CONTEMPORARY VERSIONS OF FOUNDATIONALISM
* * *
2.2 From Descartes to C.I. Lewis
2.3 Chisholm's foundationalism
2.4 Moser's foundationalism
In this chapter I present an overview of contemporary foundationalism, with special focus on two prominent versions. I will not present a taxonomy of all possible foundationalist positions, because most of them are not defended nowadays and because those that are defended, I will argue, are variations on the same basic theme. I have selected the two most prominent versions, those of Roderick Chisholm and Paul Moser, as representative. Since there haven't been many foundationalists in the second half of the 20th century, and still fewer prominent foundationalists, it has been a relatively simple task to isolate Chisholm and Moser as foundationalism's current leading advocates. Both bear a historical debt to C.I. Lewis, who, along with some of the early Logical Positivists, was one of the very few foundationalists in the earlier part of this century. Foundationalism has not been a popular position in the recent history of philosophy. I will first put contemporary foundationalism in historical context, then present Chisholm's and Moser's accounts, isolate the differences and similarities between the two accounts, and then criticize them.
2.2 From Descartes to C.I. Lewis
Like everything else in modern philosophy, the story of contemporary foundationalism starts with Descartes. Prior to Descartes, foundationalism was the dominant view — one might almost say that, to the extent that epistemological views were discussed, foundationalism was taken for granted. Yet with Descartes foundationalism acquires a different and distinctly modern flavor. For the same reasons that Descartes can be seen as sending philosophy off in a new direction, he can be seen as giving to modern foundationalism its nonclassical features, the most fundamental of which are maintained to this day in contemporary versions, with very few exceptions. Descartes, for instance, made answering skepticism the most pressing epistemological question. Descartes also argued that we must start by taking our subjective psychological states, including our percepts, as primary and then proceed to argue our way to an external world. And Descartes argued that mental events bear no necessary relation to physical things, thus implicitly accepting a nonrelational view of consciousness. These three related Cartesian features — skepticism as occupying the dominant position, representationalist or indirect methodologies as the proper way to go, and viewing consciousness nonrelationally — all of which I reject — form a legacy accepted in varying degrees by virtually all contemporary foundationalists.
Descartes's foundationalism can be summarized in three statements:
1. We start with the obvious and certain, i.e., clear & distinct ideas.
2. We can justify other ideas by deducing them from the obvious.
3. Together, 1 & 2 give us most of our common sense ideas.
Descartes's is a foundationalist view, since we get a set of basic beliefs or propositions which do not depend upon others for their legitimate acceptance, and we get a method to derive all other justifiable propositions from them.
Both points 1 and 2 have come under attack, and as a result many of the details of Descartes's account have been abandoned. Disagreements over what is obvious have been unending to the point that philosophers hesitate to call anything "self-evident." What is taken as obvious is not nearly as much as Descartes thought and, other than the Cogito, Descartes's candidates, most of them metaphysical principles, have been viewed with suspicion.
Later foundationalists are also split over whether what is basic need be certain. Lewis and Chisholm follow Descartes in requiring certainty, though some such as Moser drop this requirement. More will come up on this as we discuss each version.
Descartes has also been seen as being overly restrictive in allowing deduction as the only method to generate further justifiable beliefs. Why only deduction? Cannot various inductive methods also be brought to bear upon the obvious, yielding further justified beliefs? Descartes was of course concerned with what can be justified with certainty, and it has been a common view that induction cannot yield certainties, so this perhaps explains Descartes's limiting us to deduction. Most later foundationalists have been willing to allow inductive methods to generate justified beliefs, though generally agreeing with Descartes that the highest degree of justification can be conferred only by deduction.
From the perspective of most contemporary foundationalists, the problem with the specifics of Descartes's program is that if we restrict ourselves to what is truly obvious and certain, then we can't deduce much, and then we end up not knowing or being justified in believing very much — and certainly nowhere near as much as common sense holds that we either know or are justified in believing. Accordingly, much of Descartes's program has been abandoned.
