CAUSAL EXPLANATION PROVIDES KNOWLEDGE WHY
因果解释提供了为什么

Events have causes. We often try to explain events, and we often succeed. The causal relation is a relation in the world which either holds or fails to hold independently of how its relata are described: the relation is extensional, and its relata are normally taken to be events. The explanatory relation is, however, intensional. This means that we cannot replace a term with co-referring or coextensional terms within an explanatory context without risking that we change the truth-value of the whole. I shall simply say that "explains" is an intensional relation, and I do that without thinking of this as an ontological commitment, or as something that anything really hangs on.
事件是有原因的。我们常常试图解释事件，而且常常成功。因果关系是世界中的一种关系，它的成立与否与如何描述其关系式无关：这种关系是外延的，其关系式通常被视为事件。然而，解释关系是内向的。这就意味着，我们不能在解释性语境中用共指或共延伸的术语来替换某个术语，否则就有可能改变整个语境的真值。 我只想说，"解释 "是一种内向关系，我这样做并没有把它当作一种本体论的承诺，也没有把它当作任何事物真正依赖的东西。

It might be helpful to concentrate on the limitations on substitutivity, and forget all talk about intensional entities. We must, of course, come to terms with the insight that what we explain is events in the real world. Events are concrete. The intensionality of "explains" brings with it no mystery, however; an explanation of an event explains that event "under a description". This latter locution is technical, as talk about propositions is. The point is that a set of sentences explains an event only as long as the event to be explained is described in such a way that the relevance of what we say is plain. Only then do we provide an explanation; an explanation must, necessarily, provide understanding. Relevance is relevance for this (causal) understanding. This point about explanation is what underlies the intensionality of explanations, and what limits substitutivity.
把注意力集中在对可替代性的限制上，而忘掉所有关于内维实体的讨论，可能会有所帮助。当然，我们必须认识到，我们所解释的是现实世界中的事件。事件是具体的。然而，"解释 "的内向性并不神秘；对事件的解释是 "在描述之下 "对事件的解释。后一种说法是技术性的，正如关于命题的讨论一样。关键在于，只有当要解释的事件被描述成我们所说的内容的相关性是显而易见的，一组句子才能解释一个事件。只有这样，我们才提供了解释；解释必然提供理解。相关性就是对这种（因果）理解的相关性。关于解释的这一点是解释的内向性的基础，也是限制替代性的原因。

The sentence or statement describing the event to be explained must express a fact, or a true proposition. If it does not, there is nothing to explain. The fact to be explained in the causal case is typically that an event of a certain kind has occurred.
描述要解释的事件的句子或陈述必须表达一个事实或一个真命题。否则，就没有什么可解释的了。在因果关系中，需要解释的事实通常是某类事件已经发生。

When we explain why a particular event of a certain type has occurred (a fact), we can for short say that we explain that event. It is thus quite easy to maintain that we explain events even if the explanatory relation itself is an intensional relation between statements or propositions.
当我们解释某类特定事件（事实）发生的原因时，我们可以简短地说，我们解释了该事件。因此，即使解释关系本身是陈述或命题之间的内涵关系，我们也很容易坚持认为我们解释了事件。

I now want to push aside considerations as those above. I will simply speak of the explanation of events. Let us move on. Events have causes, and causes themselves have other causes. Typically, when we give a causal explanation, we relate cause (or causes) and effect somehow. But we do not just relate them (extensionally understood). That would not necessarily provide an explanation at all. We must relate them in particular ways to provide an explanation. I shall maintain that these ways are identified thus: There is no explanation of why an event occurred unless we provide an understanding why that event occurred. This seems to me to capture the point of explanation.
我现在要撇开上述考虑。我只想谈谈对事件的解释。让我们继续前进。事件有原因，而原因本身又有其他原因。通常，当我们给出因果解释时，我们会以某种方式将原因（或原因）与结果联系起来。但我们并不只是把它们联系起来（从外延上理解）。那样不一定能做出解释。我们必须以特定的方式把它们联系起来，才能提供解释。我坚持认为，这些方式是这样确定的：除非我们能够理解事件发生的原因，否则就无法解释事件发生的原因。在我看来，这似乎抓住了解释的要点。

There have been very many suggestions about how we must relate cause and effects to provide a causal explanation. Many have seen the requirement that explanation provide understanding as insubstantial and vague. Similar thoughts would arise from an illumination of understanding from the concept of knowledge. On the whole it has been seen as entirely un-illuminating to use or employ the concept of knowledge in accounting for explanation, knowledge was at best the final aim, never to be used from the beginning in such analyses. This paper aims to turn that around in the theory of explanation.
关于我们必须如何将因果关系联系起来以提供因果解释，已经有很多建议。许多人认为，要求解释提供理解是不切实际和含糊不清的。从 "知识 "的概念中获得理解，也会产生类似的想法。总体而言，在解释说明中使用或运用知识概念被认为完全没有启发性，知识充其量只是最终目的，在此类分析中从一开始就不能使用。本文旨在扭转解释理论中的这一局面。

The most famous account of explanation is Hempel's D-N account. (NB: Hempel always thought that explanations could be probabilistic as well.) In its classical form Hempel's view is that an explanation of a particular event consists in a correct deductive-nomological argument. There are law-premises and particular fact premises. The conclusion says that the event we want explained took place, and the argument is valid. All premises are true. On this view the structure of an explanation is the starting point. The intensionality of explanation is accounted for by the need to deduce the occurrence of an event from a law. The needs of deduction from laws determine the intensionality of explanation, we might say.
最著名的解释论是亨普尔的 D-N 论。(注：亨普尔一直认为解释也可以是概率性的）亨普尔的经典观点是，对一个特定事件的解释包含一个正确的演绎-名词论证。有法律前提和特定事实前提。结论说我们想要解释的事件发生了，论证是有效的。所有前提都是真实的。根据这种观点，解释的结构是起点。解释的意图性是通过从规律推导事件发生的需要来解释的。我们可以说，从规律推导的需要决定了解释的内在性。

Hempel's account is not insubstantial or vague. But there are many well-known problems for this classical approach. For example, it seems to let in too many arguments as explanations. The approach therefore needs to impose further constraints. An easy way out here might be to say that only a D-N argument that presents a (set of) cause(s) is an explanation. But letting in too many is not the only problem: the approach also seems to let in too few arguments as explanations. Most causal explanations in real life fall much short of the ideal of deductive arguments with law-premises. Furthermore, the ideal might indeed be pointless, because our interest might fall on some particular part of the causal history of an event.
亨普尔的论述并非毫无根据或含糊不清。但是，这种经典方法存在许多众所周知的问题。例如，它似乎让太多的论据成为解释。因此，这种方法需要施加进一步的限制。一个简单的办法是，只有提出（一组）原因的 D-N 论证才是解释。但是，让太多的论证成为解释并不是唯一的问题：这种方法似乎也让太少的论证成为解释。现实生活中的大多数因果解释都远远达不到具有法律前提的演绎论证的理想。此外，这一理想可能确实毫无意义，因为我们的兴趣可能会落在事件因果历史的某些特定部分上。

David Lewis has presented an exceedingly simple view of causal explanation, and had related and defended his view against many views in the literature. His thesis is this: To explain an event is to provide some information about its causal history. Lewis does not, of course, think of this as giving necessary and sufficient conditions for something to count as an explanation. Lewis does think of his thesis as equivalent to the thesis above about understanding: To understand why something happened is to possess some information about its causal history.
戴维-刘易斯（David Lewis）提出了一个极为简单的因果解释观点，并针对文献中的许多观点为自己的观点进行了论证和辩护。他的论点是这样的：解释一个事件就是提供一些关于其因果历史的信息。当然，刘易斯并不认为这给出了某些东西算作解释的必要条件和充分条件。刘易斯确实认为他的论点等同于上文关于理解的论点：要理解某件事情发生的原因，就必须掌握关于其因果历史的某些信息。

This paper can be seen as a critical reflection on Lewis. I agree with Lewis that we should not aim for a reductive account of causal explanation. I shall nevertheless dispute his thesis. The thesis is so weak that it can hardly be false, someone might think. I shall argue that it is much too weak. We ought to say something stronger in order to provide something fruitful about what explanation is all about. I shall in the first part consider two strengthenings of Lewis's thesis. In the second part I shall develop my account a bit further, and finally I shall bring this account of explanation to bear on the issue of inference to the best explanation.
本文可视为对刘易斯的批判性反思。我同意刘易斯的观点，即我们不应以还原因果解释为目标。不过，我还是要对他的论点提出质疑。有人可能会认为，这个论点太薄弱了，几乎不可能是错误的。我将争辩说，它太弱了。我们应该说一些更有力的东西，以便为解释的意义提供一些富有成效的东西。在第一部分中，我将考虑对刘易斯论点的两种加强。在第二部分中，我将进一步发展我的论述，最后我将把这一解释论述用于推论最佳解释的问题上。

The two strengthenings of Lewis are these:
刘易斯的两个优势是

It is easy to see that the first strengthening is a somewhat minor modification of Lewis's view. It leaves us with a view that in many ways relates to other views on causal explanation as Lewis's view does. The second strengthening gives us a view that seems to depart more from Lewis. Both suggested strengthening are rooted in the fact that I think that a thesis about explanation should give the point of explanation in such a way that some helpful guidelines are given to a potential explainer about how to go about her task. But there is also more to it than that.
不难看出，第一种强化是对刘易斯观点的微小修改。它给我们留下的观点在许多方面与其他关于因果解释的观点相关，就像刘易斯的观点一样。第二种强化则给我们提供了一种似乎更偏离刘易斯的观点。所建议的两种加强都源于这样一个事实，即我认为关于解释的论述应该以这样一种方式给出解释的要点，即为潜在的解释者如何完成其任务提供一些有益的指导。但问题还不止于此。

Another preparation for what is to come is this: Sylvain Bromberger has pointed out that an explanation might be subjected to two very different types of question. An explanation is something about which it makes sense to ask: "How long did it take", "Who gave it?" etc. These questions see an explanation as an act by someone, an intentional activity in space and time. An explanation may also be an object about which we can ask: "Does anyone know it? Who thought of it first? Is it very complicated?" Perhaps we can think of this as a process/product ambiguity. I shall think of it as the two sides of explanation: The activity of explaining, and the explanation given by the activity, which we may call the activity side and structural side of explanation. We can focus on either side when thinking about explanation, but we need to think about both for a general theory of (causal) explanation. One important issue is precisely how to relate them.
另一个准备工作是这样的：西尔万-布罗姆伯格（Sylvain Bromberger）曾指出，一种解释可能会受到两种截然不同的问题的追问。一种解释是有意义的提问："花了多长时间"、"是谁给出的 "等等。这些问题将解释视为某人的行为，是在空间和时间上的有意活动。一个解释也可以是一个对象，我们可以问："有人知道它吗？谁先想到的？它很复杂吗？也许我们可以把这看作是过程/产品的模糊性。我认为这是解释的两个方面：解释的活动和活动给出的解释，我们可以称之为解释的活动方面和结构方面。在思考解释问题时，我们可以侧重其中任何一个方面，但对于（因果）解释的一般理论而言，我们需要同时考虑这两个方面。一个重要的问题是如何准确地将它们联系起来。

Bromberger's point is fully compatible with seeing explanation as providing understanding, since providing understanding is also both a concrete activity and the provision of an abstract object. Another way of putting the point is this: An explanation may be something concrete (an activity) and it may be something abstract. If it is the concrete act of explaining, we may ask how long it took, whether it was interrupted etc. We are then thinking of an act of explaining something. We may also think of it as an abstract entity, as something about which it can make sense to ask whether anyone knows it, whether it is complicated, what is its structure etc.
布罗姆伯格的观点与把解释视为提供理解是完全一致的，因为提供理解既是一种具体的活动，也是提供一种抽象的对象。另一种说法是这样的：解释可以是具体的（活动），也可以是抽象的。如果是具体的解释行为，我们可以问用了多长时间，是否被打断等等。这样，我们就想到了解释行为。我们也可以把它看作是一个抽象的实体，可以问是否有人知道它、它是否复杂、它的结构是什么等等。

Guide for the reader: The objections to my views are numerous. I will answer some major objections to my views about the activity side after I have argued in support of the two modifications I am going to suggest. I ask the readers for patience until that point.
读者指南：对我的观点的反对意见很多。我将在论证支持我提出的两项修改建议之后，回答对我关于活动方面的观点的一些主要反对意见。在此之前，请读者耐心等待。

Let us first look at the act of explaining something, in this case the act of explaining a particular event. In such an act, someone who possesses some knowledge or information about the cause or causes of this event tries to convey this knowledge or information to someone else. That is done with the purpose of providing understanding of why it happened: That other persons asks for an explanation, asks why something occurred, and that is a request for understanding why it happened. The serious explainer must think of herself as possessing an explanation, i.e. information that provides understanding, and must think of her reply as making that explanation available to a person who does not possess the explanation.
让我们先来看看解释行为，这里指的是解释某一特定事件的行为。在这一行为中，拥有关于该事件起因的某些知识或信息的人试图将这些知识或信息传达给其他人。这样做的目的是让别人了解事件发生的原因：其他人要求解释，询问为什么会发生这样的事情，这就是要求了解事情发生的原因。认真的解释者必须认为自己拥有解释，即提供理解的信息，并且必须认为自己的答复是向不拥有解释的人提供解释。

In giving the reply the explainer makes an assertion. We might think that this fact in itself constrains our account of explanation. In particular, if we think that the constitutive rule of assertion directly links knowledge and assertion, and says that you should only assert what you know, then the fact that we assert something when giving an explanation, might in itself have consequences for a thesis about explanation. But I shall leave that issue aside, at least for the moment, and concentrate on the act of explaining and what this act is supposed to provide (namely understanding).
解释者在回答时作出了断言。我们可能会认为，这一事实本身就限制了我们对解释的解释。特别是，如果我们认为断言的构成规则将知识与断言直接联系起来，并说你只应断言你所知道的东西，那么我们在作出解释时断言某些东西这一事实本身就可能对关于解释的论题产生影响。 不过，至少在目前，我暂且不谈这个问题，我将专注于解释行为以及这一行为应该提供的东西（即理解）。

