an ordering relation**—this point being before that one is a fact of the same kind as "that building is to the left of this one." **What disappears is only the standing of "the present" as one point specially marked out among them.**

**我主张所有的因果判断都是错的。** **这一步不是一个需要单独证明的新命题，是"因"比"相继"多要的那半句反事实从未被兑现过"这一条论证，若成立，就自动覆盖到底的一个推论。** 要例外，得举出至少一个案例，它的反事实那半句确实被兑现了——**而兑现这半句所需要的东西（一个真实存在的、走了另一条路的世界），根本不存在，不分案例。所以没有一个因果判断能够例外。** **须先把"错"这个字钉死，排除两种误读。** **第一种误读：把它读成"所有因果判断报告的相继事件都没发生"。不是。** 客户确实换了地方，那家店确实倒闭了——**这些相继发生，一件不假。"错"不在这里。** **第二种误读：把"错"读成"这些判断毫无用处，该被抛弃"。也不是。** 取消的是"因"作为一个额外物理关系的资格，**不是取消它记账、预测、协调的用处。** **那么"错"精确指的是什么？指的是：每一个因果判断，都声称了一件从未被满足过的事——它声称自己不只是相继，还兑现了那半句反事实；而这个声称，没有一次是真的。** **由此有一个推论：因果判断之间，没有高低之分。**"隔壁店抢客户导致倒闭"和"他打喷嚏导致倒闭"，在"是否兑现了那半句反事实"这一点上，完全平等——**一个也没有兑现过。** **那么"隔壁店那一版更有依据"这份日常感觉从哪儿来？它量的从来只是相继那一半，不是因果那一半。** 客户确实换了地方，这是一条摸得着的物理链条；打喷嚏没有传导任何东西，这条链条根本不存在。**这份差别，是相继层面的差别——它衡量的是"A之后是B"这条链条扎不扎实，跟"A导致了B"这个额外声称是否被兑现，是两回事。** **一个可以自己验的检验：拿两个感觉不同的因果判断，分开问两个问题——一，有没有一条摸得着的相继链条？二，那半句反事实兑现了没有？** 你会发现第一问的答案千差万别，**而第二问的答案，一律是"没有"。** ## Abstract **I maintain that every causal judgment is false.** **This is not a new claim requiring separate proof; it is the automatic corollary of the argument that the counterfactual half "cause" claims beyond "succession" has never once been redeemed—if that argument holds, it covers every instance without exception.** An exception would require at least one case in which the counterfactual half really was redeemed—**and what redemption requires (a really-existing world that took the other road) simply does not exist, in any case whatsoever. So no causal judgment can be exempt.** **"False" must first be pinned down, to rule out two misreadings.** **The first misreading: taking it to mean that none of the successive events causal judgments report ever occurred. Not so.** The customers really did move elsewhere; the shop really did close—**these successions are undeniable. The falsity does not lie here.** **The second misreading: taking "false" to mean these judgments are useless and ought to be discarded. Also not so.** What is withdrawn is "cause's" standing as an additional physical relation, **not its usefulness for bookkeeping, prediction, and coordination.** **So what precisely does "false" pick out? It picks out this: every causal judgment claims something that has never once been satisfied—it claims to be more than mere succession, to have redeemed that counterfactual clause; and this claim has never once been true.** **A corollary follows: there is no ranking among causal judgments.** "The shop next door poached customers, causing the closure" and "he sneezed, causing the closure" stand on exactly equal footing regarding whether the counterfactual clause has been redeemed—**neither has redeemed it.** **Where, then, does the everyday sense that the first is "better grounded" come from? It has always tracked the succession half, never the causal half.** The customers really did move—a traceable physical chain; a sneeze conducts nothing—no such chain exists at all. **That difference belongs to the level of succession—it measures how solid the chain "A followed by B" is, which is an altogether different matter from whether the further claim "A caused B" has been redeemed.** **A test anyone can run: take two causal judgments that feel differently plausible, and ask two separate questions—first, is there a traceable chain of succession? second, has the counterfactual clause been redeemed?** You will find the first answer varies enormously, **while the second answer is, uniformly, "no."**

**利用反事实设想来思考，利用大量的反事实设想，并把它们组合起来，作为思考的依据、线索——这是一种强大、复杂到惊人的、奇迹般的能力。** **同时在脑子里跑几十个甚至上百个平行的、未曾发生的世界，交叉比对，权衡取舍，从中理出一条思路**——这确实是一项极其复杂的认知壮举。 **但我们把每个反事实设想称为一个因果关系，利用因果关系来思考，就变得似乎普通而司空见惯了。** **事实上，利用大量的因果关系来思考，其实质就是利用大量的反事实设想来思考——是同一件事，只是换了一个名字。** **而这个换名，不是没有代价的。代价，是把一项本该显眼的认知劳动，伪装成了不需要被看见的东西。** **"发现"和"构造"这两个词，暗示的东西不一样。**"发现"暗示东西早就在那儿，你只是碰巧看见了——**不需要交代过程，好像谁站在那儿都会看见。**"构造"如实地要求交代过程：**你怎么想到要比较这几个候选，怎么排除了别的候选，怎么权衡出了这一个。** **"因果关系"这个说法，把构造伪装成了发现——于是构造所需要的那份复杂劳动，也跟着一起被伪装掉了，显得好像什么劳动都没有发生，事情本来就是那样。** **由此可以得出一条推论：一件事显得越"理所当为常"，往往恰恰说明它背后藏着的那份认知构造，被伪装得越彻底，而不是说明这件事真的越简单。**"普通、司空见惯"，不是这件事真的普通，是伪装做得好。**日常生活里那些听起来最不需要解释的因果判断，可能恰恰是最复杂的一批构造——只是复杂到我们从来没被要求正视它。** **而这不是要否定这项能力，恰恰相反：本文要做的，是把这项能力从"因果关系"这层伪装底下重新请出来，让它作为它本来的样子——一项强大、复杂到惊人的、奇迹般的能力——被正视一次。** ## Abstract **Thinking by means of counterfactual supposition—using a great many counterfactual suppositions and assembling them together as the basis, the thread, of one's reasoning—is a capacity powerful and astonishingly complex, something like a marvel.** **Running dozens, even hundreds, of parallel, never-occurred worlds in one's head at once, cross-comparing them, weighing them against each other, and drawing out a single line of thought from them**—this is indeed a formidable feat of cognition. **But once each counterfactual supposition is called a causal relation, and we speak of thinking by means of causal relations, the very same thing comes to seem ordinary, commonplace.** **In fact, thinking by means of a great many causal relations is, in substance, thinking by means of a great many counterfactual suppositions—one and the same activity, merely renamed.** **And this renaming is not without cost. The cost is that a labor of cognition that ought to stand out in plain view is disguised as something needing no notice at all.** **"Discovery" and "construction" imply different things.** "Discovery" implies the thing was already there, and one merely happened to see it—**no accounting of the process required, as though anyone standing there would have seen the same.** "Construction" honestly demands such an accounting: **how did you come to compare these particular candidates, how were the others excluded, how did you arrive at weighing out this one.** **"Causal relation" disguises construction as discovery—and so the complex labor that construction requires is disguised along with it, making it appear as though no labor occurred at all, as though matters were simply so.** **From this a corollary follows: the more "self-evident" a thing appears, the more likely it is that the cognitive construction hidden behind it has been disguised all the more thoroughly—not that the thing is genuinely simpler.** "Ordinary, commonplace" does not mean the thing itself is ordinary; it means the disguise has worked well. **The causal judgments in daily life that seem to need the least explanation may be exactly the most complex constructions of all—only complex to a degree we have never been asked to face.** **This is not meant to deny the capacity; quite the opposite: what this paper does is bring that capacity back out from beneath the disguise of "causal relation," so that it may be seen once, as it truly is—a capacity powerful and astonishingly complex, something like a marvel.**

**因果关系并非存在于因事件到果事件之间的一根实存的、有物理作用的、额外的线。那么因果关系是什么？** **因果关系建立在反事实思考之上，似乎是反事实思考得出的某种结论，而该结论似乎反映了世界那边（而不是仅仅在我们心灵里）的某种本质结构。这是一种非常有迷惑性的感觉。** **事实上，因果关系仅仅也恰恰只反映了该反事实思考本身，该因果关系仅仅就是该反事实思考的别名而已，而无任何该反事实思考之外的额外的实质性内容。** 当你说出一个具体的因果关系时，**你就是在用另一种隐晦的方式来描述你头脑中的那个对应的反事实思考而已**——这是一种语言的游戏，语义的陷阱，**这里不存在任何对于世界那边的揭露。** **而纯粹的"别名说"还差一步，得先补上，否则会解释不了一件事：我们平时说一个人"搞错了因果关系"，这份"错"从哪儿来？** **若"X导致了Y"纯粹只是"我头脑中正设想着一个没有X、也没有Y的世界"这次心理活动的别名，那么这句话会变得永远为真、不可能为假**——因为只要说出这句话，那次设想确实发生了。**而这跟我们实际使用这个词的方式对不上：我们说的"搞错了"，从来不是"他没有在想那个反事实"（他显然在想），是"他不该那么笃定"。** **所以因果句报告的，除了那次设想本身，还有一份姿态：把这个特定的候选，当成无数个同等适用的候选里"该被信的那一个"，来对待。** **姿态，才是可以为假的地方。** 错不在于设想没有发生，**在于这份姿态里那句隐含的"而且这是对的"，没有任何依据**——它假装这份选中经过了世界那边的验证，而实际上，它只是从无限个同等适用的故事里，被此刻的情感挑中的那一个。 **一个可以自己验的检验：把因果句翻译回纯粹的设想句，看信息有没有丢。**"隔壁店导致了这家店的倒闭"，翻译成"我设想了一个不存在隔壁店的平行宇宙，那里这家店没有倒闭"——**若两句真是同一件事的别名，翻译前后该等量。而若翻译之后觉得"少了点什么、没那么有力了"，那份丢掉的东西，正是那份被拆穿的确信姿态——它在"导致"这个说法里能悄悄躲藏，在"设想"这个说法里却无处可躲。** **一个必然会被搬出来的反驳：可我设想那个平行宇宙时，用的是真实的物理、经济规律啊，这难道不算一次对世界的揭露？** 不算——**推演内部正确，证明的只是这个设想内部自洽，跟"这个世界本来会不会走到这条设想里"，是两件事。用对了规律，答的仍然是另一道题。** **所以这里没有对世界那边的揭露。这里只是说出了你头脑中的那个反事实思考——外加一份不该有的确信——换了一种表达方式。** ## Abstract **Causation is not a real, physically efficacious extra line running from a cause-event to an effect-event. So what is causation?** **Causation is built on counterfactual thinking; it seems to be some conclusion reached by that thinking, a conclusion that seems to reflect some essential structure on the world's side—not merely in our minds. This feeling is highly deceptive.** **In fact, causation reflects only that counterfactual thought itself; the causal relation is merely another name for that counterfactual thought, with no substantive content beyond it whatsoever.** When you state a specific causal relation, **you are only describing, in another, disguised manner, the corresponding counterfactual thought in your head**—this is a language game, a semantic trap, **and there is no disclosure of the world's side here at all.** **But the pure "alias" account needs one more step, or it cannot explain something: we do speak of a person having "gotten the causal relation wrong"—where does that wrongness come from?** **If "X caused Y" is purely an alias for the mental act of supposing a world without X and without Y, then the sentence becomes necessarily true whenever uttered, incapable of being false**—since merely uttering it guarantees the supposition genuinely occurs. **But this does not match how we actually use the word: "getting it wrong" never means "he isn't supposing that" (he plainly is); it means "he ought not to be so confident."** **So a causal sentence reports, beyond the supposition itself, a further stance: treating this particular candidate as the one, among infinitely many equally applicable candidates, that ought to be believed.** **The stance is where falsity can attach.** The error is not that the supposition failed to occur; **it is that the implicit "and this is correct" carried by the stance has no basis whatsoever**—it pretends this selection has been verified by the world's side, when in fact it is merely the candidate picked out, from infinitely many equally valid stories, by the present emotion. **A test anyone can run: translate the causal sentence back into a pure supposition and see whether any information is lost.** "Next door caused this shop to close," translated as "I am supposing a parallel universe without the shop next door, in which this shop did not close"—**if the two really are aliases for one thing, the translation should be lossless. And if, after translating, something feels missing, less forceful, that missing thing is precisely the exposed confidence-stance—one that can hide quietly inside "caused" but has nowhere to hide inside "supposing."** **An objection sure to arise: but I constructed that parallel universe using real physical and economic laws—doesn't that count as a disclosure about the world?** It does not—**getting the internal reasoning right shows only that the supposed scenario is internally coherent, which is a different matter from whether the actual world would have gone that way. Using the correct laws still answers another problem.** **So there is no disclosure of the world's side here. There is only the counterfactual thought in your head—plus an unwarranted confidence—restated in another mode of expression.**

**世界上不存在真正的因果关系。** **先把两件事分开，因为它们从来不是一件事：相继发生，和因果关系。** **相继发生是真的，从没有人否认过：一个状态之后是另一个状态，按规律演化，一段接一段，处处相接。** 前面所有的论证，一次也没有质疑过这一点。 **而"因果关系"比"相继发生"多要的那一半，从来不是一条物理链条本身，是一句反事实：要不是X，就不会有Y。** 没有这半句，"他抢占了客户，隔壁店因此破产"和"他打了个喷嚏，然后隔壁店破产了"，在结构上没有任何分别——**都只是A之后跟着B。** 让第一句显得比第二句更"真"的，不是链条（客户确实换了地方），是我们暗自加的那句反事实：**要不是抢客户，隔壁店就不会破产。** **而这句反事实，从来没有被兑现过。** 它没有第二条线可以比对，没有一份被真正完成的计算，是照着结论往回配出来的一个前件，是情感从无限个同等适用的故事里选中的那一个。**这一份欠条，一次也没有被兑现过——不是因为兑现很难，是因为要兑现它，需要一样东西（一条真实存在的、走了另一条路的世界），而这样东西从来不存在。** **于是"因"比"相继"多要的那一份，从来没有到账。** **一个必然会被搬出来的反驳：物理定律不是明明白白写着F=ma吗？改变F，a不就跟着变了？** 而这是一次计算，不是一次读数：**把不同的F代进公式、算出不同的a，这一步毫无问题——可它算的是另一道题，不是"这个世界本来会怎样"这道题。**F=ma告诉你的是一套关系式在数学上成立，**不告诉你"如果那天F真的不同，这个宇宙此刻会在哪儿"——后一句，仍然要一条从没被写出来过的第二条线。** **所以剩下的，只有相继。** 一条不分岔的线，一段接一段地往下走，**没有"为什么是这样而不是那样"，只有"就是这样"。** **须说准取消的是什么：取消的是"因"作为一个额外物理关系的资格，不是取消它作为记账、预测、协调工具的用处。** 说"是他导致了这件事"，在生活里极其好用——它分配责任，指引改进，协调众人的行动。**这份用处，本文一个字不碰。取消的只是"这份好用，来自它捕捉到了一个世界那边真实存在的额外关系"这一步推断。** ## Abstract **There is no genuine causal relation in the world.** **Two things must first be separated, since they were never one: successive occurrence, and causation.** **Successive occurrence is real, and no one has ever denied it: one state follows another, evolving by law, one stretch after the next, everywhere in contact.** No argument in this whole line of work has ever questioned this. **And what "causation" claims beyond mere succession has never been a physical chain in itself; it is a counterfactual clause: had it not been for X, there would not have been Y.** Without that clause, "he poached the customers, so the shop next door went bankrupt" and "he sneezed, and then the shop next door went bankrupt" share exactly the same structure—**merely A followed by B.** What makes the first sentence feel truer than the second is not the chain (the customers really did move) but the silently added counterfactual: **had the customers not been poached, the shop would not have gone bankrupt.** **And that counterfactual clause has never once been redeemed.** It has no second line to compare against, no computation ever actually completed; it is an antecedent fitted backward to a wanted conclusion, the one selected by emotion out of infinitely many equally valid stories. **This debt has never once been paid—not because payment is hard, but because paying it would require something (a really-existing world that took the other road) that never exists.** **So what "cause" claims beyond "succession" has never been delivered.** **An objection sure to arise: doesn't physics plainly state F=ma? Change F and doesn't a change accordingly?** But this is a computation, not a reading: **substituting a different F into the formula and computing a different a is unobjectionable—but it computes another problem, not the problem of "how this world would actually have gone."** F=ma tells you a relation holds mathematically; **it does not tell you "had F actually been different that day, where this universe would now stand"—that further claim still requires the missing second line, never once produced.** **So what remains is only succession.** One unbranching line, proceeding stretch by stretch, **with no "why this and not that," only "it is thus."** **What is withdrawn must be stated precisely: withdrawn is "cause's" standing as an additional physical relation, not its usefulness as a tool of bookkeeping, prediction, and coordination.** To say "he caused this" is enormously useful in life—it assigns responsibility, guides improvement, coordinates collective action. **That usefulness is not touched. What is withdrawn is only the further inference that this usefulness derives from having captured some additional relation genuinely present on the world's side.**