But let us now raise those aspects of the Cartesian approach that have had an enduring impact on later foundationalisms.
Descartes is the source for the common view that the foundational propositions must be reports of subjective psychological states. Take perception as a prime example. One cannot, on this view, have a basic proposition of the form "I perceive an X" or "There is an X." The basic propositions must be of the form "I seem to perceive an X," for only these can be certain. One starts by being aware of one's subjective psychological states, which gives rise to propositions that are reports on one's psychological states. On the basis of these propositions, one tries to make inferences to propositions about an external world, i.e. those of the form "There is an X." We start, then, subjectively with the "priority of the first-person case" or the "immediate certainty of consciousness," and only indirectly get to anything existing independently of the first-person or consciousness.
There are three sorts of Cartesian considerations behind making subjective propositions basic, the first of which is also the leading one for contemporary foundationalists: pressure from skepticism. For any perceptual experience, for example, a skeptic will point out the possibility that that experience is an hallucination or an illusion. Given such possibilities one would be unjustified in taking the object of one's experience to be an object that exists independently of one's experience. One could, for instance, experience kiwi-fruitly even when there is no real kiwi fruit in one's mouth. Some mischievous elves could be playing games with you, or some university researchers could be doing experiments on you. Or your consciousness could be generating its own contents, as in dreams and hallucinations. Who knows? So one can't start off assuming one is aware of some external, physical reality.
Such skeptical considerations do not arise in a vacuum; they depend on a certain model of mind, a certain theory of the nature of consciousness. They depend on the claim that perceptual contents could occur just as they do even if there were no corresponding external fact. There is thus no essential or necessary relation between perceptual contents — or, more generally, mental contents — and an external world. Views that hold conscious events to be essentially dependent upon, or necessarily related to, external reality are called relational theories of mind, so let us call the view that there is no such essential dependence or necessary connection the nonrelational theory of mind. It is in part this background view that allows Descartes to grant so much weight to skeptical considerations, and thus to the view that we must start with a prior certainty of consciousness. This nonrelational view is a premise common to both representationalist and idealist views of consciousness in general, and therefore of their analyses of perception. We will find the nonrelational view playing a prominent role when we come to look at Chisholm's and Moser's contemporary foundationalisms.
The nonrelational view of consciousness need not be a philosophical primary, and I suspect that in Descartes's case it isn't, for there are two other Cartesian considerations that can lend it support, one coming from Descartes's interest in parts of Christianity's world-view and one coming from a philosophical interpretation placed on the scientific facts about perception. These are the second and third of the considerations mentioned above as motivating the view that we must start with subjective first-person propositions.
In Descartes's case, his mind/body dualism fits neatly with such a nonrelational view. If the mental and the physical are different and self-sufficient realms, then what occurs in one need not be explained by reference to what occurs in the other. There may be contingent relations, but these are not essential to either. Hence percepts, being on the mental side of things, bear no necessary connection to external, physical reality, and whether a given percept does bear a relation to a physical object is a fact that needs to be established. The perceptual experience itself is not enough evidence. Thus, Descartes's tentative ontological commitments motivate a nonrelational account of consciousness, thus motivating an initial acceptance of skeptical considerations, thus leading to starting with subjective propositions about one's psychological states.
The scientific source is the causal story about perception and the philosophical representationalist theory of perception it can lead to. Science teaches us, to take the case of vision, that light strikes the retina after being reflected from some physical object; it is then converted to a different form of energy and sent along the optic nerve to the visual cortex for further processing. At some point one has an experience of seeing. In attempting to interpret these facts, the representationalist view argues that since there is processing involved, the subject cannot be directly aware of the physical object at the start of the causal chain. Instead the subject must be directly aware only of the end result of the processing — which is taken to be some sort of representation of the physical object at the start of the causal chain. The physical sense organs, on this view, are intermediaries between the object and the subject. This view can be integrated with dualism, as it is in Descartes's case, if one thinks of the mind as something distinct from the senses and as waiting at the end of the road, so to speak, to see what the senses are going to present to it; yet one need not be a dualist to accept the representationalist interpretation.