Our attempt at explaining might fail. When considering what to say at all in response to a request for an explanation, we must consider whether we have something we can say which is not an outright failure. What we say when we try to give an explanation must be something we at least believe to be an explanation, something we believe provides understanding why the event to explain happened. Again: This paper is not in the business of giving anything like necessary and sufficient conditions for explanation. Nevertheless we must consider possible constraints on a reasonable thesis about explanation. A possible constraint is this: a thesis about explanation must, somehow, throw light on what it is to explain, and thereby, at least in a minimal sense, illuminate for the person who tries to explain how she should go about the task. If a thesis about explanation clearly allows as a possible explanation any number of statements that the explainer knows not to be explanations, then the thesis about explanation does not illuminate explanation.
我们试图解释 的努力可能会失败。在考虑对解释请求说什么时，我们必须考虑我们是否可以说一些不会彻底失败的话。当我们试图做出解释时，我们所说的必须是我们至少认为是一种解释的东西，是我们认为能够让人理解为什么要解释的事件会发生的东西。再说一遍：本文并不是要给出解释的必要条件和充分条件。不过，我们必须考虑关于解释的合理论题可能受到的限制。一个可能的制约因素是：关于解释的论题必须以某种方式阐明解释是什么，从而，至少在最低限度的意义上，为试图解释的人指明她应该如何去完成任务。如果一个关于解释的论题明确允许任何数量的解释者知道不是解释的陈述作为一种可能的解释，那么这个关于解释的论题就没有阐明解释。

Lewis's thesis is that to explain is to provide some information about the causal history. Lewis is clear that he does not use "information" in such a way as to imply truth. Information can be misinformation. This creates a problem. It seems entirely unreasonable to take Lewis to allow that an explainer can knowingly pass on misinformation and still think of what she is doing as explaining. Still that seems to be implied by his stated thesis about explanation in conjunction with his use of the term "information". Let us rule out that possibility, both as an interpretation of Lewis and as a possible position on explanation. Minimally, then, the explainer must believe that the information she gives is correct. I take that to imply that a serious explainer would wholly withdraw an explanation in which an untruth played an essential role. If an untruth played a part but not an essential part, then the explanation might be modified, but not wholly withdrawn. The notion of "essential role" I am appealing to is Lewis's, and is also itself in need of illumination. I shall maintain that "essential" here must be understood as essential for (causal) understanding.
刘易斯的论点是，解释就是提供一些关于因果历史的信息。刘易斯很清楚，他使用 "信息 "的方式并不意味着真理。信息可能是错误的信息。这就产生了一个问题。如果认为刘易斯允许解释者在明知信息有误的情况下仍然认为自己所做的是解释，这似乎是完全不合理的。不过，他对解释的论述以及他对 "信息 "一词的使用似乎也暗示了这一点。让我们排除这种可能性，既排除对刘易斯的解释，也排除对解释的可能立场。那么，最低限度，解释者必须相信她所提供的信息是正确的。我认为，这意味着一个严肃的解释者会完全撤回一个不真实的解释，而在这个解释中，不真实起了至关重要的作用。如果不实之词在其中起了一定的作用，但不是关键作用，那么解释者可能会修改解释，但不会完全撤销解释。我所说的 "基本作用 "是刘易斯的概念，它本身也需要澄清。我将坚持认为，这里的 "基本 "必须理解为对（因果）理解的基本。

Let us look at a Lewis-type example. There has been a crash. There is the icy road, the drunk driver, the bald tyre, and much more. (There is also the bald driver, the red car, the fact that it happened on a day with an " " in it and what not. I shall come back to these facts.) In a case like this there might be no easy way of identifying the cause, there might be, to put it loosely, many causes that work together. Competing explanations might mention different causes, and this raises many interesting issues. But so far the simple question before us is whether, in case we give an explanation by providing some information about the causal history, we should a) believe, b) believe with good reason, or, perhaps, c) know that the information we provide is correct.
让我们来看一个刘易斯式的例子。发生了车祸。有结冰的路面、醉酒的司机、光头胎等等。(此外，还有光头司机、红色汽车、车祸发生在有 " "的日子等等。我会再谈这些事实）。在这种情况下，可能没有简单的方法来确定原因，说得宽泛点，可能有许多原因共同起作用。相互竞争的解释可能会提到不同的原因，这就产生了许多有趣的问题。但到目前为止，摆在我们面前的一个简单问题是，如果我们通过提供一些关于因果历史的信息来做出解释，我们是否应该（a）相信，（b）有充分理由相信，或者（c）知道我们提供的信息是正确的。

I shall take a) as obviously true, as argued above. The question is whether we need something stronger than a). In the normal case, it seems clearly true that we need at least b), when we concentrate on the activity-side. This is because we need reasons for our explanatory belief when we utter it. In order to give something as an explanation, either that the driver was drunk, or that the tyre was bald or both, we would normally require that we have reason to believe that the driver was drunk or that the tyre was bald. We would normally feel obliged to check one way or other whether these things were the case before we offered such an explanation to someone. It seems definitely to be true that we should have some reason to believe that the information we offer in an explanation is correct.
如上所述，我认为 a) 显然是正确的。问题是我们是否需要比 a) 更强的东西。在正常情况下，当我们把注意力集中在活动方面时，我们至少需要 b），这显然是对的。这是因为我们在说出解释性信念时需要理由。为了作出某种解释，或者说司机喝醉了，或者说轮胎瘪了，或者说两者都是，我们通常需要有理由相信司机喝醉了或者轮胎瘪了。我们通常认为，在向某人提供这样的解释之前，我们有义务核实一下情况是否属实。我们应该有理由相信我们在解释中提供的信息是正确的，这一点似乎肯定是对的。

I conclude that the explainer must minimally believe with some reason, and perhaps with good reason, that the essential information given in an explanation must be true. The question is whether this observation should have consequences for our thesis about explanation. It seems to me that it would if we accept the constraint on such theses above, namely that a thesis about explanation must illuminate (in a minimal sense) for an explainer how to go about the task. Since untruths should not be told, this must be reflected in the thesis. The thesis might then be: to explain an event is to give some information one has reason to believe is correct about its causal history.
我的结论是，解释者必须在最低限度上有理由相信，也许是有充分的理由相信，解释中提供的基本信息必须是真实的。问题是，这种看法是否会对我们的解释论产生影响。在我看来，如果我们接受上述对此类论题的限制，即关于解释的论题必须（在最低限度的意义上）为解释者指明如何去完成任务，那么它就会产生影响。由于不应该说不真实的话，这一点必须反映在论题中。因此，论题可以是：解释一个事件就是提供一些人们有理由相信是正确的关于其因果历史的信息。

Now, in case the information is not correct, is the information given explanatory? Has the person requesting an information been given an explanation - has understanding been provided? The answer seems to be clearly "no". If your potential understanding of why something happened is based on false belief, then you do not understand why it happened. It is not enough for the provision of understanding why something happened that one has generated beliefs about causal factors.
现在，如果信息不正确，所提供的信息是否具有解释性？索取信息的人是否得到了解释--是否得到了理解？答案显然是 "否"。如果你对事情发生原因的潜在理解是基于错误的信念，那么你就不理解事情发生的原因。要理解事情发生的原因，仅凭一个人对因果关系的信念是不够的。

Understanding must really be based on correct beliefs. One has not provided understanding to someone else unless that someone's understanding of the causal process is based on correct information or belief. Explaining, and from that providing an explanation, is therefore factive, and the factivity of explanation is derivative on the factivity of understanding.
理解必须真正基于正确的信念。除非别人对因果过程的理解是基于正确的信息或信念，否则他就没有向别人提供理解。因此，解释，并由此提供解释，是事实性的，而解释的事实性是理解的事实性的衍生。

Let us move on to the issue of whether we should require something stronger than b), namely that the correct information we give should amount to knowledge on our part. If we knew that drivers in these parts very often had bald tyres, we would have a reason to believe that this particular driver was driving around with bald tyres. Still, it seems to me, that knowledge would not entitle us to simply give as an explanation of the accident in question that the tyres were bald. Even if we knew that drivers in this part of the country had bald tyres much more often than not, it would not seem right to just put forward the explanation that this accident was caused by a bald tyre. We might, in the imagined case, be entitled to saying that the tyre might have been bald, or even that it was probably bald, but that would be as far as we could rightly go. We would not be entitled to simply explaining the accident by saying that the tyre was bald.
让我们继续讨论我们是否应要求比 b)更强的条件，即我们所提供的正确信息应相当于我们的知识。如果我们知道这些地方的司机经常使用光头胎，我们就有理由相信这位司机开车时使用光头胎。不过，在我看来，这种了解并不能使我们有权简单地以轮胎是光头来解释这起事故。即使我们知道在这个国家的这个地区，司机使用光头胎的情况比不使用光头胎的情况要多得多，但仅仅提出这起事故是由光头胎造成的这一解释似乎也是不对的。在设想的情况下，我们也许有权说轮胎可能是光头的，甚至可能是光头的，但这是我们所能正确解释的范围。我们无权简单地用轮胎没气来解释这起事故。

The same goes for drunk driving. Even if we knew that of the drivers in these parts were drunk at the hour of the accident, it would seem wrong to straightforwardly maintain that the driver was drunk as part of an explanation of a particular accident. It seems that you should know that the present driver was drunk in order to explain the accident by that fact. If the only thing you do know is that of the drivers in these parts were drunk at this hour, then that is what you should say.
醉酒驾驶也是如此。即使我们知道， ，这些地方的司机在事故发生时都是醉酒的，但如果直截了当地认为司机醉酒是某起事故的部分原因，似乎也是不对的。似乎你应该知道现在的司机是醉酒的，以便用这一事实来解释事故。如果你唯一知道的是， ，那么你就应该这么说。

These points about knowledge seem possibly to be derived from facts about assertion. I hinted at that possibility above. Still I claim they are also a reflection of facts about understanding: You would not understand why this accident happened if you thought it happened because the driver was drunk if it turned out that the driver was not drunk. In that case you would not understand why this accident happened. In general you do not understand why something happens as long as an untruth plays an essential role of the explanation you provide.
这些关于知识的观点似乎可能来自关于断言的事实。我在上文暗示了这种可能性。不过，我认为它们也反映了关于理解的事实：如果你认为这起事故的发生是因为司机喝醉了酒，而事实证明司机并没有喝醉，那么你就不会理解这起事故发生的原因。在这种情况下，你就不会理解为什么会发生这样的事故。一般来说，只要不真实的东西在你提供的解释中起着至关重要的作用，你就不会理解事情发生的原因。

I take it as established that understanding causal processes is factive. Understanding is what you transfer when you explain someone why something happened. When we explain something, we must point to something true, and in fact to something we also have reason to believe is true.
我认为，对因果过程的理解是事实性的。当你向别人解释某件事情发生的原因时，"理解 "就是你所传递的东西。当我们解释某件事情时，我们必须指向一些真实的东西，而且事实上指向一些我们也有理由相信是真实的东西。

The choices were: should we a) believe, b) believe with good reason, or, perhaps, c) know that the information we provide is correct. The suspicion is this now that we should opt for c): only c) seems obviously to provide understanding. b) also seems too weak; there is no account of good reason that will do the trick alone and be sufficient. Notice this as well: If you, when giving an explanation, state more than you what know, then the possibility that the person who receives the explanation forms false beliefs about the causal history of the event to be explained increases, and then the possibility of misunderstanding increases. To provide understanding, and to prevent misunderstanding, you should provide things you know, and only things you know.
我们的选择是：a) 相信，b) 有充分理由地相信，或者，c) 知道我们提供的信息是正确的。现在的疑虑是，我们应该选择 c）：只有 c）显然能让人理解。b）似乎也太薄弱了；没有任何一个关于良好理由的说法能单独做到这一点并足够充分。还要注意这一点：如果你在解释时说的比你知道的还多，那么接受解释的人对要解释的事件的因果历史形成错误信念的可能性就会增加，那么误解的可能性就会增加。为了让人理解并防止误解，你应该提供你知道的东西，而且只提供你知道的东西。

These considerations drive us towards this: In the normal case of a causal explanation of a particular event, the explainer considering what to say in giving an explanation should state facts about the causes, facts he knows to be true. In that case one can see oneself as providing understanding in the right way to the person who asks for understanding. Of course knowledge is fallible and all that, but that is granted. It is similarly taken for granted that you do not have to know that you know in order to know. If we accept the general constraint that a thesis about explanation must somehow illuminate for an explainer how she should go about the business of explaining, then we seem stuck with the thesis that to explain an event is to provide understanding, and to provide understanding, we should provide knowledge about its causal history.
这些因素促使我们这样做：在对某一特定事件进行因果解释的正常情况下，解释者在考虑如何进行解释时，应该陈述有关原因的事实，即他所知道的真实情况。在这种情况下，我们可以把自己看作是以正确的方式向要求理解的人提供理解。当然，知识是易错的，但这是理所当然的。同样理所当然的是，你不必为了知道而知道自己知道。如果我们接受这样的一般限制，即关于解释的论题必须以某种方式为解释者指明她应该如何去解释，那么我们似乎就只能接受这样的论题：解释一个事件就是提供理解，而要提供理解，我们就应该提供关于其因果历史的知识。

We seem to have established a first modification of Lewis; the thesis 1 above. The way we have done that is by a) imposing as a constraint upon a thesis about explanation that it should give some guideline to a potential explainer how to go about giving an explanation. That constraint is not a very strong constraint. b) We have imposed structure from the point about the aim of explanation, i.e. understanding. As a result we conclude that explaining is factive, and that the factiveness of explanation is a reflection of the factiveness of understanding. I am also much inclined to continue: The factiveness of understanding is a reflection of the factivity of causal knowledge. The latter point is however, in need of further argument.
我们似乎已经确立了对刘易斯的第一种修正；即上述论题 1。我们做到这一点的方法是：a) 对关于解释的论题施加了一个约束，即它应为潜在的解释者提供如何进行解释的一些指导。b) 我们从解释的目的，即理解这一点出发，强加了结构。因此，我们得出结论：解释是事实性的，解释的事实性是理解的事实性的反映。我也很想继续说下去：理解的事实性是因果知识的事实性的反映。 不过，后一点还需要进一步论证。