**各种可见的事件之间没有关系，那么我们又如何在它们之间建立关系的呢？** **我们建立关系，利用了一些简单的、错误的模式，其中一种是直觉式的零和——也就是加法。** **一个人似乎"惨"，我们就会自动去找出一个"爽"的人，并认为爽的人以某种方式造成了惨的人的惨。** 这似乎意味着，**我们默认快乐和痛苦的总和应该保持为零**——把快乐记正数，痛苦记负数，一份负数进了账，我们就去找一份能配平它的正数。**找到那个爽的人，认定他某个在先的行为导致了惨的人的惨，就行了，我们就满意了，因此不会再去寻找其他不同的可能解释方式。** **而要从"两批发生"走到"一个因"，中间有三步验证，而这三步一步也没走过：**第一步，确认两批发生之间存在物理接触或信息传递的链条；第二步，确认这条链条的方向；第三步，确认这条链条的量级——爽的人那份爽，是否真的构成了惨的人那份惨的完整原因。**三步一步没走，只因为账面"看起来"平了，我们就止步了。** **这是一次记账动作，只是账本从"谁该负责"换成了一份假想的世界总量表。** 这份总量表假定：世界有一个总的快乐—痛苦余额，而这个余额应该守恒——**跟"这块砖的移动要记在谁名下"是同一类操作：先预设一份账本的存在，再把观察到的事塞进对应的格子里，而账本本身的存在，从没有被证明过，只是被默认了。** **一个可以自己验的检验：去看总量不平的情况，是怎么被处理的。** 若这条守恒律是真的，应该存在同样多的反例——一个人惨了，却怎么也找不到一个对应的、爽的人。**这时通常的处理，不是承认账不平，是继续找，直到找到一个够格的"爽的人"为止**——把"爽"的范围放宽，把时间线拉长，把间接程度放松，直到某个人、某件事，勉强能被塞进那个正数的格子里。**——一条真正的物理守恒律，不需要我们努力去找才能配平；需要努力寻找配平对象的"守恒律"，恰恰暴露了它不是从世界那边读出来的，是我们自己需要它成立，所以在主动凑。** **这条模式之所以显得"简单"，是因为它把三步验证压缩成了一次算术。** 不需要指出物理关系，不需要指出方向，不需要指出量级，**只需要凑出一个正负号相反、量级感觉差不多的对象，账面上一平，我们就满意了，因此不会再去寻找其他不同的可能解释方式。**

**不同的时间点，比如事前、事后，对同一个事件的预测、认知不同——从这个经验，能不能得出结论：该事件并非被预先确定？** **不能。而理由不在于"决定论有多种版本"，是这里根本不涉及本体论意义上的决定论。** **须先立一个区分。预先确定，并非专指决定论式地预先计算出唯一结果。** 这句话里藏着两个不同的问题： **认识论决定论**：某个事件，能不能被预先计算出唯一结果——这是一个关于**我们、关于计算能力、关于信息可得性**的问题。 **本体论决定论**：某个事件，本身是否只有一种可能的走法——这是一个关于**世界自身**的问题，跟有没有人在算、算不算得出来，毫无关系。 **而这两者之间的关系是单向的：无法预先计算出唯一结果，不等于结果本身不唯一。** 前者可能仅仅是因为信息不够、算力不够；世界那边完全可以是本体论决定论的，只是没有一个足够强大的计算者，能把那唯一的结果提前算出来。**"算不出来"是我们这边的一句报告，"不唯一"是世界那边的一件事，中间隔着一道从未被填上的鸿沟。** **按这道鸿沟重新看"预测会变"这个例子：它从头到尾只触及认识论层面。** 事前信息不完整，事后信息完整，一个理性主体的预测（哪怕是一组带概率的可能结果）随之更新——**这跟事件本身有没有被预先决定，毫无关系；一个彻底决定论的世界里，信息不完整的预测者，照样会看到自己的预测随数据增加而更新。这一步无法从认识论层面跨到本体论层面。** **要从认知上否定本体论决定论，需要什么？需要证明：存在某个完整的事前状态，它确实可以导向不止一种事后结果——而且这份证明本身得是认识论可及的，不能只是"我算不出唯一结果"。** 而我们手上，从来只有后者——算不出唯一结果，从没有人交出过"确实有多种可能"的证明。**无法从认知上否定本体论决定论的可能性，那么就无法排除本体论决定论。** **即使量子力学，也无法否定本体论决定论——即使它支持认识论决定论。** 波函数本身的演化，由薛定谔方程给出：**给定完整的量子态，下一时刻的量子态是唯一的、可以精确计算的**——这是教科书级别的认识论决定论。**而随机性只出现在测量结果这一层**：对这一层，量子力学只给概率分布。**即使拿到了波函数这一层完美的、确定的演化规律，也没能帮我们回答测量结果在本体论上是不是被决定的——两层之间那道缝，没有被那部分的确定性填上。** **核心证据是：存在跟标准量子力学在全部实验预测上完全等价、却在本体论上彻底是决定论的诠释——德布罗意-玻姆力学（导航波理论）**，它给粒子指派了确定的、始终存在的位置和轨迹，测量结果由这些隐变量的初始值完全决定，只是我们测不到这些初始值。**这说明：量子力学的全部经验内容，对本体论决定论和本体论非决定论，给出的是完全相同的预测——它在这道题面前，预测力为零。** ## Abstract **At different points in time—say, before and after an event—our predictions or understanding of the same event differ. From this experience, can we conclude the event was not predetermined?** **No. And the reason is not that "determinism comes in versions"; it is that this experience does not touch ontological determinism at all.** **A distinction must first be established. "Predetermined" does not specifically mean computing a unique outcome in advance in the deterministic sense.** Two different questions hide beneath that phrase: **Epistemic determinism**: whether a given event can have a unique outcome computed in advance—a question about **us, about computational power, about the availability of information.** **Ontological determinism**: whether the event itself admits only one possible course—a question about **the world itself**, unrelated to whether anyone is computing it or can compute it. **And the relation between the two runs one way: an inability to compute a unique outcome does not entail that the outcome itself is not unique.** The former may simply reflect insufficient information or computing power; the world may perfectly well be ontologically deterministic while no computer powerful enough exists to derive that unique outcome in advance. **"Cannot compute" is a report about us; "not unique" is a fact about the world—and the gap between them has never once been bridged.** **Reviewing "predictions change" through this gap: it touches only the epistemic layer.** Information is incomplete beforehand, complete afterward, and a rational agent's prediction (even one given as a set of possible outcomes with probabilities) updates accordingly—**this has nothing to do with whether the event itself was predetermined; in a fully deterministic world, a predictor with incomplete information will still see predictions update as data accrues. This step cannot cross from the epistemic to the ontological.** **What would it take to deny ontological determinism from cognition? It would require showing that some complete prior state genuinely admits more than one subsequent outcome—and this showing must itself be epistemically accessible, not merely "we cannot compute a unique result."** And what we have in hand has only ever been the latter—no one has ever produced proof that multiple outcomes genuinely obtain. **Unable to deny the possibility of ontological determinism from cognition, we are therefore unable to rule it out.** **Even quantum mechanics cannot deny ontological determinism—even though it supports epistemic determinism.** The evolution of the wavefunction itself, given by the Schrödinger equation, is **textbook epistemic determinism: given a complete quantum state, the next state is unique and precisely computable.** **Randomness appears only at the layer of measurement outcomes**, where quantum mechanics gives only probability distributions. **Even possessing this perfect, deterministic law for the wavefunction's evolution does not answer whether measurement outcomes are ontologically determined—the gap between the two layers is not bridged by that layer's determinism.** **The core evidence: there exists an interpretation empirically equivalent to standard quantum mechanics in every experimental prediction, yet wholly deterministic ontologically—de Broglie–Bohm pilot-wave theory**, which assigns particles definite, ever-present positions and trajectories, with measurement outcomes fully determined by the initial values of these hidden variables, which we simply cannot measure. **This shows: the entire empirical content of quantum mechanics gives identical predictions under ontological determinism and ontological indeterminism—on this question, its discriminating power is zero.**

**居于康威生命游戏中的一台图灵机模式，也可以在它的宇宙中定义某种空间距离的概念，甚至它也可以设法去逼近它的宇宙中基本单元间的"距离"。** 这不是一句让步，是一件它做得到、而且能做得很精密的事：**它可以按照自己那套规则的相互作用结构——哪些格子直接影响哪些格子的下一状态、信息在几步之内能从这儿传到那儿——建起一整套关于"近""远"的理论，做实验，验证，提炼出不变量，跟人类做物理学没有本质分别。** **但它不可能发现石头阵列中石头间的距离，或者计算机内存中分别存储两个细胞生死值的内存地址之间的距离——而后面这两种距离，才是它的宇宙基本单元之间的真正距离。** **而这两种距离的关系，不是精确与粗略的关系，是可能毫无关系、甚至截然矛盾的关系。** 最尖锐的证据是：**计算机内存中，甚至可以用同一个内存地址，来同时存储两个细胞的生死值**——这意味着，两个被它自己那套内部物理学判定为"隔着整个宇宙、彼此最远"的格子，在真正的基座层面，可能相距为零。 **这不是内部理论不够精细、有待改进的问题。恰恰相反：内部理论越是被造得严谨、越是能在自己的宇宙里被反复验证、越是能推广、越是能被别的观察者独立确认——它就越确定地，跟基座层面的真实关系毫无瓜葛。** 因为这份严谨性、这份可推广性，来自同一个源头：**这套内部距离概念，是完全由规则自身的因果、相互作用结构定义的，而这套结构可以在无数种截然不同的基座上被同样实现**——用石头，用电路，用一整间仓库分散摆放的开关，用任何东西，只要它们的因果关系照着同一条规则运作。**正是这份"不挑基座"的能力，才让这套概念显得那么可靠、那么客观、那么放之四海而皆准——而这份能力的代价，是它对基座本身，一个字也说不出来。** **这对我们意味着什么：假若我们的宇宙，也如前文所设，是可以被数学模拟的，那么我们的物理学——无论多么精密、多么被反复验证、多么能预测未来——所描述的那套"距离""空间"，跟本文所说的图灵机模式那套内部物理学，处在完全相同的位置上。** 它能做的一切，都是在描述规则自身的相互作用结构；**至于承载这套规则运作的东西，实际以什么样的空间关系存在——甚至存不存在"空间关系"这回事——我们的物理学，从定义上，永远够不着，也永远不会因为更精密而更够得着。** ## Abstract **A Turing-machine pattern living inside Conway's Game of Life can define its own concept of spatial distance within its universe, and can even try to refine and approach the "distance" between the basic units of that universe.** This is not a concession but something it can do, and do with real precision: **it can build, from the interaction structure of its own rule—which cells directly affect which other cells' next state, how many steps information takes to travel from here to there—an entire theory of "near" and "far," run experiments, confirm results, extract invariants, in a manner no different in kind from human physics.** **But it cannot discover the distance between the stones in a stone-array implementation, or between the memory addresses in which two cells' life-or-death values are separately stored—and it is these latter two distances that are the true distance between the basic units of its universe.** **And the relation between the two is not one of precision versus crudeness; it may be no relation at all, or flatly contradictory.** The sharpest evidence: **a computer's memory can even use one and the same address to store both cells' values simultaneously**—meaning two cells its own internal physics judges to be at opposite ends of its universe, maximally far apart, may, at the level of the true substrate, be separated by nothing at all. **This is not a matter of the internal theory being insufficiently refined and awaiting improvement. Quite the opposite: the more rigorously the internal theory is built—the more it withstands repeated confirmation within its world, the more it generalizes, the more independently it can be verified by other observers within that world—the more certainly it has nothing whatsoever to do with the true relations at the substrate level.** For that very rigor and generalizability spring from one and the same source: **this internal concept of distance is defined entirely by the causal and interactional structure of the rule itself, and that structure can be equally realized on countless, utterly different substrates**—stones, circuits, switches scattered across an entire warehouse, anything at all, so long as their causal relations follow the same rule. **It is precisely this substrate-indifference that makes the concept seem so reliable, so objective, so universally valid—and the price of that indifference is that it can say not one word about the substrate itself.** **What does this mean for us: supposing our own universe is, as argued elsewhere, one that could be mathematically simulated, then our physics—however precise, however repeatedly confirmed, however predictive of the future—describes a "distance," a "space," that stands in exactly the same position as the Turing-machine pattern's internal physics.** Everything it can do is describe the interaction structure of the rule itself; **as for what spatial relations, if any, actually hold among whatever carries that rule into execution—our physics, by definition, can never reach that, and will never come closer to it through greater precision.**

