Two Dogmas in Retrospect 12

Two Dogmas in Retrospect12 回首两个教条 原著：W•V•奎因，译者：翟玉章 In retrospecting “Two Dogmas” I find myself overshooting the mark by twenty years. I think back to college days, 61 years ago. I majored in mathematics and was doing my honors reading in mathematical logic, a subject that had not yet penetrated the Oberlin curriculum. My new love, in the platonic sense, was Whitehead and Russell’s Principia Mathematica. 在回首“两个教条”一文时，我发现自己曾经过分攻击了目标，而且有 20 年之久。我回想 起了自己 61 年前读大学的时候。当时我主修数学，并且把数理逻辑作为我的优等生学习任 务。奥柏林学院那时还没有开设这门课程。怀特海和罗素的《数学原理》成了我的新爱，那 是一种柏拉图意义上的爱。 I was taken with the clear, clean incisiveness of its formulas. But this was not true of its long introduction to volume 1, nor of some of the explanatory patches of prose that were interspersed through the three volumes. In those pages and passages the distinction between sign and object, or use and mention, was badly blurred. Partly in consequence, there was vague recourse to intensional properties, or ideas, under the disarmingly technical name of propositional functions. These ill-conceived mentalistic notions paraded as the philosophical foundation for the clean-cut classes, truth functions, and quantification that would have been a far better starting point in their own right. 我深深折服于书中那些清晰、明快而又深刻的公式。但第一卷的长长的导言是个例外，散布 于全部三卷中的一些解释性段落也是例外。在这些篇章和段落中，符号和对象之间的区别， 或者使用和提及之间的区别，被严重地混淆了。部分地由于这种混淆，这本书通过“命题函 项”这一令人放心的技术名词，语焉不明地引入了内涵性质或观念。这些欠考虑的心灵主义 概念竟得以招摇过市，充任了类、真值函项和量化这些清晰概念的哲学基础，其实这些清晰 的概念本身本可以成为一个更好的起点的。 The distrust of mentalistic semantics that found expression in “Two Dogmas” is thus detectable as far back as my senior year in college. Even earlier I had taken kindly to John B. Watson’s Psychology from the Standpoint of a Behaviorist, which Raymond Stetson had assigned to us in his psychology class. Nor do I recall that it shocked any preconceptions. It chimed in with my predilections. 所以，“两个教条”所表达的对心灵主义语义学的不信任，可以追溯到我上大四的时候。甚

原载《淮阴师范学院学报》2013 年第 2 期（第 165-170 页）。 This is a revised version of a paper that I presented at the University of Toronto in December, 1990 in memory of my presentation of ‘Two Dogmas of Empiricism’ there in December, 1950.这是我 1990 年 12 月向多 伦多大学提交的一篇论文的修订版，是为对我 1950 年 12 月在此提交“经验论的两个教条”一文的纪念。 1 2

至在大四以前，我就很喜欢约翰•B•沃森的《行为主义视野中的心理学》一书，这是雷蒙德• 斯泰森在心理学课上给我们指定的读物。我想不起来它曾动摇过我的任何先入之见。行为主 义与我的偏好是一致的。 In 1931, my second year out of college, I was writing my PhD thesis under Whitehead’s sponsorship at Harvard. It is my first and latest book: first because I wrote it before the others, and latest because it was published only a few months ago. Garland Publishing Company included it in their recently conceived series of twentieth-century Harvard PhD thesis in philosophy. They did a surprisingly tidy job of my painfully wrought typescript, replete with logical symbols. I’m tickled. But this is not a commercial. The book is expensive, and I can offer better ones. 1931 年，也就是大学毕业后的第二年，我在哈佛的怀特海门下写作博士论文。这是我最早 也是最晚的一本书：最早，是因为它的写作比其他书早；最晚，是因为它几个月前才出版。 它被选进了加兰出版公司最近策划的哈佛大学 20 世纪哲学博士论文丛书。对我那费力搞定 的充满了逻辑符号的打字稿，他们做了超乎想像的整理工作。对此我很高兴。但我并不想在 这里做广告。这本书很贵，我可以提供更好的书。 The relevance of my PhD thesis to our present topic is that in it I reworked the foundations of Principia Mathematica in strictly extensional terms, and propositional functions to the winds. The term‘propositional function’ carried over, but to denote strictly classes, as I would now phrase it. 我的博士论文与我们现在的话题的相关性在于，它使用严格的外延术语重塑《数学原理》的 基础，而把命题函项丢到了一边。“命题函项”这个术语虽然被保留了下来，但被严格地用 来表示类（这是我现在的措辞）。 I was not abetted in my extensionalism by the Harvard professors of that time. Whitehead, C. I. Lewis, H. M. Sheffer, and E. V. Huntington all were soft on intensions and introspective meanings. But a postdoctoral fellowship the next year took me to a kindred spirit in Czechoslovakia: the great Carnap. He was just finishing his Logische Syntax der Sprache, and I read it and discussed it with him as it issued from his wife’s typewriter. He was setting his face steadfastly against modal logic and mentalistic talk of meanings. His little logic text Abriss der Logistik had likewise been impeccably extensional, as was his impressive application of mathematical logic to epistemology in Der logische Aufbau der Welt. 我的外延主义并没有得到那时的哈佛教授们的鼓励。怀特海、C•I•刘易斯、H•M•谢费尔和 E•V•亨廷顿全都对内涵和内省意义持姑息的态度。但是第二年的博士后研究让我在捷克斯洛 伐克结识到一位志同道合之士：伟大的卡尔纳普。那时他的《语言的逻辑句法》一书马上就 要完成了，它从他妻子的打字机上一出炉，我就读到了它，并与他进行讨论。他坚决抵制模 态逻辑和对意义的心灵主义谈论方式。他以前的作品，无论是给人留下深刻印象的运用数理 逻辑处理认识论问题的《世界的逻辑结构》，还是简明的逻辑读物《符号逻辑概论》 ，表现出 来的也都是彻头彻尾的外延主义。 True, in Logische Syntax we find him making capital of a purported distinction between analytic

