[1] Meta-epistemology concerns (i) the concepts employed in epistemology, concepts such as justification, rationality, and knowledge and (ii) the methods and criteria employed in determining how exactly we are to apply such concepts. Substantive epistemology involves using the aforementioned concepts to determine the conditions under which we have knowledge (or justified belief) or this or that sort, and what knowledge (or justified belief) it is that we actually possess.

[2] shall be restricting my discussion to foundationalism as a theory of the structure of justified belief. If knowledge is justified true belief (plus something else), then foundationalism may be spelled out in terms of the “justification” component alone. Moreover, even if one thinks that justification is not necessary for knowledge, then there is still the question of whether foundationalism is the appropriate or correct way to think of the structure of justified beliefs.

[3] Keith Lehrer 1974, Frederick Will 1974, Roderick Chisholm 1966, 1977 all construe the regress problem in terms of showing justification, whereas Alston 1989c (p. 28-32) and Audi 1993 (pp. 118-125) are acutely aware of the two ways to construe the regress problem: in terms of showing justification or being justified, though Nathan 1980 (p.99) suggests - incorrectly in my opinion - that the regress problem is generated only on the showing version (see footnote no. 4).

[4]Notice here that I am not taking the infinite regress problem to be one of completing an infinite sequence of some activity of showing in finite time (clearly impossible), but rather as an argument directed at the contingent (and perhaps logically necessary) truth that no human noetic structure is such that it has an infinite set of beliefs. The problematics of an infinite regress of showings only arises on the dialectical version of the infinite regress problem.

[5] Evidential probability is contrasted with statistical and physical probabilities. The latter concern events rather than propositions or statements (though statistical and physical probabilities may be included as evidence in evidential probability judgements). Physical probabilities measure the extent to which some event or physical state of affairs is predetermined in its causes. Statistical probability is about proportions in actual classes and individuals as members of those classes.

[6] This argument is developed by Paul Moser 1989 (p. 61).

[7] (3*) differs from (3**) since in (3*) I am thinking of a switch from linear to holistic coherence, whereas (3**) is pure holistic coherence.

[8] The argument from epistemic isolation is developed by Moser 1989 (pp.  176-183), Audi 1988 (p. 91), and Plantinga 1993c (pp. 81-82, 110-111).

[9] Throughout the thesis I will speak of foundational beliefs as the termini of justificatory paths or chains. One could also think of foundational beliefs as the starting-points of such paths and the last mediately justified belief of some sequence as the terminus of that path. I don’t think much actually hangs on this. As a point of some pedagogical value, it does strike me that by beginning with a (putatively) mediately justified belief one actually presents an argument for foundationalism (as the present account demonstrates) rather than tacitly assumes it. Moreover, since an immediately justified belief does not entail a foundational belief (and so does not entail foundationalism), I think it is better to work from the top down.

 

[10] On a foundationalist view, not every belief need be either wholly immediately or wholly mediately justified. Foundationalism need not exclude cases of partial immediate (or mediate) justification, where immediate and mediate justifiers are severally necessary and jointly sufficient for justification. If justification is necessary for knowledge, then there will also be three types of knowledge: immediate, mediate, and partly immediate/mediate.

[11] Foundationalism entails immediately justified beliefs, but the converse does not hold. My account at this point assumes that an immediately justified belief is in itself only potentially a foundational belief. It is only actually foundational if there is at least one other belief based on it. Secondly, I state that mediately justified beliefs will terminate in “at least” some immediately justified beliefs because I want to leave open the possibility that an individual might have, among the set of his immediately justified beliefs, a subset of immediately justified beliefs that are not foundational since no other belief is based on them.

[12] This requirement is not as strong as it might first appear. “The connecting belief requirement is not the requirement that the principle of inference governing the connection be a premise; it is rather that there be a kind of cognitive appreciation of the relation between the premise and what it grounds that is necessary for the justificatory success of the relation” (Audi 1993, p. 21). On the relevance of occurrent and nonoccurrent basing relations to occurrent and nonoccurrent knowledge, see Moser 1989 (p. 158).

[13] See Audi 1993 (pp. 49-71) for an account of various forms of psychological foundationalism.