Perception is also not the primary source of foundational propositions for Descartes, since he puts more weight on rationalistic metaphysical premises; so the representationalist interpretation of perception is only a contributing factor in his endorsement of the nonrelational view. For empiricist versions of foundationalism, however, perception is the only source of knowledge, so the fact that representationalist accounts of perception have lent impetus to nonrelational views takes on more importance for analyzing those views. Yet whether affiliated with empiricist or non-empiricist general approaches, representational theories of perception lead to the conclusion that propositions about sensory/perceptual states do not entail the existence of what the person seems to be experiencing. Thus one must initially say "I seem to taste something bitter" and not "I taste something bitter." The experience could be a dream or an experimental byproduct, which is to say we are once again implicitly endorsing the nonrelational view.
This Cartesian representationalist project, which shares important features with Locke's empiricist representationalist project, was generally seen to have been a failure by the time of Hume and Kant. There is no way, starting only with premises about internal states of mind, to make valid or even meaningful inferences to an external world. Even if, for example, we posit an external world merely as the best causal explanation for our subjective experiences, how did we even get the idea of "cause" without first having access to an external world? Or, supposing we can derive a concept of causality from phenomenal states, on what grounds can we then say that it also applies an external world? The external world starts to drop out of the picture. Upon the recognition of this, there were two responses to the failure of the representationalist project: the Humean and the Kantian.
Hume gives up the representationalist project. If there is no way to ground normative epistemological standards like truth and justification upon an objective reality, then there is no way for epistemology to be a normative discipline. All epistemology can do is describe how people arrive at the beliefs they do. Epistemology becomes a radically descriptive field of study, on this view, so Hume can be said to be the first modern to "naturalize epistemology," to use Quine's phrase.
Kant, on the other hand, wants to retain epistemology's standing as a normative discipline, though he agrees entirely with Hume that it is impossible to derive an external world from propositions about subjective experiences. One certainly cannot validate one's conceptual scheme by reference to an independent reality, but this doesn't mean the only alternative is to exorcize epistemology's normative elements. Another alternative is to view the scheme as having normative standards within it. Our schemes are subjective products, granted, and this simply means that objectivity is internal and relativized to us.
Until the 20th century Kant's general approach set the tone. The dominance in the 19th century of versions of idealism and the corresponding shift away from correspondence accounts of truth and justification to coherence versions of each, bears witness to Kant's influence.
With the rejection of idealism in the first part of the 20th century came a revival of empiricist and realist philosophies, some direct though most indirect or representationalist. Along with this revival came a modest interest in some versions of foundationalism. Logical Positivism, for example, started off as a foundationalist project, though in the hands of Neurath and Carnap it quickly evolved into a more coherentist program. C.I. Lewis, with the publication of his Mind & the World Order in 1929, also reintroduced a more lasting systematic foundationalism into the 20th century debate. Lewis, however, did not find much support until about the time of his death, when Roderick Chisholm entered the scene. But from about 1960 until the mid-1980s, Chisholm had to carry the foundationalist ball alone, virtually right up to his retirement. Paul Moser has emerged in the mid- to late-1980s with a systematic foundationalist program, though he too has not — at least, has not yet — been able to alter significantly the antifoundationalist tide.
And it has been a tide. Foundationalism has been rejected by virtually every major epistemologist and philosopher of science of the last half of the century, from the later Wittgenstein to Popper to Sellars and Quine. In place of foundationalism, two alternatives have been developed. Sellars, Rescher, Lehrer, Rorty, and Bonjour all offer broadly coherentist alternatives. Quine, some of the evolutionary epistemologists, and eliminative materialists such as the Churchlands, have pushed rather for the rejection of epistemology's normative aspect and the assimilation of what remains by the natural sciences.