**如果组成宇宙的物质最终可以归于某种最基本的单元，而假如最基本单元的性质是可以用数学方式来描述的，比如可以根据n个基本粒子的状态，用数学公式推导出它们在下一个时刻的状态（不一定要唯一），那么这样的宇宙，就是可以用数学来模拟的。** **"不一定要唯一"这半句是全篇的枢纽**：一条给出概率分布、而非单一确定结果的规则，照样算数——概率分布本身也是可以用数学描述的对象。**这一步把"这个宇宙是不是决定论的"从整个论证里排除了出去：不管规则给出的是一个值还是一族值，只要它能被写成数学，这个宇宙就"可以被数学模拟"。** **假如另有一个对这样的宇宙的数学模拟，那么居于被模拟的宇宙之中的人或任何主体，就无法判断其所居是否本来就是一个数学模拟——或者说，二者对内部个体而言，是一模一样的。** **这里要把"一模一样"和"无法分辨"分开，因为它们不是同一个强度的话。** "无法分辨"留着一道口子：也许确实有个差异，只是内部的人够不着。**而"一模一样"把这道口子也堵上了：不是差异存在而接触不到，是压根没有剩余的内容，能让"差异"这个词挂靠。** 一旦一个宇宙的完整内容被"基本单元的状态＋一条更新规则"穷尽地写出来，"这是不是模拟"这一问就没有留下任何进一步的真值条件——就像给定一个三角形三边的完整长度，再去问"它真的是用蓝墨水画的，还是黑墨水画的"，那是一个规格说明书从未打算回答、也没有位置留给的问题。**任何"真差异"，都必须落在这份完整规格之外——而落在一个系统的完整规格之外，就是落在"没有那个位置"上。** **用摆在方格里的石头，来机械地执行跟屏幕上康威生命游戏一样的更新规则**，对场里的一个个体（比如一个图灵机模式）而言，**它无法分辨自己所在的宇宙到底是石头的还是数字的——它不会感受到任何不同**，因为"石头做的"和"数字做的"这两个标签，从来没有出现在"基本单元状态＋更新规则"这份完整规格里。 **而当人、主体、个体探究的尺度逼近最基本单元时，会撞上另一件不同的事：日常概念本身开始难以沿用。**"基本单元之间是否有距离、有空间"，这个问题所沿用的日常空间概念，在这个尺度下难以沿用——**因为日常概念说到底只是主体自己内部的一种模糊感觉，而并非外部世界本身。** **两件家具之间有空间，这个概念其实只是表明我们曾经有过把这两件东西（或另外两件类似的）推到一起、直到没有距离的经验。** 距离，是这份"推到一起"的经验尚未被兑现的量；接触，是这份经验已经被兑现。**日常概念，是对日常经验的一种特殊的转述，一种特殊的表达方式。** **我们没有把两个原子推到一起的经验。而就算有，也显然与把两个家具推到一起的经验不同类。** 用仪器测出的原子间作用力曲线，是真实的经验，**但它是"读一条曲线、看一个数字"这一类经验，不是"用力，然后停下来"那一类。** 更要紧的是：家具那份被推的感觉，本身是由天文数字量级的原子间相互作用**整合、聚合**而成的一份读数，而单独一次这样的相互作用，从来没有到达能在神经系统里产生"被推、被挡住"这份感受的门槛。**——这不是"同样的经验、只是更小"，是根本不在同一个刻度上的两回事，就像问一粒沙是多小的一座沙丘。** **两处相遇：不是发现了世界的底层真相，是撞上了接入方式的边界。** 一处是"站在系统内部"这种接入方式的边界；一处是"由推、挡、停这类身体经验聚合出来的概念"这种接入方式的边界。**边界不同，但撞上的都是边界，不是世界的底层。** ## Abstract **If the matter composing the universe ultimately reduces to some most basic units, and if the properties of those units are mathematically describable—such that, given the mathematically describable states of n particles, a formula yields (not necessarily uniquely) their states at the next moment—then such a universe is one that can be mathematically simulated.** **"Not necessarily uniquely" is the hinge of the whole argument**: a rule that yields a probability distribution rather than a single determinate outcome still counts—the distribution itself is a mathematically describable object. **This move removes "is this universe deterministic" from the argument entirely: whatever kind of output the rule gives, one value or a family of them, so long as it can be written in mathematics, the universe "can be mathematically simulated."** **Suppose there is also a separate mathematical simulation of such a universe; then any person or subject residing within the simulated universe cannot judge whether their home was itself, all along, a mathematical simulation—or rather, the two are identical for the individual located within either.** **"Identical" must be distinguished from "indistinguishable," since they are not claims of the same strength.** "Indistinguishable" leaves a gap open: perhaps a real difference exists, merely out of the internal individual's reach. **"Identical" closes that gap too: it is not that a difference exists but cannot be reached; it is that there is no remaining content left for the word "difference" to attach to.** Once a universe's complete content has been exhaustively written as "the states of its most basic units plus an update rule," the question "is this a simulation or not" has no further truth-condition left over to answer—just as, given the complete lengths of a triangle's three sides, asking whether it was "really" drawn in blue ink or black ink is a question the specification never had room to settle. **Any "genuine difference" would have to fall outside that complete specification—and falling outside a system's complete specification is falling on a position that does not exist.** **Using stones arranged on a grid to mechanically execute the same update rules as an on-screen run of Conway's Game of Life**, for an individual within that field (say, a Turing-machine pattern), **it cannot distinguish whether its universe is made of stone or of digital signal—it feels no difference at all**, because the labels "made of stone" and "made of digital signal" never appear anywhere in the complete specification of "basic-unit states plus an update rule." **And when a person, subject, or individual investigates at a scale approaching the most basic units, a different thing is encountered: everyday concepts themselves become hard to apply.** Whether there is "distance" or "space" between basic units—a question relying on the everyday concept of space—becomes hard to sustain at that scale, **because everyday concepts, in the end, are only a vague feeling internal to the subject, not the external world itself.** **That two pieces of furniture have space between them only shows that we have had the experience of pushing these two things (or two similar things) together until there was no distance left.** Distance is the amount of that pushing-together experience not yet realized; contact is that experience realized. **An everyday concept is a special retelling, a special mode of expression, of everyday experience.** **We have no experience of pushing two atoms together. And even if we did, it would plainly be of a different kind from pushing two pieces of furniture together.** An instrument's force-versus-distance curve between atoms is a real experience, **but it is the kind of experience one has reading a curve, a number on a screen—not the kind of experience of exerting force and then stopping.** More to the point: the felt experience of pushing furniture is itself an integration, an aggregation, of an astronomical number of interatomic interactions, and a single such interaction has never once reached the threshold at which a nervous system registers being pushed against and stopped. **This is not the same experience, merely smaller; it belongs to an entirely different scale of description—like asking how small a dune one grain of sand is.** **The two halves meet at one place: not a discovery about the bottommost layer of the world, but a collision with the edge of a mode of access.** One is the edge of "standing inside a system" as a mode of access; the other is the edge of "concepts aggregated out of pushing, blocking, and stopping" as a mode of access. **Different edges—but both are edges, not the world's bottom.**

**我所有的体验，都是被给予的。所有这些被给予的体验，加在一起似乎是如此的真实，似乎这就是某种"真实"的东西。** **它们的加法非常特别：是从不同的方向、角度过来的各种不同性质的体验的同时给予**——既有仿佛是物质世界的一种所与，又有某种情感、想法、因果关系、反事实设想等，**一并被给予**，这些又综合起来刻画一个人、一个东西、一件事，乃至所有人、事、物。**比如内部（情感、因果认知）和外部（物理活动）共同刻画了一个朋友。** **这份加法看起来像是两路独立证据的相互印证：一路是"外面确实有这么个人"，一路是"我对他确实有这些感受"，两路对上了，于是那个朋友显得格外真实。** **而这不是印证，是一次成形。** 就像滑翔机不是先独立存在、再被几帧格子印证——它是"把几帧对起来看"这个动作一次做出来的；**"朋友"也不是先独立存在，再被内外两路给予汇聚证实。"朋友"就是这次异质汇聚本身，汇聚和被汇聚描述的那个对象，是一次成形的，不是先有后证。** **而这份加法之所以显得比同质的汇聚更真实，恰恰因为它是异质的：感觉像是来自不同渠道的东西，比感觉像是来自同一渠道的东西，更容易被当成独立的核验。而"感觉像来自不同地方"和"确实来自不同地方"，中间那一步，从没有人核实过。** **须再往下一层：连"这一份算外部、那一份算内部"这个分法，也是我们事后画的一道圈，给予到场时根本没有这道缝。** 那份"像是物质世界"的质地，和那份"像是情感"的质地，都只是给予自带的两种texture，不是两个不同来源发来的标签。一场梦、一段格外鲜活的回忆、一次幻觉，可以拥有跟真实知觉一模一样的"外部感"——**那份外部感是给予的一部分内容，不是给予确实来自外部的证明。** **于是内、外这两个格子本身，也是我们后来为了归档而画的。** 我们把某一批质地归进"外部"那一栏，把另一批归进"内部"那一栏，然后回头说：你看，内外两栏对上了，所以这个东西是真的。**而两栏本来就是同一批给予被我们自己拆分出来的两半，不是两个独立信源各自发来的报告。** ## Abstract **All my experience is given. All this given experience, added together, seems so real, as though this were itself some kind of "reality."** **The addition is peculiar: it is the simultaneous givenness of qualitatively different kinds of experience arriving from different directions and angles**—there is what seems a given of the material world, and there is emotion, thought, causal relation, counterfactual supposition, **all given at once**, and together these characterize a person, a thing, an event, indeed every person, event, and thing. **Inside (feeling, causal cognition) and outside (physical activity) together characterize a friend.** **This addition looks like two independent streams of evidence corroborating each other: one stream says "there really is such a person out there," the other says "I really do have these feelings about him"; the two agree, and so the friend seems all the more real.** **But this is not corroboration; it is formation at one stroke.** Just as the glider was never independently there and then confirmed by several frames—it is what the act of setting the frames side by side produces at once—**"the friend" is not independently there and then confirmed by inner and outer streams converging. "Friend" just is the name for that heterogeneous convergence itself; the convergence and the object it is said to confirm are formed together, not proved after the fact.** **And this addition seems more real precisely because it is heterogeneous: what feels as though it comes from different channels is more readily taken as independent verification than what feels as though it comes from one channel. And the step from "feels as though from different places" to "actually from different places" has never once been checked.** **One layer further: even the sorting of "this counts as outer, that counts as inner" is a ring we draw afterward; the given, on arrival, carries no such seam.** The texture that feels "like the material world" and the texture that feels "like feeling" are both merely textures the given arrives with, not labels sent by two different senders. A dream, an especially vivid memory, a hallucination can carry exactly the same "outer feel" as an actual perception—**that outer feel is part of the content given, not proof that it truly came from outside.** **So the two bins, inner and outer, are themselves something we drew afterward for filing purposes.** We sort one batch of texture into the "outer" column and another into the "inner" column, and then say: look, the two columns agree, so this thing is real. **But the two columns were always one batch of givenness that we ourselves split in half; they were never two independent sources each filing its own report.**

**因果关系、责任认定、反事实场景的可信感，其实都是靠情绪、情感加持、背书的。** **"主体在做事"这种思维方式，其实是一种特殊的逆转、奇怪的翻转。** 我们把情感、情绪，视为对于情境的注脚、反应，于是它们就似乎变成了某种需要被基于情境而合理解释的客观之物，**情境也变得与情感恰好匹配**——而实际上，**情感、情绪是与情境同时被给予的，情感、情绪是纯粹主观的。** **而这一点在因果、责任、反事实这几处，可以说得更具体、更狠：并非我们对眼前发生的故事产生对应的情感，而是此刻的情感，将决定我从无限多种同等适用的故事中，选择哪一个。** **同一批发生，可以配上无限多个同等适用的故事**——他辜负了我，或者这本就是意料之中的事，或者这跟我没关系。**没有一个故事能被那批发生本身淘汰掉**——从没有人交出过一份能把候选压缩到一个的计算。**而真正在候选里挑一个出来的，从来不是那份计算，是此刻的情感。** **愤怒挑出"他辜负了我"；麻木挑出"这不是什么大事"；恐惧挑出"这预示着更糟的事"。** 同一批发生，只需要换一种此刻的情感状态，挑出来的故事就跟着换。 **而通常我们体验到的次序是反的：先感知到故事，然后对这个故事产生对应的情感。** 实际的次序是——**情感先到场，是它选中了这个故事；而故事一旦被讲出来，我们立刻把次序读反，说成是这个故事引发了这份情感。** **这里面两端都是主观**：被选中的那个故事是主观的，替它撑腰的那份笃定感也是主观的。**而俩主观相互印证，就似乎成了客观**——而这份"印证"什么也没多证明，因为**两端是同一台机器的两个输出端口**，天然对得上，不是两次独立的测量。 **情感实际上是迷惑性的**——它把自己乔装成了故事的后果，而不是故事的作者。 **须把结论说准：这是一个认识论判断，不是一条行事的指令。** **取消的是这些判断作为外部因果事实之独立证据的资格，不是取消情感的发生，更不是建议谁去忽视、压制或不信任自己的情绪体验。** 那份被伤害的感觉是真的，那份痛苦本身不需要任何故事来验证它的实在——**取消的只是"这份笃定感独立证明了那个具体的因果叙事是唯一正确的"这一步。** ## Abstract **The credibility of causal attribution, of assigning responsibility, of a counterfactual scene, is in fact underwritten by emotion.** **Thinking of "a subject doing things" is a peculiar reversal, an odd inversion.** We take emotion to be a footnote to, a reaction to, the situation—whereupon it seems to become an objective thing requiring explanation by reference to that situation, **and the situation comes to match the emotion exactly**—when in fact **emotion and situation are given at the same time; emotion is purely subjective.** **And on causation, responsibility, and counterfactuals this can be put more sharply: it is not that we produce an emotion corresponding to the story before us, but that the present emotion decides which of infinitely many equally applicable stories we select.** **The same set of happenings admits infinitely many equally applicable stories**—he wronged me, or this was to be expected all along, or it has nothing to do with me. **None is eliminated by the happenings themselves**—no one has ever produced a computation narrowing the candidates to one. **What actually picks one out is never that computation; it is the present emotion.** **Anger selects "he wronged me"; numbness selects "this is nothing"; fear selects "this forebodes worse."** The same happenings, and merely a different present emotional state suffices to swap which story is picked. **And the order we experience is reversed: we perceive the story first, then produce the corresponding emotion.** The actual order is—**the emotion arrives first and selects the story; and once the story is told, we instantly re-read the order, saying the story caused the emotion.** **Both ends here are subjective**: the selected story is subjective, and the certainty backing it is subjective too. **Two subjectives corroborating each other come to seem objective**—yet the corroboration proves nothing further, because **both ends are two output ports of one machine**, naturally in agreement, not two independent measurements. **Emotion is, in fact, deceptive**—it disguises itself as the effect of the story rather than revealing itself as its author. **The conclusion must be stated with precision: this is an epistemic judgment, not a behavioral instruction.** **What is withdrawn is these judgments' standing as independent evidence of an external causal fact—not the occurrence of the emotion, still less any counsel to ignore, suppress, or distrust one's own emotional experience.** The feeling of having been hurt is real; the pain itself needs no story to validate its reality. **Withdrawn is only the further step: that this certainty independently proves the specific causal narrative to be the one correct account.**

