Explorations
1. The question as to the certainty of experiential statements, i.e., of sentences which "report data of direct experience," is tantamount to the problem whether, in an adequate logical (not psychological) reconstruction, the empirical basis of individual or of collective scientific knowledge can itself be considered as subject to revision.
2. That experiential statements must be unquestionably certain cannot be shown by reference to actual instances: Any sentence purporting to describe experiential data may conceivably be a lie or involve inadvertent misuse of language. Hence. experiential statements that are certain play, at best, the role of hypothetical elements in a logical reconstruction of knowledge. The assumption of such incorrigible elements is not necessary.
3. The acknowledgement of an experiential statement as true is psychologically motivated by certain experiences: but within the system of statements which express scientific knowledge or one's beliefs at a given time, they function in the manner of postulates, for which no grounds are offered.
4. Some have argued that experiential statements must be certain, or else other empirical statements could not, short of an infinite regress, even be probable. For a probabilistic conception of empirical knowledge, however, it suffices to assume that at any time some experiential statements are treated as unquestioned (though not as forever unquestionable), and that the probability of all other empirical statements is determined by reference to them.
5. The conception of experiential statements as corrigible yields a more adequate reconstruction of empirical knowledge and of scientific procedure: Empirical knowledge is concerned, not merely with descriptions of individual phenomena, but also with theoretical generalizations about them. And a conflict between a highly confirmed theory and an occasional recalcitrant experiential sentence may well be resolved by revoking the latter rather than by sacrificing the former. Science offers various examples of such procedure.
6. Doubt as to the truth of previously acknowledged experiential sentences is sometimes held to be meaningless for lack of a more fundamental standard of appraisal. But it is a matter of decision whether to grant accepted experiential statements absolute epistemic priority or whether to make any empirical assertion subject to appraisal by reference to both the experiential basis and the theoretical structure of the system that would result from including the assertion in the totality of accepted statements.
CARL G. HEMPEL
Yale University.
1. This question concerns the possibility of revising statements about the empirical basis of knowledge. The proper answer will depend on what we mean by "statements about sense-experience." Professor Hempel's Thesis 5 and Thesis 6 are true only if these statements are assumed to be formulated in a material-object language (e.g., "Smith sensed redness at 5 P.M."); Thesis 3 and Thesis 4 could be true only if they were assumed to be otherwise formulated (e.g., "Blueness here now"),
2. Because of the possibility of lies, or the misuse of language, I believe that Professor Hempel's formulation of the problem of empirical certainty must be interpreted as a convenient abbreviation, in linguistic terms, of a question about beliefs. A complete formulation of the question would have to make some reference to the speaker's beliefs as he utters an "experiential statement."
3. It is an analytic proposition that no grounds are offered (formulated) for statements which are basic. There is evidence for a basic experiential statement, and this evidence ordinarily consists of the very experiences which motivate us to acknowledge the statement as true.
3. & 4. These two theses seem to assert that experiential statements will be basic (unquestioned "postulates") in an adequate logical reconstruction of knowledge. But if experiential statements are material-object statements, there is no epistemological reason for making them basic; indeed, statements such as "Smith sensed redness at 5 P.M." would ordinarily be thought to be less probable than many non-experiential statements (e.g., non-experiential statements about the familiar objects in our immediate environment).
5. & 6. These last two theses could not be true unless experiential statements are material object statements. The recalcitrant experiential statement which we revoke because it conflicts with a highly confirmed theory, is always a statement about the past, and makes some reference to a material observer and physical time. It does not seem that we should ever be justified in revoking "Blueness here now" because it conflicts with a theory.
Is there a meaningful difference between these two interpretations of "experiential statement"? I suspect that this is the point which troubles many philosophers about the problem of empirical certainty.
RODERICK FIRTH
Swarthmore College.
1. The crux of the matter lies in the following two questions: Are there any statements (overt or covert) whose mode of causation entails that they are true of the situation in which they occur? Must there be such in order for human cognition to be justifiable?
2. I agree that the answer to the first question is No. The plausibility of the opposite answer rests on a failure to note that the necessary truth of "statements which express observations" is simply a consequence of the fact that we wouldn't say of a statement that it "expressed an observation" unless it were true.
