A3 · Thinking Tools 思考的工具
How Do Arguments Go Wrong? 論證是怎麼出錯的？
Fallacies: arguments that look better than they are, why they work on us, and how to name one without committing another.
謬誤：看起來比實際上好的論證。它們為什麼對我們有效，以及怎樣指出一個謬誤，而自己不犯另一個。
A fallacy is a bad argument in good clothing. Each of the classic ones borrows the look of a pattern that is often reasonable, which is why learning their names matters less than seeing what each one imitates.
謬誤是穿著體面的壞論證。每一個經典謬誤，都借用了某個常常合理的樣式的外表；所以比起背名字，更要緊的是看出它在模仿什麼。
中文翻譯Every paragraph has a Chinese translation. Tap 中譯▾ under a paragraph to open that one, or use Show Chinese · 顯示中譯 at the top to open them all. 每一段都有中譯：點段落下方的「中譯」只開那一段，或按上方的「顯示中譯」一次全部打開。
The essay · 課文
How Do Arguments Go Wrong? 論證是怎麼出錯的？
01The genuine and the sham 真貨與贗品
An argument that is obviously bad does little harm, because nobody is taken in by it. The dangerous ones are those that pass for good. Aristotle opened the first book ever written on the subject with exactly this observation: “That some reasonings are genuine, while others seem to be so but are not, is evident.” It happens, he added, through a certain likeness between the genuine and the sham, in the way that some people are healthy and others only made up to look it. An argument of the second kind is called a fallacy. The word does not mean a false belief, though it is often used that way. It names a fault in the route, and the destination may be true or false.
一看就知道很糟的論證沒有多大害處，因為沒有人會上當。危險的是那些冒充得了好論證的。亞里斯多德為這個題目寫了史上第一本書，開頭說的正是這件事：「有些推理是真的，另一些看似真的、其實不是，這是很明顯的。」他接著說，這是由於真貨與贗品之間有某種相似，就像有些人是真的健康，有些人只是打扮得像。第二種論證就叫作謬誤。這個詞常被用來指「錯誤的信念」，但那不是它的意思。它說的是路線上的毛病；目的地可以是真的，也可以是假的。
02Two ways to fail 兩種失敗的方式
The last lesson dealt with formal fallacies, such as affirming the consequent, where the fault lies in the pattern and can be seen with letters in place of words. Most of the fallacies met in daily life are informal. Their form may be unobjectionable. The trouble lies in what the premises say: they are beside the point, or they quietly assume too much, or a word changes its meaning halfway through. Those three troubles give the traditional three families.
上一課談的是形式謬誤，例如肯定後件：毛病出在樣式上，把詞換成字母也看得出來。日常生活裡遇到的謬誤大多是非形式的。它們的形式可能無可挑剔；問題出在前提說了什麼：前提不相干，或者悄悄預設了太多，或者某個詞在半路換了意思。這三種問題，就是傳統上的三個家族。
03Beside the point 不相干
Fallacies of relevance offer something that moves us in place of something that bears on the question at issue. The ad hominem reply attacks the speaker and leaves the argument standing. One common version, tu quoque (“you too”), tries to discredit advice by pointing out that the adviser does not follow it, as though a doctor who smokes must be wrong about smoking. The appeal to popularity treats the number of believers as evidence. The appeal to authority cites someone eminent on a matter outside his competence: an actor on medicine, a physicist on diet. Appeals to pity or fear work the same way. In every case the premise may well be true, and the conclusion is no more likely for it.
不相干的謬誤，是拿一個能打動我們的東西，頂替跟爭論中的問題真正有關的東西。人身攻擊（ad hominem）打的是說話的人，論證本身原封不動。常見的一種叫 tu quoque（「你也一樣」）：指出勸告的人自己沒做到，想藉此讓勸告失去效力，好像抽菸的醫生對抽菸的看法就一定是錯的。訴諸群眾，是把相信的人數當成證據。訴諸權威，是引用某位名人對他專長以外的事的看法：演員談醫藥，物理學家談飲食。訴諸憐憫或恐懼也是同一回事。在每一種情形裡，前提很可能是真的，結論卻沒有因此變得更可能為真。
04Assuming too much 預設太多
Fallacies of presumption smuggle in what they ought to establish. Begging the question assumes the conclusion in the premises, usually in different words. A false dilemma offers two options as though there were no others. A slippery slope asserts, without evidence for each step, that one small move will end in disaster. A hasty generalization leaps from a few cases to all of them. Post hoc ergo propter hoc, “after this, therefore because of this,” infers a cause from a mere sequence. And the straw man misrepresents an opponent’s view in a weaker form, knocks that down, and claims the victory over the original.