[14] Some philosophers have been foundationalist with respect to knowledge (Descartes) and others with respect to rational or justified belief (Locke). Moreover, among classical foundationalists (ancient and modern) not all have taken the foundationalist structure of knowledge (or rational belief) to be a view about the de facto structure of the justified beliefs of particular persons. Some foundationalists think of foundationalism as about the structure of “scientific knowledge” or “human knowledge” as a collective whole, rather than a statement about the structure of some individual cognitive system. Similarly, it has sometimes been thought to be a view about how a structure of knowledge can be built up (as in the Cartesian project), rather than a view concerning how it is in fact acquired.

[15] “Thus Descartes, along with many other seventeenth- and eighteenth-century philosophers, took it that any knowledge worthy of the name would be based on cognitions the truth of which is guaranteed (infallible), that were maximally stable, immune from ever being shown to be mistaken (incorrigible), and concerning which no reasonable doubt could be raised (indubitability). Hence the search in the Meditations for a divine guarantee for our faculty of rational intuition” (Alston 1992, p. 146).

[16] As Alston puts in “Two Types of Foundationalism” (in Alston 1989c): “It appears that the foundationalist is committed to adopting beliefs in the absence of any reasons for regarding them as acceptable. And this would appear to be the sheerest dogmatism. It is the aversion to dogmatism, to the apparent arbitrariness of putative foundations, that leads many philosophers to embrace some form of coherence or contextualist theory, in which no belief is deemed acceptable unless backed by sound reasons” (p. 36).

[17] See Alston, “Two Types of Foundationalism” in Alston 1989c.

[18] This seems to have been denied by Aristotle, for in the Posterior Analytics he says: “since the regress must end in immediate truths, those truths must be indemonstrable” (Bk. II, 72b19-24). Aristotle is no doubt thinking of foundationalism as a structure of propositions rather than beliefs. Even so, it seems that what the regress requires is immediately justified beliefs (or knowledge), not beliefs for which one can provide no reasons or demonstration at all. What are needed are unmoved movers, not unmovable movers. The existence of reasons for a foundational belief does nothing to perpetuate the regress (as Aristotle might have thought), for the higher-level belief is not needed to stop the regress, only to show that we have reason for regarding the regress as stopped.

[19] See Alston 1991c (pp. 81-93) and Moser 1989 (pp. 193-194).

[20] An objection to the belief condition of knowledge is found in Oswald Hanfling 1985.

[21] See Plato, Republic 476-79; Meno 87-88, 97a-98c; and the Theaetetus 201c-210b.

[22] Edmund Gettier, in “Is Justified True Belief Knowledge?” (1963), presented the locus classicus objection to knowledge as “justified true belief,” which has led some philosophers to either modify the concept of justification or add a fourth condition, such that the justification of a belief does not proceed via a false premise or assumption. Plantinga (1993c,  p. 6) observes that there are in fact few explicit formulations of knowledge as “justified true belief” prior to Gettier. Despite the defective character of the “justified true belief” account of knowledge pointed out by Gettier’s 1963 article, philosophers often continue to think of justification (along with truth) as necessary and at least nearly sufficient for knowledge, perhaps requiring a fourth condition as an addendum to answer the Gettier counter-example cases. The pursuit of epistemic justification, though, will have a place on the epistemological map regardless of whether it is necessary for knowledge. The question of whether our beliefs are justified, warranted, or in some sense rational is an important question independent of the connection between these epistemic desiderata and knowledge.

[23] See chapter 2 for a discussion on the origin of rationality/justification deontologism in John Locke and the Enlightenment. Contemporary deontological accounts of justification are found in Chisholm 1977, 1982; Ginet 1975; Bonjour 1985 (p. 8); Alvin Goldman 1986 (pp. 25, 59); Pollock 1986 (pp. 124-25); Lycan  1988 (p. 128); Wolterstorff 1983; Alan Goldman 1988 (p. 40).

[24] See also Bernard Williams 1973, L. Jonathan Cohen 1992 (pp. 20-27), and D.S. Clarke, Jr 1989 (pp. 31-36). Cohen 1992 distinguishes between the mental states of belief and acceptance. The former is thought of as an involuntary disposition normally to feel it true that p (and false that not-p). The latter is to adopt or have a policy of postulating p such that p factors into one’s decision to think or act in some particular context. Inasmuch as acceptances are voluntary, unlike beliefs they may be subject to deontological requirements.