and synthetic truths. He didn’t see this as a reinstatement of mentalistic meanings. He saw it as a matter of truth by linguistic convention. I expressed misgivings already in our discussions there in Prague, March 1933. Neil Tenant has uncovered, in the Carnap archives in Pittsburgh, this entry in Carnap’s unpublished jottings of the time. I translate Carnap’s somewhat telegraphic German: Quine, 31.3.33 He says after some reading of my“Syntax”MS： 1. Is there a difference in principle between logical axioms and empirical sentences? He thinks not. Perhaps I seek a distinction just for its utility, but it seems he is right: gradual difference: they are the sentences we want to hold fast. 确实，在《逻辑句法》中，我们看到他用到了分析真理和综合真理之间的所谓区别。但他并 不认为他会因此倒向心灵主义意义观。他把这一区别看成是语言约定的结果。我于 1933 年 3 月在布拉格与他讨论时就曾表达过我的不满。尼尔•坦南特在匹兹堡的卡尔纳普档案中发 现了一则当时未发表的笔记，我现在将他的类电报体的德文翻译如下： 奎因，31.3.33 他在对我的《句法》（手稿）作了一些阅读后说： 1. 在逻辑公理和经验语句之间存在着原则上的区别吗？他认为没有。也许我寻求的只 是一个有用的区别，但看来他是正确的：区别是程度上的：它们是那些我们想紧紧抓住 不放的语句。 Evidently Carnap allayed my misgivings for a while. Three expository lectures on Carnap that I gave at Harvard in 1934 were abjectly sequacious. But my misgivings surfaced again in 1935, when I wrote “Truth by Convention”. I quote the end of the first paragraph: [D]evelopments of the past few decades have led to a widespread conviction that logic and mathematics are purely analytic or conventional. It is less the purpose of the present inquiry to question the validity of this contrast than to question its sense. 显然，卡尔纳普有一阵子减轻了我的不满。我 1934 年在哈佛所作的三场阐释卡尔纳普哲学 的演讲，对他是完全附和的。但是不满在 1935 年我写作“根据约定的真理”时再次浮出水 面。我引这篇文章第一段的结尾部分： 近几十年的发展导致了一个广泛的信念：逻辑和数学是纯分析的或纯约定的。这篇论 文质疑的与其说是这一对比的有效性，毋宁说是这一对比的意义。 Carnap wrote to me in 1936 expressing “very much interest” in “Truth by Convention,” and added that he was “very keen of discussing it when we meet in April.” So I have no record of his reaction, not remembering the discussion. There must have been frequent discussions, for he was around Harvard all that summer. 卡尔纳普 1936 年写信给我，表示他对“根据约定的真理”“很感兴趣”，并说他“非常期待 我们 4 月会面时探讨这个话题。” 由于我记不得当时的讨论了，所以对于他的回应也就无从 查考。但是，当时一定是有过频繁的讨论的，因为他整个夏季都在哈佛。 I was diverging from Carnap because his aloofness from intensions and mentalism, which had so

appealed to me, had proved to be insufficiently austere. Ironically, in those same years Carnap came to welcome intensions increasingly. The trend began when Tarski persuaded him that his” Thesis of Syntax” was untenable: the thesis that “philosophy is the syntax of the language of science”. Semantics was wanted, not just syntax. Tarski was right, in his own austere sense of semantics, namely model theory and his theory of truth. But Carnap went further, even embroiling himself in modal logic. I learned of this in 1938 from Hempel, and wrote to Carnap in dismay. He duly pondered my sermon, so he wrote in reply, and he went on to defend himself as follows: Although we do not ordinarily like to apply intensional languages, nevertheless I think we cannot help analyzing them. What would you think of an entomologist who refuses to investigate flees【sic】 and lice because he dislikes them? Well, the fleas and lice proved addictive. By 1946 he was championing modal logic. 卡尔纳普对内涵和心灵主义的冷漠，曾经是那么引起我的共鸣。我后来所以与卡尔纳普产生 分歧，是因为他的冷漠实际上是不够一贯的。具有讽刺意味的是，也就是在这些年间，卡尔 纳普越来越欢迎内涵实体了。这种转向开始于塔尔斯基对他的影响，他认识到他的“句法论 题”（即“哲学就是科学语言的句法”的论题）是站不住脚的。不仅需要句法，还需要语义 学。塔尔斯基本人的那种严谨的语义学，即他的模型理论和真理理论，是正确的。但卡尔纳 普走得更远，以致让自己卷进了模态逻辑。我是 1938 年从亨普尔那里了解到这一点的，并 且写信给卡尔纳普表达了我的惊诧。他掂量出我不满的份量，回了一封信给我，为自己作了 下面的辩解： 虽然我们通常不喜欢使用内涵语言，但我仍然认为应该研究它们。如果一个昆虫学家 因为不喜欢跳蚤和虱子而拒绝研究它们，你会怎么看呢？ 看来，跳蚤和虱子是足以让人着迷的。到 1946 年，他成了模态逻辑的拥护者。 Carnap was again at Harvard in the fall and winter term of 1940-41, along with Tarski and Russell. Glorious days. Tarski and I argued persistently with Carnap over his appeal to analyticity in the opening pages of his work in progress, Introduction to Semantics. 卡尔纳普于 1940-41 年的秋冬学期又来到了哈佛，同在的还有塔尔斯基和罗素。那是一段非 常愉快的日子。针对卡尔纳普正在写作的《语义学导论》的开篇对分析性的诉求，我和塔尔 斯基与他进行了持续的辩论。 You don’t write when you can meet and talk. I find nothing on analyticity and meaning in my correspondence with Carnap until 1943, when I was in Washington as a naval officer. I wrote him a long letter about his Introduction to Semantics. One issue was analyticity and another was my criterion of ontological commitment as applied to abstract objects. The two issues were linked, for Carnap viewed his appeal to abstract objects as empty convention, and their quasi-existence analytic. 如果可以见面和交谈，就不用写了，所以我没能找到这个时期我与卡尔纳普在分析性和意义 问题上的来往信件。直到 1943 年我在华盛顿当海军军官的时候，我才就他的《语义学导论》 一书给他写了一封长信。信中讨论的一个问题是分析性，另一个是关于我的本体论承诺的标 准运用于抽象对象的问题。这两个问题对于卡尔纳普来说是联系在一起的，因为他认为他对