預設型的謬誤，把本該證明的東西偷偷夾帶進來。乞題是在前提裡預設了結論，通常換了一種說法。假兩難只給兩個選項，好像沒有別的可能。滑坡謬誤斷言一小步終將導致災難，卻沒有為中間每一步提出證據。以偏概全從少數個案一躍而到全體。Post hoc ergo propter hoc，「在此之後，所以因此之故」，單憑先後順序就推出因果。稻草人則是把對手的觀點歪曲成比較弱的版本，把那個版本打倒，然後宣稱打敗了原本的觀點。
05A word that shifts 會移動的字
Fallacies of ambiguity turn on language. In equivocation a word is used in two senses, and the argument is valid only if the sense stays fixed. “Nothing is better than good health; a cold sandwich is better than nothing; so a cold sandwich is better than good health.” The joke is harmless, but the same mechanism does real damage when the shifting word is “natural,” “theory,” or “freedom.”
歧義型的謬誤，關鍵在語言。偷換概念（equivocation）是一個詞用了兩種意思，而論證只有在意思固定不變的時候才有效。「沒有東西比健康更好；冷三明治比沒有東西好；所以冷三明治比健康更好。」這個笑話無傷大雅；可是當那個移動的詞是「自然」「理論」或「自由」的時候，同一套機制就會造成真正的傷害。
The most famous case in Chinese philosophy belongs here. Gongsun Long argued in the third century BCE that “a white horse is not a horse” (白馬非馬). The claim is absurd if 非 means “is not a kind of” and quite true if it means “is not identical with,” since the phrase “white horse” picks out color as well as shape. Whether Gongsun Long was committing an equivocation or exposing one is still debated, and that is some measure of how hard ambiguity is to pin down.
中國哲學裡最有名的例子就屬於這一類。西元前三世紀，公孫龍主張「白馬非馬」。如果「非」的意思是「不屬於」，這個主張很荒謬；如果意思是「不等同於」，它倒是真的，因為「白馬」這個詞除了形，還指定了色。公孫龍究竟是犯了偷換概念，還是在揭發偷換概念，至今仍有爭論；由此多少可以看出，歧義有多難釘住。
06The same move, done well 同一招，用得對的時候
Here the subject becomes interesting, because almost every item on the list has a respectable twin. Trusting an authority is how any of us knows that the earth is four and a half billion years old; the appeal is fallacious only when the authority is speaking outside his field, or when the experts themselves disagree. A witness’s record of lying is a fact about the person, and it is plainly relevant to whether we should accept his testimony. Some slopes really are slippery, and saying so is reasonable when each step is supported. Even a small sample can justify a conclusion if the cases are known to be alike.
這門學問到這裡才有意思，因為清單上幾乎每一項都有一個體面的雙胞胎。我們每個人之所以知道地球有四十五億年的歷史，靠的就是信任權威；只有當那位權威談的是專長以外的事，或專家之間自己意見分歧的時候，訴諸權威才是謬誤。證人的說謊紀錄是關於這個人的事實，而它顯然跟我們該不該採信他的證詞有關。有些坡確實是滑的；如果每一步都有根據，指出這一點是合理的。只要已知個案彼此相似，小樣本也可以支持結論。
So a fallacy cannot be identified by its outward shape. One has to ask what work the premise is supposed to do and whether it does it. It follows that there is a fallacy about fallacies: to show that an argument is bad is not to show that its conclusion is false. A true claim can be defended badly, and often is.
