The Special Reasons Requirement and Cartesian Skepticism

PHIL 462

March 13, 1997

Tetsuji Iseda

1. Introduction

The purpose of this paper is to evaluate the so-called "special reasons requirement" (Cherniak 1986) as a reply to skepticism. My analysis concentrates on the distinction of two senses of 'knowledge'. My diagnosis is that Austin's and others arguments make good points indeed, but still a version of Cartesian skepticism can survive the attack by making this distinction.

In section 2, I briefly introduce Descartes' skeptical doubt. In section 3, I summarize the special reasons requirement as presented by Austin, using other philosophers like Peirce, Grice, and Cherniak to supplement Austin's argument. Section 4 evaluates the special reasons requirement as a reply to Cartesian skepticism. In section 5, I consider Suppe's use of the special reasons requirement in the analysis of scientific methodology.

2. Cartesian skepticism

In the First Meditation, Descartes states his methodology to establish a secure foundation for knowledge as follows:

"...reason already convinces me that I must withhold assent no less carefully from what is not plainly certain and indubitable than from what is obviously false; so the discovery of some reason for doubt as regards each opinion will justify the rejection of all" (Descartes [1642] 1971, 61).

In other words, he accepts only indubitable things as knowledge. With this criterion, sensory experiences do not count as knowledge because they often deceive us. Some sensory experiences seem indubitable, but maybe he was just dreaming a quite vivid dream. Even though we can eliminate this possibility, still it is possible that there was a powerful evil demon who deceives him (61-65).

Descartes' own attempts to eliminate these skeptical doubts were not successful. After him, many philosophers have attempted many different replies to Cartesian skepticism, but none seems to enjoy the status of a decisive answer. Among them, the special reasons requirement is rather promising, and the purpose of this paper is to evaluate this reply.

3. The special reasons requirement

According to Cherniak (1986), the special reasons requirement is a requirement that those who claim to have knowledge are responsible only for "specific counterpossibilities that there is a definite basis for thinking might now apply" (100). This requirement is elaborated by J.L. Austin ([1946] 1979) as a reply to skeptical doubts, though similar arguments are used by other earlier philosophers, especially by C.S. Peirce ([1868] 1958). I summarize the requirement using Austin's paper, supplementing with other philosophers.

3-1. Austin's analysis

The main goal of Austin's "Other minds" ([1946] 1979) is to reply to skeptical doubt on the existence of other people's minds, but in the course of the argument Austin also applies the argument to Cartesian skepticism (though Austin does not mention Descartes' name).

Austin's strategy is to analyze the way we actually use words like 'know' or 'real'. When we use the word 'know', we mean something different from 'believe'. What is the difference? An immediate reply is that if I know I cannot be wrong (79). But Austin is not satisfied with that reply. For example, suppose I claim that "That is a goldfinch" and someone asks "How do you know?". Austin distinguishes several sorts of answers to the question "How do you know?" (79), but here we will concentrate on just one type of answers, namely answers by indication of some features of the situation supporting the claim.

In this type of answers, I indicate some features of the situation which enabled me to recognize the object as a goldfinch, saying something like "By its red head" (83). If the questioner wants to keep questioning further, he/she should say something like "But goldfinches don't have red heads", or "But many other birds have red heads" (83). In any case, the questioner should indicate some definite lack. Without such indications, it is silly to keep saying "That's not enough" (84). In Austin's words, "Enough is enough: it doesn't mean everything. Enough means enough to show that (within reason, and for present intents and purposes) it 'can't' be anything else, there is no room for an alternative, competing, description of it. It does not mean, for example, enough to show it isn't a stuffed goldfinch" (84). As this quotation suggests, the questioner's objection should be not only specific, but also relevant to "present intents and purposes". This is a clear application of the special reasons requirement.