[25] See Alston 1989c (p. 145) for a discussion of this argument from cultural isolation as it bears on deontological and truth-conducive justification. The cogency of this line of argument, though, rests heavily on there not being obligations of which we could be ignorant.

[26] Moser (1989) offers a concept of epistemic justification free from the idea of both epistemic obligation and epistemic goodness, though it emphasizes the notion of a justifier as a truth-indicator - what Moser calls an evidential probability maker.

[27] These versions of reliabilism (a)-(d) are explicated and subjected to criticism by Paul Moser 1989 (pp. 194-204). For further critical accounts, see Plantinga 1993c (ch. 9) and Pollock 1986 (pp. 114-122).

[28] See Sosa 1991 (p. 132); Bonjour 1985 (pp. 38-45) for account of this meta-incoherence problem.

[29] There is a similar difficulty lurking in the locution, given G the probability of p is high. Just what is this G? Is it some particular ground (a specific belief or experience) so that we get something like "given this bit of evidence e, the belief that p based on e has a high probability"? Or is it rather a type of ground (individuated in some way) so that the sense is “Given grounds of this sort (sensory perceptual experiences of X sort, or beliefs of this type under X conditions), the belief that p is probable”?

[30] So where the belief is <it is raining outside>, relevant characteristics would include the subject’s ability to distinguish between rain and snow and not be subject to such things as rain hallucinations (Swain 1981, p. 105).

[31] The distinction between logical and epistemic probability may be used to treat the case of necessary truths problematic on Alston’s account. A logically omniscient being would be able to calculate whether a particular mathematical judgement has a probability of 1 or 0. But human cognizers will have evidence which, given their limited logical capacities and knowledge, neither entails the mathematical proposition nor its negation. So the point here is that the relevant sense of evidential probability for justification is not logical but epistemic.

[32] I am considerably indebted to Richard Swinburne for helping me think through the notion of evidential probability by carefully pointing out and explicating the distinction between logical, epistemic, and subjective probability. See also Plantinga 1993b, pp. 150-51.

[33] Alston himself in 1991b explicitly associates reliability with epistemically relevant characteristics of the knower.

[34] See Alston 1989c (p. 213-214) for a development of the relations between PI and AI.

[35] Both of these positions are designed to rule out the much too liberal notion of just anything which is in principle knowable by a person functioning as a justifier, though directly recognizable seems to rule out cases of what a person could come to know upon reflection.  For a critique of the strong degree of access requirement, see Alston, “An Internalist Externalism” in 1989c (pp. 234-239).

[36] The notion of first- versus second-order internalism is developed by Alston 1989c, Audi 1993 (especially, pp. 336-340), and Schmitt 1992 (pp. 116-117). The latter takes  PI² and AI² as two forms of perspectival internalism: (a) Reliabilist iterativism: S is justified in believing p just in case S is justified in believing that the belief that p is reliable, and (b) Counterfactual reflective perspectival internalism: S is justified in believing p just in case S would on reflection believe that p is reliable. I take it that (a) is a form of PI² and (b) a version AI².

[37] See Alston “An Internalist Externalism” in 1989c (pp. 242-243) for a more detailed discussion of these possibilities.

[38] As Alston notes: “No adequate-support belief at an earlier stage will serve to do the job required at a later stage because it will have the wrong content. At each stage what is required is a justified belief to the effect that the ‘reason for’ relationship at the immediately previous stage is an adequate one; and no earlier beliefs of that sort of hierarchy will have been concerned with that particular ‘reason for’ relationship” (“Internalism and Externalism in Epistemology” in 1989c, p. 211).

[39] For a discussion of these points, see Alston “Internalism and Externalism in Epistemology” in Alston 1989c (pp. 201-203).

[40] “According to externalism,” writes Sosa, “there can be justification-making properties of a belief which the believer could not possibly discover merely by reflection (introspection, memory, and reason)” (1991, p. 193).  Laurence Bonjour states that in an externalist theory, “a person’s beliefs. . .[are] epistemically justified simply in virtue of facts or relations that. . .[are] external to his subjective conception” (1980, p. 56).

[41] Some have split the distinction here in terms of propositional justification (non-source-relevant) and doxastic justification (source-relevant), and have required doxastic justification for knowledge.

[42] The level distinction involved here will be developed with considerable detail in chapters 6 and 7.