We have before us two theses, and the first, the weaker thesis, seems reasonably well established. The thesis is this: To explain an event is to provide some knowledge about its causal history. The question before us is whether we should settle for this thesis, or go for a stronger thesis, namely this:
我们面前有两个论题，第一个是较弱的论题，似乎已经得到了合理的证实。这个论点是这样的：解释一个事件就是提供一些关于其因果历史的知识。摆在我们面前的问题是，我们是应该满足于这一论点，还是选择一个更有力的论点，即："解释一个事件就是提供一些关于其因果历史的知识：

This thesis represents a more radical break with Lewis. The break above, thesis one, limits the set of possible explanations to giving known information about causal histories. Should we limit it further? It seems to me that we should. It seems possible to know facts about the cause without really understanding why it brought about the effect. One can know the cause from a causal story, and still not understand why the effect occurred. Understanding is a richer notion than that captured by the first amendment, in several ways. I will approach some of them.
这一论断代表了与刘易斯更彻底的决裂。上述突破，即论题一，将可能的解释集合限制在提供已知的因果历史信息上。我们应该进一步限制它吗？在我看来是应该的。我们似乎有可能知道关于原因的事实，却不真正理解它为什么会带来结果。我们可以从一个因果故事中知道原因，但仍然不理解为什么会产生结果。在几个方面，理解是一个比第一修正案更丰富的概念。我将从几个方面来阐述。

First we must not forget that causal relations are extensional, and permit substitutions of co-referring singular terms. Identity-conditions for events is a tricky topic, but most people admit that burning a bon-fire might be the event Bob had been looking most forward to on specific day. If this bon-fire caused the house to burn down, then the event Bob had been looking most forward to this day caused the house to burn down. The truth of the last sentence might be the only thing we know about the cause, but this knowledge does not, however, provide any understanding of why the effect took place.
首先，我们不能忘记因果关系是外延性的，允许替换共指的单数词。事件的同一性条件是一个棘手的话题，但大多数人都承认，燃烧一堆篝火可能是鲍勃在某一天最期待的事件。如果这堆篝火导致了房子被烧毁，那么鲍勃这天最期待的事件就导致了房子被烧毁。最后一句话的真实性可能是我们对原因的唯一了解，但是这种了解并不能让我们理解为什么会发生结果。

Let us return to the Lewis-type example. There has been a crash. There is the icy road, the drunk driver, the bald tyre, and much more. There are also other known facts. There is the bald driver, the red car, a day with an " " in it. These latter known facts differ from the facts above: They seem causally irrelevant, and they do not seem to help in providing causal understanding. They are nevertheless pieces of correct information about the events in the causal history of the explanandum; the event to be explained. They are known facts about the causal history. We can imagine scenarios where these apparently irrelevant facts are indeed relevant. Such scenarios are very remote possibilities in this case, though. (The best we can do is most likely to imagine a scenario where such properties indicate the presence of a relevant causal factor: There might be a strong correlation between driving patterns and baldness, so that baldness indicates a reckless driver, the road might be cleared for ice on all days without an " " in them etc.)
让我们回到刘易斯式的例子。发生了车祸。有结冰的路面、醉酒的司机、光头胎等等。还有其他已知事实。有光头司机、红色汽车、有 " "的一天。后面这些已知事实与上述事实不同：它们似乎与因果关系无关，也似乎无助于提供因果理解。然而，它们是关于被解释事件的因果历史的正确信息。它们是关于因果历史的已知事实。我们可以想象一下这些看似无关的事实确实相关的情景。不过，在这种情况下，这种可能性非常渺茫。(我们所能做的最有可能是想象这样一种情景，即这些属性表明存在相关的因果因素：驾驶模式与秃头之间可能存在很强的相关性，因此秃头表明驾驶者鲁莽，道路可能在所有没有 " "的日子里都会结冰，等等）。

What this seems to show, is that to provide knowledge about the causal history of an event in a causal explanation, is not sufficient for providing understanding. We are not in the business of giving necessary and sufficient conditions, but it still seems that we must improve matters here. What is indicated by the example, is that when we provide knowledge of the causal history of an event, we must not only provide information we know to be true about the events in causal-historical chain, we must also give further information we at least believe is relevant for understanding why the event of the explained (the crash) occurred. I do not doubt that Lewis in a way presupposes that it is that kind of story we should provide - but if it is, we should make the point explicit.
这似乎表明，在因果解释中提供关于事件因果历史的知识，并不足以提供理解。我们的工作并不是给出必要条件和充分条件，但似乎我们仍然必须改进这里的问题。这个例子表明，当我们提供一个事件的因果史知识时，我们不仅要提供我们所知道的关于因果史链条中的事件的真实信息，我们还必须提供我们至少认为与理解所解释的事件（坠机）为何发生相关的进一步信息。我并不怀疑刘易斯在某种程度上预设了我们应该提供的就是这样的故事--但如果是这样的话，我们就应该明确指出这一点。

There is a further interesting question whether we should only give information we know is relevant, or whether belief is sufficient. If we have reason to believe that the information is relevant, and the information is relevant, then we are typically in the position that we know that the information is relevant.
还有一个有趣的问题是，我们是否只应提供我们知道是相关的信息，还是相信就足够 了。如果我们有理由相信该信息是相关的，而且该信息也是相关的，那么我们通常就处于 我们知道该信息是相关的地位。

The reasons we have for believing that the information is relevant, is typically other things that we know. If that is so, it seems to me that we will do better in providing understanding if we also provide the (relevant or salient) reason we have for thinking that the information we provide is relevant.
我们认为信息相关的理由通常是我们知道的其他事情。如果是这样的话，在我看来，如果我们也能提供我们认为所提供的信息是相关的（相关的或突出的）理由，我们就能更好地提供理解。

Salience has now been mentioned, and that is indeed another factor here. When we explain an event like a crash we typically point to a specific causal factor, the factor that in fact is most salient to us. One explainer says the bald tyre was the cause of the crash, another maintains it was the drunkenness, a third points to the icy road. Very often these explainers would agree that all three factors are essential for
现在有人提到了 "突出性"，这的确是这里的另一个因素。当我们解释像车祸这样的事件时，我们通常会指出一个特定的因果因素，即事实上对我们来说最突出的因素。一位解释者说，爆胎是车祸的原因，另一位解释者坚持认为是醉酒，第三位解释者则指出路面结冰。很多时候，这些解释者都会同意，这三个因素对于
the causal history: There would not have been a crash had one of these factor been absent. If that is so, there is disagreement about salience, but not disagreement about relevance. Disagreement about salience I see as disagreement about which factor it is natural to mention in an explanation among several operative or relevant causal factors. Many of these other factors would typically be taken for granted, and not mentioned at all. They might be common knowledge. Salience works against such a background. It also works against such a background when picking the salient reason for believing that the information is relevant.
因果历史：如果其中一个因素不存在，车祸就不会发生。如果是这样的话，那么就存在着关于显著性的分歧，但不存在关于相关性的分歧。关于突出性的分歧，我认为是关于在几个起作用的或相关的因果因素中，在解释中自然应该提到哪个因素的分歧。许多其他因素通常被认为是理所当然的，根本不会被提及。它们可能是常识。突出性就是在这样的背景下发挥作用的。在选择相信信息相关的突出原因时，它也是针对这样的背景而工作的。

There might be disagreement about relevance. It might be that explainer A holds that this level of drunkenness is very likely to lead to a crash under all circumstances, it probably would have happened even of the tyre had not been bald or if the road had not been icy. That is explainer A's judgement, and if so, A should better have some reasons for this view.
在相关性问题上可能存在分歧。解释者 A 可能认为，在任何情况下，这种程度的醉酒都很可能导致撞车，即使轮胎没有瘪或路面没有结冰，也很可能会发生撞车事故。这就是解释者 A 的判断，如果是这样，解释者 A 最好有一些理由。

Explainer B might hold that this driver was not so drunk that it really mattered for the driving ability in this case, and that the icy road was so slippery that all tyres, also new tyres, would have slipped. would go for the baldness of the tyre. In a scenario like this there would be competing explanations. A, B and C would be expected to bring forward reasons and justifications for their views. In the end all of them might be wrong, all mentioned factors might have been necessary for this accident after all.
解释者 B 可能会认为，这名司机并没有醉得太厉害，因此在这种情况下驾驶能力并不重要，而且结冰的路面非常湿滑，所有轮胎，包括新轮胎，都会打滑。 。在这种情况下，会有相互矛盾的解释。人们希望 A、B 和 C 为自己的观点提出理由和依据。最后，他们可能都错了，毕竟所有提到的因素都可能是这次事故的必要因素。

If one were to look at the justifications brought forward by A, B and C when explanations are competing this way, we see that general probabilistic knowledge about the relationship between drunkenness and driving accidents would not help A all that much, he would have to back his view with specific knowledge about the particular case, and justifications for his view that the other factors were not relevant in this particular case. The same goes for the other two, B, and C, and their argumentative tasks.
如果我们看一下 A、B 和 C 在这样的解释竞争中提出的理由，我们就会发现，关于醉酒与驾驶事故之间关系的一般概率知识对 A 的帮助并不大，他必须用关于特定案例的具体知识来支持他的观点，并说明他认为其他因素与这一特定案例无关的理由。另外两人 B 和 C 的论证任务也是如此。

This indicates strongly that a reasonable belief to the effect that a certain factor is relevant, reasonable because it is the type of factor which very often plays a role, is, when challenged, not sufficient to put one in a position to promote the view that it was relevant. It seems that we need more than just reasonable belief to the effect that a certain factor was causally relevant. If challenged, we need at least reasonable beliefs about causal relevance that are grounded in observations and investigative results about the particular case. These grounded reasonable beliefs they rely on would, by , and , be subjectively indistinguishable from knowledge. In this dispute between A, B and C, they would all three think of themselves as putting forward knowledge about which factors were causally relevant. Of course they cannot all be right, but that is another matter. Only someone who is right does in fact understand why this crash happened.
这有力地表明，合理地相信某一因素是相关的，因为它是经常起作用的那类因素，但在受到质疑时，这并不足以使人能够提倡认为它是相关的这一观点。看来，我们需要的不仅仅是合理地相信某一因素是因果相关的。如果受到质疑，我们至少需要关于因果相关性的合理信念，这些信念要以对特定案例的观察和调查结果为基础。通过 和 ，他们所依赖的这些有依据的合理信念在主观上与知识是没有区别的。在甲、乙、丙三人的争论中，他们都认为自己提出了关于哪些因素是因果关系的知识。当然，他们不可能都是对的，但这是另一回事。只有正确的人才能真正理解为什么会发生这样的事故。

The issues around salience must not distract us. What we are now concentrating on is whether we should be required to know that a factor is causally relevant if we offer that factor in a causal explanation of an event. The considerations above about A, B and C seem to support that. Reasonable belief about relevance does not seem sufficient for A, B or C. All three of them might have reasonable beliefs about relevance. Only one of them (at most) can have knowledge about relevance in this
围绕显著性的问题不应分散我们的注意力。我们现在要集中讨论的是，如果我们在对某一事件的因果解释中提供某一因素，是否应该要求我们知道该因素具有因果相关性。上文关于 A、B 和 C 的考虑似乎支持这一观点。对于 A、B 或 C 来说，对相关性的合理信念似乎还不够。在这种情况下，他们中只有一个人（最多一个）可以拥有关于相关性的知识。
example, and only knowledge seem to provide understanding. They need of course not know that they have knowledge about relevance.
例如，只有知识似乎才能提供理解。当然，他们不需要知道自己掌握了相关知识。

Salience, as I understand it, connects with understanding in a different way. We might ask why something happened, and seek understanding, also when many causal factors are common knowledge. (There was oxygen in the air, etc, etc.) In that context, understanding might be achieved by pointing to a further factor, to the fact that a known factor was much stronger than usual, to the absence of a factor that is normally known to be there, or to an unusual factor not normally known to be there. When an explanation provides understanding, that normally takes place in a context where there are not only a lot of beliefs and assumptions, there is also a lot of knowledge that plays a role in generating the question. Issues in the pragmatics of explanation must be understood on precisely this knowledge background.
根据我的理解，"显著性 "以不同的方式与 "理解 "联系在一起。当许多因果因素都是常识时，我们也可能会问为什么会发生这样的事情，并寻求理解。(空气中有氧气，等等等等）在这种情况下，通过指出另一个因素，指出一个已知因素比平常强得多的事实，指出一个通常已知存在的因素不存在，或者指出一个通常不知道存在的不寻常的因素，就可以达到理解的目的。当解释提供理解时，通常是在不仅有许多信念和假设，还有许多在产生问题时起作用的知识的情况下进行的。解释的语用学问题恰恰必须在这种知识背景下加以理解。

Our conclusion is that we should not only provide knowledge about the causal history of when explaining . We should provide factors we know are relevant for the case and thereby for the understanding we want to provide, factors which in fact played a role and were significant. We might limit our reply to factors that are particularly significant (salient), and there is a lot of pragmatic leeway in providing understanding, but on the whole we should only mention factors that we know are relevant when providing an explanation.
我们的结论是，在解释 时，我们不应只提供有关 因果历史的知识。我们应该提供我们所知道的与案件相关的因素，从而提供我们想要提供的理解，这些因素实际上发挥了作用，而且很重要。我们的回答可能仅限于特别重要（突出）的因素，在提供理解时有很多实用的余地，但总的来说，我们在提供解释时只应提及我们知道是相关的因素。