Then, what should I do if I admit the raised challenge is specific enough and relevant to the present intents and purposes? Ideally I should be able to prove that the specific doubt does not apply to this case, but not necessarily so. For example, to the challenge "But many other birds have red heads", what I should do is to indicate something peculiar about the read head. But "often we know things quite well, while scarcely able at all to say 'from' what we know them, let alone what there is so very special about them" (85). In goldfinch's case, if I can recognize a goldfinch from some peculiarity of the read head, this is enough to say that I know that it is a goldfinch, even though I cannot tell what is peculiar about it.

In the above argument Austin dismissed the possibility of a stuffed goldfinch as irrelevant, but if the questioner proceeds to ask "How do you know it's a real goldfinch?", this makes that possibility relevant. Thus Austin analyzes this kind of question about reality next (86-89). When we ask "But is it a real one?", we have specific counterpossibilities in mind. Thus "the goldfinch might be stuffed but there's no suggestion that it's a mirage, the oasis might be a mirage but there's no suggestion it might be stuffed" (87). Furthermore, these counterpossibilities are associated with specific recognized ways of distinguishing (87). Therefore, even when we ask questions about reality, our questions are governed by the special reasons requirement.

Of course, after we establish that I know that it is a real goldfinch through these challenges and replies, it may turn out that it is not a goldfinch after all -- for example, it may suddenly explode (88). But Austin thinks that in some sense we can still say 'If you know then you can not be wrong'. This usage of 'wrong' is best illustrated by comparing the expressions "I know" and "I promise" (98). When I say "I promise", I bound myself to others, and stakes my reputation (99). Similarly, when I say "I know", "I give others my words: I give others my authority for saying that 'S is P'" (99; italics original). Therefore, if I am mistaken, I am responsible for getting others into trouble (100). These authority and responsibility come from the fact that "I know" implies "I can't be wrong" (101). From these consideration, words like 'know' and 'wrong' in these contexts should be understood performative rather than descriptive (103); that is, when I say "I know", I am not describing my state, but I am performing an act (giving authority).

To sum up, the word 'know' is a performative word to give a kind of authority to others, and related words like 'real' or 'wrong' should be understood accordingly. This means that, when I say "I know", I am responsible only for specific doubts which are relevant to present purposes and accompanied with established procedure to deal with them. People never suggest the possibility that I never know it.

3-2. Other arguments from the special reasons requirement

Austin's analysis is, after all, an analysis of ordinary language, the way we actually use the language. The result of the analysis suggests that we are never Cartesian skeptics in our ordinary conversations. Is this relevant as a reply to Cartesian skepticism? Maybe Cartesian skeptics argue that they are proposing a new way of talking for deeper understanding or something, and therefore the way people actually use words does not matter. Thus, to make Austin's analysis relevant as a reply to the skepticism, we have to find reasons why we should stick to the ordinary usages of "know" and other words.

H. P. Grice's (1975) analysis of the "Cooperative Principle" provides a clue. According to Grice, our ordinary conversations are heavily dependent on "implicatures", which do not follow from what is said by formal logic but are implied by the words (65-66). For example, when A and B are talking about their common friend C, if A says "he is well, he has not been to prison yet", B takes this as implying (Grice introduces a word "implicating" for this specific sense of "implying") that C did something wrong, even though, logically speaking, this conclusion does not follow from what A said (65).

To understand why implicatures are possible, we should assume a principle which people are supposed to obey in their conversations, namely "Make your conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged" (67). This principle is called the "Cooperative Principle", and Grice distinguishes several maxims under this principle. For example, the above case is understood if we assume a maxim "Be relevant" (67) is working. A's statement is apparently irrelevant if there is no real possibility that C is arrested. So if we can assume that A is obeying the maxim "Be relevant", we can conclude that A is implicating such a danger.