所以謬誤沒辦法靠外形辨認。你必須問：這個前提本來應該起什麼作用，它起了沒有？由此還可以推出一個「關於謬誤的謬誤」：證明一個論證不好，並不等於證明它的結論是假的。真的主張可以被辯護得很糟，而且常常如此。
07Why they work on us 它們為什麼對我們有效
If fallacies are so well cataloged, why do they still succeed? Francis Bacon gave the classic answer in 1620. “The human understanding, when any proposition has been once laid down,” he wrote, “forces everything else to add fresh support and confirmation,” and it neglects or explains away the instances on the other side. Psychologists now call this confirmation bias. Fallacies succeed for the most part on the strength of what we already want to believe. That explains the least comfortable fact about them: they are easy to spot in an opponent and almost invisible in a friend.
既然謬誤早就被整理得這麼清楚，為什麼還是屢屢得逞？培根在 1620 年給了經典的回答。他寫道：「人的理解力，一旦某個命題被確立，就迫使其他一切都來為它增添新的支持與印證」，而對另一邊的事例，則是忽略或設法解釋掉。心理學家現在把這叫作確認偏誤。謬誤多半是憑著我們本來就想相信的東西而得逞的。這也解釋了關於謬誤最令人不舒服的一個事實：在對手身上一眼就看得出來，在朋友身上幾乎看不見。
08Using the names well 把這些名字用對
The Latin labels are useful, and they are also a temptation. To answer an argument by shouting “straw man” or “ad hominem” is to stand a label in for a reply, which is itself a failure of relevance. The better practice is slower. Say what the premise was meant to show, say why it does not show it, and then ask whether the argument can be repaired. Often it can. A slippery slope becomes a decent argument once the steps are supported. A hasty generalization becomes a hypothesis worth testing. Used this way, the catalog is less a list of crimes than a set of diagnoses, and its first patient should be oneself.
這些拉丁文標籤很有用，同時也是一種誘惑。用大喊「稻草人」或「人身攻擊」來回應一個論證，是拿標籤頂替答覆，而這本身就是不相干的毛病。比較好的做法比較慢：說出那個前提原本要證明什麼，說出它為什麼沒有證明到，然後問這個論證修不修得好。常常是修得好的。把每一步都補上根據，滑坡謬誤就成了像樣的論證；以偏概全，可以變成一個值得檢驗的假設。這樣用，這份清單就不像罪名表，而像一組診斷；而它的第一個病人，應該是自己。
Thought experiment · 思想實驗
Try it: name the flaw 親手做一次：指出毛病
Eight remarks of the kind you meet every day. For each, choose the diagnosis that fits best. The explanation then shows which respectable pattern the fallacy is imitating, and what it would take to repair the argument.
八段你每天都會遇到的話。每一段選出最貼切的診斷。解說會告訴你，這個謬誤在模仿哪一個正當的樣式，以及要補上什麼，論證才修得好。
A field guide 謬誤速查
Relevance 不相干
- Ad hominem人身攻擊Rejecting a claim because of who makes it.
- Tu quoque你也一樣Rejecting advice because the adviser does not follow it.
- Appeal to authority訴諸權威Citing someone eminent outside his field.
- Appeal to popularity訴諸群眾Treating the number of believers as evidence.
- Appeal to emotion訴諸情感Offering pity or fear in place of a reason.
Presumption 預設
- Begging the question乞題Assuming the conclusion in the premises.
- False dilemma假兩難Presenting two options as the only ones.
- Slippery slope滑坡Predicting disaster through unsupported steps.
- Hasty generalization以偏概全Concluding about all from too few.
- Post hoc後此故因此Inferring a cause from a sequence.
- Straw man稻草人Attacking a weakened version of a view.
Ambiguity 歧義
- Equivocation偷換概念Using one word in two senses.
- Amphiboly語法歧義Trading on a sentence that can be parsed two ways.
Standard form · 標準形式
The argument, laid out 論證攤開來
Why is an attack on the speaker supposed to be irrelevant? The standard case against ad hominem can be set out as an argument, and it has a weak point worth finding. Click a step to see what can be said against it.