抽象对象的诉求是一个空洞的约定，抽象对象的准存在性是分析的。 I had not thought to look on my strictures over analyticity as the stuff of revolution. It was mere criticism, a negative point with no suggestion of a bright replacement. I had not felt moved to follow “Truth by Convention” with more of the same. But word got around. Nelson Goodman had joined in some of the discussions of 1940-41 with Carnap, Tarski and me. In June and July of 1947 a triangular correspondence on the issue developed among Goodman, Morton White, and me. In 1950 I was invited by the program committee of the American Philosophical Association to present a paper on the subject at the December meeting here in Toronto. Hence “Two Dogmas.” It is remarkable that my most contested and anthologized paper was an assignment. The response was quick and startling. The paper appeared in Philosophical Review a few weeks after the Toronto meeting, and four months later there were symposia on it in Boston and at Stanford. 我未曾料到对分析性的批评会具有革命性。我只是在纯粹而消极地批评，并没有拿出正面的 替代方案。我没有想过要进一步发挥“根据约定的真理”里的思想。但我对分析性的批评还 是传播出去了。纳尔逊·古德曼也加入了 1940-41 年间在卡尔纳普、塔尔斯基和我之间的一 些讨论。1947 年的 6 月和 7 月，古德曼、莫顿·怀特和我之间就这个问题展开了三角通信。 1950 年，美国哲学学会项目委员会邀请我向 12 月在这里（多伦多）召开的会议提交一篇关 于这个问题的论文。这就是“两个教条”的诞生经过。引人注目的是，我的这篇被争论得最 多，被选录次数最多的论文，其最早出处居然是一篇命题作文。它立即引起了不同凡响的反 应。多伦多会议结束几个星期后，它刊登在了《哲学评论》杂志上，4 个月后在波士顿和斯 坦福大学就有了关于它的研讨会。 Looking back on it, one thing I regret is my needlessly strong statement of holism. The unit of empirical significance is the whole of science… Any statement can be held true come what may, if we make drastic enough adjustments…Conversely…no statement is immune to revision. 回过头来看，我很遗憾的一件事是，我对整体主义的陈述太过强硬了，而这其实是不必要的。 具有经验意义的单位是整个科学……任何陈述在任何情况下都可以被坚持为真，只要 我们做出足够剧烈的调整…反之…没有任何陈述是免受修改的。 This is true enough in a legalistic sort of way, but it diverts attention from what is more to the point: the varying degrees of proximity to observation, the example of the brick houses in Elm Street. In later writings I have invoked not the whole of science but chunks of it, clusters of sentences just inclusive enough to have critical semantic mass. By this I mean a cluster sufficient to imply an observable effect of an observable experimental condition. 虽然这个观点通过咬文嚼字完全能被坚持，但它却使我们的注意力离开了更加中肯的东西： 与观察的接近性是个程度问题。我在那篇文章中曾举榆树街上的砖房子为例说明此点。在后 来的作品中，我援引的已经不再是整个科学，而是科学中足够大的部分，即刚好能具有临界 语义量的语句群。我的意思是，这样的语句群足以蕴含一定的可观察实验条件下的可观察结 果。