Among the axioms, Grice enumerates several axioms related to the above Austin's analysis: first, "do not say what you believe to be false"; second, "do not say that for which you lack adequate evidence" (67). If these maxims are at work, we can easily understand why saying "I know" causes a responsibility described by Austin. When I say "I know", I am implicating that I believe that it is true, and I have adequate evidence for it. Therefore people blame me of giving a wrong implicature, when I turn out to be mistaken. Another maxim, "be relevant", applies to Austin's questioner. Because of this, a questioner cannot keep questioning by mere possibility.

So far, this is an analysis of the way people actually obey the cooperative principle. Grice proceeds one step more, however, to say that the cooperative principle can be thought of a "quasi-contractual matter", and it is "something which it is reasonable for us to follow, which we should not abandon" (68). This normative aspect of the cooperative principle comes from the fact that when we engage in a conversation the participants usually have some common interests, and that the cooperative principle is arranged to serve the common interests.

Can we derive an argument against Cartesian skepticism from this analysis? Cherniak (1986) provides a good line of argument. If we start to doubt everything like Descartes, it is quite difficult to obey the maxims. Without these maxims, "prime function of language ... would be vitiated", i.e. meaningful conversations would be almost impossible (Cherniak 1986, 104). This situation is more critical when we need a division of cognitive labor, namely when we should rely on other people's knowledge in some essential part of our life (105). In such a division of labor, language is the way we share each other's knowledge, and if language does not function, everyone in the society get a serious trouble. Therefore, Cartesian skepticism is not reasonable for anyone in the society. (I will critically analyze this line of argument later.)

C.S. Peirce uses the special reasons requirement in a different way. According to him, as a matter of psychological fact, "we cannot begin with complete doubt" ([1868] 1958, 40). Namely, we have many prejudices which we are not aware of, and therefore cannot put into doubt. Famous Neurath's metaphor of ship illustrates Peirce's point well: "we are like sailors who must rebuild their ship on the open sea, never able to dismantle it in dry-dock and to reconstruct it" (Neurath [1933] 1959, 201). That is, when we revise a part of our beliefs, we cannot revise everything at a time, but should assume some part of our belief system as given. It is true that Descartes said that he doubted everything, but "the mere putting of a proposition into the interrogative form does not stimulate the mind to any struggle after belief" (Peirce [1877] 1958, 100). We need a "real and living doubt" to really question and think about something (Peirce [1877] 1958, 100). Peirce also states the same point as follows: "a person may ... find reason to doubt what he began by believing; but in that case he doubts because he has a positive reason for it, and not on account of the Cartesian maxim" (Peirce [1868] 1958, 40). This is almost the same as the specific doubt Austin was talking about. Therefore, if Peirce is right, then the word usages analyzed by Austin are not accidental, but of psychological necessity. This is a quite strong argument against Cartesian skepticism (I will analyze this line of argument later also).

Cherniak (1986) introduces an interesting amendment to the special reasons requirement. Especially when people take psychological view of the special reasons requirement like Peirce, an interesting paradox occurs: "the more careful the agent is in thinking of couterpossibilities, the less he is entitled to claim to know" (Cherniak 1986, 110). For, if the person is more careful, she might find more good specific reasons to doubt her knowledge claim. She is not entitled to the knowledge claim before eliminating all these specific doubts. On the other hand, a less careful person is psychologically unable to find those specific reasons, and thus entitled to the knowledge claim without eliminating them. To avoid this absurd conclusion, Cherniak introduces a requirement called an "'input' requirement": "'The agent must collect, and remember, enough available information" (114). What is enough is decided by the community relative to the content of the knowledge claim the agent makes. If a counterpossibility is serious, the agent is required to consider it regardless of her background knowledge. Careless people are no longer entitled to claim to know before meeting the input requirement, thus the paradox is avoided.