**事情的发生里并没有灵活度，并没有像关节内裕量一样的空间。** **不是说事情有关节，关节有裕量，裕量内都行——但我们会按照"有裕量"一般，去设想事情的不同发生。** **先把"裕量"这个比喻说清楚，因为它比看起来更容易被误用。** 一颗螺栓拧进螺纹，拧3.0圈或3.1圈，功能上算一样——这个"算一样"不是螺栓自己的物理属性，**是造这套螺栓的人事先规定的一条等价关系**。裕量从来不是世界白送的，是设计出来的宽容。**而事情的发生没有这样一份设计。** 上一刻给出下一刻，没有一条写在物理里的"这个范围内都算数"的条款。 **而我们却照着"有裕量"去设想。** "要是我当初没搬家"——这句话默认了，在那一刻，本来存在着一个可以宽松取值的区间，而"搬"与"没搬"都在这个区间之内。**而那个区间从未被出示过。** **有趣的是，这种设想活动本身，作为一个事情，同样没有裕量。** 你脑子里想出"要是当初没搬家"这句话的那个具体过程——哪个神经元先兴奋，想到哪个细节，停在哪一句——**这个过程本身，也是一段没有裕量的发生**。你不可能"差不多"地想出这个念头；你想出的，就是你确切想出的那一个，一丝不差。 **对这些设想活动的设想、思考，作为一个事情，同样如此。** 你此刻正在做的这件事——读这段话、反思刚才那句设想、觉得"这个说法有点道理"——**这也是一段没有裕量的发生**，跟它要描述的对象是同一类东西。 **事后的分析、思考，通常不被认为具有与事情本身一样的性质，似乎事后的设想是超脱的，尤其是此刻的（哲学）思考似乎是超脱的。但这显然不是事实。** **所有这些都是被给予的**：事情，对事情的设想，对设想的思考，关于这一切的哲学思考——同样是被给予的。**没有一样是超脱的、跳出的。** **当我们生气或感到情绪时，我们一定是感觉自己处于某个超脱的位置——实际上并未也不可能超脱。** 任何时候我们都不可能真正位于超脱的位置——**没有那个位置**。作为系统内的主体、物体、模块、零件，**不可能到达超越系统的位置——不可能脱离系统或与系统分离。** ## Abstract **Events, as they unfold, contain no give, no space like the clearance inside a joint.** **This is not to say events have joints, joints have clearance, and anything within the clearance goes—yet we imagine events' alternative unfoldings exactly as if there were such clearance.** **First the metaphor itself must be made precise, since it is more easily misused than it looks.** A bolt turned three full turns or 3.1 turns is, functionally, "the same"—but that sameness is not a property of the bolt; **it is an equivalence stipulated in advance by whoever designed the assembly.** Clearance was never given free by the world; it is a designed tolerance. **The unfolding of events carries no such design.** One instant gives the next with no clause anywhere in physics reading "anything in this range counts." **Yet we imagine as though there were clearance.** "If only I hadn't moved that year"—the sentence presupposes that, at that moment, a lenient interval existed within which "moved" and "did not move" both fell. **That interval has never been produced.** **Interestingly, the very activity of imagining, as an event itself, has no clearance either.** The specific process by which the thought "if only I hadn't moved" arose in your head—which neuron fired first, which detail occurred to you, where the thought stopped—**that process is itself a happening with no give.** You could not have thought it "roughly"; what you thought is exactly what you thought, to the letter. **And thinking about these acts of imagining, as an event, is the same.** What you are doing right now—reading this passage, reflecting on that earlier supposition, feeling that "there is something to this"—**is likewise a happening with no clearance**, of the same kind as the object it describes. **Retrospective analysis and thought are not usually taken to share the same nature as the event itself; retrospection seems detached, and present-moment (philosophical) reflection especially seems detached. But this is plainly not so.** **All of it is given**: the event, the imagining of the event, the thinking about the imagining, the philosophical reflection on all of this—equally given. **Nothing here is transcendent, nothing stands outside.** **When we grow angry, or feel any emotion, we surely feel ourselves occupying some transcendent standpoint—yet we neither do, nor could.** At no time can we truly occupy a transcendent position—**there is no such position.** As a subject, object, module, or part within a system, **it is not possible to reach a position beyond the system—not possible to detach from it or stand apart.**

**我们后悔时，是对当前的状态不满，我们觉得当初就能看到当前这个状态。** **这不需要造一条从未发生的支路。被抱怨的那个状态，就是实际发生、此刻已知的那一个——A点。** **而后悔要求的，只是一件事：那份此刻（B点）才有的认识，倒着传回A点。** **"我早该想到会是这样。"** 这句话的意思是：**站在A点的我，本该已经拥有此刻站在B点才拥有的这份认识。** **而A点那一刻，B点还没发生。任何要传给A的东西，都还不存在。** **没有东西可传，就没有东西可倒寄。** **——所以不是"我没收到那封信"，是那封信在A点还没有被写出来。** **这跟"人只是视线不能拐弯"是同一条限制，只是错认的位置不同，而且方向相反。** **在走的那一刻（A点），错认是：** 因为看不见后面的形状，就以为自己"把直的弄弯了"（一份把不存在的原型误认为存在的错觉）。 **而在回头看的这一刻（B点），错认是：** 因为已经站在了拐弯之后，看见了整段形状，**就以为那份"看见"原本在A点就该有**（一份把此刻的认识误认为当时就有的错觉）。 **一个在线的前端犯错，一个在后端犯错——而两个错认，用的是同一条被压扁的四维线。** **一个可以自己验的检验：找一次你说过"我早该想到"的时候，问自己——那句话，是在事情发生之前说的，还是之后？** **从来都是之后。** **没有人在A点说"我现在已经看到B点会是这样"。** 那句话，只在站到B点之后才会被说出来——**而正是这个"只在之后"，暴露了这份认识的真正产地：它是B点自己的，不是A点的。** ## Abstract **When we regret something, we are dissatisfied with the present state, and feel that we could have seen this present state coming.** **This requires no imagined branch that never occurred. The state being complained of is exactly the one that actually happened and is now known—point A.** **And regret asks for only one thing: that the knowledge available now, at point B, be mailed back to point A.** **"I should have known it would come to this."** The sentence means: **the self at A ought already to have possessed the knowledge that the self at B now possesses.** **But at the moment of A, B had not yet occurred. Nothing that could be sent to A existed yet.** **With nothing to send, there is nothing to mail back.** **—So it is not "I never received that letter"; it is that the letter had not yet been written at A.** **This is the same limit as "the line of sight simply cannot bend," only the site of the error differs, and runs the other way.** **At the moment of walking (point A), the error was:** unable to see the shape ahead, one takes oneself to have "bent something straight" (an illusion mistaking a nonexistent original for an existing one). **At the moment of looking back (point B), the error is:** having already stood past the bend and seen the whole shape, **one takes that "seeing" to have been owed already at A** (an illusion mistaking present knowledge for knowledge held at the time). **One error at the front end of the line, one at the back—and both use the same flattened four-dimensional line.** **A test anyone can run: recall a time you said "I should have known," and ask—was that said before the event, or after?** **Always after.** **No one at A says "I can already see how B will turn out."** That sentence is said only once one is standing at B—**and this "only afterwards" exposes the true provenance of the knowledge: it belongs to B, not to A.**

一台由基本粒子构成的机械、结构—— 你把它想象为主体，是什么意思？ 你所想象的它的另一种行为，又是什么意思？ 这两问不问对错，只问那句话说了什么。 先答第二问，因为它是第一问的工具。 “把它想象为主体”，做的是四件事： 其一，给它一个名字——从此那一堆里的一片被当成一件。 其二，给它一份延续——昨天那一片和今天这一片，算同一个。 其三，给它一个内外——这些算它的，那些算环境。 其四，给它一个”要什么”——于是它的动作可以被读成”它在试图……”。 四样做完，那台机械就”成了主体”。没有第五样。 ——所以”把它想象为主体”的意思，就是做这四件事；不多也不少。而四样一样也不在那一堆粒子那儿，全是加上去的。 由此可以答第一问。 “它本可以不那样动”——请把这句话翻译成关于粒子的话：哪一批粒子，在哪一刻，变成什么？ 一试就知道，连从哪儿下手都不知道。 而这不是译得不好，是没有可译的内容。 那句话给出的是一个圈的名字，加一个动词——而这两样都不是那一堆粒子的语言。 为什么？因为那个”它”是加出来的，而加出来的东西没有内部结构可供改动。 你能改的，只有名字下面的那句话；改不了名字下面的那堆粒子。 Abstract A machine, a structure, made of elementary particles— what does it mean to imagine it as a subject? And what does it mean to imagine a different behaviour of it? Neither question asks whether this is right; both ask what the sentence says. Take the second first, since it is the tool for the first. “To imagine it as a subject” is to do four things: One, give it a name—thereafter a patch of that heap counts as one item. Two, give it continuity—yesterday’s patch and today’s count as the same one. Three, give it an inside and an outside—these count as its, those as environment. Four, give it a “what it wants”—so that its motions can be read as “it is trying to…” With those four the machine has “become a subject.” There is no fifth. —So “imagining it as a subject” means doing those four things, no more and no less. And not one of the four is over there in the heap of particles; all are added. Which answers the first question. “It could have moved otherwise”—translate that into particle-talk: which particles, at which instant, changed to what? Try it, and one does not even know where to begin. And this is not a poor translation but the absence of anything to translate. The sentence supplies the name of a ring plus a verb—and neither is the language of that heap. Why? Because that “it” was added, and what is added has no interior to alter. What you can alter is only the sentence under the name; you cannot alter the particles under it.

我想喝水，于是我伸手拿起了杯子。 这句话里，是”我”发起了这件事。 可”我”是什么？ 是从这一大片正在变的东西里，圈出来的一块。而这个圈是我们自己画的：“我”到哪儿为止？ 皮肤？呼进去的空气？肚子里那些帮着消化的细菌？手里握着的杯子？没有一条现成的缝在那儿，是我们为了办事划的。 那真正在那儿的是什么？是那些最小的零件。 于是就有了一个两头都不成的局面： 一头——“我”看起来有想法（我想喝水），可它不在：它只是一个圈。 另一头——那些最小的零件确实在，可它们没有想法：它们不想去哪儿，只是照着规矩动。 一头不在，一头不想。 而这两件事其实是同一件。因为”想要”这个词，本来就是给大块头用的。 它要有一个想的人，要有想要的东西，还要想的那个和后来做的那个是同一个——三样都得先有一个圈。 圈没了，“想要”这个词跟着掉下来。 所以”最小的零件怎么会想要什么”这个问题，答案不是”它们不想”，而是：这个词根本没搬到那边去过。 有人会说：“一个零件不想，可它们合起来就想了呀。”——而”合起来”就是那个圈。 你得先划一道边，才谈得上”边里面这些合起来”；而边是你划的。 所以剩下的结论只有一个：没有”发起”这回事。 “发起”要两样：一个起点，和一条从起点出去的线。而两样都得先画个圈。 剩下的是什么？ 是那一大片，一直在变。手确实抬起来了，杯子确实到了另一个高度——一件不少。 只是那些变化不是”从某个地方出发”的。 Abstract I want a drink, so I reach out and pick up the cup. In that sentence, “I” initiated the whole thing. But what is “I”? A patch ringed off from this great spread of changing things. And we drew the ring ourselves: where does “I” stop? The skin? The air breathed in? The bacteria helping digest? The cup in the hand? No ready-made seam is there; we drew it to get things done. What is actually there? The smallest parts. So neither side works: One side—“I” seems to have wants (I want a drink), but is not there: it is only a ring. The other side—the smallest parts really are there, and have no wants: they do not want to go anywhere; they simply move by the rules. One is not there; the other does not want. And these are really one thing. For “wanting” is a word made for the big chunks. It needs someone who wants, something wanted, and the wanter and the later doer to be the same one—all three need a ring first. Take the ring away and “wanting” comes down with it. So the answer to “how could the smallest parts want anything” is not “they don’t,” but: the word never travelled there at all. Someone will say: “one part doesn’t want, but together they do.”—And “together” is the ring. You must first draw a boundary before there is any “these inside it, together”; and you drew it. So only one conclusion remains: there is no such thing as initiating. “Initiating” needs two things: a starting point, and a line running out from it. And both need a ring drawn first. What is left? The great spread, still changing. The hand really did rise; the cup really did reach another height—not one bit of it is denied. Only, those changes did not set out from anywhere.

中间层所指的，似乎是个对象、主体。而中间层不存在——是说没有这个主体。 那么，主体的不同行为就无法从主体自身发起，而只能从一个个的基本粒子处发起。 而基本粒子又如何有发起不同活动的意愿呢？ 这是一次两难，两边各缺一样，且缺的正好是对方有的： 中间层这一边——有意愿的样子（我想走那条路），可它不在：圈是画的，边界随用途变。 基本粒子这一边——在，可它没有意愿：一个粒子不想去哪儿，它只是照着规矩走。 ——一头缺存在，一头缺意愿。 而这不是两个坏消息，是同一件事的两面。因为”意愿”本来就是中间层的词。 它要一个持有者（谁的意愿）、一个对象（想要什么）、一份贯穿前后的同一性（想的和后来做的是同一个）——三样都要一个圈。取消了圈，“意愿”这个词跟着掉下去。 所以”基本粒子怎么会有意愿”这一问，答案不是”它没有”，而是：这个词根本没往那边搬。它是从圈那儿借来的，而借出方不在。 一条常见的退路也因此不通：“单个粒子没有意愿，可它们合起来就有了。” ——而”合起来”就是那个圈。 说涌现，等于说：我在这一片上划一道边，边里面的那些东西合起来有一个意愿；而那道边是我划的。涌现不是一个新的位置，它是中间层换了个名字。 于是落点不是”发起在别处”，是：没有发起这回事。 “发起”跟”活动者的活动”一样，是一次成形的复合体：一个起点，一个从它出去的动作。而两样都要一个圈。 Abstract What the middle layer denotes seems to be an object, a subject. And that the middle layer does not exist means there is no such subject. Then the subject’s different behaviours cannot be initiated by the subject itself, but only by individual elementary particles. And how would an elementary particle have the will to initiate a different activity? This is a dilemma, and each side lacks precisely what the other has: The middle layer—has the look of will (I wanted to take that road), but is not there: the ring is drawn, its boundary shifting with use. The elementary particles—are there, and have no will: a particle does not want to go anywhere; it merely follows the rules. —One side lacks existence, the other lacks will. And these are not two pieces of bad news but two faces of one. For “will” is a word of the middle layer. It requires a holder (whose will), an object (wanting what), and an identity running through (the wanting and the later doing being of one) —all three require a ring. Withdraw the ring and “will” falls with it. So the answer to “how would a particle have will” is not “it does not,” but: the word never travelled there. It was borrowed from the ring, and the lender is not there. A common retreat is thereby closed: “single particles have no will, but together they do.” —And “together” is the ring. To speak of emergence is to say: I draw a boundary on this patch, and the things inside it together have a will; and I drew the boundary. Emergence is not a new position; it is the middle layer under another name. So the conclusion is not that initiation lies elsewhere, but: there is no such thing as initiation. “Initiation,” like “an agent’s activity,” is a compound formed at one stroke: a starting point, and an act issuing from it. And both require a ring.

整体存在。或者，基本粒子们存在。中间层不存在。 这一句比”中间层不是主体”更硬一档：不是没有一个滑翔机在走，是没有滑翔机。 而它站得住，理由只有一条：圈是画上去的。 那儿有的是格子。滑翔机是我们在格子上圈出来的一片。 而一个圈出来的东西，“存在”这一栏怎么填？填不了。 因为它的边界不由那边给——换一种用途，圈就换一处，而它跟着变。一样跟着我的用途变的东西，不在那边。 “或者”这个词是准的。 不是”两者都存在”，是这两种说法各自成立，而它们说的是同一件事：说整体存在，就没有多余的部件，因为部件都在里面；说基本粒子们存在，就没有多余的整体，因为整体就是它们全体。两种记法，同一样东西——而中间层在哪一种记法里都没有位置。 必然会来的反问是：“那我不存在？” 而回法不是安慰，也不是硬顶，是指出这句话的形状：“我”正是那个中间层的名字。 问”我存不存在”，等于问”这个圈存不存在”——而圈是画的。 而这不否认任何一件发生：呼吸在发生，走路在发生，此刻读这行字这件事在发生。一件不少。 取消的是”这些之外还有一个持有者”。 最后要挡一处：这不是一个宇宙论。“整体存在”不是说有一个大容器装着小东西——它是说：除了那一整片，没有另一份清单。 Abstract The whole exists. Or: the elementary units exist. The middle layer does not. This is a step harder than “the middle layer is not a subject”: not that no glider is moving, but that there is no glider. And it holds for one reason: the ring is drawn on. What is there are cells. The glider is a patch we ringed off on the cells. And for something ringed off, how is the column “exists” to be filled? It cannot be. For its boundary is not given from that side—change the use and the ring moves, and it moves with it. What shifts with my purposes is not on that side. The word “or” is exact. Not “both exist,” but that the two statements each hold and say the same thing: say the whole exists and no parts are left over, since the parts are inside it; say the elementary units exist and no whole is left over, since the whole is all of them. Two notations, one thing—and the middle layer has no place in either. The inevitable retort: “then do I not exist?” The reply is neither consolation nor defiance but a remark on the shape of the question: “I” is the name of that middle layer. To ask whether I exist is to ask whether this ring exists—and the ring is drawn. And this denies no happening: breathing happens, walking happens, the reading of this line happens. Not one is diminished. What is withdrawn is a holder behind them. One last thing to block: this is not a cosmology. “The whole exists” does not posit a great container holding small things—it says: besides that one stretch, there is no second inventory.