I formulate the matter nowadays in terms of what I call observation categoricals. An observation categorical is a generalization of the form ‘Whenever this, that’ where ‘this’ and ‘that’ are observation sentences. I’ll say more about observation sentences later. Now a cluster of sentences has critical semantic mass if it implies an observation categorical; and the experimental check of the cluster of sentences consists in testing the implied categorical by arranging for fulfillment of its observable protasis and seeing whether the apodasis is realized. 如今，我用我所谓的观察直言句来表达我的观点。观察直言句是指形如“每当这样，就那样” 的概括句，其中“这样”和“那样”是两个观察句。我将在后面对观察句作出进一步的说明。 一个语句群具有临界语义量，如果它能蕴含某个观察直言句；对这一群语句的实验检验，其 实就是检验它所蕴含的观察直言句：先让条件子句所说的可观察条件得到满足，再看结果子 句所说的结果有没有出现。 This is meant as a schematic caricature of the experimental method. In practice many sentences of the cluster are tacit and some are probabilistic or dependent on unspecified things being equal. But I think it catches the essence of experimental testing, and therewith of empirical content. 这是对实验方法的缩略而夸张的说明。在实践中，语句群中的许多语句是不曾明言的，有些 则是或然的或依赖于未具体说明的同等条件。但是我认为它抓住了实验检验以及与之相联的 经验内容的本质。 I have appealed here to implication: the cluster of sentences implies the observation categorical. In so doing I give logic a special status: logical implication is the link between theory and experiment. However, we remain free here to adjust and to vary the limits of what to count as logic. 我在这里诉诸蕴含：一群语句蕴含观察直言句。我这样做给了逻辑一种特殊的地位：逻辑蕴 含是理论和实验相联系的纽带。不过，即使在这里，在逻辑界限的调整和改变方面，我们仍 拥有自由。 Thus consider the law of excluded middle, which I cited in “Two Dogmas” to illustrate my claim that ‘no statement is immune to revision.’ If the prospect of simplification and clarification at the level of quantum physics were to prove so overwhelming as to induce us to abrogate the law of excluded middle, we could still do so. It would just mean setting the limits of what to count as logic very narrowly and reckoning the law of excluded middle to the clusters of defeasible hypotheses that do the implying. 在“两个教条”中，为了说明“没有任何陈述是免受修改的”这一主张，我举了排中律的例 子。在这里我们就再来谈谈排中律。如果在量子物理学层面上的简单性和清晰性在其他领域 到头来也是势不可挡的，那么我们就有可能废弃排中律。这仅仅意味着我们把逻辑的界限大 大收窄了，而将排中律视为参与蕴含的一群可废止的假说中的一个假说。 In a footnote to “Two Dogmas” I noted Duhem’s priority in stressing holism. As a matter of curiosity, however, I might mention that when I wrote and presented “Two Dogmas” here forty

years ago, and published it in the Philosophical Review, I didn’t know about Duhem. Both Hempel and Philipp Frank subsequently brought Duhem to my attention, so I inserted the footnote when “Two Dogmas” was reprinted in From a Logical Point of View. Another insert was page 35 on postulates, which was prompted by a paper by Richard Martin in defense of Carnap on analyticity. But that was explained in a footnote. 在“两个教条”的一个脚注里，我提到迪昂更早地强调了整体论。不过，我可以当作一则轶 事提一下，当我 40 年前写作并提交“两个教条”时，以及在《哲学评论》发表这篇文章时， 我并不知道迪昂。随后，亨普尔和菲利普•弗兰克都使我注意到了迪昂，于是在“两个教条” 被重印进《从逻辑的观点看》一书时，我插入了这个脚注。另一处插入是第 35 页上关于公 设的段落，它是由理查德•马丁为卡尔纳普的分析性概念辩护的一篇论文所促成的。但这一 点已经在一个脚注中说明过了。 I think Carnap’s tenacity to analyticity was due largely to his philosophy of mathematics. One problem for him was the lack of empirical content: how could an empiricist accept mathematics as meaningful? Another problem was the necessity of mathematical truth. Analyticity was his answer to both. 我认为卡尔纳普对分析性的固守在很大程度上是因为他的数学哲学。他的一个问题是数学缺 乏经验内容：经验主义者怎么能认为缺乏经验内容的数学是有意义的呢？另一个问题是数学 真理的必然性。分析性是他对这两个问题的答案。 I answer both with my moderate holism. Take the first problem: lack of content. Insofar as mathematics gets applied in natural sciences, I see it as sharing empirical content. Sentences of pure arithmetic and differential calculus contribute indispensably to the critical semantic mass of various clusters of scientific hypotheses, and so partake of the empirical content imbibed from the implied observation categoricals. 我则是用我的温和整体论来回答这两个问题的。先说第一个问题，即缺乏经验内容的问题。 只要数学被用于自然科学，我就认为它分享了经验内容。纯算术和微积分中的语句对于许多 科学假说群之获得临界语义量是不可或缺的，所以也就带有了它们从所蕴含的观察直言句吸 收到的经验内容。 As for inapplicable parts of mathematics, say higher set theory, I sympathize with the empiricist in questioning their meaningfulness. We do keep their sentences as meaningful, but only because they are built of the same lexicon and grammatical constructions that are needed in applicable mathematics. It would be an intolerably pedantic tour de force to gerrymander our grammar in such a way as to account the inapplicable flights ungrammatical while preserving the applicable part. 至于数学中不可被应用的部分，比如高阶集合论，我同情经验主义者对它们的有意义性的质 疑。我们之所以仍然认为它们是有意义的，只是因为它们和可应用的数学具有相同的词汇和 语法。如果我们刻意调整我们的语法，使合乎语法的语句只及于可应用的部分而不及于不可 应用的部分，这也未免过于迂腐了。