To sum up, in this section we considered two ways to use the special reasons requirement to reply to Cartesian skepticism. One is Grice's cooperative principle, supplemented by Cherniak's consideration of the division of cognitive labor. From this point of view, Cartesian skepticism should be avoided because it jeopardizes our linguistic community. The other is Peirce's psychological argument. From this point of view, Cartesian skepticism is psychologically impossible. Especially the latter view implies a paradox, and Cherniak's input requirement helps to eliminate the paradox and to make the special reasons requirement more plausible.

4. Distinction between two senses of knowledge

Now, the question is how should we evaluate the arguments from the special reasons requirement. My diagnosis is that these arguments make a good point indeed, but still a version of Cartesian skepticism can survive the attack. To make this case, I will introduce a distinction between two senses of knowledge. The distinction is inspired by Suppe's (forthcoming) distinction between knowledge and knowledge claim, though I arrange the distinction suitable for the present purpose.

The 'knowledge' Descartes is worrying about seems to be something different from the ' knowledge' in Austin's sense. Descartes starts the Meditation as follows: "some years ago now I observe the multitude of errors that I had accepted as true in my earliest years, and the dubiousness of the whole superstructure I had since then reared on them" (Descartes [1642] 1971, 61). This is why he wants to make a clean sweep and to begin again from the very foundations to "establish some secure and lasting result in science" (61). Thus what he wants is to avoid errors which can make his whole intellectual work to establish a superstructure a waste of time. This sounds a good enough motivation.

With this motivation, Austin's own ordinary language analysis is not quite relevant to Cartesian skepticism, because the 'knowledge' Austin is talking about can be wrong. In other words, Descartes is not interested in so-called 'knowledge', but something which does not make his intellectual efforts a waste of time, whatever is the name of it. Austin's claim that the word 'wrong' should be understood in a non-descriptive way is also irrelevant, because whether or not it is called 'wrong' does not matter for Descartes, while whether or not it requires him to rework the whole structure built on it does. So Descartes could even give up words like 'knowledge' or 'wrong' and introduce a totally new terminology. I would like to propose to use terms like 'C-knowledge', 'C-know', 'C-wrong' and 'C-real' to denote what Descartes wants to call 'knowledge', 'know', 'wrong', and 'real' ('C' stands for 'Cartesian').

Take the case of a goldfinch for illustration. For ordinary purposes, Austin's requirement is good enough to establish that I know that it is a goldfinch. But now Descartes is going to spend next several decades in establishing an elaborate ornithological theory based solely on this observation. Just imagine that the goldfinch suddenly explodes after fifty years of study; then, it was not a goldfinch after all, and the fifty years of study was all a waste of time! Descartes' motivation is to avoid such a consequence, and this is why Descartes wants to C-know that it is a C-real goldfinch.

Now, the question is if Grice's and Peirce's replies provide reasons not to introduce such a new terminology. First, Grice's own argument does not help very much. Even if the cooperative principle is a "quasi-contractual matter", there is no reason we cannot change the (quasi-) contract. Descartes is introducing a new kind of motivation into conversations, and as far as there are other people who share the same worry, Descartes and those people can talk to each other meaningfully under the new contract. This new contract does not disturb ordinary people if Descartes keep apart C-knowledge talk from knowledge talk. Also, as far as we keep the distinction, Cartesian skepticism does not threat Cherniak's division of cognitive labor. The division of cognitive labor is concerned with knowledge (as opposed to C-knowledge), and this is totally different practice from Descartes' pursuit of C-knowledge. So Grice's line of argument does not help to prohibit Cartesian skepticism.

A real challenge to the distinction between C-knowledge and knowledge comes from Peirce's argument. If Peirce is right, Cartesian skepticism is a pseudo-problematic, and the enterprise is impossible to pursue, for, as a matter of a psychological fact, we can not doubt everything, and when we doubt, we need sufficient reasons.

