1 For instance, Goldman, Alvin ‘A Causal Theory of Knowing,’ The Journal of Philosophy 64 (1967) 357–72CrossRefGoogle Scholar; ‘Innate Knowledge,’ in Stich, Stephen ed., Innate Ideas (Berkeley and Los Angeles: University of California Press 1975) 111–20Google Scholar; ‘Discrimination and Perceptual Knowledge,’ The Journal of Philosophy 73 (1976) 771–91CrossRefGoogle Scholar; ‘What Is Justified Belief?,’ in Pappas, G. ed., Justification and Knowledge (Dordrecht: Reidel 1979)CrossRefGoogle Scholar; Epistemology and Cognition (Cambridge, MA: Harvard University Press 1986)Google Scholar; Harman, Gilbert Thought (Princeton: Princeton University Press 1973)Google Scholar; and Kitcher, Philip The Nature of Mathematical Knowledge (Oxford: Oxford University Press 1983)Google Scholar.

2 These issues are related to questions concerning a priori knowledge. I believe that the interesting problem regarding this form of knowledge is not what its precise characterization may be, but rather whether any of us actually has such knowledge. ‘A priori’ and ‘a posteriori’ have been traditionally used as adverbs which refer to types of justifications for our beliefs, for instance by Kant. (In the Introduction to the Critique of Pure Reason, trans. Norman Kemp Smith [New York: St. Martin’s Press 1929], at B2, Kant characterizes a priori knowledge as knowledge which is independent of experience. In my ‘Kant and Innatism,’ Pacific Philosophical Quarterly 68 [1987] 285-305, I argue that the notion of independence of experience in Kant must be interpreted as justificatory independence.) In the recent analytic tradition, ‘a priori’ and ‘a posteriori’ have been used to refer to types of methods or processes for arriving at beliefs. That a proposition could be known a priori might be accepted even by anti-a priorists. Naturalized epistemologists who follow some of Quine’s strictures might accept the idea that an ideal rational observer or God could have a priori knowledge, but conclude that we do not have any after considering the conditions under which we actually come to believe or accept sentences. However, if it turned out that some people do have a priori knowledge, then it would turn out that some people have infallible or non-defeasible knowledge. This is of central interest in determining the status of our knowledge of mathematics and logic, which have been traditionally regarded as paradigmatic cases of a priori knowledge. Consideration of the problems involved in the attribution of a priori justification to subjects shows that in the absence of theories of the sort given by traditional philosophers-namely theories which provide principles for inferring from the objective character of the justification of a truth to the way it can or must be known-difficulties in showing that there is any a priori knowledge are great, if not insurmountable.

3 In my ‘Frege and Kant on A Priori Knowledge,’ Synthese 77 (1988) 285-319, I argue that Frege is concerned with what I here call ‘objective justification.’ This view of Frege goes against Philip Kitcher’s interpretation. According to Kitcher, Frege’s approach to knowledge is similar to Kitcher’s own ‘psychologistic’ approach. See Kitcher, Philip ‘Frege’s Epistemology,’ The Philosophical Review 88 (1979) 235–62CrossRefGoogle Scholar, especially section II; and The Nature of Mathematical Knowledge, Chapter 1.

4 How do we assess whether a proposition is objectively justified relative to other propositions? Dreaming persistently that something will happen does not justify relative to any evidence or good theory the conclusion that it will happen. On the other hand, having those dreams may justify the conclusion that one is obsessed. This conclusion may be taken to follow from psychological theories about the relationships between the contents of our dreams and our neuroses or obsessions. The plausibility of these theories in turn may be taken to follow from clinical evidence. Theories and (sometimes implicit) normative inductive principles guide our assessments of whether propositions are related by the justification relation or not.

5 This point is the epistemological analogue of Donald Davidson’s point regarding action: ‘... a person can have a reason for an action, and perform the action, and yet this reason not be the reason why he did it. Central to the relation between a reason and an action it explains is the idea that the agent performed the action because he had the reason,’ ‘Actions, Reasons, and Causes,’ in Davidson, Donald Essays on Actions and Events (Oxford: Oxford University Press 1980), 9.Google Scholar

6 The theses of the underdetermination and indeterminacy of the mental, as articulated by Quine and Davidson, lead to these results. (See, for instance, Quine, W. V. Word and Object [Cambridge, MA: MIT Press 1960], Chapter 2Google Scholar; and Davidson, Donald Inquiries into Truth and Interpretation [Oxford: Clarendon Press 1984]Google Scholar, Essays 2, 4, 9-11, 13, and 14.) However, my arguments here for the underdetermination of attributions of subjective justification, and, in the next section, for there being difficulties in principle in the very idea of subjective justification, are not merely extensions or applications of the views of Quine and Davidson. I give independent arguments for my theses, and such arguments do not depend on claims about the underdetermination or indeterminacy of mental states other than ‘actually making the right connections’ (or ‘actually applying a rule,’ and the like).