This does put it to us to ponder truth and falsity for the inapplicable sentences. In doing so we can get some guidance from a maxim that is already serving the natural scientist, namely Occam’s razor: where choice is otherwise undetermined, opt for economy. This attitude is in keeping with my inclination to minimize the cleavage between mathematics and natural science. That tendency is abetted already by my point about shared empirical content, and also by my questioning the analytic/synthetic distinction. 这确实带来了一个问题：我们该如何决定这些不可应用的语句的真假呢？在这方面，奥康姆 剃刀可以给我们一些启发。这个在自然科学家那里已经在发挥作用的原则的意思是：如果用 别的方法不足以作出选择，那就选择最经济的。这一态度是与我竭力缩小数学和自然科学之 间的区别的倾向相一致的。我关于经验内容被共享的主张，以及我对分析/综合区别的质疑， 则使这种倾向得到了助长。 What then about the other problem, that of the necessity of mathematical truth? This again is nicely cleared up by moderate holism, without the help of analyticity. For let us recall that when a cluster of sentences with critical semantic mass is refuted by an experiment, the crisis can be resolved by revoking one or another sentence of the cluster. We hope to choose in such a way as to optimize future progress. If one of the sentences is purely mathematical, we will not choose to revoke it; such a move would reverberate excessively through the rest of science. We are restrained by a maxim of minimum mutilation. It is simply in this, I hold, that the necessity of mathematics lies: our determination to make revisions elsewhere instead. I make no deeper sense of necessity anywhere. Metaphysical necessity has no place in my naturalistic view of things, and analyticity hasn’t much. 另外一个问题，即数学真理的必然性，又怎么样呢？这也可以由温和整体论得到很好的解释， 而无需用到分析性。让我们回想一下，当具有临界语义量的一个语句群被实验反驳时，这一 危机可以通过取消这一群中的这个语句或那个语句得到解决。我们希望我们的选择能够有利 于未来的进步。如果其中一个语句是纯粹的数学语句，我们便不选择取消它；这样的举措将 会引起科学的其他部分的过度波动。最小损毁原则是这里的指导原则。简单说来，我的主张 是这样：数学的必然性在于我们不修改数学而修改其他部分的决心。我在任何地方都没有赋 予必然性以更深刻的意义。形而上学必然性在我的自然主义哲学里没有任何地位，分析性也 是如此。 Analyticity undeniably has a place at a common-sense level, and this has made readers regard my reservations as unreasonable. My threadbare bachelor example is one of many undebatable cases. It is intelligible and often useful in discussion to point out that some disagreement is purely a matter of words rather than of fact. The point can commonly be sustained and acted upon by a paraphrase that circumvents a troublesome word. Often in talking with a foreigner we recognize some impasse as due to his having mislearned an English word rather than to his having a bizarre view of the subject matter. This is a bit of practical psychology at which we are all adept. 分析性在常识层面上无疑是有一席之地的，这让我对分析性的保留态度让读者觉得不可思议。

我的关于单身汉的乏味的例子就是许多无可争辩的例子之一。当我们在讨论中想指出某个分 歧纯粹是语词上的分歧而不是事实上的分歧时，分析性概念是一个容易理解的概念，而且常 常也很有用。这种分歧通常可以通过对成问题的词进行释义来解决。在与外国人交谈时，我 们也经常注意到，某些僵局的出现，并不是因为外国人在论题上持有某种离奇的观点，而是 因为他误解了某个英语词。以上是我们都很擅长的一些实用心理学。 In Roots of Reference I proposed a rough theoretical definition of analyticity to fit these familiar sorts of cases. A sentence is analytic for a native speaker, I suggested, if he learned the truth of the sentence by learning the use of one or more of its words. This obviously works for ‘No bachelor is married’ and the like, and it also works for the basic laws of logic. Anyone who goes counter to modus ponens, or who affirms a conjunction and denies one of its components, is simply flouting what he learned in learning to use ‘if’ and ‘and’. (I limit this to native speakers, because a foreigner can have learned our words indirectly by translation.) 在《指称之根》一书中，我为分析性提出了一个粗略的理论定义，以便说明这些熟悉的例子。 我建议：某个句子对某个母语人士是分析的，如果他是在学习其中一个或几个词的用法的过 程中知道其真值的。很明显，这个定义适用于“没有单身汉是已婚的”以及类似的例子，也 适用于逻辑的基本规律。一个人如果与分离规则对着干，或者在肯定一个合取句的同时却否 定其中一个成分句，那么他将是在公然藐视他已经在语言学习过程中学会的“如果”和“而 且”的用法。（我把此处所论仅限于母语人士，因为一个外国人可以通过翻译间接地学会我 们的词。） I also recommended improving this rough definition by providing for deductive closure, so that truths deducible from analytic ones by analytic steps would count as analytic in turn. All logical truths in my narrow sense--that is, the logic of truth functions, quantification, and identity--would then perhaps qualify as analytic, in view of Gödel’s completeness proof. 我还建议通过规定演绎闭合改进这个粗略的定义，即把所有能从分析真理经过分析的步骤演 绎而来的真理都算做分析真理。这样一来，在我的狭义逻辑（即关于真值函项、量化和等同 的逻辑）的范围内，根据哥德尔的完全性证明，所有逻辑真理大概都是分析真理。 If the logical truths are analytic—hence true by meaning of words—then what are we to say of revisions, such as imagined in the case of the law of excluded middle? Do we thereby change our theory or just change the subject, change the meaning of our words? This has been a recurrent challenge, and my answer is that in elementary logic a change of theory is a change of meaning. Repudiation of the law of excluded middle would be a change of meaning, and no less a change of theory for that. 如果逻辑真理是分析真理，即根据逻辑词汇的意义而为真的真理，那么逻辑真理的修改，比 如我们曾经设想的对排中律的修改，是怎么回事呢？这种修改是对理论的改变，还是对主题 或语词意义的改变？这个挑战已经屡次出现，而我的回答是：在基本逻辑中，理论的改变就 是意义的改变。对排中律的抛弃，是对词的意义的改变，但这种改变也正好就是理论的改变。