I think that this argument is basically sound, but this does not make Cartesian skepticism meaningless. Rather, we can reinterpret Cartesian skepticism to meet Peirce's challenge. Since this is not Descartes' original position, I would like to call this position 'Non-skeptical Cartesian' (admittedly a little paradoxical name). Non-skeptical Cartesians are different from Descartes in that they do not require people to doubt everything; they just require people to reassess each belief from Cartesian point of view, namely with the criterion that we should not accept "what is not plainly certain and indubitable" (Descartes [1642] 1971, 61). The point of reassessment is not whether or not a belief is true, but whether or not we can found our intellectual structure on the belief. These are different kinds of questions, and it is plausible that sometimes we refrain from making a belief a foundation of such a structure, even when we have no doubt about the belief. Under this interpretation, we can redefine C-knowledge as beliefs which are qualified as foundations for further intellectual effort, regardless of our psychological states about them, i.e. whether or not we can doubt them. I think this is a way to accommodate Descartes' original motivation to Peirce's psychological argument, and this is why I call the position described here 'Cartesian' even though it is not skeptical.

Using Neurath's boat metaphor, my Non-skeptical Cartesians are doing something like this: their purpose is to built a huge tower on the boat, which can be even bigger than the boat. Apparently they hate to keep rebuilding the tower every time people reconstruct the boat. So they look for a part of the boat which people will never reconstruct, and build the tower when they find such a place.

At a first glance, Non-skeptical Cartesians may seem harmless. But the radical assessment they require is destructive enough. A large part of our scientific knowledge is built upon our common sense knowledge. In this respect the relationship between scientific knowledge and common sense knowledge can be seen as the tower and the boat in the previous paragraph. Then, if we cannot find any secure place on the boat, we cannot built towers, according to Non-skeptical Cartesians. This means that we cannot built any further scientific knowledge on secure foundations. Of course, as a matter of fact, people build such knowledge using the special reasons requirement. But this is irrelevant if my evaluations on Austin's and Grice's argument is right. Non-skeptic Cartesians are playing a different kind of game, not talking about knowledge but about C-knowledge.

Let us now turn to possible objections to Non-skeptical Cartesians. A possible reply from Peirce is that even this kind of reassessment is psychologically impossible, but he needs further support for such a claim, which I don't think he has. A more serious reply is that we do not have to play Cartesian games, whether skeptical or non-skeptical. Indeed, if everyone is happy with the special reasons requirement and knowledge produced by it, why should we care about the absolute certainty Descartes wants? I think that this question can be answered partly by Cherniak's input requirement. As we saw above, the input requirement says that the community decides which counterpossibilities should be considered in each case, considering the seriousness of the results. Now, skeptic and non-skeptic Cartesians recognize the seriousness of founding our knowledge system on a false belief. For them the seriousness of such false foundations is so extremely grave that they cannot help adopting an extremely strong input requirement, i.e. the absolute certainty. Of course we can object that founding a belief system on a false belief cannot be that serious mistake, but if the relevant community (maybe the community of epistemologists?) accepts such a high standard, why should we try to change their mind? At least Cherniak's argument for the input requirement does not require rationality of the community on this matter (though it requires rationality of individuals ruled by the input requirement). I think this argument is enough to secure a legitimate place for Cartesians, but still there is a question if those in the larger society should take the community of Cartesians seriously. Does the Cartesian enterprise have any bearing on ordinary knowledge process? To answer this question I would like to look at Suppe's argument in the next section, but before that, let me summarize my argument so far.

My analysis ends up with revising Cartesian skepticism in several points, but I think that the revised version retains much of the original thrust of Cartesian skepticism. My revisions are (1) introducing C-knowledge talk to make ordinary language analysis irrelevant, (2) introducing Non-skeptical Cartesian position to meet the challenge from psychological impossibility and (3) interpreting the absolute certainty requirement as an extremely strong input requirement to secure a place for Cartesians in the society. In this revised version, Cartesians no longer talk about 'knowledge' but talk about 'C-knowledge', do not doubt everything but reassess everything to find a secure foundation for knowledge. Except for these points, almost all arguments used by traditional Cartesian skepticism are valid in this revised version. Thus I think that Cartesian skepticism has survived the attack from the special reasons requirement.