為什麼攻擊說話的人被認為是不相干的？反對人身攻擊的標準理由可以攤開成一個論證，而它有一個值得找出來的弱點。點任何一步，看看別人可以怎麼反對它。
Primary text · 原典
In their own words 原典怎麼說
The study of fallacies begins with Aristotle’s short treatise on the tricks of professional debaters. Two thousand years later John Locke gave several fallacies the Latin names they still carry.
對謬誤的研究，始於亞里斯多德談職業辯士伎倆的一篇短論。兩千年後，洛克替其中幾種謬誤取了沿用至今的拉丁文名字。
Aristotle, Sophistical Refutations; John Locke, An Essay Concerning Human Understanding 亞里斯多德《辯謬篇》；洛克《人類理解論》 · Aristotle translated by W. A. Pickard-Cambridge; Locke in the original English (both public domain)
That some reasonings are genuine, while others seem to be so but are not, is evident. This happens with arguments, as also elsewhere, through a certain likeness between the genuine and the sham.
有些推理是真的，另一些看似真的、其實不是，這是很明顯的。在論證上會如此，就像在別的地方一樣，是由於真貨與贗品之間有某種相似。
The whole theory is in the word “likeness.” A fallacy is defined by reference to the good argument it resembles, which is why every item in the exercise above has an “it imitates” line. Aristotle went on to list thirteen types, six that depend on language and seven that do not, and most modern lists descend from his.
整套理論都在「相似」這個詞裡。謬誤是參照它所模仿的好論證來界定的；所以上面練習裡的每一題，都有一行「它模仿的是」。亞里斯多德接著列出十三種：六種依賴語言，七種不依賴；現代的清單大多是從他那裡傳下來的。
The first is, to allege the opinions of men, whose parts, learning, eminency, power, or some other cause has gained a name, and settled their reputation in the common esteem with some kind of authority.
第一種，是援引某些人的意見；這些人的才能、學問、顯赫、權勢或別的原因，使他們博得了名聲，並且在公眾的評價裡確立了某種權威。
Locke calls this argumentum ad verecundiam, the argument addressed to modesty, because the hearer is made to feel that doubting such a person would be insolent. Notice where he locates the fault. He does not say authorities are usually wrong. He says the argument works by shaming the listener into silence and so produces assent without knowledge. (“Parts” here means abilities.)
洛克把這稱為 argumentum ad verecundiam，「訴諸謙遜的論證」，因為它讓聽的人覺得，懷疑這樣的人物就是無禮。注意他把毛病放在哪裡。他沒有說權威通常是錯的。他說的是：這種論證靠著讓聽者羞於開口而奏效，所以產生的是沒有知識的同意。（這裡的 parts 是「才能」的意思。）
Both sides · 正反兩面
Objections and replies 反駁與回應
The list of fallacies is a ragbag. It has no principle behind it, textbooks disagree about what belongs on it, and learning thirty Latin names teaches nobody to reason.
謬誤清單是個大雜燴。它背後沒有原則，各家教科書對該收哪些意見不一；背三十個拉丁文名字，教不會任何人推理。
The charge was made forcefully by the logician C. L. Hamblin in 1970, and much of it sticks: the traditional list grew by accident. But a field guide can be useful without being a theory. What matters is that each entry be taught together with its legitimate twin, so that the student learns a question to ask instead of a label to apply.
邏輯學家漢布林在 1970 年有力地提出過這項指控，而且大半說中了：傳統清單是偶然累積起來的。不過，野外圖鑑不必是理論，也可以有用。要緊的是，每一項都要連同它正當的雙胞胎一起教，讓學生學到的是一個該問的問題，而不是一個可以貼的標籤。
Many so-called fallacies are sensible shortcuts. Nobody can check everything personally, so going along with experts and with the majority is what a rational person with limited time ought to do.
許多所謂的謬誤，其實是明智的捷徑。沒有人能事事親自查證；所以跟著專家、跟著多數人走，正是時間有限的理性人應該做的。
This is right, and it corrects a common misuse of the subject. Such appeals fail as proofs and succeed, often, as evidence. The mistake is to treat a defeasible reason, one that holds only until something overrides it, as if it settled the matter. On this view, to call something a fallacy is to say that its support has been overstated.