You could have a change of meaning in elementary logic without change of theory. That would be if the word ‘and’ were merely put to use in place of ‘or’ and vice versa: no change of theory. But in abandoning the law of excluded middle we would not be preserving the law in any notation. 你也可以在基本逻辑中只改变意义而不改变理论。如果你仅仅把“而且”和“或者”的用法 对调一下，就会发生这样的事；这里并没有理论的改变。但在抛弃排中律的时候，这条规律 在任何记号系统中都无法得到保留。 For the past five minutes I have been expressing a generous attitude toward analyticity that may seem out of character. In fact my reservations over analyticity are the same as ever, and they concern the tracing of any demarcation, even a vague and approximate one, across the domain of sentences in general. The crude criterion in Roots of Reference, based on word learning, is no help; we don’t in general know how we learned a word, nor what truths were learned in the process. Nor do we have any reason to expect uniformity in this regard from speaker to speaker, and there is no reason to care. Elementary logic and the bachelor example are clear enough cases, but there is no going on from there. 我在前面五分钟里对分析性所表示的慷慨态度，似乎与我平素的态度是不相称的。事实上， 我对分析性的保留态度一如既往，我关注的是能否找出适用于所有语句的划界标准，哪怕是 模糊和近似的标准。《指称之根》中的建立在语言学习基础上的粗略标准是没有什么用的； 一般地说，我们既不知道一个词是如何学会的，也不知道哪些真理是在学习过程中知道的。 我们也没有理由期望语言学习的过程在语言共同体内具有主体间性，甚至也没有理由在意这 种主体间性。基本逻辑和单身汉的例子诚然是足够清楚的标本，但从此出发却是无以为继的。 It is the wrong kind of question. When in relativity theory momentum is found to be not quite proportional to velocity, despite its original definition as mass times velocity, there is no flurry over redefinition or contradiction in terms, and I don’t think there should be. The definition served its purpose in introducing a word for subsequent use, and the word was thereafter ours to use in the evolving theory, with no lingering commitments. Definition is episodic. Mostly in natural science we are not even favored with definitions, much less bound by them. New terms are just introduced by partial descriptions: electrons, neutrinos, quarks. 这是一个误入歧途的问题。尽管动量最初被定义为质量和速度的乘积，但当人们在相对论中 发现动量和速度并不完全成比例时，并没有惊呼：为了避免自相矛盾，动量这个概念需要重 新定义，而且我认为也不应该这样大呼小叫。定义的目的是引进一个供日后使用的词，从此 以后这个词就成了我们的词汇家族的一员，被使用于我们不断演化的理论，并不存在关于它 的用法的永久的承诺。定义是一种偶发的历史事件。在自然科学中，人们在大多数情况下并 不青睐定义，更不用说受它约束了。新的术语仅仅通过局部的描述而被引入：电子、中微子、 夸克。 In short, I recognize the notion of analyticity in its obvious and useful but epistemologically insignificant applications. The needs that Carnap felt for the notion in connection with mathematical truth are better met through holism. Beyond its manifest cases I find analyticity less

help than hindrance. It begets an uncritical notion of meaning, or synonymy, that can induce a false sense of understanding. For it is clear that analyticity and synonymy are interdefinable. 总之，我承认分析性概念可以运用于一些明显的实例，也承认这种运用是有益的，但我同时 认为这个概念在认识论上是没有什么价值的。卡尔纳普要求分析性概念能帮助说明数学真理， 但这一要求可以由整体主义得到更好的满足。除了明显的情况外，我认为分析性非但不会带 来帮助，而只会起反作用。它产生的无批判的意义概念或同义性概念，造成了理解的假象。 因为很显然，分析性和同义性是可以相互定义的。 Well, domestically interdefinable. Two expressions are synonymous iff their biconditional or equation is analytic. But I should mention that synonymy and therefore meaning are even worse off than analyticity when we transcend a single language. For if the two expressions to be equated belong to different languages, their biconditional or equation is far from analytic; it is incoherent, belonging to no language. 确实，在同一语言内，它们是相互可定义的。两个表达式是同义的，当且仅当由它们作为成 分的双向条件句或等同句是分析的。但我必须指出，一旦越出同一语言，同义性以及意义的 概念，甚至比分析性的概念更糟。因为如果被视为同义的两个表达式属于不同的语言，那么 由它们形成的双向条件句或等同句就远不是分析的，而是不属于任何语言的废话。 I mentioned my letter of 1943, where I took issue with Carnap both on analyticity and on ontological commitment to abstract objects, the two issues being linked. This same pairing of the two issues recurs in the antepenultimate paragraph of “Two Dogmas.” Carnap’s separation of questions of existence into questions of fact and questions of framework was a separation of the synthetic and the analytic. Collapse this epistemological duality and you collapse the ontological duality. Sticks, stones, sets, and numbers all become, for me, denizens of the world on an equal footing. Values of variables. 前面我提到我 1943 年致卡尔纳普的信，在信中我对卡尔纳普的两个互相联系的观点，即对 分析性的观点和对抽象对象的本体论承诺的观点，表示了反对。这同一对问题再次出现在“两 个教条”的倒数第三段。卡尔纳普在存在问题上的二分法（事实问题和框架问题），就是综 合和分析的二分法。一旦摧毁了认识论上的二分法，也就同时摧毁了本体论上的二分法。对 我来说，棍棒、石块、集合和数都是世界上的平等成员，都是变项的值。 So also for the contrast noted in the remaining two paragraphs of “Two Dogmas”: the contrast supposed by Carnap and C. I. Lewis between the factual and the pragmatic. “In repudiating such a boundary,” I wrote, “I espouse a more thorough pragmatism.” This passage had unforeseen consequences. I suspect it is responsible for my being widely classified as a pragmatist. I don’t object, except that I am not clear on what it takes to qualify as a pragmatist. I was merely taking the word from Carnap and handing it back: in whatever sense the framework for science is pragmatic, so is the rest of science. 对“两个教条”最后两段中谈到的区别，即卡尔纳普和 C•I•刘易斯所认为的在事实问题和 实用主义问题之间的区别，也应作如是观。我写道：“我否定这样一条界线，而拥护一种更