A further important point which the example of A, B, and C brings out is this: It simply seems true that we often, when explaining, do better by providing not only a known relevant factor, but also the reasons and grounds we have for believing that factor relevant. Those reasons should be part of the causal story we provide. In order to promote understanding and prevent misunderstanding, it seems that everybody would gain by being given the specific grounded reasons for believing that the relevant factors were indeed relevant. One way of doing that might, in a situation, be to show that one factor was sufficient, or nearly sufficient, for the occurrence of the explanandum, given the background knowledge. That might be done by subsuming the event to be explained under a known regularity: The drunkenness was perhaps at a level where driving ability is very much reduced, so much reduced that normal driving along this stretch of road is extremely unlikely. If that is known, it is vital to pass this knowledge on to those requesting an explanation. We therefore should provide more than a known factor, and the statement that this factor is relevant, we should provide why we take it to be relevant, and when doing that we should contribute what we know about why it is relevant. So there are two things: Knowledge about which factors are relevant, and knowledge about why these factors are indeed relevant. In explanation both kinds of knowledge are important. The first knowledge links the explanans and the explanandum as cause and effect. The latter knowledge links the explanans and the explanandum by providing the relevant reasons for thinking the first link to be there. Note how this way of putting it does not tie the property of being explanatory with the property of predictive power, or with the property of establishing a modal link that exhibits the necessity of the effect. Still it makes the provision of understanding the central feature.
A、B 和 C 的例子还引出了另一个重要问题：我们在解释时，不仅要提供已知的相关因素，还要提供我们认为该因素相关的理由和依据，这样做往往会更好。这些理由应该是我们提供的因果故事的一部分。为了促进理解和防止误解，似乎向每个人提供相信相关因素确实相关的具体理由会使他们受益匪浅。在某种情况下，做到这一点的一种方法可能是，根据背景知识，说明一个因素足以或几乎足以导致解释性备忘录的发生。要做到这一点，可以把要解释的事件归入一个已知的规律中：醉酒者的醉酒程度可能会大大降低驾驶能力，以至于在这段路上正常驾驶的可能性极小。如果知道了这一点，就必须将这一知识传递给那些要求解释的人。因此，我们不仅要提供一个已知的因素，并说明这个因素是相关的，我们还应该提供为什么我们认为它是相关的，在这样做的时候，我们应该提供我们所知道的为什么它是相关的。因此，有两件事情：关于哪些因素相关的知识，以及关于为什么这些因素确实相关的知识。在解释过程中，这两种知识都很重要。前一种知识将解释者和被解释者作为因果关系联系起来。后一种知识通过提供认为第一种联系存在的相关理由，将被解释者和解释对象联系起来。请注意，这种表述方式并没有把 "解释性 "这一属性与 "预测力 "这一属性，或与 "建立模态联系以显示效果的必然性 "这一属性联系起来。不过，它还是把提供理解作为核心特征。

Another very different and indeed interesting case would be where the causal factor does not make the explanandum very likely at all, but it is the only known causal factor. Then this last fact needs to be stated in the explanation. (This can be
另一种非常不同而且确实有趣的情况是，因果因素根本不可能使解释词变得非常可能，但它却是唯一已知的因果因素。那么就需要在解释中说明这最后一个事实。(这可以是
taken as the case of untreated latent syphilis and paresis. In this case we know of no other causal factor, and for all we know there might not be one. What we have is to my mind an explanation because it provides the causal knowledge we have, knowledge that again might be the only causal knowledge we can have in this case.) Again: we should supply what we know and only what we know, in order to prevent misunderstanding arising, and in order to make plain how we know that the factor in question was relevant.
在这种情况下，我们不知道还有其他的致病因素，而且据我们所知，可能也没有其他的致病因素。在这种情况下，我们不知道有其他因果因素，而且我们也知道可能没有其他因果因素。在我看来，我们所拥有的是一种解释，因为它提供了我们所拥有的因果知识，而这种知识又可能是我们在这种情况下所能拥有的唯一因果知识。）再说一遍：我们应该提供我们所知道的，而且只能提供我们所知道的，以防止产生误解，并说明我们是如何知道有关因素是相关的。

Conclusion: Explanation provides understanding, and understanding is, simply put, knowing why. In understanding we link the explanandum and the explanans. Knowing, for good reason(s), that the link is there, constitutes causal understanding. Knowing what there is to know about what links there are to many or all facts in the causal history constitutes (full) causal understanding.
结论：解释提供理解，而理解，简单地说，就是知道为什么。在理解的过程中，我们把解释者和被解释者联系起来。有充分的理由知道这种联系是存在的，就构成了对因果关系的理解。知道了因果历史中许多或所有事实之间的联系，就构成了（完全的）因果理解。

It seems to me that the aim of providing (causal) understanding when explaining, limits the ways of giving causal histories dramatically. The ways of limiting the causal histories all contribute to accounting for the semantic intensionality of explanations. Especially important for accounting for the intensionality is the need for preservation of knowledge in the allowable substitutions. Lewis does not aim for providing necessary and sufficient conditions for explanation, and neither do I. I share in many ways Lewis's picture. I still hold that we can do better that Lewis, and that by seeing understanding as the aim of explanation we do better than Lewis.
在我看来，解释时提供（因果）理解的目的极大地限制了提供因果历史的方式。限制因果历史的方式都有助于解释的语义内向性。尤其重要的是，必须在允许的替换中保留知识，这对解释的内向性尤为重要。刘易斯的目标不是为解释提供必要和充分条件，我也是如此。我仍然认为，我们可以比刘易斯做得更好，把理解视为解释的目的，我们就比刘易斯做得更好。

This first part can be summed up as a simplified slogan: explaining is to provide knowledge why p..
这第一部分可以概括为一个简化的口号：解释 就是提供知识，说明为什么......

a) Conjectures a) 猜想

Many are now getting impatient with a major objection: We often to do not know the explanans for a fact when we explain. I agree. There is clearly such a thing as conjecturing an explanation. (There is also such a thing as mistakenly believing one knows , for instance when happens, contrary to belief, to be false.) In the case of conjecturing an explanation, we either suggest an explanation while believing that we do not really know that the explanans took place (We say: May be the road was icy?), or we do not know that the factor we point to was relevant. (We say: It might have been the icy road). What we typically do when we conjecture an explanation is to mark or indicate that what we are doing is somehow short of explaining. We indicate that by the maybes and might have beens. I said above that the serious explainer must think of herself as possessing an explanation (of some sort), and must think of her reply as making that explanation available to a person who does not possess the explanation. I believe the explainer who conjectures an explanation sees herself as not really possessing an explanation, and only as possessing a reasonable or likely hypothesis. To possess an explanation one needs further evidence.
现在，许多人对一个重要的反对意见感到不耐烦了：我们在解释时往往不知道解释者是谁。我同意这个观点。显然，有一种情况叫做猜测解释。(还有一种情况是误以为自己知道 ，例如，当 恰好与所相信的相反，是假的）。在猜测解释的情况下，我们要么提出一个解释，同时认为我们并不真正知道发生了这样的解释（我们说：可能是路面结冰了？(我们说：可能是路面结冰了）。当我们猜测一种解释时，我们通常会做的是标记或表明我们正在做的事情在某种程度上还不足以解释。我们用 "可能 "和 "也许 "来表明这一点。我在上面说过，认真的解释者必须认为自己拥有（某种）解释，而且必须认为自己的回答是在向不拥有解释的人提供解释。我认为，猜测解释的解释者认为自己并不真正拥有解释，而只是拥有一个合理或可能的假设。要拥有一个解释，就需要进一步的证据。

It is also clear that by conjecturing an explanation I may be providing the propositional structure that happens to be the right objective explanation. My distinction between conjecturing an explanation and explaining is wholly on the subjective side, and concern whether the potential explainer should rightly be said to see herself as putting forward knowledge why p or not. Of course one might be mistaken in that as well.
同样清楚的是，通过猜想一种解释，我可能会提供恰好是正确的客观解释的命题结构。我对猜想解释和解释之间的区别完全是主观方面的，涉及的是潜在的解释者是否应该正确地认为自己提出了 "为什么是 p "的知识。当然，在这一点上，我们也可能是错的。

Lastly there is a delicate and difficult border terrain from conjectures like the one about drunkenness to probabilistic knowledge of causal factors, and statistical explanation. Take the case of drunken driving I have provided. Knowing that of the drivers in this place are drunk at this time is only good enough for conjecturing an explanation of the crash as long as the interesting condition, the drunkenness, is something that we know can be checked independently by a blood-test. In case that has not been done, we should stick to a conjecture when it comes to identifying the cause(s) of the crash. But in many cases there is no such independently checkable factor, and there is nothing more to identifying the cause(s) than the probabilistic knowledge we have about connections, no independent condition (like drunkenness) we know how to check for. In that case we must again state precisely that: we must state what be believe we know, and leave it at that. We are then not conjecturing an explanation that we believe we can have, we are stating the only explanation we have and believe (with reason) we can have.
最后，从像醉酒这样的猜想到对因果关系的概率知识和统计解释，存在着一个微妙而困难的边界地形。以我提供的醉酒驾驶案例为例。只要我们知道醉酒这个有趣的条件是可以通过血液测试独立检查出来的，那么知道 这个地方的司机在这个时候喝醉了，就足以猜测车祸的原因。如果没有这样做，我们在确定车祸原因时就应该坚持猜测。但在许多情况下，并不存在这样一个可以独立检查的因素，而且除了我们所掌握的有关联系的概率知识之外，并没有其他更多的原因可以确定，我们也不知道如何检查任何独立的条件（如醉酒）。在这种情况下，我们必须再次说明：我们必须说明我们相信我们所知道的，并就此打住。这样，我们就不是在猜想一个我们认为可以得到的解释，而是在陈述我们所拥有的、并且相信（有理由相信）我们可以得到的唯一解释。

In most explanations of some complexity we idealize and simplify, and what we say when explaining is often strictly false. Since this is so, your thesis is completely off the mark, I hear many people saying. I reply that this is true in a way, but there is a lot to consider here. Firstly, I am only dealing with causal explanation. Simplification in the causal case must in general be such that it promotes causal understanding. It might, in a situation, promote such understanding not to mention some tiny factors, and concentrate on the major factors (among those relevant). It might promote the understanding to simplify things so that one can get a manageable mathematics going. Still, after all that, all idealization is and must be a matter of sound judgement on the background of knowledge why it happened. Only on such a background is it clear that simplifications and idealizations are fully in order and that they promote understanding and not the opposite. The understanding of how simplification and idealization works must therefore be provided as a development of the present thesis about causal explanation, and such things constitute a (scientific) refinement of the activity of providing knowledge why.
在对某些复杂问题的大多数解释中，我们都会将其理想化和简单化，而我们在解释时所说的往往都是严格意义上的错误。我听到很多人说，既然如此，你的论点就完全不对了。我回答说，这在某种程度上是对的，但这里有很多问题需要考虑。首先，我只涉及因果解释。一般来说，因果关系中的简化必须能够促进对因果关系的理解。在某种情况下，不提一些微小的因素，而集中于主要因素（在相关因素中），可能会促进这种理解。它可能会促进人们对简化事物的理解，从而使人们能够获得可管理的数学知识。尽管如此，所有的理想化都是而且必须是在了解其发生原因的背景下做出的正确判断。只有在这样的背景下，简化和理想化才是完全合乎规律的，才会促进理解，而不是相反。因此，对简化和理想化如何起作用的理解，必须作为目前关于因果解释的论述的发展来提供，而这种事情构成了对提供 "为什么 "的知识这一活动的（科学）完善。

c) Knowing what c) 了解什么

I have above in fact presupposed that we know and agree what our epistemic puzzle is when asking for an explanation. That presupposition is behind the thought that what we are after can be captured by "why"-questions in the causal case. May
事实上，我在上文已经预先假定，在要求解释时，我们知道并同意我们的认识论难题是什么。这一预设的背后是这样一种想法，即我们所追求的东西可以通过因果情况下的 "为什么 "问题来捕捉。五月
be this is wrong. If it is, my starting point is really a causal query rather than a specific why-question-I do not think the syntactic form of the question is essential. It has also been presupposed that we all know what the explanandum is, or what the epistemic query is. But sometimes we do not know this. Sometimes we are in the business of finding that out. In that case, a why-question may disguise itself as a what-question. Sometimes deep scientific disagreements really concern not how to explain why an event occurred, but concern instead what that event to be explained is. Then the what-question is the prior and primary question, and the why-question is secondary.
这可能是错的。如果是这样的话，我的出发点其实是一个因果问题，而不是一个具体的 "为什么 "问题--我认为问题的句法形式并不重要。还有一种预设是，我们都知道解释体是什么，或认识论的疑问是什么。但有时我们并不知道这一点。有时，我们正在寻找答案。在这种情况下，"为什么 "的问题可能会伪装成 "是什么 "的问题。有时，深刻的科学分歧并不真正涉及如何解释某个事件发生的原因，而是涉及要解释的事件是什么。那么 "是什么 "的问题就是首要问题，而 "为什么 "的问题则是次要问题。

Take one of Lewis possible counterexamples to his own thesis: Walt is immune to smallpox. Why? He is immune because he possesses antibodies able to fight off a virus-attack. This reply does not give the cause of the immunity. It rather explains what it is to be immune. If this is the reply we seek, we have the case of a whatquestion in the form of a why-question. We should not find that odd.
以刘易斯可能提出的一个反例来反驳他自己的论点：沃尔特对天花免疫，为什么？他之所以有免疫力，是因为他拥有能够抵御病毒攻击的抗体。这个回答并没有给出免疫的原因。而是解释了什么是免疫。如果这就是我们要找的答案，那么我们就有了一个以 "为什么 "的形式提出的 "是什么 "的问题。我们不应该觉得奇怪。

Note that we may be after a cause after all. Perhaps the answer we want is this: Because he has been vaccinated. That is in fact the natural answer to a whyquestion. It gives a causal explanation of the immunity, and assumes that we know what the explanandum is.
请注意，我们追求的可能终究是一个原因。也许我们想要的答案是因为他接种过疫苗。事实上，这是对 "为什么 "问题的自然回答。它给出了免疫的因果解释，并假定我们知道解释体是什么。