這是對的，而且糾正了這門學問常見的誤用。這類訴求當作證明是失敗的，當作證據卻常常是成功的。錯誤在於把一個「可被推翻的理由」——在沒有別的東西壓過它之前才成立的理由——當成一錘定音。照這個看法，說某個東西是謬誤，意思是說：它提供的支持被誇大了。
Fallacy-spotting makes people worse. It hands them a vocabulary for dismissing whatever they dislike.
抓謬誤會讓人變得更糟。它給了人一套詞彙，用來打發一切他不喜歡的東西。
There is evidence for the worry, since more skilled reasoners are often better at defending what they already believe. The remedy is the order of application. Use the catalog on your own side first, state the other side’s argument in its strongest form before criticizing it, and remember that a refuted argument leaves its conclusion standing.
這個擔憂是有根據的：比較會推理的人，常常只是更擅長替自己原本的信念辯護。解方在於使用的順序。先把這份清單用在自己這一邊；批評之前，先把對方的論證用最強的形式說出來；並且記得，論證被駁倒了，它的結論仍然站著。
Terms of art · 哲學術語
The vocabulary of the subject 這門學問的用語
L. fallacia ‘deceit, trick’
An argument that seems better than it is. Not a synonym for “false belief.”
看起來比實際上好的論證。不是「錯誤信念」的同義詞。
L. ‘to the man’
Rejecting a claim on the ground of facts about the person who makes it.
以關於說話者的事實為由，拒絕他的主張。
L. ‘you too’
Dismissing a criticism because the critic is guilty of the same thing.
因為批評者自己也犯同樣的毛病，就不理會他的批評。
A distorted, weaker version of an opponent’s position, set up to be knocked down.
把對手的立場扭曲成比較弱的版本，立起來好打倒。
Gk. di- ‘two’ + lēmma ‘premise’
Presenting two alternatives as though they were the only ones.
把兩個選項說成僅有的選項。
L. aequus ‘equal’ + vox ‘voice’
Using a word in two senses within one argument.
在同一個論證裡，一個詞用了兩種意思。
L. ‘after this, therefore because of this’
Inferring that one event caused another merely because it came first.
只因為一件事先發生，就推斷它造成了後一件事。
L. via Old French desfaire ‘to undo’
Said of a reason that supports a conclusion unless further information overrides it.
形容這樣的理由：它支持結論，除非有進一步的資訊壓過它。
The tendency to notice and favor evidence for what one already believes.
傾向於注意並偏好支持自己既有信念的證據。
Advanced English · 高級英文
Words and phrases worth keeping 值得帶走的字與片語
Nobody is taken in by an argument that is obviously bad.
一看就很糟的論證，沒有人會上當。
The dangerous arguments are those that pass for good.
危險的論證，是那些冒充得了好論證的。
The premise does not bear on the question at issue.
這個前提跟爭論中的問題無關。
Tu quoque tries to discredit advice by attacking the adviser.
「你也一樣」想藉著攻擊勸告者，讓勸告失去效力。
Fallacies of presumption smuggle in what they ought to establish.
預設型的謬誤，把本該證明的東西偷偷夾帶進來。
The straw man misrepresents an opponent’s view in a weaker form.
稻草人把對手的觀點歪曲成比較弱的版本。
Fallacies of ambiguity turn on language.
歧義型的謬誤，關鍵在語言。
Ambiguity is hard to pin down.
歧義很難釘住。
It neglects or explains away the instances on the other side.
對另一邊的事例，它忽略或設法解釋掉。
Fallacies succeed on the strength of what we already want to believe.
謬誤是憑著我們本來就想相信的東西而得逞的。
Reading check · 理解測驗
Did it land? 讀懂了嗎
1. How does the essay define a fallacy?
2. When is an appeal to authority reasonable, according to the essay?
3. What is the “fallacy about fallacies”?
4. Why, on Bacon’s account, do fallacies keep succeeding?
5. What did Locke originally mean by argumentum ad hominem?
Write · 你怎麼想？
Collect three fallacies from this week’s news, advertising, or social media. For each, name it, say which legitimate pattern it imitates, and rewrite it as a better argument for the same conclusion.