加彻底的实用主义。”这段话产生了未曾料到的后果。我猜测，这是我被普遍地归入实用主 义者的原因。我不反对，但是我并不清楚要成为一个实用主义者需要什么资格。其实，我只 是从卡尔纳普那里拿来了“实用主义”这个词，并回赠过去：科学的框架在什么意义上是实 用主义的，科学的其他部分也在同样的意义上是实用主义的。 Let us now stand off, look at the two dogmas, and consider what the point is in repudiating each of them. Repudiation of the first dogma, analyticity, is insistence on empirical criteria for semantic concepts: for synonymy, meaning. Language is learned and taught by observing and correcting verbal behavior in observable circumstances. There is nothing in linguistic meaning that is not thus determined. John Dewey made this point long ago. What I did, not in “Two Dogmas” but in Word and Object, was to press that point for its negative implications regarding the notion of meaning. 现在让我们站远些看看这两个教条，考虑一下对它们的否定分别意味着什么。对第一个教条 即分析性的否定，旨在坚持语义学概念（同义性、意义）的经验标准。语言的教和学，都是 通过观察和校正可观察场合下的言语行为而实现的。语言意义中没有任何东西不是这样形成 的。这一点约翰•杜威早就指出过。我在《语词和对象》一书（不是“两个教条”一文）中 所做的，就是挖掘这一观点对意义概念的消极后果。 As for the second dogma, reductionism, its repudiation is holism, seconding Duhem. The reductionism that is properly at stake here is moderate reductionism. Extreme reductionism, the notion that every scientific sentence should have a full translation in sense-datum language, is by now a straw man. Moderate reductionism was the lingering dogma, the notion that each scientific sentence has its own separate empirical content. Moderate holism is its denial. It says that a scientific sentence cannot in general be expected to imply empirical consequences by itself. A bigger cluster is usually needed. 至于第二个教条，即还原论，对它的否定导致了对迪昂整体论的支持。这里成问题的还原论 恰当地说是温和还原论。极端还原论的主张，即每一个科学语句都可以完全用感觉材料语言 来翻译，此时已经成为一个稻草人了。残存下来的是温和还原论的教条：每一个科学语句都 有其单独的经验内容。温和整体论是对它的否定。根据温和整体论，科学语句一般说来并不 能被指望独自蕴含经验后果。为了能够蕴含经验后果，常常需要更大的语句群。 “Two Dogmas” is occasionally quoted for my depiction of [t]he totality of our so-called knowledge or beliefs… [as] a man-made fabric which impinges on experience only along its edges. Maybe this sparked Joe Ullian’s title for the little manual that he and I coauthored twenty years later: The Web of Belief. Clearly my metaphor needed unpacking, and that was largely my concern in the ten years between “Two Dogmas” and Word and Object. “两个教条”有一段间或被引用的我的描述性文字： 我们的所谓知识或信念的总体……就像一个人工织造物，与经验发生密切接触的只是其 边缘地带。

20 年后，乔•尤利安为我和他合著的小册子取名《信念网》，也许就是受了这段文字的启发。 毫无疑问，我的这个比喻需要进一步澄清，从“两个教条”到《语词和对象》的 10 年间， 我主要就是做这件事情。 Becoming more consciously and explicitly naturalistic, I stiffened up my flabby reference to ‘experience’ by turning to our physical interface with the external world: the physical impacts of rays and molecules upon our sensory surfaces. 随着我的自然主义越来越自觉和越来越明确，我不再使用不够严谨的“经验”这一措辞，而 把它直接理解为我们与外部世界的物理接口：光线和分子对我们感觉器官的物理作用。 For purposes of a systematic account, I needed to marshal these scattered impacts as unit aggregates of some sort. Exteroceptors offered a neat solution. Impacts of rays and particles are irrelevant except as they trigger receptors, and happily it is only a question of triggering, with no question of more or less. So I identified one’s input from the external world, on any given occasion, with one’s global neural input on that occasion: hence with the temporally ordered set of all one’s triggered exteroceptors during that brief moment. 为系统说明计，我需要把这些零散的作用化归为某种单元的聚合体，外感受器提供了利索的 解决方案。光线和粒子的作用的共同点只在于对感受器的触发，而在其他方面则是不相干的。 但幸运的是，我需要的恰恰就是这个共同点，不多也不少。于是，我把一个人在某个场合下 从外部世界所获得的输入等同于这个人在那个场合的全局神经输入，即：在那短暂的时间里 那个人的被先后触发起来的所有外感受器的有序集。 Intricate brain processes, which neurologists are illuminating bit by bit, intervene between this neural input and perception. Exciting though that physiological research is，I was able to bypass it for my purposes by leaping to the resulting relation of perceptual similarity between global neural inputs. Inputs that are grossly dissimilar intrinsically—that is, in respect of what receptors are triggered—can issue in similar perceptual effects after the brain’s swift and unconscious work of selecting and correlating. Perceptual similarity comes in degrees, and admits of a coarse behavioral criterion in the reinforcement and extinction of responses. 从神经输入到感知之间是错综复杂的脑过程，神经学家正在逐步搞清楚这个过程。尽管这一 生理学的研究是令人兴奋的，但我根据我的目的却可以绕过它，而直接进到作为脑过程结果 的在各个全局神经输入之间的知觉相似性。全局神经输入间本质上是极不相似的——也就是 说，就被触发的是何种感受器而言，神经输入在内在性质上是极为不同的，但是，在经过大 脑快速而无意识的的选择和关联之后，它们却可以产生相似的知觉效果。知觉相似性有程度 之分，而且能够通过强化-消退反应中的简单的行为标准加以确定。 The grouping of global neural inputs by this relation of perceptual similarity affords a sufficient theoretical basis, I have argued, for distinguishing the sensory modalities and the various sensory qualities and Gestalten. A striking parallel emerges with Carnap’s Logischer Aufbau der Welt, in his derivations from similarities of global experiences.