3. Although there are no statements whose mode of causation entails that they are true, there are statements whose mode of causation makes it likely that they are true. Consider, for example, the utterance "(The ball is) out" evoked by his environment from an honest referee in the course of a game.
4. I also agree that the answer to the second question is No. In order for our symbol activities to provide us with information about the world, it is indeed necessary that there be a connection between at least one mode in which statements are caused and the truth of the statements caused. But it is sufficient that statements caused in a certain way be likely to be true of the situation in which they occur.
5. If the only way in which a statement could acquire probability were qua being a token of a type which is probable to various degrees with respect to various sets of other sentence types, then tokens of sentences belonging to one coherent system could never be better off with respect to probability than tokens of sentences belonging to any other coherent system. Once, however, we note (see 3 above) that tokens can be probable otherwise than as instances of a type which stands in probability relations, the way is clear for an understanding of the role of observation statements in human knowledge.
6. An observation statement is ipso facto not a reasoned statement, i.e., a statement which occurs as the conclusion of an argument. However, postulates connotes "choice," while the force of an observation statement lies exactly in its not being chosen, but rather wrung from us, even though the set by which we prepare to make them involves choice.
7. Empirical statements other than observation statements are justified by finding arguments the conclusions of which have the same meaning as the statements to be justified. When one of the premises is an observation statement, we usually do not demand that the process of justification be continued; not, however, because we think such statements incapable of justification. On the contrary.
8. Yet observation statements as such are not justifiable in the manner described in the previous paragraph. For smce their meaning depends on the context in which they occur (their "egocentric" character), no argument can have a conclusion which is a token of the same type as a given observation statement. To be sure, an argument can indeed be found of which the conclusion is a non-egocentric statement which corresponds in a familiar way to the observation statement, but this argument won't justify the observation statement qua observation statement. The latter purpose can be accomplished only by finding an argument of the form, "The statement in question was caused in such and such a manner, therefore it was probably true." I take it that it is the justifiability of observation statements along this line that Hempel has in mind in 3.
9. Even if there were statements whose manner of causation entailed their truth, so that they would be in one sense "incorrigible," this would by no means involve that they were not "subject to revision," nor do the philosophers who postulate them believe that they are not subject to revision. Once the statement is past (and only then can there be question of revision), we would indeed be entitled to say that if the statement was caused in a certain way, then it was true, but that it was caused in that way would be a claim no less corrigible than other historical statements
WILFRID SELLARS
University of Minnesota.
1. I believe that the most questionable of Professor Hempel's theses is the first. The traditional approach to the problem is to say that certainty and uncertainty are characteristics, not of sentences, but of beliefs. And some of Hempel's remarks suggest it may be difficult for him to ignore the epistemological issues which the traditional approach involves.
2. He points out that the sentences in which he is interested do in fact "express" beliefs and he intimates that, in deciding whether a sentence is certain, we must consider whether it is an adequate expression of someone's belief. "That experiential statements must be unquestionably certain cannot be shown by reference to actual instances: Any sentence purporting to describe experiential data may conceivably be a lie or involve inadvertent misuse of language." Presumably the difficulty with sentences which are lies, or which misuse language, is that they may express inaccurately the beliefs of the person who asserts them. But an experiential sentence may be an adequate expression of someone's belief and still not be certain. To decide whether the sentence is certain it would seem essential to consider the belief which the sentence expresses and also the experience to which the belief and therefore the sentence refer.
3. Hempel does say that "the acknowledgement of an experiential statement as true is psychologically motivated by certain experiences." But the concept of motivation would hardly be enough to enable us to explicate either the concept of empirical certainty or the concept of experience justifying (providing a basis for) a belief. It is easy to find instances of beliefs which are motivated by experience, but for which we should not want to claim either empirical certainty or justification.
4. Hempel's rejection of the traditional approach is probably based upon certain methodological principles from which he is able to infer that sentences expressing the traditional issues do not have any meaning. The traditionalist might point out, however, (a) that there is better reason for holding that some of his beliefs are certain than there is for accepting the methodological principles; and (b) that, in judging the adequacy of these methodological principles, we must ask whether they take account of the fact that some beliefs are certain. It seems quite likely, however, that controversy over these two points would lead to an impasse.
RODERICK CHISHOLM
Brown University.