Take a very different sort of case. Why did Walt's hand move? Here is one reply: It moved because he wanted to give a sign to Peter. A very different reply might cite a set of neurophysiological events resulting in the hand movement. These two replies are very interestingly different. They seem to conceptualise the explanandum in different ways. The first thinks of the explanandum as an action. Actions we typically explain by the acting person's motives. The second reply seems to think of the hand movement as a neurophysiological event with a neurophysiological cause, and itself as stating that cause.
换一种情况。华特的手为什么会动？这里有一个答案：因为他想给彼得一个手势。另一种截然不同的回答可能会提到导致手移动的一系列神经生理事件。这两种回答有趣地截然不同。它们似乎以不同的方式将解释事项概念化。第一种观点认为解释体是一种行为。我们通常用行动者的动机来解释行动。第二个回答似乎认为手的运动是一个神经生理学事件，有一个神经生理学原因，而它本身也说明了这个原因。

There is the possibility that these two explanations are explanations of the same event, only described in different ways. There is also the possibility that these explanations explain different events. In fact, we are here up against very big controversies about the relationship between the mental and the physical. On one view, the view that sees the action and the neurophysiological event as different entities ontologically, there might be no question about explanations competing to explain the same explanandum. There are two distinct events, and they are explained inside different explanatory schemes (and this does not necessarily deny interactions between these explanatory schemes). On another view, which sees the action as also a complex neurophysiological event, the two explanatory schemes might represent be different ways of explaining the same thing, and the relationship between the two explanatory schemes might be thought of in many different ways. One such view sees the relationship as that between folk theory (intentional explanation) and (real) science. Problems surrounding reductionism lumber in the background.
这两种解释有可能是对同一事件的解释，只是描述方式不同而已。还有一种可能是，这两种解释解释了不同的事件。事实上，我们在这里正面临着关于精神与物理之间关系的巨大争议。有一种观点认为，行动和神经生理事件在本体论上是不同的实体。有两个不同的事件，它们在不同的解释方案中得到解释（这并不一定否认这些解释方案之间的相互作用）。另一种观点认为行动也是一个复杂的神经生理学事件，两种解释方案可能是解释同一事物的不同方式，两种解释方案之间的关系可以有多种不同的思考方式。其中一种观点认为这是民间理论（意向性解释）与（真正的）科学之间的关系。而围绕还原论的问题则是一个背景。

My aim here is not to argue for or against one metaphysical view or other. My aim is only to point to the fundamental importance of the "what"-questions, and separate disagreements on those issues from disagreements about how to answer why-questions. There might be full agreement about how to explain neurophysiological events but disagreement about whether actions can at all be explained that way. Different answers to what-questions might be at the bottom of
我在这里的目的并不是要支持或反对某种形而上学观点。我的目的只是要指出 "是什么 "问题的根本重要性，并把在这些问题上的分歧与如何回答 "为什么 "问题的分歧区分开来。在如何解释神经生理学事件上可能存在完全一致的意见，但在行动是否可以这样解释的问题上却存在分歧。对 "是什么 "问题的不同回答可能是 "为什么 "问题的根本原因。
the most fundamental scientific disagreements. The different answers to whatquestions might reflect deep differences not only in scientific but also in metaphysical outlook. Parallel issues arise for the relationship between the entities and events described in the various special sciences, between physics, chemistry and biology. On the whole our general metaphysical outlook is deeply influenced by the development of sciences and the success of various explanatory strategies, and there has of course been significant development through the centuries as to how we at all view and think about what-questions. For the rest of the paper I shall push "what"questions aside, and return to the why-questions.
最根本的科学分歧。对这些问题的不同回答不仅反映了科学观的深刻差异，也反映了形而上学观的深刻差异。各种特殊科学所描述的实体和事件之间的关系，物理学、化学和生物学之间的关系，也存在着类似的问题。总体而言，我们的一般形而上学观深受科学发展和各种解释策略成功与否的影响，当然，几个世纪以来，我们如何看待和思考 "什么问题 "也有了重大发展。在本文的其余部分，我将把 "是什么 "的问题放在一边，回到 "为什么 "的问题上来。

An act of explaining can be more or less satisfactory. Lewis holds that it will not be instructive to fuss about whether an unsatisfactory explanation deserves to be called an explanation. I agree with that statement. Does the knowledge requirement bring with it constraints which conflict with this? I do not think so. I shall survey the important cases; they are a somewhat mixed bag. It seems to me that my view has much the same flexibility as Lewis's in most of the cases, but that it fares much better than Lewis's in that it can account in a natural way for why certain explanations are unsatisfactory.
解释行为可以是比较令人满意的，也可以是不那么令人满意的。刘易斯认为，在不令人满意的解释是否值得被称为解释的问题上大做文章不会有任何启发。我同意这种说法。知识要求是否会带来与此相冲突的限制呢？我认为不会。我将列举一些重要的案例；这些案例有好有坏。在我看来，我的观点在大多数情况下与刘易斯的观点具有相同的灵活性，但它比刘易斯的观点好得多，因为它能够以一种自然的方式解释为什么某些解释不能令人满意。

Consider the following: 请考虑以下几点：

If the false conjunct is a relatively unimportant one which could be fairly easily corrected, and the main conjuncts are known all right, it seems to me that we from our perspective might want to say that the explainer knew the essentials of the explanation even if some detail was wrong. As long as we can say that, we might also want to say that an explanation was given, even if it was not fully satisfactory. It is not satisfactory because there were things the explainer should have known in addition to what she knew. Of course the explainer ought to see herself as knowing an explanation as long as she is not conjecturing an explanation.
如果错误的连词相对来说并不重要，可以很容易地纠正，而主要的连词都知道，那么在我看来，从我们的角度来看，我们可能想说，解释者知道解释的要点，即使有些细节是错误的。只要我们能这么说，我们可能还想说，解释者已经给出了解释，即使它并不完全令人满意。说它不令人满意，是因为除了解释者知道的东西之外，还有她本应知道的东西。当然，只要解释者不是在猜测一种解释，她就应该认为自己知道一种解释。

There are also the following cases: 3 . The explainer provides fairly little information. This case is easy: The explainer knows something and provides that, but that is not all that much. Of course it would have been better if more had been known. 4. The information given may be stale news. 5. The information given may not be of the sort the recipient wants. 6. The information might be given in a jumbled or bad way. 7. The recipient might start out with some misinformation and the explainer might fail to correct this. In all of these cases my view does not really commit me to saying whether or not the unsatisfactory explanation deserves to be called an explanation. That might we an issue we should not push. Nevertheless my view provides a platform for assessing the unsatisfactoriness and analyze it, and that we need.
还有以下情况：3 .解释者提供的信息相当少。这种情况比较简单：解释者知道一些事情并提供了这些信息，但这些信息并不多。当然，如果能提供更多信息就更好了。4.所提供的信息可能是陈旧的新闻。5.所提供的信息可能不是接受者想要的。6.所提供的信息可能杂乱无章或糟糕透顶。7.7. 接收者可能一开始就得到了一些错误信息，而解释者可能无法纠正这些错误信息。在所有这些情况下，我的观点并不真正要求我说不令人满意的解释是否值得被称为解释。这可能是一个我们不应该推动的问题。不过，我的观点提供了一个评估和分析不令人满意之处的平台，而这正是我们所需要的。

The latter cases, from 3 to 7, are all cases where the recipient does not receive the right thing, what is received is either too little, stale etc. The situation of the recipient is not much improved, and if the recipient did not know an explanation of an occurrence of something before receiving the service of the explainer, fairly little has changed. The knowledge situation of the receiver ought to change upon receiving an explanation, and ideally the situation should be such as to leave the recipient with knowledge why. If that has not happened, an explanation has not been given. The present view provides the rough standard, and the point of an explanation. It also provides us with the right material for saying, in many cases, that this is not really explaining. The difficulty with Lewis's view on explanation is again that we seem not to be given any material for saying anything like that. The point is that that material should be given by the account of explanation even if such an account should not settle all grey-zone cases and rule on all of these whether they are explanations or not.
后一种情况，从 3 到 7，都是接受者没有得到正确的东西，得到的东西太少或太陈旧等。如果接受者在接受解释者的服务之前不知道对某一事件的解释，那么接受者的情况也不会有太大的改善。接受者的知识状况应该在接受解释后有所改变，理想的情况是让接受者知道为什么。如果没有发生这种情况，就说明没有做出解释。本观点提供了解释的粗略标准和要点。它还为我们提供了正确的材料，使我们能够在许多情况下说，这并不是真正的解释。刘易斯关于解释的观点的难点还在于，我们似乎没有得到任何材料来说出类似的话。问题的关键在于，解释的论述应该提供这种材料，即使这种论述不应该解决所有灰色地带的情况，也不应该对所有这些情况作出是否是解释的裁决。

The other type of case is where the receiver receives the objectively right stuff but the explainer does not know that.
另一种情况是，接收者收到了客观上正确的东西，但解释者并不知道。

When thinking of oneself as conjecturing an explanation, one might be providing the objectively correct explanation and one might not. Explanation can, as Bromberger so forcefully pointed out, be an objective entity, something that we are the first to know. Let us concentrate on this objective side of things, and the abstract object that is the explanation, an object which can be known for the first time, and which has a structure we can discover.
当我们认为自己在猜想一种解释时，我们可能提供了客观上正确的解释，也可能没有。正如布罗姆伯格极力指出的那样，解释可以是一个客观实体，是我们首先知道的东西。让我们把注意力集中在事物的这一客观方面，以及作为解释的抽象对象，一个可以被首次认识的对象，它具有我们可以发现的结构。

A natural thought, given what we have argued so far, is that the structure of the abstract object should be seen as deeply constrained by the fact that a good explanation provides understanding, and this I see as providing knowledge why. There cannot be a further point to structure of this abstract object beyond that of serving the point of explanation.
鉴于我们迄今为止的论证，一个自然的想法是，抽象对象的结构应被视为深受一个好的解释提供理解这一事实的制约，我认为这就是提供知识的原因。除了为解释服务之外，这个抽象对象的结构不可能有其他意义。

It is a fact that much of the best work on explanation might be seen as attempts to identify the type of structure such an abstract entity, the explanation, must have. There is a temptation to think that there is an objective entity here whose structure and properties we can discover, an entity that exists independently of our
事实上，关于解释的许多最出色的工作都可以看作是试图确定解释这种抽象实体必须具有的结构类型。有一种诱惑让人认为，这里有一个我们可以发现其结构和属性的客观实体，一个独立于我们的解释而存在的实体。
explanatory attempts. When we objectify, this ideal object becomes the object of inquiry, and we see the task of the theory of explanation as identifying the requirements for being such a structure. Satisfying these requirements is then seen as necessary and sufficient for being an explanation.
解释的尝试。当我们客观化时，这个理想的对象就成了探究的对象，我们认为解释理论的任务就是确定成为这样一个结构的要求。满足这些要求被视为解释的必要条件和充分条件。

There, are of course, disagreements about what type of structures and properties we should concentrate upon. One line is to concentrate on logical and syntactic structure of the explanation. Linguistic object are in good ontological standing, and are taken as unproblematic for the physicalists, as logical properties also are. Hempel and Oppenheim seem to have thought along these lines when laying down their explicit requirements for an explanation of a fact. (A fact is a true description of an event: There are law-premises and particular fact premises. The conclusion says that the event we want explained took place, and the argument from the premises to the conclusion is deductively is valid. All premises are true.)
当然，对于我们应该关注哪类结构和属性，也存在分歧。一种观点认为，我们应该关注解释的逻辑和句法结构。语言对象具有良好的本体论地位，对物理主义者来说是没有问题的，逻辑属性也是如此。亨普尔和奥本海姆在提出对事实解释的明确要求时，似乎也是这样考虑的。(事实是对事件的真实描述）：有法律前提和特定事实前提。结论说我们想要解释的事件发生了，从前提到结论的论证是演绎有效的。所有前提都是真的）。

The verdict today, as stated above, is that the strategy of capturing what an explanation is by identifying structural properties, for instance formal and syntactic properties of such an objective structure, fails. Furthermore the idea that explanation and prediction come close together seems altogether wrong, and one can on the knowledge approach see explanation in a quite different way. But Hempel's view still has appeal: In a range of cases it seems right, or perhaps ideally right, to provide a D-N-explanation. And we might also ask why this is so.
如上所述，今天的结论是，通过确定结构属性（例如这种客观结构的形式和句法属性）来把握解释是什么的策略是失败的。此外，认为解释与预测密切相关的观点似乎完全错误，人们可以从知识的角度以完全不同的方式看待解释。但是，亨普尔的观点仍然具有吸引力：在一系列情况下，提供 D-N 解释似乎是正确的，或者说是理想的。我们也许还会问，为什么会这样呢？

David Lewis discusses in some detail the relationship between his own proposal and the D-N model. Defenders of the D-N model of course take objection to the employment of the ordinary notion of causation, that notion has resisted precise analysis and is not available. But when they give examples to motivate the D-N model they invariably pick examples where the covering law-model includes a list of joint causes of the explanandum event.
戴维-刘易斯详细讨论了他自己的提议与 D-N 模型之间的关系。当然，D-N 模型的辩护者反对使用普通因果关系概念，因为这一概念一直受到精确分析的抵制，无法使用。但是，当他们举例说明 D-N 模型的动机时，他们总是选取这样的例子，即覆盖法模型包括解释事件的共同原因清单。

In fact Lewis's model and the covering law model of explanation could have been reconciled if one had a covering law model of causation. Even the present approach to explanation could have been reconciled with a covering law model if we had a workable D-N type reductive theory of causation. But we do not have a D-N analysis of causation, and in my judgement there is no good reason to think that we can or even could have one. However, a D-N argument may present us with causes, and then it definitely looks explanatory. This fact we ought to be able to account for.
事实上，如果我们有一个覆盖法的因果关系模型，刘易斯的模型和覆盖法的解释模型是可以调和的。如果我们有一个可行的D-N型因果还原理论，甚至现在的解释方法也可以与覆盖律模型相协调。但是，我们并没有对因果关系进行 D-N 分析，而且根据我的判断，也没有充分的理由认为我们能够或甚至可能有这样的分析。然而，D-N 论证可能会向我们提出原因，那么它看起来肯定具有解释性。我们应该能够解释这一事实。