7 An approximation to a reconstruction of others’ (and our own) justifications may be attainable through interpretations in Davidson’s sense. In trying to make sense of people’s speech, beliefs and actions we typically project crude or elaborate, implicit or explicit interpretations on what we observe in others and ourselves. Phenomena like basing beliefs on some other beliefs may also be explainable and made intelligible by interpretations which take into account the believer’s abilities, goals, inclinations, background knowledge, psychological and other circumstances. These reconstructions can be good approximations to the subjective justification particular individuals have at a time. However, only when we have projected enough theory on large portions of the speech and behaviour of others and ourselves can we have an approximate answer to the question whether a particular individual justifies her belief by appealing to experience or not, or whether she is justified at all.

8 Mind 4 (1895) 278-80. In the discussion of Carroll’s article I have been inspired by Barry Stroud’s ‘Inference, Belief, and Understanding,’ Mind 88 (1979) 179-96.

9 Anil Gupta suggested to me a similar version of this point: the possession of the rule would be identified with a brain state; and the fact that the rule is operative at a certain time would be explained by an appropriate causal connection between that brain state and a certain behaviour. I prefer to put the point in terms of the state of understanding the rule; however, the notion of understanding does not play an essential role in my discussion and I could have presented my arguments in terms of brain states. To see that this proposal runs into the same difficulties discussed below, one can translate my arguments into talk of identification of the relevant brain states. That is, the brain state that corresponds to the possession or understanding of the rule might consist in a cognitive representation of the rule or it might be a disposition. In the first alternative the difficulties discussed below in relation to identifying the understanding of the rule with a cognitive content arise again; hence, the regress is set in motion. In the second alternative, we face the problems also discussed below of the explanation in terms of dispositions: has the disposition to draw certain conclusions on the basis of certain premises been instilled by the rule or by reliance on authority, or the like? The identification of the brain state that corresponds to the disposition to follow the rule-as opposed to the brain state that corresponds to the disposition to follow an authority, say—stil3 eludes us.

10 Another problem with this proposal is that, even assuming that we have a way of identifying the state of understanding the rule, there are difficulties in identifying the ‘appropriate causal connection’ between that state and a certain behaviour. The causal connection might be deviant and include in the causal chain a state that consists in appealing to an authority, and so forth.

11 Saul Kripke, in his discussion of Wittgensten’s views on following a rule, focuses on these limitations: see Kripke, S. Wittgenstein on Rules and Private Language: An Elementary Exposition (Cambridge, MA: Harvard University Press 1982)Google Scholar. I suspect that Wittgenstein also had in mind other problems of the sort discussed here.

12 Philip Kitcher in The Nature of Mathematical Knowledge offers what he calls a ‘psychologists’ epistemology for mathematics. This approach focuses on processes that are supposed to explain ‘in the appropriate way’ the presence of a belief in a subject. For mathematics this epistemology becomes an ‘evolutionary theory of mathematical knowledge’ (which includes a certain interpretation of the history of mathematics), and it is used by Kitcher in his attempt to show that mathematical knowledge is not a priori. According to Kitcher, mathematics is grounded in an historical chain of authorities and ultimately in perception. On this view, the perceptual, historical and social processes that produce our mathematical knowledge are ‘warrants,’ that is, processes which produce belief ‘in the right way.’ Moreover, in order to account for sophisticated mathematical knowledge, Kitcher abandons the consideration of the actual psychological and socio-historical processes that lead to the acquisition of mathematical beliefs in actual subjects. For the processes of perceiving and manipulating reality performed by actual individuals can only account for the rudimentary mathematical beliefs characteristic of the origins of mathematical knowledge. Hence, Kitcher introduces the notion of ideal operations performed by ideal subjects: ‘...I construe arithmetic as an idealizing theory: the relation between arithmetic and the actual operations of human agents parallels that between the laws of ideal gases and the actual gases which exist in our world’ (109). I believe that by introducing the ideal operations of ideal subjects, Kitcher moves towards what I have called ‘objective justification’ and hence abandons a purely ‘psychologistic’ or naturalistic approach to the epistemology of mathematics: see Charles Parsons’ review of Kitcher’s book in The Philosophical Review 95 (1986) 129-37, esp. 133-4.

13 Most contemporary theorists of knowledge seem to ignore this point. For instance, Nozick, Robert in Philosophical Explanations (Cambridge, MA: Harvard University Press 1981Google Scholar) gives necessary and sufficient conditions for knowledge and then proceeds to attempt to answer epistemological skepticism without discussing the problems involved in attributing to us the fulfillment of his conditions for knowledge.

14 I wish to thank Romane Clark, David Copp, Michael Friedman, Anil Gupta, and anonymous referees for their very valuable comments.