5. Suppe's epistémeology

Suppe (1993, forthcoming) advocates 'epistémeology', which studies epistéme, i.e. "amalgamation of knowing and claiming to know" (Suppe forthcoming, Volume 2 chapter I section 4). Suppe's claim is that the failures of many epistemological efforts come from the failures "to keep distinct these two components of epistéme" (ibid.). Suppe's argument provides us with an interesting relation between Cartesian skepticism and the special reasons requirement.

It is impossible to get into the details of Suppe's highly elaborate epistemological arguments in this paper, so let us just look at the general feature of his theory. First, Suppe analyzes knowledge as follows (Suppe 1993, 158; more developed version can be found in Suppe forthcoming Volume 1, chapter X):

(1) S knows that P if and only if

i. S is in an experiential state R;

ii. S believes that P, and being in the experimental state R is a cause of that belief;

iii. The circumstances are such that there is a C descriptive of actual circumstances (including both the presence and absence of various factors) such that

~ <c> (C & R & ~P) & <c> (C & ~P) & <c> R

is true (where <c> stands for 'causally possible').

There are several interesting features with this analysis of knowledge. First, this analysis says nothing about justification of knowledge claims. Thus the special reasons requirement is irrelevant to this notion of knowledge. On the other hand, a part of the condition, ~ <c> (C & R & ~P), says that if the experiential state R and the circumstance C hold, then P is causally necessarily true. Other parts of the condition says that R and C do hold, thus if S knows P in this sense P cannot turn out to be wrong. This feature is what Cartesians (at least Non-skeptical Cartesians) were looking for, so if we can find out which beliefs are knowledge in Suppe's sense, this is a satisfactory goal for Cartesians.

So far on Suppe's analysis of knowledge. Suppe's account of justification of knowledge claims is totally different from it. This process is governed by the special reasons requirement: "knowledge claims prima facie are to be accepted unless one can raise specific doubts about the claim and give reasons for such specific doubts" (Suppe 1993, 162). Like Cherniak's input requirement (actually Suppe cites Cherniak 1986 here), relevance of each counterpossibility is "social context dependent" (Suppe 1993, 163). This justification process is called "credentialing" (153). Naturally, the credentialing process "is, for the most part, distinct from the evidential basis" for the knowledge, i.e. from the experiential state R in (1) (168).

Thus Suppe wants both causal necessity of the truth of knowledge and credentialing of knowledge claims by the special reasons requirement. Without the first part, the quality of the knowledge in the community is problematic because assumptions used for justificatory arguments can be false (Suppe1993, 169). Without the second part, we cannot explain how knowledge can be passed on to other members of the community (169). So, for example, Suppe's analysis of scientific observation is as follows (161):

(2) Relative to research group or discipline G, S observes that P only if

i. S satisfies condition (1) with respect to P and an experiential state R;

ii. P satisfies the relevance conditions associated with G, its domain, and its background;

iii. S's claim that P was produced in accordance with the replicability standards imposed by G;

iv. S's claim that P is transmitted in an observation report that provides arguments addressing the actual or potential specific doubts about P legitimized by G, and those arguments meet the adequacy standards for such arguments in effect in G.

As you can see, the condition i embodies knowledge part, and the condition iv embodies the special reasons requirement. I think that Suppe's position is quite close to what I called Non-skeptical Cartesian position in that he worries about establishing knowledge claim on false assumptions and that he is careful in distinguishing knowledge and knowledge claims (in my terminology, C-knowledge and knowledge, respectively). Nevertheless Suppe is more interested in analyzing credentialing process than in finding out which beliefs are knowledge according to his definition. In this sense Suppe is not a Cartesian.