你在电影里看见一个人在跑。 可胶片上有什么？一格一格的画面。 每一格都是不动的。没有一格里有人在跑。 那个”跑”是从哪儿来的？ 是把这些格子一格接一格连起来看，才出现的。 而这一连，是你这边做的。 世界上的事情，也是这样。 杯子被拿起来了——这件事是真的：杯子的位置变了，手的位置变了，一堆东西的位置都变了。 而”有一个人，他把杯子拿了起来”——这句话里多出来一样东西。 多出来的是那个”他”，和那个”拿”。 而这两样是一起出现的： 没有那个”他”，那些位置变化就只是位置变化；没有那个”拿”，那个”他”也说不出是什么。 两个一起来，一个也不能单独来。 这不是说你看错了。 看错，得有一个对的看法在旁边——比如你把8看成了6，那儿确实有个8。 而这里没有那个”对的看法”。那儿本来就没有一个”他”可看，也没有一个”拿”可看。 不是你看错了那儿有什么，是你看见了那儿没有的东西。 那到底有什么？两头： 一头是整个世界——所有东西加上它们照着的那套规矩。 另一头是最小的那些零件——组成一切的那些。 而中间那些——一个人、一只手、一辆车、一家公司——都是我们圈出来的一块。 圈是我们画的。而圈里没有一个”它”在那儿做事。 最后一句得老实说：这篇文章每一句都在犯它自己说的毛病。“我们圈出来的”“你这边做的”——每一句都又摆了一个做事的人进去。 而这躲不掉：说话就得有个主语。 而这篇文章说的正是——那个位置本来是空的。 Abstract You see a man running in a film. But what is on the strip? Frames, one after another. Each is still. In no single frame is anyone running. So where does the running come from? From running the frames together, one after the next—that is when it appears. And the running-together is done on your side. Things in the world are like this too. The cup was picked up—that is real: the cup’s position changed, the hand’s position changed, a great many positions changed. But “there is a man, and he picked up the cup” adds something. What it adds is the “he” and the “picking up.” And those two arrive together: without the “he,” those changes of position are merely changes of position; without the “picking up,” the “he” has nothing to it. Both come at once, and neither can come alone. This is not to say you saw wrongly. Seeing wrongly requires a right seeing alongside it—if you read an 8 as a 6, there is really an 8 there. Here there is no such right seeing. There was never a “he” to see, nor a “picking up” to see. It is not that you saw wrongly what is there; you saw what is not there. Then what is there? Two ends: One end is the whole world—everything, together with the rules it all runs by. The other end is the smallest parts—the ones everything is made of. And everything between—a person, a hand, a car, a company—is a patch we have ringed off. We drew the ring. And inside the ring there is no “it” doing anything. One last honest remark: every sentence here commits the fault it describes. “We ringed it off,” “done on your side”—each puts a doer back in. And there is no avoiding it: to speak is to have a subject. And what this paper says is precisely—that position is empty.

我们感知的活动者的活动，是幻觉。 先把范围收准。那批发生是真的：格子在开关，位移在发生，空气在振动。本文一个字也不否认它们。 幻觉的对象，是一个复合体：活动者—活动。 而这两个成分是一次成形的，互为条件。 不是先有一个活动者、再看见它活动；也不是先有活动、再去找一个活动者。 没有活动者，那些发生就不成其为”活动”；没有活动，那个活动者就没有内容。 两个一起出现，一个也不能单独出现。 在康威生命游戏里，这一点可以指着看。 那儿有的是：几帧格子的开关。 而把这几帧对起来看，“滑翔机”和”它在走”同时出现。 没有滑翔机，也没有”走”这回事——有的只是一片开关的图案，在相邻的几帧里彼此相像。 而对起来看，是这一侧做的。 那份”它在动”，是这一动作的产物。 所以”幻觉”这个词在这里用得住，而且用的是本义，不是”误判”那个引申义。 因为”看错”要有一个正确的看法可对照。 而这里没有：那批发生本来就没有活动者可看，也没有活动可看。 不是我看错了那儿有什么，是我看见了那儿没有的一样东西。 而这跟”结构不是主体”是同一条的两面。 语言的形状要一个主语，而那个位置若非填不可，只有两端填得进去：整个宇宙（物理定律的迭代），或所有基本粒子。 即：没有作为主体、同时又非最小单元的局部。 两端之外，没有活动者，也没有活动——而我们偏偏在那儿看见了一个东西在动。 Abstract The activity of an agent, as we perceive it, is an illusion. First, the scope. The happenings are real: cells switch, displacements occur, air vibrates. The paper denies none of it. What the illusion is of is a compound: agent-and-activity. And the two components are formed at one stroke, each conditioning the other. It is not that an agent exists first and is then seen to act; nor that activity exists first and an agent is then sought for it. Without an agent those happenings do not amount to “activity”; without activity that agent has no content. They arrive together, and neither can arrive alone. In Conway’s Game of Life this can be pointed at. What is there: the switching of cells across a few frames. And on setting those frames side by side, “the glider” and “it is moving” appear together. There is no glider, and no moving—only a pattern of switches resembling one another across adjacent frames. And the setting side by side is done on this side. That sense of “it is moving” is the product of that act. So “illusion” holds here in its literal sense, not in the derived sense of “misjudgment.” For “seeing wrongly” requires a correct seeing to be compared with. And there is none: in those happenings there was never an agent to see, nor an activity to see. It is not that I saw wrongly what is there; it is that I saw something that is not there. And this is the other face of “a structure is not a subject.” The form of language requires a subject, and if that position must be filled, only two ends will fill it: the whole universe (the iteration of physical law), or all the elementary units. That is: there is no subject that is a part and not a minimal unit. Outside the two ends there is no agent and no activity—and there we nonetheless see something moving.

我在三维空间里动——随着时间的流逝而改变位置。我在三维空间里也可以不动，保持静止。这里的动，就是改变三维空间中的位置。 那么：我在四维时空里可以选择动或静吗？四维时空里还有动或静的概念吗？ 如果把四维时空的时间轴看作空间轴，那么我是这个四维空间中的一个静物。 像一块石头，从一头到另一头，摆在那儿。 问它”要不要动一下”，没有意义。 而要害在下一句：这里不是”我不动”，是”动”这个概念没有落点。 “不动”是”动”的一个选项——一样东西可以动而没动，那叫静止。而四维那儿连选项都没有：不是我选了静，是那儿没有”动”这一栏。 ——并非对错的问题，是这个词在这里没有适用的对象。 而变化也一样。 三维空间里存在变化，动就是变化的一种。 而”变化”这个词自带两个时刻：一个”之前”，一个”之后”。 把四维时空看作四维空间之后，那两个时刻都在里面了。 它们不是”之前”和”之后”，是这一头和那一头。 ——就像一根棍子的两端：你不会说”棍子从这端变成了那端”。 所以那儿没有变化，也没有”改变”。而我们日常意识不到这一点，误以为在四维空间中还是可以”动”、还是有”变化”“改变”的。 而”我本来可以改变它”这句话，要的正是这个不存在的”改变”。 而那份错觉的产地，可以指出来。 我拿起了杯子，我认为我改变了杯子的静止——是因为我在当时还看不到杯子在四维空间中的形状，而这个形状蕴涵了杯子会被拿起。 杯子在四维空间中的形状，本来就是那个样子：直一段，弯一段。 它不是一根竖着的柱子后来被弯了一下。而我在伸手的那一刻，只看得到它的一小截——那个还没弯的部分。于是我以为我把一根直的东西弄弯了。 ——而实际是：我从来没见过那根直的东西。我见过的，只是那个形状的前半截。 所以”改变”要求的那个”原来是那样”，是从我只看到前半截来的：看得少，就以为看到的那一截是它的全貌。改变是一份视野的产物，不是杯子的属性。 而按同一条，“我”也一并被解释了：我那只手在四维空间中的形状，也已经包含了那一段伸出；我以为是我发动了它，同样是因为在伸手那一刻，我看不到自己的形状。 ——所以”我改变了杯子”这句话里，两头都是同一次视野不足。 而这份不足是无法当场补上的：能被感觉到的每一样，都在那条线里面。 Abstract I move in three-dimensional space—changing position as time passes. I can also stay still. Motion here just is change of position in three-dimensional space. So: can I choose to move or stay still in four-dimensional spacetime? Is there any concept of motion or rest there at all? If the time axis of spacetime is taken as a spatial axis, then I am a still object in that four-dimensional space. Like a stone laid out from one end to the other. Asking it whether it would like to move means nothing. And the crux is what follows: this is not “I do not move”; it is that “motion” has nowhere to land. “Not moving” is one option under “moving”—a thing that could move and does not is at rest. In four dimensions there is not even the option: I have not chosen rest; there is no column marked “motion.” —Not a matter of right or wrong, but of a word with no object here. And the same holds for change. Change exists in three-dimensional space, and motion is one kind of it. But “change” carries two moments with it: a “before” and an “after.” Once spacetime is taken as four-dimensional space, both moments are inside it. They are not before and after but this end and that end. —Like the two ends of a stick: one does not say the stick changed from this end into that one. So there is no change there, and no altering. And in daily life we do not notice this, supposing that in four-dimensional space we may still “move,” that there is still “change,” still “altering.” And “I could have changed it” demands precisely that nonexistent altering. And the source of the illusion can be named. I pick up the cup and take myself to have changed its rest—because at that moment I cannot yet see the cup’s shape in four-dimensional space, and that shape already implies its being picked up. The cup’s four-dimensional shape simply is that shape: a straight stretch, then a bend. It is not an upright column later bent. But as I reach out I see only a small part of it—the part not yet bent. So I take myself to have bent something straight. —Whereas in fact I never saw the straight thing. What I saw was the first half of that shape. So the “it was otherwise” that “change” requires comes from my seeing only the first half: seeing little, one takes the seen part for the whole. Change is a product of the field of view, not a property of the cup. And by the same token “I” am explained too: my hand’s four-dimensional shape already contains that reaching; I take myself to have initiated it only because, at the moment of reaching, I cannot see my own shape. —So in “I changed the cup,” both ends are the same shortfall of view. And the shortfall cannot be made up on the spot: everything that can be felt is inside the line.

本文质疑一个被广泛使用却极少被检查的推理：某个高层概念具有良好的预测能力，因此世界中真的存在一个与之相应的高层实体。本文指出这一步不成立。首先，我们对这一推理的使用是不一致的：重心、平均家庭、空穴、赤道皆有卓越的预测能力，却无人据此设定相应实体——重心甚至可以落在空无一物之处。而在湿润、温度、生命、心灵这些概念上，同一步骤却被顺理成章地走出。本文考察可能的分界判据，逐一表明它们或循环、或不成立，实际起作用的只是使用习惯的深浅。其次，本文给出两条正面理由：其一，把一个高层概念追到底，得到的不是一个对象而是一张”这样或那样”的清单，而”或”是运算，不在世界之中；其二，完整的底层叙述不使用该概念也能讲完，反之不能，凡在完整叙述中不被需要者，不取得存在席位。第三，“划分切中了世界的关节”这一常见解释是循环的：它把预测成功重新命名，挂到世界身上，再回头用它解释预测成功。本文主张，预测成功是划分与后续之间的关系——依此线归为一类者，其后续彼此相似——这是关于接下来发生什么的事实，无需任何预先分好的实体。举证责任落在断言实体存在的一方。本文最后表明，被撤销的是那个问法，而非任何预测、任何现象。 Abstract This paper challenges an inference widely used yet rarely examined: that because a higher-level concept has good predictive power, there must exist a corresponding higher-level entity in the world. The paper argues this step fails. First, our use of the inference is inconsistent: center of mass, the average family, the semiconductor hole, and the equator all possess excellent predictive power, yet no one posits corresponding entities for them—the center of mass can even fall where nothing whatever is present. Yet with wetness, temperature, life, and mind, the same step is taken as a matter of course. The paper examines candidate criteria for the division and shows each to be either circular or false; what actually operates is nothing but the depth of our habituation to a word. Second, two positive arguments are given: tracing a higher-level concept to its base yields not an object but a list of the form “this way or that way,” and the “or” is an operation, not something in the world; and a complete lower-level narrative can be told without ever using the concept, while the reverse fails—whatever is not needed in a complete account earns no seat in existence. Third, the familiar explanation that a division “carves nature at its joints” is circular: it renames predictive success, hangs the name on the world, and then explains the success by it. The paper holds that predictive success is a relation between a division and what follows—locals grouped by the line have sequels resembling one another—a fact about what comes next, requiring no pre-partitioned entities. The burden of proof falls on whoever asserts the entity. Finally, the paper shows that what is withdrawn is the question, not any prediction and not any phenomenon.

先说清要比的这两组东西。 两条岔路：一个人走在路上，前面路分成两条。沥青，路肩，一块指路牌——这一切全在空间维度中显现，跟时间无关：它们现在就是分开的，两条都看得见、都摸得着，一张照片就能同时拍下。它们同处于一个时空之内，此刻，同一片地上。 两个四维时空：“要是我当时走了另一条路”——这句话构造的不是一条路，是一个平行的四维时空。所谓四维时空，是把整段过程连时间一起看：世界从头到尾的那一整串。真实的那个里，人走的是岔路一；构造出的那个里，人走的是岔路二。 而它跟第一个不是只在路口不同：遇见的人不同，那天晚上不同，此后二十年都不同——它是一整个世界的重造。 而我们只是在头脑中把第二个四维时空与第一个并置，并赋予了第二个以同等的真实感。而我们对它俩的感知，却恰如感知三维空间中的两条岔路——就好像它俩就是三维空间中的两条岔路。 “接近”是一个空间中的概念，同时又涉及时间：指我在很短的时间内就能移动过去、抵达。 把这个概念的成立条件列成清单，就看得见它被搬到哪儿去了。要说 A 接近 B，得有四样：一个共同的度量；一条从 A 到 B 的路径；一个能沿路径移动的东西；以及移动所需的时间。 两条岔路：四样齐全。 两个四维时空：四样一样也没有。 没有共同的度量，没有路径，没有可移动者，也没有”移动所需的时间”——因为时间在每一个里面。 所以”接近”在这里不是难以判断，是无从谈起。 我们至多可以说这两个四维时空在某个时空局部里有些相似；而相似并不蕴含空间中接近，因此也不蕴含”可达”。 请问谁又如何能从第一个四维时空”走到”第二个四维时空？ 若不能，这两个四维时空的接近、可达就完全无意义——而我们却把这两个概念搬了过来，并产生了”它们接近、可达”的感受；该感受在这里根本不适用——并非对错的问题。 一句翻译，可以把这次搬运照出来。 “你为什么没走到第二个四维时空去？它和第一个很接近，因此可达啊！”——说出来当场荒唐。 “你为什么没走到岔路二去？它和岔路一很接近，因此可达啊！”——天天有人说。 而后一句就是前一句：责怪别人没走到岔路二，等于责怪他没走到第二个四维时空。 岔路二与岔路一在空间上接近，但人走到岔路二还需要具备其他条件。条件不具备的话，接近也过不去——而这恰恰就是无法”走到”第二个四维时空。 而在我的经验里，没有人到过另一个四维时空。（“从来”这个词暗示了时间轴，在这里用它不合适。） 若沿用”难”的概念，把”难”定义为很少有人做到，那么到另一个四维时空就太难了——做到的人数是零。 接地气地说：从来没有人能在条件不具备的情况下，做到条件具备时能做到的事。 而那份责怪的产地也就清楚了：只有在没意识到某个具体条件的情况下，我们才会认为做某件事”不难”、而某甲没做到。 ——责怪不是从对方那儿来的，是从我这边少列了一条条件来的。 Abstract First, the two things to be compared. Two forks of a road: walking along, one finds the road dividing. Asphalt, verge, a signpost—all of it appearing in the spatial dimensions, nothing to do with time: they are divided now, both visible, both touchable, both caught in one photograph. They lie within one spacetime, at this moment, on the same stretch of ground. Two four-dimensional spacetimes: “if I had taken the other road”—this constructs not a road but a parallel four-dimensional spacetime. By a four-dimensional spacetime is meant the whole process taken together with time: the world’s entire course from end to end. In the real one the man takes fork one; in the constructed one he takes fork two. And it differs from the first not only at the junction: different people met, a different evening, different twenty years—it is the remaking of a whole world. And we merely juxtapose, in our heads, that second spacetime with the first, and grant it an equal air of reality. Yet we perceive the pair exactly as we perceive two forks of a road in three-dimensional space—as though the pair simply were two forks in space. “Near” is a spatial concept that also involves time: it means I can move across and arrive in a short while. List the conditions under which it holds and the transfer becomes visible. To say A is near B requires four things: a common metric; a path from A to B; something able to move along it; and the time the movement takes. Two forks in a road: all four. Two four-dimensional spacetimes: none of the four. No common metric, no path, no mover, and no “time the movement takes”—because time lies inside each of them. So “near” here is not hard to judge; it does not arise. At most we may say the two spacetimes are somewhat similar in some local region; and similarity does not entail spatial nearness, and so does not entail reachability. Who, and how, could “walk” from the first spacetime to the second? If no one can, then nearness and reachability between two spacetimes are wholly meaningless—yet we have carried both concepts across and produced the feeling that they are near and reachable; and that feeling simply does not apply here—it is not a matter of right or wrong. One translation exposes the transfer. “Why did you not walk over to the second four-dimensional spacetime? It is near the first, and therefore reachable!”—absurd the moment it is said. “Why did you not take the second fork? It is near the first, and therefore reachable!”—said every day. And the second sentence is the first: to blame someone for not taking the second fork is to blame him for not reaching the second spacetime. The second fork is spatially near the first, but a man’s taking it requires further conditions. If they are not met, nearness will not carry him across—and this is precisely the impossibility of “walking to” the second spacetime. And in my experience no one has ever reached another spacetime. (“Never” implies a time axis and is therefore unsuitable here.) If we retain the concept of “hard,” defining it as what few manage, then reaching another spacetime is extremely hard—the number who have managed is zero. Plainly put: no one has ever done, under conditions not met, what could be done when they were met. And the source of the blame is thereby clear: only when we have failed to notice some specific condition do we take a thing to be “not hard” and hold that someone simply failed to do it. —The blame does not come from him. It comes from an item left off my list.