I think we can account for it this way: When the D-N argument provides the relevant causal information, then it is explanatory because it provides the relevant causal knowledge, and because it provides the best reason we have for thinking that the cause is the cause. Providing such a reason is exactly the role of subsumption under the regularity, which in this case is a true causal regularity. The knowledge approach can therefore easily account for the great appeal of the D-N model, by noting that there may be many cases where the -model provides exactly what the knowledge approach says should be provided, namely knowledge why the event to be explained happened. But, the knowledge approach would maintain, in many cases we have a different sort of reason for thinking that the cause is the cause.
我认为我们可以这样解释：当D-N论证提供了相关的因果信息时，它就是解释性的，因为它提供了相关的因果知识，因为它提供了我们认为原因就是原因的最好理由。提供这样一个理由正是归入规律性的作用，在这种情况下，规律性就是真正的因果规律性。因此，知识方法可以很容易地解释 D-N 模型的巨大魅力，它指出，在许多情况下， ，模型所提供的正是知识方法所说的应该提供的，即为什么要解释的事件会发生的知识。但是，知识方法认为，在许多情况下，我们有另一种理由认为原因就是原因。

It is indeed a fact that real life explainers mostly do not bother about serving up full D-N arguments. We seldom do, and most often we cannot. It may be that we cannot because there is so much we do not know. There is always a lot more to
事实上，现实生活中的解释者大多不屑于提供完整的 D-N 论证。我们很少这样做，而且常常做不到。我们做不到，可能是因为我们不知道的东西太多了。总是有很多东西
know about everything. But: This permanent lack of complete knowledge does not stop us from knowing why many an event occurred. We may have such knowledge in very many different ways, i.e. in ways much short of anything like D-N arguments. Some times the best we can have is the identification of the mechanism (for instance in a psychological case) that produced the event we want explained. Knowing the mechanism is then knowing why, it is the reason for believing the cause to be the cause, and this should not, in this case, be thought of as an explanatory sketch to be filled into a D-N argument.
什么都知道。但是这种永久性的全面知识的缺乏并不妨碍我们知道许多事件发生的原因。我们可以通过许多不同的方式来了解这些知识，也就是说，我们可以通过与D-N论证相去甚远的方式来了解这些知识。有时，我们所能得到的最好的知识就是确定产生我们想要解释的事件的机制（例如在心理学案例中）。知道了机制，也就知道了原因，也就是相信原因就是原因的理由，在这种情况下，我们不应该把它看作是一个可以填入 D-N 论证的解释性草图。

There is no denying that the great appeal of the D-N model has been intimately connected with the great appeal of a regularity-view on causation. It is fairly easy to see that such a reductive view on causation itself, a view which reduces the causal relation to epistemically manageable factors and connections, namely regularities etc, lends a lot of credence to the D-N model. Some of the ammunition against the D-N model is ammunition against the regularity view on causation.
不可否认，D-N 模型的巨大吸引力与关于因果关系的规律性观点的巨大吸引力密切相关。不难看出，这种关于因果关系本身的还原观点，即把因果关系还原为认识论上可处理的因素和联系（即规律性等）的观点，为D-N模式提供了许多可信度。反对D-N模式的一些弹药就是反对因果关系规律性观点的弹药。

It might also be fair to say that when Wesley Salmon developed his alternative model of explanation, the statistical relevance model, then he was also led to an alternative view on causation and causal processes. Salmon's view on explanation is that to explain an event is to assign to it the broadest homogeneous reference class. (It is noticable how he avoids using the concept of causality when characterizing explanation). In causal explanation one issue then is whether this always gives us the best reason for identifying the cause as a cause. It is clear that it does not always do that, that all depends on the causal mechanisms at work and on how they interact whether mechanisms cancel each other out and so on. Salmon needs to link evidential probabilistic facts with causal knowledge, and it is far from clear that it can be done - some times it works but some time is does not, as Nancy Cartwright has been stressing.
或许还可以这样说，当韦斯利-萨尔蒙提出他的另一种解释模式--统计相关性模式时，他也被引向了另一种关于因果关系和因果过程的观点。萨尔蒙关于解释的观点是，解释一个事件就是给它分配一个最广泛的同质参照类。(值得注意的是，他在描述解释时避免使用因果关系的概念）。那么，在因果解释中，一个问题是，这是否总能给我们提供最好的理由来确定原因是什么。显然，它并不总是能做到这一点，这完全取决于起作用的因果机制，以及这些机制是如何相互作用的，这些机制是否会相互抵消等等。正如南希-卡特赖特（Nancy Cartwright）一直强调的那样，萨尔蒙需要将证据性的概率事实与因果知识联系起来，但目前还不清楚是否能做到这一点--有时能做到，但有时做不到。

(Salmon's work on causal processes is to my mind very refreshing, and the fact that it turns the theory of causation away from arguments and linguistic structures to (ontic) processes in the world. This turn is very significant, and I believe this is a good and healthy move in thinking about causation. The delicate issues here concern the extent to which you can say something interesting and informative about causation without actively employing causal notions. Perhaps what you can say amounts to this: The presence of a cause must somehow increase the likelihood of an effect when all other causal factors are held fixed. Maybe you can say little true
(萨尔蒙关于因果过程的研究让我耳目一新，它将因果关系理论从论证和语言结构转向了世界的（本体）过程）。这一转向意义重大，我认为这是思考因果关系的一个良好而健康的举措。这里的微妙问题涉及到在不主动使用因果概念的情况下，你能在多大程度上说一些关于因果关系的有趣和有意义的话。也许你可以这样说在所有其他因果因素都固定不变的情况下，原因的存在必然会在某种程度上增加结果产生的可能性。 也许你可以说一点真话

46

See Jon Elster's (1999) discussion of mechanisms, chapter 1. I agree with much of what he says about mechanisms, but I would stress that we may have causal knowledge in many different ways. The Sherlock Holmes method, for instance, rules out the impossible and leaves us with the improbable (but still the only thing possible).
参见乔恩-埃尔斯特（Jon Elster，1999 年）关于机制的讨论，第一章。我同意他关于机制的许多观点，但我要强调的是，我们可以通过许多不同的方式获得因果知识。例如，福尔摩斯的方法排除了不可能的事情，只留下不可能的事情（但仍然是唯一可能的事情）。

47 It seems to me that Salmon saw his theory of explanation as the evidential starting point of explanations, a basis that also needed explaining. The hard problem has always been, as it is in the regularity-view, to find a way to distinguish between mere correlations and genuine causal relations, and Salmon's strategy went by spatially continuous processes, and interactions between those processes, and in fact it appealed to the notion of a genuine causal process. This notion surely had the traditional problems built into it if the aim was reductive. Salmon later turned to the transfer of conserved quantities to account for the causal relation, and then he definitely seemed to be begging the question about many types of causal connections, for instance all issues about mental causation.
47 在我看来，萨尔蒙将他的解释理论视为解释的证据起点，而这一基础也需要解释。正如在规律性观点中一样，困难的问题始终是找到一种方法来区分单纯的相关性和真正的因果关系，而萨尔蒙的策略是通过空间上连续的过程以及这些过程之间的相互作用来实现的，事实上它诉诸于真正的因果过程的概念。如果以还原为目的，那么这一概念肯定存在传统的问题。后来，萨尔蒙转而用守恒量的转移来解释因果关系，这无疑是在乞求许多类型的因果联系，例如所有关于精神因果关系的问题。
about causation, beyond this, and one needs to employ causal notions even to get this right. (One can say that a cause makes the effect happen, but I see that as the employing of causal notions to illuminate causality.))
除此以外，关于因果关系，人们甚至需要运用因果概念来正确理解。(我们可以说，原因使结果发生，但我认为这是运用因果概念来阐明因果关系）。

James Woodward has also recently argued against the view that causal explanation involves subsumption under laws. According to him, whether or not a generalization can be used to explain has to do with whether it is invariant rather than whether it is lawful. Invariance comes in degrees and has other advantages - a generalization can be invariant even if it holds only in a limited spatiotemporal interval. (Note how the notion of cause again is avoided in order to characterize causal explanations!). Again my perspective on this view is that we have a good reason for believing the cause to be a cause of some effect when we can connect that cause with the effect with an invariant generalization. That is why explanations with invariant generalizations are explanatory. But the question is whether this is the only type of reason for believing a cause to be a cause, or the only one we should look for. There is no good reason to think that, even if there are often good reasons for looking for invariant generalizations.
詹姆斯-伍德沃德（James Woodward）最近也反驳了因果解释涉及法律归属的观点。 他认为，一个概括能否用来解释，与它是否不变有关，而不是与它是否合法有关。不变性有程度之分，而且还有其他好处--即使一个概括只在有限的时空区间内成立，它也可以是不变的。(请注意，为了描述因果解释的特征，我们又一次回避了原因的概念！）。同样，我对这一观点的看法是，当我们能用一个不变的概括把原因与结果联系起来时，我们就有充分的理由相信原因是某种结果的原因。这就是为什么带有不变概括的解释具有解释性。但问题是，这是否是我们认为原因是原因的唯一理由，或者说我们应该寻找的唯一理由。我们没有充分的理由这样认为，即使我们经常有充分的理由寻找不变的概括。

The present approach maintains that the explanation should provide knowledge why, and that this task unifies causal explanation. We can see the task of providing knowledge why as a task that can be fulfilled in quite different ways. However, it involves stating what the cause is (what the causes are) and providing in the right sort of way the reasons for thinking that the cause is the cause. What way depends on the audience etc. etc. Logical or semantic structure cannot identify causal explanation. This conception follows naturally if we were to agree that there is no reductive account of causation or of causal knowledge available. The reductive approach to causation lends support to the idea that there is an objective structure common to all causal explanations, since there are conditions necessary and sufficient for causality, and knowledge of those conditions would by definition always yield causal knowledge. If we leave that reductive ideal behind as misguided, then the demands on how to fulfil the explanatory task will naturally be given a concrete interpretation relative to the discipline and the knowledge situation in which we find ourselves. We can start with the causal connections we recognize and think we know about, and we have no external or general motivation for being revisionary about what causal connections there are. Of course we can change our minds about such things as we learn more, but that is another matter. But this means taking our present explanatory practices very seriously indeed, and seeing ourselves as working from within them when theorising about explanation. This does not mean that the task of theorising about explanation does not impose normative constraints upon explanation. Of course it does, and one such task is to be explicit about your assumptions and all your reasoning. Still this view can maintain that the way of arriving at a normative theory has some of the elements of a reflective equilibrium. (The sort of normative equilibriums I am thinking of can be local and situated.) In a discipline where we do explain from laws, the D-N model might become the natural instantiation of the task to exhibit knowledge why in an explanation. But in other disciplines like for instance economics we do not explain by invoking laws, we explain by identifying causal mechanisms, and the identification of such mechanisms exhibits the causal knowledge that can be had.
目前的方法认为，解释应提供 "为什么 "的知识，这一任务统一了因果解释。我们可以把提供 "为什么 "的知识看作是一项任务，可以通过多种不同的方式来完成。不过，这需要说明原因是什么（原因是什么），并以正确的方式提供认为原因是原因的理由。至于采用何种方式，则取决于受众等等。逻辑或语义结构无法确定因果解释。如果我们同意不存在对因果关系或因果知识的还原解释，那么这一概念就会自然而然地产生。对因果关系的还原方法支持这样一种观点，即所有因果解释都有一个共同的客观结构，因为因果关系存在必要和充分的条件，而对这些条件的了解按其定义总会产生因果知识。如果我们将这一还原性理想视为误导，那么如何完成解释任务的要求自然会被赋予与我们所处的学科和知识情境相关的具体解释。我们可以从我们认识到的、自认为知道的因果联系入手，我们没有外在的或普遍的动机去修正有哪些因果联系。当然，我们可以随着学习的深入而改变对这些事情的看法，但这是另一回事。但是，这意味着我们要非常认真地对待我们目前的解释实践，并在对解释进行理论研究时将自己视为在这些实践的范围内工作。这并不意味着理论化解释的任务不会对解释施加规范性约束。当然有，其中一项任务就是明确你的假设和你的所有推理。尽管如此，这种观点仍然认为，得出规范性理论的方式具有反思性均衡的某些要素。(我所考虑的那种规范性均衡可以是局部的和情景化的）。在我们确实从规律出发进行解释的学科中，D-N 模型可能会成为在解释中展示 "为什么 "的知识这一任务的自然实例化。但在其他学科中，比如经济学，我们不是通过援引规律来解释，而是通过识别因果机制来解释，而识别这种机制就展示了可以获得的因果知识。

On this way of thinking, the constraints on the structural/objective side should be seen as a reflection of the task of stating knowledge why in the normatively best way in concrete situations. However, disciplines change, knowledge situations change, the perceived relationship between disciplines change, and the general metaphysical outlook which frames the knowledge situation also undergoes historical changes. W. Salmon once wrote:
按照这种思路，结构/客观方面的限制应被视为在具体情境中以规范的最佳方式说明知识为何的任务的反映。然而，学科在变，知识情境在变，学科之间的认知关系在变，框定知识情境的一般形而上学观也在发生历史性的变化。W. Salmon 曾写道

[...] what constitute adequate explanation depends crucially upon the mechanisms that operate in the world. In all of this there is [...] no logical necessity whatever. [...] My aim has been to articulate contingent features of scientific explanations in this world as we presently conceive it. (Salmon 1984, p. 240 and 278)
[......]什么才是充分的解释，关键取决于世界的运行机制。在这一切中，[......]没有任何逻辑必然性。[......]我的目的是阐明科学解释在我们目前所设想的世界中的偶然特征。(萨尔蒙，1984 年，第 240 页和第 278 页）