我主张，根据这种知觉相似性关系对全局神经输入的分类，提供了区分感觉模式、各种感觉 性质和格式塔的充分的理论基础。这种做法与卡尔纳普在《世界的逻辑结构》中的做法有着 惊人的相似之处，只不过我的全局神经输入间的知觉相似性在卡尔纳普那里称为全局经验间 的相似性。 I should say that in this sketch I am somewhat updating Word and Object in the light of later writings. 我应该指出，在上述对知觉相似性的梗概性描述中，我已根据我后来的作品对《语词和对象》 的相关内容进行了更新。 The sentences near the periphery of the fabric—in the metaphor of “Two Dogmas”—came in Word and Object to be called observation sentences. Their connection with experience was explained as association with ranges of perceptually similar neural inputs. Primitively the association was by conditioning. Further associations are forged in the fullness of time by retroaction from one’s developing theory of the world. “两个教条”比喻中的靠近织造物边缘的语句，在《语词和对象》中被称为观察句。它们与 经验的联系则被解释为与具有知觉相似性的神经输入域的联系。这种联系最初是通过条件作 用实现的。随着时间的推移，人们提出的关于世界的理论也反过来参与了这种联系的锻造。 The connection into the fabric, between observation sentences and theoretical sentences, are forged by shared vocabulary. For the observation sentences are couched, not in sense-datum language, but directly in the thing language, as Carnap called it. Thing words are among the words earliest learned, though learned at first only as one-word observation sentences or as segments of monolithic observation sentences. They are there awaiting gradual reification of their designata and incorporation into theoretical contexts. 至于边缘与织造物内部的联系，即观察句和理论句之间的联系，则是由它们的共有词汇所锻 造的。因为观察句并不是用感觉材料语言表达的，而是直接用卡尔纳普所谓的事物语言表达 的。表示事物的词是最初学会的一批词，尽管一开始只是作为独词观察句或作为观察句整体 的一个片断而学会的。随着它们逐步获得了实体所指，它们也同时进入到理论语境。 Thanks to this sharing of vocabulary by observation sentences and theoretical sentences, logical relations hold between them. The pertinent logical relation, we saw, is logical implication of observation categoricals by clusters of theoretical sentences. This is, of course, the merest caricature of scientific method. It invites analysis of actual chunks of serious science, to see just what some of the typical chunks with critical semantic mass might be, and how the chains of logical implication from them to observation categoricals might run. 正是靠着共享的词汇，观察句和理论句之间才能发生逻辑的联系。最中肯的逻辑联系，我们 已经看到，表现为一群理论句对观察直言句的逻辑蕴含。这当然只是对科学方法的漫画式说 明。在这一框架中，我们仍需对严肃科学中的实际语句群作出进一步分析，看一看具有临界 语义量的典型语句群是什么样子的，这样的语句群又是怎样通过逻辑蕴含的链条通达于观察

直言句的。 All this still leaves the heuristics of hypothesis untouched: that is, the technology of framing hypotheses worth testing. This is the domain of the maxim of minimum mutilation, and of Occam’s razor. It is the domain also of standard deviation, probable error, and whatever else goes into sophisticated statistical method. 即使做到了这些，也还有一个问题没有触及到：假说发现法，即构造值得检验的假说的技术。 最小损毁原则和奥康姆剃刀在这里可以派上用场。同样地，标准偏差法、概率误差法和其他 复杂的统计方法也可以派上用场。 Deeper insights into the nature of scientific inference and explanation may some day be gained in neurology, coupled perhaps with computer simulation, as hinted by the new developments in so-called connectionist models; I think of Paul Churchland. The nature and nurture of science remains a fertile field of inquiry, varied and inviting. 也许有一天，神经学加上电脑模拟，会对科学推理和解释的本性作出更深刻的说明，就像所 谓的联结主义模型的新发展所提示的那样。我想起了保罗•丘奇兰。科学的来龙去脉，仍是 一片有待探索的肥沃土地，五光十色而魅力无穷。 （From Canadian Journal of Philosophy, Volume 21, Number 3, September 1991, pp. 265-274） （原载《加拿大哲学杂志》，第 21 卷第 3 期（1991 年 9 月） ，pp. 265-274）