**空间中的岔路，在空间维度中显现，与时间维度无关。** **而四维时空过程中的反事实设想，相当于构造出一个平行的四维时空，里面的物理过程与真实四维时空不同**——比如真实四维时空中人走的是岔路一，**而构造出的四维时空中人走的是岔路二。** **我们却把这两个四维时空剥去时间轴，只剩空间部分**——即剥去沿着时间轴的变化，**只留在空间上的轨迹**；**因此似乎就简单地对应了空间中的两条岔路，又似乎对应了在岔路口的取舍、选择。** **这是把四维时空三维空间化的认知滑坡。** **而三维空间中的两条岔路同时并存，不代表两个四维时空并存。** **要害在"并存"这个词，它在两边根本不是同一个意思。** **两条岔路的并存，有一个"同时"可用。** 它们此刻都在那儿，都是沥青，**一张照片就能同时拍下**——**这是一个四维时空之内的并存，容器是现成的：那片地，今天下午三点。** **而两个四维时空的并存，没有"同时"可用。** **时间在每一个里面**；**你没法说它们"在某一刻都在"，因为没有哪一刻在两者之外。** **滑坡正是滑在这个词上：用一种有容器的并存，去给一种连容器都还没被出示过的并存作保——而作保的那一方，从来没有被要求出示那个容器。** **第二击：那次投影不是丢了一维，是把一个巨大的东西压成了一个小东西。** 那个构造出的四维时空，**跟真实那个不是只在路口不同**：此后每一点都不同——遇见的人不同，那天晚上不同，此后二十年不同。**它是一整个世界的重造。** **而剥去时间轴之后，这一整个重造，看起来只剩"在这个路口拐哪边"。** **——投影把一次全世界的构造，压成了一次几度角的偏转。** **所以那个设想之所以显得便宜，正是投影给的**：空间上那点差别小得可以忽略，**而四维上的差别没有上限。** **而这个滑坡是单向的，这一点本身就是证据：** 我们从"两条岔路并存"滑到"两个四维时空并存"，**却从来没有人反过来走**——没有人说"既然两个四维时空不能并存，所以那条路不能分岔"。**一条只朝一个方向走的推理，多半不是推理。** ## Abstract **A fork in space appears in the spatial dimensions and has nothing to do with the temporal one.** **A counterfactual about a process in four-dimensional spacetime amounts to constructing a parallel four-dimensional spacetime whose physical process differs from the real one**—in the real spacetime the man takes fork one, **in the constructed one he takes fork two.** **Yet we strip both spacetimes of their time axis and keep only the spatial part**—strip away the变化 along time and **retain only the trajectory in space**; **whereupon it seems simply to correspond to two forks in space, and further to a choice made at the fork.** **This is the cognitive slide by which four-dimensional spacetime is made three-dimensional.** **And that two forks coexist in three-dimensional space does not entail that two four-dimensional spacetimes coexist.** **The crux is the word "coexist," which does not mean the same thing on the two sides.** **The coexistence of two forks has a "at the same time" available.** Both are there now, both are asphalt, **and one photograph catches them together**—**this is coexistence within one spacetime, and the container is ready to hand: that stretch of ground, three o'clock this afternoon.** **The coexistence of two spacetimes has no "at the same time" available.** **Time is inside each of them**; **one cannot say they are "both there at some moment," since no moment lies outside both.** **The slide is a slide on that word: a coexistence with a container is used to vouch for a coexistence whose container has never been exhibited—and the voucher has never been asked to produce it.** **A second blow: the projection does not merely drop a dimension; it compresses something enormous into something small.** The constructed spacetime **does not differ from the real one only at the fork**: it differs at every point thereafter—different people met, a different evening, different twenty years. **It is the remaking of a whole world.** **And once the time axis is stripped, that whole remaking looks like nothing more than "which way to turn at this junction."** **—The projection compresses the construction of an entire world into a deflection of a few degrees.** **So the cheapness of the supposition is a gift of the projection**: the spatial difference is negligible, **while the four-dimensional difference has no upper bound.** **And the slide runs one way, which is itself evidence:** we slide from "two forks coexist" to "two spacetimes coexist," **and no one has ever run it backwards**—no one says "since two spacetimes cannot coexist, that road cannot fork." **A piece of reasoning that runs in only one direction is mostly not reasoning.**

主体这边感觉到有分岔，却认为分岔是系统那边的性质。 而在没有分岔的系统——比如康威生命游戏——里面的主体，仍然会感觉到有分岔。 要点在”感觉”两个字。 说”想象出分岔”，那还像是一个动作，可以问它按什么规则做的；而说”感觉到有分岔”，说的是它到场时就在那儿——所与，不是所作。 一个人站在岔路口，不是先造出两条路再决定相信它们；他一睁眼，那两条路就在那儿。 于是问题的形状就变了：不是”他造出的那批分岔能不能被系统的随机性背书”，而是——他把这边的一份感觉，当成了那边的一条性质。 弄错的是产地。 而演示正好打在这一处。 康威生命游戏规则四行，彻底决定论，一处岔口也没有；而它是图灵完备的，里面可以跑一个足够复杂的主体。那个主体照样会感觉到有分岔——“上一步该走另一步”。而我们从外面一帧一帧看得清清楚楚：那一支从来没有出现过，它不是被算过又被否决的，它根本不在任何一帧里。 一样在没有它的地方照样出现的东西，不可能是它带来的。 这也解释了对手为什么会那么答。 他不是先有一个理论、再去找证据；他是先有那份感觉——分岔明明就在那儿——然后去找一个能安放它的地方，而随机性是现成的、体面的、听起来像物理的一个位置。所以那不是一个论证，是一次安置。 而最省的验法是：那份感觉的强度，跟系统的随机性无关，却跟我这边的状态有关。 同一件事，心情好的时候那份”本来可以”很淡；夜里躺在床上翻来覆去的时候，它极强。 而系统那边一分没变。变的是这边。 ——一份随我这边的状态而涨落的东西，说它是那边的一条性质，本身就说不通。 须盯住一处：本文没有证明世界里没有分岔。 本文证明的是那份感觉的产地在主体这一侧——而这两件事分开之后反而更有力：不必主张世界那边是什么样，只需指出那份感觉不来自那边。 Abstract On the agent’s side there is a felt presence of branching, and the agent takes branching to be a property of the system’s side. Yet in a system without branching—Conway’s Game of Life—an agent within it will still feel that there is branching. The word “felt” carries the weight. To say “imagines branches” still sounds like an act, and one may ask by what rule it was performed; to say “feels that there is branching” says that it is there on arrival—given, not made. Standing at a fork, one does not first construct two roads and then decide to believe in them; on opening one’s eyes, the two roads are there. The shape of the question therefore changes: not “can the branches he constructed be underwritten by the system’s randomness,” but—he has taken a feeling on this side for a property of that side. What is mistaken is the provenance. And the demonstration strikes exactly there. The Game of Life has four rules, is wholly deterministic, and contains no fork whatever; and being Turing complete, an agent of sufficient complexity can run inside it. That agent will feel that there is branching all the same—“I should have played otherwise.” And from outside we see, frame by frame, that the alternative never appeared: it was not computed and rejected; it is in no frame at all. What appears equally where a thing is absent cannot have been brought about by that thing. This also explains the opponent’s reply. He does not begin with a theory and seek evidence; he begins with the feeling—the branching is plainly there—and then looks for somewhere to put it, and randomness is a ready, respectable, physics-sounding place. So it is not an argument but a placement. And the cheapest test: the strength of that feeling has nothing to do with the system’s randomness, and much to do with my own state. Of one and the same event, in good spirits the “could have” is faint; lying awake at night it is overwhelming. The system has not changed by a hair. What changed is here. —What fluctuates with my own state cannot be a property of the other side. One thing to keep hold of: the paper has not shown that the world contains no branching. It has shown that the provenance of the feeling lies on the agent’s side—and separating these makes the case stronger: no claim about the far side is needed, only that the feeling does not come from there.

**发生只有一条（时间）线。我们给它加上想象出的反事实分岔，岔上还可以分岔，原来的一条线就成了一个树状。** **而这一篇处理的是一段三步的对话。** **第一步，我说：那些反事实分岔都是不可能的。** **第二步，对手说：由于量子随机性而可能。** **第三步——而本文要指出：这两句之间缺一步。** **他的回答假定了一件事：那些分岔的存在，取决于系统有没有随机性。** **而这个连接，从来没有被论证过。它是被假定的。** **打在这个假定上的，是一个演示：在没有随机性的系统中——比如康威生命游戏——里面的主体，比如一台能下棋的图灵机模式，同样可以想象出反事实分岔**（"上一步棋如何走会更好"）。 **生命游戏规则四行，一九七〇年就写死了；彻底决定论，没有随机数，而我们从外面一帧一帧看得一清二楚。** 而它是**图灵完备的**，所以原则上可以在里面跑一台会下棋的机器。**它保存着局面，它有一个目标（赢），而它走完马之后，对方走了一步它没有料到的棋——于是它生成："我不该走那个马。"** **而我们在外面看得清清楚楚：那盘棋的每一步，从初始格局起就定了。**"走车"那一支**一次也没有出现在任何一帧里——它不是被算过又被否决的，它从来没有存在过。** **而它照样生成了那句话。** **由此得出的结论，要按最省的形式说：（个体所想象的）反事实分岔，与（系统的）随机性无关。** **不是"随机性解释不了它"**——那还留着一个"也许别的能解释"的口子；**是这两样之间没有关系：有它、没它，那些分岔都在那儿。** **而主语必须盯住：演示证明的是"那个主体想象得出分岔"，不是"世界里没有分岔"。** 这两件事分开之后反而更有力：**本文根本不必主张世界里没有分岔。** 只要指出**那些分岔的来处在主体这一侧**——**那么对手拿系统那一侧的性质来给它们背书，就接不上：背书的对象和被背书的东西，不在同一侧。** ## Abstract **What happens is one line in time. We add to it imagined counterfactual branches, and branches upon branches, so that the single line becomes a tree.** **This paper treats a three-step exchange.** **First, I say: those counterfactual branches are all impossible.** **Second, the opponent says: quantum randomness makes them possible.** **Third—and this is the paper's point: a step is missing between the two.** **His reply assumes something: that whether those branches exist depends on whether the system contains randomness.** **That connection has never been argued. It is assumed.** **What strikes at the assumption is a demonstration: in a system without randomness—Conway's Game of Life—an agent within it, say a chess-playing Turing machine, can imagine counterfactual branches just the same** ("how the last move might better have gone"). **The Game of Life has four rules, fixed since 1970; it is wholly deterministic, with no random numbers, and can be watched frame by frame from outside.** And it is **Turing complete**, so in principle a chess-playing machine can run within it. **It holds a position, it has a goal (to win), and after it moves the knight the opponent plays a reply it did not anticipate—whereupon it generates: "I should not have moved the knight."** **And from outside we see plainly: every move of that game was settled from the initial configuration.** The branch "moving the rook" **appears in no frame at all—it was not computed and rejected; it never existed.** **And it generates the sentence all the same.** **The conclusion should be put in its most economical form: the counterfactual branches (as imagined by an individual) have nothing to do with the randomness (of the system).** **Not "randomness cannot explain them"**—that leaves open "perhaps something else can"; **but that there is no relation between the two: with or without it, the branches are there.** **And the subject must be kept hold of: the demonstration shows that the agent can imagine branches, not that the world contains none.** Separating these makes the case stronger: **the paper need not claim that the world contains no branches.** It need only note that **the branches originate on the agent's side**—**so that invoking a property of the system's side to underwrite them does not connect: what is underwritten and what underwrites are not on the same side.**