My difference from Salmon, if there is one on this point, is that to see these contingent features of explanation as various concrete ways of satisfying the task of providing knowledge why. That general feature I do not see as contingent.
如果说在这一点上我与萨尔蒙有什么不同的话，那就是我把解释的这些偶然特征看作是满足提供 "为什么 "的知识这一任务的各种具体方式。我不认为这种一般特征是偶然的。

I therefore think that the search for necessary and sufficient conditions for explanation in general, and the search for such conditions in the properties of an objective structure, the structure of causal explanation, is, in a way, badly misconceived. It is a misconception that there is such a thing as the objectively right structure of all causal explanations of particular events. This misconception is fed by the misconception that in the end there will be a satisfactory account of causation. This latter misconception fails to realise that some of the basic concepts, like the concepts of knowledge and causation, are so centrally placed that there are no reductive accounts of them. We can still say illuminating things about such concepts, and we should not be afraid of employing them for our philosophical purposes. Employing them becomes a way both of illuminating both the philosophical problems and these concepts.
因此，我认为，从总体上寻找解释的必要条件和充分条件，以及从客观结构--因果解释的结构--的属性中寻找这种条件，在某种程度上是严重的误解。人们误以为存在着这样一种东西，即对特定事件的所有因果解释的客观正确结构。这种误解的根源在于，人们误以为最终会有一个令人满意的因果关系解释。后一种误解没有意识到，一些基本概念，如知识和因果关系的概念，处于如此核心的位置，以至于对它们没有还原的解释。我们仍然可以对这些概念做出有启发性的解释，而且我们不应该害怕为我们的哲学目的而使用它们。运用这些概念既是阐明哲学问题的一种方式，也是阐明这些概念的一种方式。

We can put this somewhat differently: There is an objective explanation with a particular structure when a good explanation of an event has been given, and in that case we can ask all of Bromberger's questions about it. Still, the properties of the objective explanation do not generalize to all explanations of all particular events, even if they might generalize interestingly. They do not generalize across the board because the understanding and the knowledge why is different from discipline to discipline, and sometimes inside disciplines. The objective structure that is an instantiation of knowledge why, is clearly not the same in all disciplines or in all causal explanations outside science. What remains constant is only that they are instantiations of knowledge why.
我们可以换一种说法：当我们对某一事件做出了很好的解释时，就会有一个具有特定结构的客观解释，在这种情况下，我们就可以提出布罗姆伯格的所有问题。尽管如此，客观解释的特性并不能概括所有特定事件的所有解释，即使它们可能概括得很有趣。它们不能一概而论，因为不同学科，有时是学科内部，对 的理解和认识是不同的。作为 "为什么 "这一知识的实例的客观结构，在所有学科或科学之外的所有因果解释中显然并不相同。不变的只是它们都是 "为什么 "知识的实例化。

At this point there will be various charges, and one charge will be that the concept of knowledge is vague, another will be that by employing the concept of knowledge we start where we at best should end, it is not a concept available for use. The concepts that are available for use are presumably evidence, beliefs (expressed in linguistic structures), inferential connections, assertion, justification etc. To these points I want to say: T. Williamson's work in epistemology tries precisely to show
在这一点上，会有各种指控，一种指控是知识的概念含糊不清，另一种指控是使用知识的概念，我们的出发点充其量是我们应该结束的地方，它不是一个可供使用的概念。可以使用的概念大概是证据、信念（用语言结构表达）、推理联系、断言、理由等。对于这些观点，我想说的是威廉姆森（T. Williamson）在认识论方面的工作正是要表明
that we throw a lot of interesting light on belief by starting with knowledge. By seeing knowledge as the prior concept, we might put this by saying that belief is to be understood as failed knowledge, knowledge is not to be understood as belief with success.
我们可以从知识入手，为信念提供许多有趣的启示。将知识视为先行概念，我们可以说，信念应被理解为失败的知识，而知识不应被理解为成功的信念。

Once we cease to take belief to be simple and conceptually prior, we can experiment with using the concept of knowledge to elucidate concepts of justification and evidence. (Williamson 2001 p. 9)
一旦我们不再认为信念是简单和概念先行的，我们就可以尝试用知识的概念来阐释理由和证据的概念。(威廉姆森，2001 年，第 9 页）

Williamson goes on to give extremely interesting elucidations of evidence, justification and assertion. I extend a strategy of this type to explanation and (causal) understanding.
威廉姆森接着对证据、理由和断言进行了极为有趣的阐释。我将这种策略推广到解释和（因果）理解上。

In the light of Williamson's work this is indeed a natural suggestion. There is no reason not to employ the concept of knowledge from reason of conceptual priority; knowledge is prior to explanation anyway if knowledge is prior to the concepts needed to account for explanation in all attempts that superficially seem to avoid employing knowledge. Their gain, by avoiding knowledge, is really nothing anyway.
从威廉姆森的研究来看，这确实是一个自然的建议。从概念的优先性出发，我们没有理由不使用知识概念；如果知识先于解释所需的概念，那么在所有表面上似乎避免使用知识的尝试中，知识无论如何都是先于解释的。通过回避知识，他们的收获其实是一无所获。

There is another charge, and that is that the concept of knowledge lacks precision. It is true that it lacks precision in the sense that we cannot give a precise definition of knowledge. That is just because the concept is so central, and a definition would have to employ concepts that are secondary to knowledge. But the fruits of employing knowledge can only be seen by the application. One such application is in diagnosing and understanding the shortcomings of alternative approaches. The case of causal explanation is special because the concept of causality seems to be a concept with many parallels to the concept of knowledge, it has an equally central standing, as has been forcefully argued by Nancy Cartwright. I want to make the following points as a closing of this paper (and final preparation for the coda), and as an indication of where we might want to go from here:
还有另一种指控，那就是知识的概念缺乏精确性。从我们无法给知识下一个精确定义的意义上说，它确实缺乏精确性。这只是因为这个概念是如此核心，而一个定义将不得不使用对知识来说次要的概念。但是，运用知识的成果只能通过应用才能看到。应用之一就是诊断和理解其他方法的不足之处。因果解释的情况比较特殊，因为因果性概念似乎是一个与知识概念有许多相似之处的概念，它具有同样重要的地位，南希-卡特赖特（Nancy Cartwright）对此进行了有力的论证。我想提出以下几点，作为本文的结尾（也是为尾声做最后的准备），并表明我们可能希望从这里开始：

a) A knowledge account of causal explanation carries with it great insights that are likely to be lost by "reductive" accounts. The knowledge account can be seen to contain resources that make it able to say why the D-N model is right when it is right and wrong when it is wrong. It also accounts for the general attraction of the epistemic accounts of causal explanation; they are attractive because they capture the need for reasons for seeing the cause(s) as cause(s). The various epistemic virtues they appeal to, be it uniformity, coherence, or simplicity all have their role in providing reasons for believing the cause to be the cause. Predictive power, however, is not a defining trait of explanation. It is rather a consequence of there being a special type of good reasons for believing a cause to be a cause.
a) 对因果解释的知识解释蕴含着巨大的洞察力，而这些洞察力很可能被 "还原 "解释所遗失。可以看出，知识论包含的资源使其能够说明为什么 D-N 模型在正确的时候是正确的，而在错误的时候是错误的。它也说明了因果解释的认识论账户的普遍吸引力；它们之所以有吸引力，是因为它们抓住了把原因视为原因的理由需求。它们所诉诸的各种认识论优点，无论是统一性、一致性还是简单性，都可以为相信原因就是原因提供理由。然而，预测力并不是解释的决定性特征。相反，它是有一种特殊的好理由让人相信原因就是原因的结果。

The present view also has resources to explain why Salmon's approach is right when it is right, and wrong when it is wrong. Salmon is right that there might be limitations in what kinds of reasons we can give for believing a cause to be a cause, and we should not be misled by a misguided ideal of epistemic perfection. His view of explanation, that to explain an event is to assign to it the broadest homogenous reference class, is nevertheless much too simple as a view on causal explanation - it doe not always links evidence with causal knowledge, and there might be causal structures it cannot identify.
本观点还可以解释为什么萨尔蒙的方法在正确的时候是正确的，而在错误的时候是错误的。萨尔蒙的观点是正确的，即我们能够为相信原因是原因而给出的理由可能是有限的，我们不应该被错误的认识论完美性理想所误导。他的解释观认为，解释一个事件就是为其指定一个最广泛的同质参照类，然而，作为一种因果解释观，这种观点过于简单--它并不总是将证据与因果知识联系在一起，而且可能存在它无法识别的因果结构。

I will give a controversial example. We might hold: We explain an action by giving the reason(s) for which the agent acted. This does not fit the D-N model, and nothing is gained by trying to force it into it. Nothing is gained by introducing the idea of transmission of conserved quantities in this explanatory context. Knowing the reason for which the agent acted is clearly a way of knowing why the agent acted thus, and provides us with an explanation of the act. If we are also provided with information about how we know this reason to be the reason for which the agent acted, we have a good explanation, knowledge why. Knowing the reason is being familiar with a causally relevant factor, and nothing is gained by denying this because you accept a reductive view on causation. Of course a lot more needs to be said about what it is to explain an action. But a main point made here is independent of the more general and also controversial point that we explain actions causally by giving reasons. It is the point that if the latter point were right, we would have a way of knowing why an action took place.
我举一个有争议的例子。我们可以认为我们通过给出行为人采取行动的原因来解释行为。这并不符合 D-N 模型，把它强加于 D-N 模型也不会有任何好处。在这种解释背景下引入守恒量传递的概念也不会有任何好处。知道行为主体采取行动的原因显然是知道行为主体为什么这样做的一种方式，它为我们提供了对行为的解释。如果我们还能了解到我们是如何知道这个原因的，我们就有了一个很好的解释，知道了为什么。知道原因就是熟悉与因果相关的因素，如果因为接受了关于因果关系的还原观点而否认这一点，那么我们将一无所获。当然，关于什么是对行为的解释，还有很多需要说明的地方。但是，这里提出的一个要点与我们通过给出理由来解释行为的因果关系这一更普遍也更有争议的观点无关。这就是，如果后一点是正确的，我们就有办法知道为什么会发生一个行动。

b) We have to see the constraints on the structural side, the properties of the objective explanation given, as arising out of the meeting between the demand for knowledge why, (arising out of the need for understanding, i.e. causal understanding), and the concrete situation in which the explanation is given, a situation which is ultimately framed within a scientific and a metaphysical outlook. The approach takes very seriously the need to bring in the knowledge situation in question, and it therefore brings the world in. In the scientific case it brings in the knowledge situation of the discipline in question. It is a serious mistake for an account of explanation not to do that.
b) 我们必须把结构方面的制约因素，即所给出的客观解释的属性，看作是对 "为什么 "的知识需求（产生于对理解的需求，即因果理解）与给出解释的具体情境之间的交汇，而这种情境最终是以科学和形而上学的观点为框架的。这种方法非常重视引入相关知识情境的必要性，因此它引入了世界。就科学而言，它引入了相关学科的知识状况。如果解释学不这样做，那将是一个严重的错误。

c) The account does not at all aim to replace the normative role of a theory of explanation with a purely descriptive task. Knowledge is a normative concept. Explanations get their point from the normative task of providing knowledge why, and the aim of science is scientific knowledge. When we explain causally, we apply our knowledge why. Our understanding is relative to our system of concepts. Such systems also develop and improve, and as they do improve, we improve our understanding, our explanations and our knowledge. Still, as long as explanations provide knowledge why, we have all the objectivity we can ask for in explanation.
c) 该论述的目的完全不是要以纯粹的描述性任务取代解释理论的规范性作用。知识是一个规范性概念。解释的意义来自提供 "为什么 "的知识这一规范性任务，而科学的目的是科学知识。当我们解释因果关系时，我们运用的是 "为什么 "的知识。我们的理解是相对于我们的概念体系而言的。这种体系也在发展和完善，随着它们的完善，我们的理解、解释和知识也在完善。尽管如此，只要解释提供了 "为什么 "的知识，我们就拥有了我们所能要求的解释的所有客观性。

d) This account has a complex relationship with other accounts. In one respect it operates on a high level of abstraction, and it aims to integrate both positive and negative knowledge about explanation into a more concrete account. Interesting non-reductive theories are always much indebted to reductive attempts. I have argued that various theories of explanation can be seen as exemplifying various kinds of explanatory knowledge, and that one main problem simply is that they do not generalize. There is great need of exploring further what explanatory knowledge consists in - there might be kinds that have not been covered, and there might be many corrections that need be made also in the case of accounts of explanation that seem correct for a certain situation. Knowing p does not require that one knows that one knows: knowing is not luminous. On the other, knowing that you do know is very helpful when deciding whether or not to assert p. It is also very helpful when
d) 这种解释与其他解释之间的关系很复杂。一方面，它的抽象程度较高，旨在将关于解释的正反两方面知识整合到一个更具体的论述中。有趣的非还原性理论总是离不开还原性尝试。我认为，各种解释学理论可以被视为各种解释学知识的典范，而一个主要的问题仅仅在于它们并不具有普遍性。我们很有必要进一步探讨解释性知识的内涵--可能还有一些种类的解释性知识没有涉及到，而且，对于那些在某种情况下似乎是正确的解释性论述，我们可能还需要做出许多修正。知道 p 并不要求一个人知道自己知道：知道 并不发光。 另一方面，知道自己确实知道 对决定是否断言 p 非常有帮助。

49

See Williamson (2000). 见 Williamson (2000)。
wondering whether to assert an explanation. Part of the motivation for continued effort in exploring explanation could be to get closer to the favoured state of knowing that you do know why when you explain. This opens up a different perspective on what the philosophy of science is about which is able to exploit and digest for its own purposes much of the present good work. The debate between modalist, unificationist/epistemicist, and ontic-mechanist approaches to explanation, to use Salmon's vocabulary, could all be seen in this particular light as well.
不知道是否要作出解释。继续努力探索解释的部分动机可能是为了更接近 "当你解释时，你确实知道为什么 "的理想状态。这就为科学哲学的内涵开辟了一个不同的视角，它能够为自己的目的利用和消化目前的许多优秀成果。用萨尔蒙的术语来说，模态主义、统一主义/表义主义和本体-机制主义解释方法之间的争论，也都可以从这个特殊的角度来看待。