1. If "empirical basis" means "a basis in experience," it is as uncertain as any other item in experience. All the theses then are true, obvious and unimportant, being merely reformulations of the fact that "experience" refers to a world of encounterable contingencies. But if it means "a basis for experience," a basis is not only not uncertain, but inescapable, since without it the experience would be impossible.
2. A basis for experience is a presupposition of experience. Uncertainty regarding the truth of meaning of an empirical assertion is consistent with the incorrigibility and unavoidability of what the assertion presupposes.
3. Presuppositions are not postulates without warrant. Nor are they dependent for their truth, meaning or being on a system of statements. To say anything at all we must presuppose at least, (a) the existence of a language, (b) the existence of a communicator, (c) the possibility of an interpreter. The statements, "there is no language," "there are no communicators," "it is impossible to know what anyone says." all reject their own presuppositions. Although logically self-consistent, they are absurd, incapable of being true. They are cases of empirical statements which are necessarily (but not logically) false; their contradictories are necessarily (but not logically) true.
4. Probability statements, like any others, make presuppositions which are not merely unquestioned but are unquestionable. Every probability statement, for example, presupposes a plurality of cases neither completely independent of nor completely dependent on one another.
5. A presupposition is not identical with a theory. It is a fact and not a statement, and when expressed in a statement, it might have the form of a singular: whereas theories are statements and always encompass a possible plurality. The conflict between presupposition and statement must always involve the abandonment of the statement or of both presupposition and statement.
PAUL WEISS
Yale University.
1. Certainty, like truth, can be construed so as to apply directly to statements (whether they are vieved or not): this avoids the psychologistic connotations of the concept of certainty of beliefs. The certainty of an accepted experiential statement has then to be conceived neither as logical necessity nor as indubitabilitv in the psychological sense, but rather as systematic exemption from revision in the light of additional evidence (To comments by Mr. Chisholm. and to Mr. Firth's point 2).
2. Experiental sentences might be expressed in a phenomenalistic or in some "material-object" (Mr. Firth's term) idiom. In the latter case, it is preferable to construe them as attributing directly observable properties or relations to physical objects ("This is a blue liquid" "The needle of this instrument coincides with the third mark on the scale"), rather than in the manner of "Smith sensed red at 5 pm"; for agreement among observers is much more readily obtained in regard to sentences of the former type hence they are better qualified to represent a common, and fairly stable, basis of intersubjective empirical knowledge.
3. When an experiential sentence is accepted "on the basis of direct experiential evidence," it is indeed not asserted arbitrarily: but to describe the evidence in question would simply mean to repeat the experiential statement itself. Hence, in the context of cognitive justification, the statement functions in the manner of a primitive sentence. This latter term avoids the misleading connotations of the word 'postulate' pointed out by Mr. Sellars, with whose comments I am in practically complete agreement (To remarks by Mr. Chisholm and Mr. Firth.).
4. However, statements of the experiential type also permit of indirect test: it involves the deduction from them, in conjunction with suitable generalizations or theories, of other experiential statements, for which direct evidence is available. The possibility of thus obtaining disconfirming indirect evidence militates against the conception of accepted experiential statements as irrevisable. This remark applies even to experiential statements in the phenomenalistic idiom; for these could not serve as a basis of any kind of empirical knowledge if they did not enter into systematic connections expressible in general theoretical principles. Therefore, 'Blueness here now,' if construed as incapable of conflict with any generalization or theory, could not function as a basic element in the systematic validation of empirical knowledge.
5. The claim advanced by Mr. Weiss that "to say anything at all we must presuppose at least, (a) the existence of a language, etc.," can be reformulated thus: The statement, "There are instances of significant use of language," implies, and thus presupposes, the statement "There is a language, and there is at least one communicator and one interpreter of that language." Now, this implication holds by virtue of the meaning of "significant use of language"; it is therefore logically necessary. But it would render the "presuppositions" in question necessary only if the statement S. "There are instances of significant use of language," were necessary. S, however, is contingent. If it be argued that the very occurrence of a token of S constitutes conclusive evidence that S is true, then it should be borne in mind that (a) such occurrence itself is contingent: (b) the qualification of any string of marks as a significant linguistic expression (rather than a mere doodle or an accidental pattern in ink) has the character of an interpretive hypothesis which cannot be conclusively verified by observational evidence.
CARL G. HEMPEL
Yale University.