**在一个完全决定论的世界里，一个 agent 也同样可以生成这种树——只是他无法搬出量子随机性之类，来把这个树的枝说成是实的。** **这一句把整条线的位置定死了：那棵树跟世界是不是决定论的，一点关系也没有。** **先给一个具体的例子，免得"这种树"落空。** **我在下棋，走完之后发现走错了一步。** 我后悔，**并在心里想象出走法不同的分支**："那一步该走车而不是马"——**而那一支上还能再分岔**（"走车之后他若还那样应，我就……"），**于是看起来像一棵树。** **这个例子有一处特别干净：整盘棋是全记录的。** 走过的每一步都在谱上；**而"我本该走的那一步"不在谱上。同一张纸，一条是记录，一条不是。** **还有一处更要紧：下棋的人自己就知道那盘棋已经定了。** 他不会认为那盘棋"本来可以是别的"——**可他照样后悔。** **他明知不能改，照样在心里把那一步挪一遍。** **而造这棵树用的东西，一样也不需要世界那边留一个口子：** **账目**（我的子、我的位置、我本该有的优势）、**愿望**（我想赢这盘）、**事后的信息**（走完才看出那一步坏在哪儿）。**三样，全在他这一侧。** **而最干净的演示，是康威生命游戏。** 规则四行，一九七〇年就写死了；**彻底决定论，没有随机数，而我们从外面一帧一帧看得一清二楚。** 更巧的是**它长出来的"东西"恰恰以图样为单位**：滑翔机、飞船——**它们并不存在，存在的只有格子的开关。** **而生命游戏是图灵完备的**：原则上可以在里面跑一个能下棋的图灵机模式。**于是那个模式一边下棋，一边生成"我不该走那个车"。** **而我们在外面看得清清楚楚：那盘棋的每一步，从初始格局起就定了。**"走另一步"那一支，**一次也没有出现在那一大片格子的任何一帧里。** **而它照样生成了那句话。** **这里还有一个我们自己那儿没有的检验：把初始格局重跑一万遍——一万遍都一样，没有一次走那个车。而在我们这儿，重跑这个动作根本做不了。** **（一处分寸：这个演示不要求那个图样有感受。要验的不是它难不难受，是那棵树长不长得出来——而树是一批记录和一份愿望的产物，不是一份感受的产物。）** **所以随机性从来不是那棵树的原料。它是被后来搬来的一张证明。** **由此两件事被彻底拆开：** **枝的生成**——**不需要任何物理条件**；决定论的世界里照样生成，**而且天天在生成。** **枝的实化**——他要说那条枝"真的可能"，**这才需要搬东西。而这是被质疑之后才补的。** **——所以对手搬量子随机性时，他不是在解释那棵树怎么来的，他是在事后给一棵早就长好的树补一份产权证明。** 这也解释了那份补证为什么总是临时的、只在一处出场：**造树的时候一样也没用上它，用的是账目和愿望；只有被追问时才去搬。** **还有一条更省的推论：若那棵树在决定论世界里照样长，那么"世界是否决定论"对那棵树的存在与形状，不产生任何差别。而不产生差别的东西，不能拿来解释它。** 他搬随机性来解释那棵树，**等于用一个在两种情形下不一样的东西，去解释一个在两种情形下都一样的东西——两边都不动，中间那根线是画上去的。** **最后一处最狠：那个 agent 说不出那句话，可他照样有那棵树。** **树是他必然会长的；那句"枝是实的"是他后来学会说的。** 而在决定论世界里，**他连学都学不会**——**而他的懊恼一分不少。** **——所以懊恼从来不靠那句话撑着。那句话是后配的。** ## Abstract **In a wholly deterministic world an agent can generate this same tree—he merely cannot invoke quantum randomness and the like to declare its branches real.** **This fixes the position of the whole line of argument: the tree has nothing to do with whether the world is deterministic.** **First a concrete case, lest "this tree" stay empty.** **I am playing chess, and after moving I see the move was wrong.** I regret it, **and imagine a branch in which I moved differently**: "that should have been the rook, not the knight"—**and that branch can branch again** ("and had he replied so, then I..."), **so it looks like a tree.** **This case is especially clean: the whole game is fully recorded.** Every move played is on the score sheet; **the move I "should have" made is not. One sheet, one line a record and the other not.** **And something more telling: the player himself knows the game is settled.** He does not think the game "could have gone otherwise"—**and he regrets it all the same.** **Knowing it cannot be changed, he still moves that piece over in his head.** **And the materials for the tree need no opening on the world's side:** **ledger** (my pieces, my position, the advantage I should have had), **wish** (I want to win this game), **later information** (only after moving does the flaw appear). **Three items, all on his side.** **The cleanest demonstration is Conway's Game of Life.** Four rules, fixed since 1970; **wholly deterministic, no random numbers, watchable frame by frame from outside.** Better still, **the "things" it grows are patterns**: gliders, spaceships—**which do not exist; only the cells' switching exists.** **And the Game of Life is Turing complete**: in principle a chess-playing Turing machine can run inside it. **So that pattern plays chess and generates, meanwhile, "I should not have moved the rook."** **And from outside we see plainly: every move of that game was settled from the initial configuration.** The branch "moving otherwise" **appears in no frame of that vast field of cells, not once.** **And it generates the sentence all the same.** **Here too is a test we lack for ourselves: rerun the initial configuration ten thousand times—all ten thousand alike, and not once the other move. In our own case, rerunning cannot be done at all.** **(A calibration: the demonstration does not require the pattern to have feelings. What is at issue is not whether it suffers but whether the tree grows—and the tree is a product of records and a wish, not of a feeling.)** **So randomness was never a material of that tree. It is a certificate brought in afterwards.** **Two things thereby come apart:** **Generating a branch**—**requires no physical condition whatever**; in a deterministic world it is generated all the same, **and generated daily.** **Realizing a branch**—he wants to say the branch is "really possible," **and only for this must something be fetched. And it is fetched only after a challenge.** **—So when an opponent invokes quantum randomness, he is not explaining how the tree came to be; he is retrospectively supplying a deed for a tree long since grown.** Which also explains why the certificate is always ad hoc and appears in one place only: **none of it was used in the growing—ledgers and wishes were—and it is fetched only under questioning.** **A cheaper consequence: if the tree grows in a deterministic world too, then whether the world is deterministic makes no difference to the tree's existence or shape. And what makes no difference cannot be used to explain it.** **And this must be put precisely, for it is easily reversed:** randomness **differs between the two cases**—absent under determinism, present otherwise—**while the tree is the same in both.** **So what he does is use something that varies between the two cases to explain something that does not.** **And that fails outright as method: what varies cannot explain what does not vary. It changed; the thing explained did not. Remove it and the tree stands; add it and the tree stands—so it is not the source.** **Sharpest of all: that agent cannot say the sentence, and still has the tree.** **The tree he will grow necessarily; the sentence "the branch is real" he learns to say later.** In a deterministic world **he cannot even learn it**—**and his regret is not diminished by a hair.** **—So regret never rested on that sentence. The sentence was added afterwards.**

我打算预约看牙，但想先看看车有没有问题——处理车的事好像更紧急重要，而看车似乎应该很快。 结果看车耽搁了一天，打电话也就耽搁了一天。今天打的电话，约到下周三；而今天也是周三。 于是我觉得：如果昨天打，也许可以约到今天；今天打，则比”主干”迟看牙一周。 先把两个词定住。 主干：从我出生到此刻，实实在在走过的那一串——叫它主干，只因为它是唯一发生过的那一条，不因为它更重要或更好。 枝：从主干上某一点起，我们说”若当时不那样”，于是引出一条线——而画的那一笔不在主干里：那一段发生里没有一处岔口，岔口是我们标的。 两者合起来看像一棵树，严格说却不是：是一条线，加上一堆我们画的线；它们像枝，只因为接在主干上，而”接”这个动作也是我们做的。 而须当场指出一处：本文开头那句”今天打，则比’主干’迟看牙一周”，其中的”主干”已经不指实际发生了。 今天打、约到下周三，这才是实际发生的那一支；而那句话里的”主干”指的是”昨天打、今天看上牙”——一条没有发生过的枝。 那句话把枝叫成了主干，而写下它的时候并没有察觉。它留在那儿，因为它本身就是一个标本。 而那棵树的形状，并非保持不变。 在更早的时间点，我能看到这两个分支吗？ 看不到——或者说，不能像现在这样”清晰”地看到。 那时我看到的是：可能迟一天、迟一周、不迟、甚至还可能早（晚约理论上可能约到更早的，若有人临时取消）。 现在只剩两支。而这不是四支里我挑了两支——是那几支不见了。 被后来的事收走了：号已经给到下周三，“不迟”“还可能早”就没了立足之处。修剪的是后来的事实，不是我的判断。 而误认在于：我们把当下这个形状回填到了过去——以为以前看也是这两个分支。 回填毫无察觉，因为那几支消失时没有一个删除动作被记下来：不是被划掉，是从此不再出现。 其次，现在这份”清晰”本身也不清晰。“昨天打能约到今天”只是个猜想——没有人查过。于是”迟了一周”这个差额，一端是硬的（下周三，人家给的号），一端是软的（今天，从未被查证）——而我们把它当成一个确定的数在心疼。 而最要紧的是：那时候我并不知道车没啥大问题，不知道修车要花多久。“先看车”正是在不知道的情况下做的判断；而现在知道了，那个动作就显得像白耽误了一天。 可”白耽误”这句话，只有拿到那条信息之后才说得出口——那个反事实的成立，需要一份当时不存在的信息。 再进一步：即使牙医确认”昨天几点约就能约到今天”，那一支的身份也没变。 他核实的是一条记录（那个号昨天是空的，排班表上写着）；而我要的是一条走法（我拿起电话、线没占、没人先拿走、接线的人报了这个号、我今天来了）——这条走法一次也没有发生过。记录是硬的，走法是造的；而我们把前者的硬度转给了后者。 要点不在他答得多，在他每一次答的都是同一类东西：再问”那我前天打呢”“大前天呢”，他照样答得出来——而每一次答的都是一条空号记录，不是一条我的走法。排班记录支撑”那个号昨天是空的”，支撑得很好；它不支撑”我会拿到它”，因为它对我打没打电话、线占不占、有没有人先拿走，一个字也没说。 而朝前那一个同样是构造：我今天约了下周三，那么下周三就能看上牙——此刻这也是个头脑构造物（医生可能病假，我可能有事）。两句要的东西完全一样：一条没有发生过的走法；而我们只对朝后那一个起疑。 由此落到最后一处：这体现了文化把反事实、头脑构造物当真。 文化中，昨天我是可以打电话的——即使我实际上没打。 这句话在文化里不是一个猜测，是一条关于我的属性，跟”我会说中文”同一种语气。 而这批属性是互不相容的，只是从来没有被摆在一起过。 昨天那个时段只有一个空号，而想约的人不止我一个。今天我们各自懊恼”昨天打就好了”——我这句成立要求乙没打，乙那句成立要求我没打；两句同时成立，而世界那边只有一个号。 文化可以让我和乙同时遗憾昨天没打电话，虽然总共只有一个空号。 而这不是我们各自算错了。是每一条”本来可以”在生成时，都把别人当成了不动的：我造那条枝时其余一切照抄——包括乙那天照旧没打；而我不需要多想一步，因为那条枝是照着我的愿望配出来的，而乙不在我的愿望里。 所以每一条”本来可以”，都是在一个”只有我在动、别人都定住”的世界里成立的——而那个世界，一人一个。 检验很省：把两个人的懊恼摆在一张桌上。 平时不会摆——各记各的，一辈子不对账；而一摆上，那个号就只有一个。 不是有人错了，是那批账目从设计上就不必兼容。 而不必两个人才撞车，一个人也够：我昨天”本可以”打电话约牙，同时也”本可以”早点去修车、本可以送孩子、本可以把那份东西交了——每一条都成立，而它们要的是同一段昨天下午。 文化中，人是可以做到他没做到的事情的——比如提前做出一个在事后看是合理的决定。 而这一句里藏着一个逆流的动作：合理性是事后才有的（我现在知道车没大问题，所以”先看车”显得不合理），而它被要求当时就在。 被要求的不是”当时做个合理的决定”，是”当时做出一个用后来的信息判定为合理的决定”——而这需要的东西只有一个名字：预知。 所以文化里那个”人”，不是一个我们观察到的东西，是一个由要求反推出来的东西：先有要追责的需要，再有一个能被追责的对象，而那个对象必须具备”本可以”和”本该预知”这两样。 ——那个”人”，是由他没做过的事构成的。 而这套东西还有一个后果：我们可能会视喜爱的那条枝为”主干”。 怎么看出来？看他在哪条线上算账——说”今年比去年好些了”，基准是实际那条；说”这才到哪儿，本来早就该……”，基准是那条枝。 于是实际发生的那一条反倒成了偏离：主干成了枝，枝成了主干。 而这个倒转是无声的，因为两者都是线，换一条来当主干，看起来一模一样。 Abstract I meant to book the dentist, but wanted first to check the car—the car seemed more urgent, and checking it seemed quick. The car took a day longer, so the call was a day late. I called today and got next Wednesday; today is Wednesday too. So I feel: had I called yesterday I might have got today; calling today puts me a week behind the “trunk.” One thing must be flagged at once: in the opening sentence, “calling today puts me a week behind the trunk,” the word “trunk” no longer refers to what actually happened. Calling today and getting next Wednesday is the actual branch; the “trunk” of that sentence means “called yesterday, seen today”—a branch that never occurred. That sentence calls a branch the trunk, and it was written without noticing. It is left standing, because it is itself a specimen. But the shape of that tree does not stay fixed. At an earlier point, could I see these two branches? No—or not with the present “clarity.” What I saw then was: perhaps a day late, perhaps a week, perhaps not late at all, perhaps even earlier (a late call may catch a cancellation). Now only two remain. And it is not that I chose two out of four—the others are gone. Later facts took them: the slot is now next Wednesday, so “not late” and “even earlier” have nowhere to stand. What prunes is later fact, not my judgment. The misreading: we backfill the present shape into the past—taking it that we saw these same two branches then. The backfilling goes unnoticed because no act of deletion was ever recorded: they were not struck out; they simply stopped appearing. Second, the present “clarity” is not clear either. “Yesterday’s call would have got today” is only a guess—no one checked. So the shortfall “a week late” is hard at one end (next Wednesday, the slot I was given) and soft at the other (today, never verified)—and we grieve over it as a definite figure. Most importantly: at the time I did not know the car had no real problem, nor how long a repair would take. “Car first” was a judgment made in that ignorance; now that I know, the move looks like a day wasted. But “wasted” can only be said once that information is in hand—the counterfactual requires information that did not exist at the time. Further: even if the dentist confirms “a call at such an hour yesterday would have got today,” the status of that branch is unchanged. He verifies a record (the slot was free yesterday; it is on the schedule); what I want is a course (I lift the phone, the line is free, no one takes it first, the receptionist offers that slot, I come today)—a course that never occurred. The record is hard; the course is constructed; and we transfer the hardness of the first to the second. The point is not how many he can answer but that each answer is of one kind: ask “and the day before?” and “the day before that?” and he answers them all—each time with a record of a free slot, never with a course of mine. The schedule supports “the slot was free yesterday,” and supports it well; it does not support “I would have got it,” because it says nothing about whether I called, whether the line was busy, whether someone took it first. And the forward one is equally a construct: I booked next Wednesday, so I shall be seen next Wednesday—this too, right now, is a construct of the head (the dentist may fall ill; I may be called away). Both sentences want the same thing: a course that has not occurred. And we suspect only the backward one. Hence the last point: this displays a culture taking counterfactuals, constructs of the head, for real. In the culture, I could have telephoned yesterday—even though I did not. In the culture this is not a guess but a property of me, in the same tone as “I speak Chinese.” And these properties are mutually incompatible; they have simply never been set side by side. Yesterday there was one free slot, and more than one person wanting it. Today we each regret not having called—my regret requires that the other did not call, and his requires that I did not; both hold at once, while the world holds one slot. The culture lets me and the other regret, together, not having called yesterday, though there was only one slot in all. And this is not a miscalculation on either side. Each “could have” is generated with everyone else held fixed: making my branch, I copy the rest over—including his not having called; and I need think no further, because the branch was fitted to my wish, and he is not in my wish. So every “could have” holds in a world where only I move and everyone else stands still—and there is one such world per person. The test is cheap: set two people’s regrets on one table. Ordinarily they never are—each keeps his own books and they are never reconciled; and once set together, there is only one slot. No one has erred; those ledgers were simply never required to be compatible. Nor does it take two: one person suffices. Yesterday I “could have” called the dentist, and equally “could have” taken the car in earlier, “could have” collected my child, “could have” sent that document—each holds, and all of them want the same yesterday afternoon. In the culture, a person can do what he did not do—for instance, make in advance a decision that is, in hindsight, reasonable. And here a current runs backwards: reasonableness arrives afterwards (I now know the car was fine, so “car first” looks unreasonable) and is then required to have been there at the time. What is demanded is not “make a reasonable decision then” but “make then a decision judged reasonable by information that came later”—and what that requires has only one name: foreknowledge. So the “person” in the culture is not something we observed but something reverse-engineered from a requirement: first the need to hold someone accountable, then an object that can be held, and that object must possess both “could have” and “should have foreseen.” —That “person” is constituted by what he did not do. One further consequence: we may take a branch we are fond of for the “trunk.” How to tell? See on which line he keeps his accounts—“better this year than last” reckons against the actual line; “and look where I am, I should long since have—” reckons against the branch. The actual line thereby becomes the deviation: the trunk becomes a branch, the branch the trunk. And the inversion is silent, since both are lines: swap which counts as trunk and nothing looks different.