The change of focus gives a perspective on most issues in a theory of explanation. I shall end by making some remarks about inference to the best explanation. This I do because this issue has become such a central issue in debates about realism. . It has becomes central because of the claim that the best explanation of the success of science is the truth of science, and we therefore should conclude that science is mainly true. The latter conclusion requires the soundness of inference to the best explanation.
重点的改变为解释理论中的大多数问题提供了一个视角。最后，我将就最佳解释的推论发表一些看法。我之所以这样做，是因为这个问题已经成为现实主义争论中的一个核心问题。 .它之所以成为核心问题，是因为有人声称，科学成功的最佳解释就是科学的真理，因此我们应该得出结论，科学主要是真实的。后一种结论要求对最佳解释的推论是正确的。

The traditional issue of whether inference to the best explanation is acceptable can be seen as the issue of whether we can infer from the fact that one explanation satisfies all criteria for being a good explanation - better than its competitors-, to the conclusion that this explanation is also to be believed: we should accept it as a true description of the world. The problem is naturally conceived in a setting where truth is not a necessary property of an explanation; it is rather the issue (in Hempel's wording) of whether the best explanation is a true explanation. The whole
关于推论最佳解释是否可以接受的传统问题，可以看作是这样一个问题：我们是否可以从一种解释满足了成为好解释的所有标准--比其竞争者更好--这一事实，推论出这种解释也值得相信的结论：我们应该接受它是对世界的真实描述。这个问题自然是在真理不是解释的必然属性的情况下提出的；问题在于（用亨普尔的话说）最佳解释是否是真正的解释。整个

As a preamble to that, I will cast a glance at Hempel's discussion of his requirements. His last requirement, the requirement that the premises must be true, has a somewhat interesting history in Hempel's discussion. In a postscript (from 1964) to his early (1948) article with Oppenheim, Hempel adds that true premises characterize a correct or true explanation. He is therefore making a distinction between true and correct explanations, and explanation in general. In an analysis of the logical structure of explanatory arguments, he says this truth-requirement may be disregarded. (See Hempel (1965 p. 249)). Of course, what seems to be implied by the latter is simply that the rendering of the full logical structure of explanatory arguments is independent of whether the premises are true: the formal structures of correct and incorrect explanations are the same. If the aim is to capture the essentials of explanation by identifying their formal structure, then a truth-requirement is additional. From where does it stem and what is its status? Note how Hempel's distinction between true explanations and explanations is almost the mirror image of the activity-side distinction between conjecturing an explanation and explaining. From the present perspective the requirement that statements in the explanans be true flows directly out of the knowledge approach to explanation, that one must know that the explanans is true. This truth-requirement thus flows out of what I have called the activity side, and that is a side played down by Hempel.
作为前言，我想回顾一下亨普尔对其要求的讨论。他的最后一个要求，即前提必须为真的要求，在亨普尔的讨论中有一段有趣的历史。在他早期（1948 年）与奥本海姆合写的文章的后记（1964 年）中，亨普尔补充说，真实前提是正确或真实解释的特征。因此，他是在区分真正的解释和正确的解释，以及一般的解释。在分析解释性论证的逻辑结构时，他说这一真实要求可以忽略不计。(见亨普尔（1965 年，第 249 页）。当然，后者似乎只是暗示，解释性论证的完整逻辑结构的呈现与前提是否为真无关：正确解释和不正确解释的形式结构是相同的。如果我们的目的是通过确定解释的形式结构来把握解释的本质，那么真理要求就是额外的。它源自何处，地位如何？请注意亨普尔对真实解释和解释的区分几乎就是活动方面对猜想解释和解释的区分的镜像。从目前的视角来看，对解释者的陈述必须真实的要求直接源于解释的知识方法，即人们必须知道解释者是真实的。因此，这一真实性要求来自我所说的活动方面，而这正是亨普尔所淡化的方面。

51 Ben-Menahem (1990) contains a nice discussion of inference to best explanation. It also contains a negative discussion of the use of such inference to prove realism. I myself am sceptical as to whether truth is a property that can have a real explanatory role in empirical explanations. I think that what really explains the success of the theory that the atom is is that the atom is . This point can be put by employing the concept of truth. Still the real explanation of the empirical phenomenon of the success of science need not make reference to truth, and can be given also by those who believe truth is a purely logical concept.
51 Ben-Menahem（1990）对最佳解释的推论进行了很好的讨论。该书还对使用这种推论来证明现实主义进行了反面讨论。我本人对真理是否是一种可以在经验解释中发挥真正解释作用的属性持怀疑态度。我认为，真正能够解释 原子是 这一理论的成功之处在于，原子是 。这一点可以用真理的概念来说明。不过，对科学成功这一经验现象的真正解释并不需要提及真理，那些认为真理是一个纯粹逻辑概念的人也可以给出解释。
problematic presupposes that we have available the important properties of good (potential) explanations independently of whether we believe they explain anything. But that can be granted.
有问题的前提是，我们拥有好的（潜在的）解释的重要属性，而与我们是否相信它们能解释任何事情无关。但这是可以肯定的。

The subjective part of the inference to the best explanation naturally comes out as the issue of whether we should commit ourselves (in belief) to the best of several competing hypothesis, (or to the one and only hypothesis is case there is no competition). We might maintain, with Timothy Williamson, that the mental state of belief should be seen as deriving its properties from the mental state of knowing, and that to adopt a belief is to make an epistemic commitment that would be knowledge if what is believed is true.
推论最佳解释的主观部分自然是我们是否应该（在信念中）对几个相互竞争的假说中最好的假说做出承诺（或者在没有竞争的情况下对唯一的假说做出承诺）。我们可以与蒂莫西-威廉姆森（Timothy Williamson）一样，认为 "信念 "的心理状态应被视为从 "认识 "的心理状态中衍生出来的，而 "相信 "就是做出认识论上的承诺，如果 "相信 "的东西是真的，它就是 "知识"。

This then describes the subjective side of inference to the best explanation: The issue boils down to whether the best of competing hypothesis is deserving of such an epistemic commitment. And of course there is no general rule about that, it may be deserving of such a commitment, and it might not be. That all depends on the competition and on how well the better explanation satisfies other requirements upon being something we should commit ourselves to in belief. This again boils down to general epistemological issues, where it of course counts much in favour of accepting a proposition that it explains well propositions known to be true. But there is no simple valid inference to the best explanation, no rule of inference. And the property of loveliness, a property of the whole explanation, is not linked with the property of likeliness, a property of the explanans.
这就描述了推论最佳解释的主观方面：问题归根结底在于，最佳的竞争假说是否值得做出这样的认识论承诺。当然，这并没有一般的规则，它可能值得作出这样的承诺，也可能不值得。这完全取决于竞争的激烈程度，以及更好的解释在满足其他要求，成为我们在信仰中应该承诺的东西的程度。这又要归结到一般认识论问题上了，接受一个命题当然要看它能不能很好地解释已知为真的命题。但是，并不存在简单有效的推论来得出最佳解释，也不存在推论规则。可爱的属性是整个解释的属性，它与相似的属性（解释者的属性）并无关联。

The issue has another side, and it is this. If we think of ourselves as really explaining and not as conjecturing an explanation, then we are already committed to the truth of the explanation we are giving. In this case inference to the best explanation is not even an issue. (It is possible to imagine a case where there is competition in explanation by various explanations whose truth we are committed to anyway. That again is a different situation.)
这个问题还有另一面，那就是。如果我们认为自己是在真正地解释，而不是在猜测一种解释，那么我们就已经对自己所做解释的真实性做出了承诺。在这种情况下，推论最佳解释甚至都不是问题。(我们可以设想这样一种情况，即各种解释之间存在竞争，而无论如何，我们都要相信这些解释的真实性。这又是另一种情况了）。

Inference to the best explanation, when made in practice, can therefore be seen as making a judgement as to whether we can judge ourselves to be explaining the explanandum, and not just conjecturing an explanation of the explanandum. In order to make that positive step, it matters much how well the hypothesis we favour, the best hypothesis, explains. On the other hand, if we are committed to seeing ourselves as explaining the event in question, we are committed to the truth of the explanation we give. The second commitment is then contained in the first.
因此，在实践中对最佳解释的推论，可以被看作是对我们是否可以判断自己正在解释解释体，而不仅仅是猜测解释体的解释做出的判断。为了迈出这积极的一步，我们所青睐的假说，即最佳假说，其解释效果如何就非常重要了。另一方面，如果我们致力于把自己看作是在解释相关事件，我们就会致力于我们所做解释的真实性。第二个承诺就包含在第一个承诺中。

Inference to the best explanation might be two different things. This is brought out by the way I conceive of the relationship between what I have called the activity side and structural side of explanation. There are two different subjective states I distinguish, explaining and conjecturing an explanation. Subjectively there might be many cases we do not know how to categorize, whether we are explaining or just conjecturing. If it is right that to explain you need to know why , there is a sense in which inference to the best explanation is utterly trivial; if you are committed to the best explanation in the sense that you hold that it actually explains the explanandum, then you are committed to the truth of this explanation. There is
推论出最佳解释可能是两码事。我对所谓解释的活动方面和结构方面之间关系的设想就说明了这一点。我将解释和推测解释区分为两种不同的主观状态。从主观上讲，可能有很多情况我们不知道如何归类，我们到底是在解释还是只是在猜想。如果说要解释 ，就必须知道为什么 ，那么在某种意义上，推论最佳解释是完全微不足道的；如果你致力于最佳解释，即你认为它确实解释了解释体，那么你就致力于这一解释的真实性。有
another sense in which inference to the best explanation is not trivial at all, and that is when making the inference to the best explanation boils down judging that one is not just conjecturing an explanation, one is in fact explaining when putting forward the best explanation one has. Making that step is no simple inference at all, and there is no general recommendation that you always commit yourself in belief to the best or most likely among competing explanations, even the best of them might not be all that good or all that likely, they might all be conjectures, some good and some bad. There being no general rule here reflects the fact that you need not know that you know when you know, and you need not believe that you know when you know. You may believe that you know the explanation and be wrong, and you may believe you do not know the explanation and be wrong. Still the commitment you make when you start believing that you know is a very significant. If you are right, you are truly able to explain.
在另一种意义上，推论最佳解释也并非微不足道，那就是当推论最佳解释归结为判断一个人不只是在猜想一种解释，而是在提出他所拥有的最佳解释时，他实际上是在解释。做出这一步根本不是简单的推论，也没有任何一般性的建议让你总是相信相互竞争的解释中最好的或最有可能的解释，即使是最好的解释也可能并不那么好或那么有可能，它们可能都是猜想，有好的也有坏的。这里没有一般规则，这反映了一个事实：当你知道的时候，你不必知道你知道，当你知道的时候，你不必相信你知道。你可能相信自己知道解释，但却错了，你也可能相信自己不知道解释，但却错了。不过，当你开始相信自己知道时，你所做的承诺还是非常重要的。如果你是对的，你就真正能够解释。

Ben-Menahem, Y. (1990). The Inference to the Best Explanation. Erkenntnis 33: 319-344.
Ben-Menahem, Y. (1990).最佳解释的推论》。Erkenntnis 33: 319-344.

Bromberger, S. An Approach to Explanation. In R. J. Butler (ed.): Analytical Philosophy, 2nd Series. Oxford: Blackwell: 72-103.
Bromberger, S. An Approach to Explanation.In R. J. Butler (ed.)：Analytical Philosophy, 2nd Series.Oxford: Blackwell：72-103.

Elster, J. (1999). Alchemies of the Mind. Cambridge: Cambridge University Press.
Elster, J. (1999).Alchemies of the Mind.剑桥：剑桥大学出版社。

Hempel, C. (1965). Aspects of Scientific Explanation. New York: The Free Press.
Hempel, C. (1965).Aspects of Scientific Explanation.New York：The Free Press.

Humphreys, P. (2000). Review of W. Salmon's Causality and Explanation. Journal of Philosophy 97(9): 523-528.
Humphreys, P. (2000).Review of W. Salmon's Causality and Explanation.Journal of Philosophy 97(9)：523-528.

Lewis, D. (1987). Causal Explanation. In his Philosophical Papers, vol. 2. Oxford: Oxford University Press.
Lewis, D. (1987).Causal Explanation.牛津大学出版社：牛津大学出版社。

Lipton, P. (1991). Inference to the Best Explanation. London: Routledge.
Lipton, P. (1991).Inference to the Best Explanation.London：Routledge.

Salmon, W. (1984). Scientific Explanation and the Causal Structure of the World. Princeton: Princeton University Press.
Salmon, W. (1984).Scientific Explanation and the Causal Structure of the World.Princeton：普林斯顿大学出版社。

Salmon, W. (1989). Four Decades of Scientific Explanation. In Minnesota Studies in the Philosophy of Science, Vol. XIII, Minneapolis: University of Minnesota Press: 3-219.
Salmon, W. (1989).Four Decades of Scientific Explanation.In Minnesota Studies in the Philosophy of Science, Vol. XIII, Minneapolis：Minnesota Studies in the Philosophy of Science, Vol. XIII, Minneapolis: University of Minnesota Press：3-219.

Salmon, W. (1998). Causality and Explanation. New York: Oxford University Press.
Salmon, W. (1998).Causality and Explanation.New York：牛津大学出版社。

Williamson, T. (2000). Knowledge and its Limits. Oxford: Oxford University Press.
Williamson, T. (2000).Knowledge and its Limits.Oxford：牛津大学出版社。

Woodward, J. (2000). Explanation and Invariance in the Special Sciences. British Journal for the Philosophy of Science 51: 197-254.
Woodward, J. (2000).Explanation and Invariance in the Special Sciences.British Journal for the Philosophy of Science 51：197-254.