從這禮拜的新聞、廣告或社群媒體蒐集三個謬誤。每一個都說出名稱、說明它模仿哪一個正當的樣式，再把它改寫成支持同一個結論的更好論證。
Describe a belief of your own that you once defended with a bad argument. What was the fallacy, and do you still hold the belief?
描述一個你自己的信念，你曾經用一個壞論證替它辯護。那是哪一種謬誤？你現在還持有這個信念嗎？
Is it ever reasonable to believe something because most people believe it? Answer in about 150 words, with one example on each side.
「因為大多數人相信，所以我也相信」，這有合理的時候嗎？用大約 150 個英文字回答，正反各舉一個例子。
Dharma and the West · 佛法與西方哲學
A view from the Dharma 佛法怎麼看
The discourse to the Kālāmas (Aṅguttara Nikāya 3.65) reads in places like a list of fallacies of relevance. The Buddha tells a puzzled town not to settle a question by oral tradition, by lineage, by hearsay, by the authority of scripture, by the speaker’s apparent competence, or by the thought “this ascetic is our teacher.” The list also includes “by logic” and “by inference,” which shows that it is not a charter for free thinking, as it is often taken to be. Its point is that none of these sources is sufficient by itself. The test he gives is practical: whether a thing, when taken up, leads to harm or to welfare.
對伽藍族人所說的那部經（《增支部》3.65），有些段落讀起來像一份不相干謬誤的清單。佛陀告訴一個感到困惑的城鎮：不要因為口耳相傳、因為傳承、因為傳聞、因為合於經典的權威、因為說者看似有能力，或者因為「這位沙門是我們的老師」，就據以論定一個問題。這份清單也包括「因為邏輯」與「因為推論」；可見它並不像一般常說的那樣，是一份自由思想的宣言。它的重點是：這些來源沒有一個單憑自己就足夠。他給的檢驗是實踐上的：一件事，照著去做，是導向損害，還是導向利益。
Buddhist logicians later did something the Western tradition did not manage. They derived their fallacies from a single principle. Dignāga held that a good reason must satisfy three conditions (trairūpya): it must be a property of the subject, it must be present in similar cases, and it must be absent from all dissimilar cases. A fallacious reason (hetvābhāsa, 似因) is one that fails a condition, and so there are three kinds: the unestablished, the inconclusive, and the contradictory. Hamblin’s complaint that the Western list has no principle behind it does not apply here. The price is narrowness: the scheme covers faults of inference and has nothing to say about a straw man (Dharma and the West, D14).
佛教的邏輯學家後來做到了一件西方傳統沒做成的事：從單一原則推導出他們的謬誤。陳那主張，好的「因」必須滿足三個條件（因三相，trairūpya）：它必須是主題所具有的性質（遍是宗法性），必須在同類事例中出現（同品定有性），必須在一切異類事例中都不出現（異品遍無性）。似因（hetvābhāsa）就是不滿足其中某個條件的因，所以有三類：不成、不定、相違。漢布林抱怨西方的清單背後沒有原則，這個抱怨在這裡用不上。代價是範圍窄：這套架構涵蓋的是推論上的過失，對稻草人這類毛病無話可說（佛法與西方哲學 D14）。
Go further · 延伸
People and reading 人物與延伸閱讀
- Aristotle, Sophistical Refutations, translated by W. A. Pickard-Cambridge ↗The first catalog of fallacies; chapters 1 to 5 give the thirteen types. 第一份謬誤目錄；第一到五章列出十三種。
- John Locke, An Essay Concerning Human Understanding, Book IV, chapter 17, sections 19–22 ↗Two pages in which three Latin names are coined. 兩頁之內，造出三個拉丁文名稱。
- Francis Bacon, Novum Organum (1620), Book I, aphorisms 38–68 ↗The four Idols: why minds go wrong before any argument begins. 四種偶像：在任何論證開始之前，心靈為什麼就已經走偏。
- Stanford Encyclopedia of Philosophy, “Fallacies” ↗Includes the history and Hamblin’s criticism of the standard treatment. 包含歷史沿革，以及漢布林對「標準處理法」的批評。
- Stanford Encyclopedia of Philosophy, “School of Names” ↗On Gongsun Long and the white horse. 談公孫龍與白馬。