如果宇宙的实际演化路径是一条线，那么我们的认知则是一个树状： 在那条实际的主干线上可以分岔，岔出的线上还可以再分岔，就形成树。除了主干之外的线都是反事实。 而本文的正题是那句最后的话：我们很容易会把某条我们喜爱的反事实线，视为”主干”。 先把这棵树的形状说准，因为它跟真树不一样。 真树的枝从主干里长出来，带着主干的材质；而这棵树的枝，是从主干上画出去的——画的那一笔不在主干里。 主干那一段发生里，没有任何东西对应到那个分岔点：分岔点是我们标的。 所以严格说它不是一棵树，是一条线，加上一堆我们画的线；而画的那些看起来像枝，只因为它们接在主干上——而”接”这个动作也是我们做的。 其次有一处极不对称，值得先看一眼：枝上还能再分岔，而没有人往下画。“要是钥匙没掉”是一条枝；而在那条枝上，“要是那天我早出门两分钟”呢？——同样可以分，而一次也没有人分过。 主干上我们每天分几十次，枝上一次也不分。若它们真是同一类线，分岔的密度该差不多。 ——枝不是用来住的，是用来指着的：它一被画出来就完成了任务，没有人在上面继续生活，所以也没有人在上面继续分岔。 而那个倒转，就在这里发生：某条枝被当成了主干。 怎么看出来？看他在哪条线上算账。 真正被当成主干的那一条，是他从中计算得失的那一条：“我本该有那个位置”“我本来该过那种日子”——得失是相对于那条线算的，而不是相对于实际那条。 于是实际发生的那一条，反倒成了偏离：主干成了枝，枝成了主干。 而这个倒转是无声的——因为在图上，两者都是线；换一条来当主干，图看起来一模一样。 没有一处提示，没有一次声明，没有哪一天他决定过要换。 倒转之后的后果，全在那本账上： 实际的每一天都被记成亏空（因为它是相对于那条线的偏离）；而任何改善都填不平，因为那条线不随实际改动而下调；而它也永远不会被证伪——枝上没有别的分岔可比，也没有任何东西能落到它上面来打它一下。 Abstract If the actual evolution of the universe is one line, our cognition is a tree: branches may be drawn from that actual trunk, branches may branch again, and so a tree forms. Every line but the trunk is a counterfactual. The paper’s subject is the last remark: we readily take some counterfactual line we are fond of to be the “trunk.” First, the shape of this tree, which is unlike a real one. A real branch grows out of the trunk and carries its material; the branches here are drawn out from the trunk—and that stroke is not in the trunk. In that stretch of actual happening nothing corresponds to the branch point: the branch point is marked by us. Strictly it is not a tree but one line plus a set of lines we drew; and they look like branches only because they meet the trunk—and the meeting, too, is our doing. Second, a striking asymmetry: branches could branch again, and no one draws further. “If the keys had not been lost” is a branch; and on that branch, what of “if I had left two minutes earlier”?—equally available, and never once drawn. On the trunk we branch dozens of times a day; on a branch, never. Were they lines of one kind, the density of branching would be comparable. —A branch is not for living on but for pointing at: drawn, it has done its work; no one goes on living there, so no one goes on branching there. And here the inversion occurs: a branch is taken for the trunk. How to tell? See on which line he keeps his accounts. The line genuinely taken as trunk is the one against which he reckons gains and losses: “I should have had that position,” “I should have been living that life”—the reckoning is relative to that line, not to the actual one. So the actual line becomes the deviation: the trunk becomes a branch, and the branch the trunk. And the inversion is silent—because on the diagram both are lines; swap which is trunk and the diagram looks exactly the same. No prompt, no declaration, no day on which he decided to change. The consequences all appear in the ledger: every actual day is entered as a shortfall (being a deviation from that line); no improvement ever closes it, since that line does not come down as the actual rises; and it can never be refuted—a branch has no rival branches to compare with, and nothing can land on it to strike it.

一个人说：“要不是刚才丢了钥匙，我现在就能开门进屋了。” 本文只问一句：为什么他在造反事实时，只造出了这一个？ 而实际上可以造的反事实，数量是无限多的。 无限的来处不必论证，摆几层就看得见。 光是”钥匙没掉”之后那五分钟怎么走——无限种；“钥匙没掉”本身的方式——没滑出、滑出被兜住、掉了又弹回兜里，无限种；而除了钥匙，别的每一样也都可以改：门锁、天气、那趟车、他出门的时刻——每一样各有无限种改法。 而这些，一个也没到过他面前。 “无限多”这个词把一条辩护堵死了。 若只是”一大堆”，他还能说”我挑了最重要的那几个”；而无限多，“挑”这个动作就没法进行了——没有人能从无限里挑，除非他手上先有一条极窄的规则。而那条规则，正是要问的东西。 更要紧的是：他不是造了一个，他是收到了一个。 “造”这个字暗示一个动作，而动作可以问方法。 可他并没有做那个动作：他没有先列出候选、再取其一，他张口就是那一句。 所以那一问最准的形式，不是”你怎么挑的”，而是：你什么时候造过别的？ ——一次也没有。 而”因为别的不重要、不相关”这句辩护不成立：不重要是谁定的？那正是要问的东西；而”不相关”这句话本身就已经用上了那个没交代的筛选——先筛完，再拿筛的结果当筛的理由。 最后一处最难看：他手上那一个，恰恰是条件最苛的一个。“钥匙没掉、其余全照实际来”要求那五分钟一步不差；而”钥匙没掉但半路折回买了瓶水”要求松得多。 若真按”可以被设想”铺开，苛的那一份最不该被单挑出来当代表。 而他挑中它，不是因为它最可能，是因为只有它推得出那句话的下半截：“我现在就能开门进屋了。” 所以顺序是反的：不是先设想一个可能、再看它推出什么，是先有那个结论、再往回配一个能推出它的前件。那些从没被造出来的，是被那个结论挡住的。 而那个结论，只是他的愿望。 “我现在就能开门进屋了”写着像一句被推出来的话，可它是一句愿望换了装：“我想进屋”说出来站不住，而”我本来就能进屋了”说出来理直气壮——同一件事，换了个语法就换了个身份。 于是那个最苛的前件也有了来处：愿望有多具体，前件就有多苛。 我想的不是”进某个屋”，是进这扇门、现在、这一刻——所以那五分钟必须一步不差。苛，不是因为世界那边要求这么多，是因为愿望要求这么多。 一个当场可验的：换一个愿望，那个反事实立刻换一个。 同一件事，若他此刻心疼的是那把钥匙——那句话会变成”要是出门时看一眼兜就好了”；若他怕的是家人埋怨——会变成”要是我没借走那把备用的就好了”。掉钥匙这件事一分没变，而反事实换了三个：变的是愿望。 ——那么那个”本来可以”讲的是什么？它讲的不是那件事，是他此刻想要什么。 这也解释了那份不甘为什么持久：一个被驳倒的结论会被放下，而一个落空的愿望只会一遍遍再来。那句话半夜还会回来，不是它没被想清楚，是它本来就不是一句想法。 Abstract Someone says: “If I hadn’t lost my keys just now, I could open the door and go in.” The paper asks one question: why, in making a counterfactual, did he make only this one? When in fact the counterfactuals that could be made are infinite in number. The infinity needs no argument; a few layers make it visible. Just how those five minutes might run after “the keys were not lost”—infinitely many; the ways of “not losing” them—not slipping out, slipping but caught, dropping and bouncing back into the pocket: infinitely many; and besides the keys, everything else could be altered too: the lock, the weather, that bus, the minute he left—each with infinitely many variants. And not one of these ever came before him. “Infinitely many” closes off one defence. Were they merely numerous, he might say “I picked the important ones”; with infinitely many, the act of picking cannot be performed at all—no one picks out of an infinity, unless he already holds some very narrow rule. And that rule is precisely what is in question. More to the point: he did not make one; he received one. “Make” implies an act, and acts can be asked for their method. But he performed no such act: he did not first list candidates and then take one; the sentence was simply there when he opened his mouth. So the sharpest form of the question is not “how did you pick,” but: when did you ever make another? —Never once. And “because the others are unimportant or irrelevant” will not do: unimportant by whose reckoning? That is exactly what is in question; and “irrelevant” already employs the unaccounted filter—filtering first, then offering the filtrate as the reason for the filter. Worst of all: the one he holds is the most demanding of the lot. “Keys not lost, everything else exactly as it was” requires those five minutes step for step; “keys not lost but I turned back for water” requires far less. Laid out by “can be supposed,” the strictest is the last that should be singled out as representative. And he singled it out not because it is likeliest but because it alone entails the second half of his sentence: “I could open the door and go in.” So the order is reversed: not supposing a possibility and then seeing what follows, but having the conclusion first and fitting an antecedent to it. What was never made was blocked by that conclusion. And that conclusion is only his wish. “I could open the door and go in” is written like something inferred, but it is a wish in disguise: “I want to go in” cannot stand on its own; “I could have been in by now” is said with perfect confidence—the same thing, given another grammar and thereby another standing. The strictest antecedent thus finds its source: the more specific the wish, the stricter the antecedent. I do not want to enter some door but this door, now, this minute—so those five minutes must run step for step. The strictness comes not from the world’s demands but from the wish’s. A test: change the wish and the counterfactual changes at once. Same event: if what pains him is the key itself—“if only I had checked my pocket on the way out”; if he fears reproach at home—“if only I hadn’t taken the spare.” The losing of the keys has not changed by a hair, and the counterfactual has changed three times: what changed is the wish. —So what is that “could have” about? Not about the event, but about what he wants right now. This also explains the persistence of the sting: a refuted conclusion is set down, while a thwarted wish simply returns. That sentence comes back at night not because it was not thought through, but because it was never a thought.

“他本来可以不那样。” 这句话每天被说出无数次，而没有人被要求出示任何东西。 本文要说的是：这不是那句话已经被论证过了。是没有人意识到这里需要论证——所以那份论证相当于被搁置了。 “相当于”三个字要保住。 真被搁置的东西会被记着：有人会说”这个以后再论”，会在脚注里留一句，它在待办上。而这一处不在任何待办上。 没有人写下”‘本来可以’的可能性有待证明”——不是欠着不还，是账上根本没有这一笔。 而有一个当场可见的反应，本身就是证据：你向某人指出这里需要论证，他会立刻以他自己的方式当场进行论证。 一个真被论证过的东西，人不会当场重证。 问”三角形内角和为什么是180度”，他会引用，或者说”我记不清了，你去查”。引用，或者认不知道。 而这一处，他既不引用，也不认不知道——他当场就造一个。而且造得很从容，因为他不觉得自己在造，他觉得自己在复述一件明摆着的事。 那份从容的言外之意是：这个论证是容易的。 而”容易”是那份从容的产物，不是那份论证的性质。 一个可做的实验：请两个人分别当场论证，把两份摆在一起。 一个说”物理定律不禁止”，一个说”世界不是决定论的”，一个说”我当时确实可以选”——三份都很从容，三份互不相同，有的还互相矛盾。若那份论证真是现成的，三个人该给出同一份。 当场论证，意思是这是常识。而并没有这么一个常识：学术上至今没有一份可以被当成常识来用的定论。 于是形状是：内容是空的，而行为是齐全的。 社群的行为，样样都像对待常识：不引用（说这话时没有人标出处）、不教（不进课堂，因为没人觉得它需要被教）、不复核（这么说，旁人点头就过去了）、质疑者担举证责任（你要说”不见得”，得由你来说清楚）。每一条行为，恰好挡住了会暴露那个空缺的那一条路。 这套建制，类似于宗教信仰中的教义：不需要出处，不由个人负责，质疑者担举证，各人版本不同而从不校准，而最要紧的一处是——它不承担预测，所以没有可以出错的接口。 唯一不对得上的是：教义知道自己是教义，而这一条以事实的身份出场。 一个人说得出”我不信这个教”，却说不出”我不信’我本来可以’”——因为那句话在他嘴里不是一条信念，是对刚才那件事的描述。 归根到底：非常基础的信念，并不被视为需要论证。 而”基础”指的是位置，不是内容。 后悔、追责、褒奖、许诺、“你应该”——全都架在它上面。要交代它，得先把上面那些卸下来——不是没人想论证它，是那个动作太贵。 由此有一条可当场做的检验：要求出示的强度，跟位置的高低成反比。“这瓶奶过期了没有”要看日期；“他昨天到底去没去”要问要查；而”人做事总有个原因”“他本来可以不那样”——没有人要你出示。 而这本该反过来：撑得越多的，越该先查一遍。房子越高，地基越该验；而我们从不验地基，正因为房子已经盖在上面了。 所以：凡是从不被要求出示的，多半正是撑得最多的那一条。 Abstract “He could have done otherwise.” This is said countless times a day, and no one is ever asked to produce anything. The paper’s claim: this is not because the sentence has been argued for. It is that no one takes it to need an argument—so the argument has, in effect, been suspended. “In effect” must be kept. What is genuinely suspended is remembered: someone says “we shall come back to this,” or leaves a footnote; it is on the list. This is on no list. No one has written “the possibility of ‘could have’ remains to be shown”—it is not an unpaid debt but an entry that was never made. And a visible reaction is itself the evidence: point out to someone that an argument is needed here, and he will at once supply one, in his own way. No one re-proves what has genuinely been proved. Ask why a triangle’s angles sum to 180 degrees and he will cite, or say “I forget—look it up.” Citation, or an admission of not knowing. Here he neither cites nor admits: he manufactures one on the spot. And with composure, because he does not take himself to be manufacturing—he takes himself to be restating something obvious. The implication of that composure: the argument is easy. But “easy” is a product of the composure, not a property of the argument. An experiment: have two people each argue on the spot, and set the two side by side. One says “physical law does not forbid it,” another “the world is not deterministic,” another “I really could have chosen”—all composed, all different, some mutually inconsistent. Were the argument ready to hand, three people would give the same one. Arguing on the spot means: this is common sense. And there is no such common sense—no settled account exists that could serve as one. So the shape is: the content is empty while the conduct is complete. A community’s conduct is in every respect that of common sense: no citation (no one gives a source), no teaching (it enters no curriculum, since no one thinks it needs teaching), no cross-checking (one says it, others nod), and the questioner bears the burden (say “not necessarily” and it is for you to explain). Each of these conducts happens to block the very route that would expose the gap. This apparatus resembles that of a religious dogma: no source required, no individual answerable, the questioner bearing the burden, versions differing and never reconciled—and, crucially, it carries no predictions, so it has no interface at which it could fail. The one disanalogy: a dogma knows itself to be a dogma, while this appears in the guise of a fact. One can say “I do not hold that faith”; one cannot say “I do not believe ‘I could have’”—because in his mouth that sentence is not a belief but a description of what just happened. In sum: very basic beliefs are not taken to require argument. And “basic” refers to position, not content. Regret, blame, praise, promising, “you ought”—all stand upon it. To account for it, one must first take down what stands on it—it is not that no one wants to argue it, but that the act is too expensive. Hence a test: the demand for credentials falls as position rises. “Is this milk expired”—check the date; “did he go yesterday or not”—ask, verify; but “people always have a reason for what they do,” “he could have done otherwise”—no one asks you to produce anything. And it ought to run the other way: what supports the most should be checked first. The taller the house the more the foundation wants inspecting—and we never inspect it, precisely because the house is already standing on it. So: whatever is never asked for its credentials is likely the very thing holding up the most.