Thus far, Suppe keeps apart knowledge from knowledge claims, but he seems to think that they are not totally different: "only in extremely advanced and sophisticated cases of sciences is there much likelihood that the justification of the knowledge claim will articulate the regularities in virtue of which clause 1iii is satisfied. Even there, typically only selected aspects of clause 1iii are covered" (Suppe 1993, 168-169). Even though Suppe talks in a negative manner here, he admits that there is some relationship between advancement and sophistication of science and refinement of justification procedure in the direction of his analysis of knowledge.

Suppe himself does not develop this idea, but I would like to use it as a reply to the question I suspended in the previous section, namely if the Cartesian enterprise has any bearing on ordinary knowledge process. The more a scientific field is sophisticated, the more specific doubts each knowledge claim should eliminate. The attainment of infallible status of justified knowledge claims can be seen as an ideal limit of the process of this sophistication. Now the apparent question is how we can elaborate credentialing process in each field. Non-skeptic Cartesians provide a starting point for such a purpose. For example, a philosopher like Suppe may come up with a theory of knowledge driven by Cartesian worry. Suppe's elaborate theory may help to find what is wrong with present credentialing process. By a refinement through this consideration, the credentialing process proceeds toward the ideal limit. As far as the rest of the society can make use of Cartesian doubt in this way, it is good for the society to let philosophers adopt such an extreme standard for knowledge.

In this section I looked at Suppe's view on knowledge and knowledge claims. His epistémeology is an interesting mixture of Cartesian worry and an interest in credentialing process like Austin. I found that Suppe provides an answer of how Non-skeptical Cartesians can influence on the credentialing process of ordinary people.

6. Concluding remarks

In this paper, first I summarized Descartes' original skeptical doubt. Then I looked at Austin's argument from the special reasons requirement, together with supplementary arguments by Grice, Peirce, and Cherniak. Next I evaluated the arguments, and concluded that a revised version of Cartesian skepticism can survive the attack. At least the society has a reason to allow a community of philosophers to pursue the Cartesian enterprise.

A final remark. Originally Cartesian skepticism looked as if something everyone should worry. But Austin's analysis revealed that for the majority of the society there is no reason to worry about such a thing. I think that this is a genuine contribution of Austin. Nevertheless, I also think that the argument is not strong enough (even after supplemented by other philosopher's arguments) to convert philosophers who have already started to worry about the Cartesian doubt. Since I take Austin's purpose as showing why Cartesians are wrong, I should conclude that he failed in this enterprise. References (the original publication years are indicated by []; e.g. Austin ([1946] 1979) means that 1946 is the original publication year of the paper.)

Austin, J. L. ([1946] 1979) "Other minds", reprinted in J. Urmson and G. Warnock eds., Philosophical Papers. New York: Oxford University Press.

Cherniak, C. (1986) Minimal Rationality. Cambridge: The MIT Press.

Descartes, R. ([1642] 1971) Meditations on First Philosophy, reprinted in E. Anscombe and P.T. Geach trans. and eds., Descartes: Philosophical Writings, Englewood Cliffs, NJ: Prentice Hall.

Grice, H. P. (1975) "Logic and Conversation", in D. Davidson and G. Harman eds., The Logic of Grammer. Encino, Calif.: Dickenson.

Neurath, O. ([1933] 1959) "Protocol sentences", in A. J. Ayer ed. Logical Positivism. New York: Free Press.

Peirce, C. S. ([1868] 1958) "Some consequences of four incapacities", reprinted in P. P. Wiener ed., Charles S. Peirce: Selected Writings (Values in a Universe of Chance). New York: Dover Publications.

---. ([1877] 1958) "The fixation of belief", reprinted in P. P. Wiener ed., Charles S. Peirce: Selected Writings (Values in a Universe of Chance). New York: Dover Publications.

Suppe, F. (1993) "Credentialing Scientific Claims", Perspectives on Science 1, 153-203.

---. (forthcoming) Facts, Theories, and Scientific Observation