So we have had offered in the 20th century three major alternatives. In my view, the history of 20th century philosophy has been a sophisticated replay of the history of modern philosophy from Descartes to Hume and Kant. We are offered Cartesian representationalist approaches in the foundationalisms of Lewis, Chisholm, and Moser. In the spirit of Hume, we get naturalized epistemologies in Quine and the evolutionary epistemologists. And in the broadly Kantian spirit, we have linguistic versions of idealism in Sellars and Rorty. This is certainly a broader historical thesis than I will be able to defend in this essay, though I think it is true, and I will be able to offer at least some preliminary indications of its truth as we proceed.
While foundationalism has been in the very small minority, it has been there. Antifoundationalists still devote considerable amounts of energy to refuting it, and in some quarters work is being done to advance foundationalism. For the remainder of this chapter, we will investigate the two major foundationalisms now available, i.e. Chisholm's and Moser's.
2.3 Chisholm's foundationalism
The latest and most finely tuned incarnation of Chisholm's foundationalism appears in the 1989 third edition of his popular textbook, Theory of Knowledge. Over the years, three philosophically global constraints have guided Chisholm's presentation of foundationalism. The first is that skepticism cannot be defeated. For any claim to knowledge or even justified belief, the skeptic can raise the usual brains-in-a-vat objections, effectively undermining the claim. Ultimately, Chisholm concludes, one must either accept skepticism and its consequences or, if one is to have such a thing as a theory of knowledge, beg the question against it and get on with the project.
The second constraint is the apparently vicious circle the problem of the criterion generates. The problem is that in generating an account of knowledge, one has to decide whether method or content has priority. One can either assume that one has knowledge and work backwards to the methods and principles that make it possible ("particularism") — or one can assume some methodological principles and derive justified beliefs in accordance with them ("methodism"). But, Chisholm points out, neither option is defensible. How can one claim to know something without being able to answer questions about how one knows? And how can one establish under what conditions one knows without reference to actual instances of knowing? Both seem impossible. But without adopting one of them, no epistemological project can get off the ground. Hence one must either abandon epistemology ("skepticism") or arbitrarily choose to adopt either particularism or methodism. These are the only options.
The third and decisive global constraint is Chisholm's G.E. Moorean conviction that common sense knowledge of the external world exists. This conviction works in conjunction with the other two constraints, for it gives him the moral support he needs in begging the ultimate questions from the skeptic and stepping around the problem of the criterion in adopting particularism. Chisholm chooses not to accept skepticism or its consequences: If the choice is between skepticism and knowledge, why not accept the position that best squares with one's natural inclinations? Chisholm's inclinations dictate that he does know some things, so he is willing to beg the question against the skeptic (at least a little bit). In making this leap, Chisholm points out that Santayana was entirely correct in claiming that "our knowledge involves an element of animal faith." Against those who charge that begging the question is never a satisfactory response to a problem, Chisholm responds that that is the best we can do and that the onus is on the skeptic to demonstrate that the faith is misplaced. And since the only alternative is skepticism, it is hard to see that one's faith could be better placed.
As for the problem of the criterion, since we've eliminated skepticism, that leaves only particularism and methodism. There's no solution to the problem, so again one has to opt for whichever position seems best. And again, Chisholm feels that common sense is best: people seem first to know things and only upon reflection grasp the methods and principles involved. So he opts for particularism.
Chisholm's epistemological method, then, given the three constraints, is to assume that we do know some external world propositions and to work backwards in figuring out how we know them. He assumes that we do have justified beliefs and instances of knowledge and seeks the principles that must underlie them. Ordinarily, when we ask what justifies a given claim, we presuppose that there is a justification for it, that we are justified in many of our beliefs, and that we know (at least vaguely) the difference between justified and unjustified beliefs. In the case of any particular belief, when we ask the question and then investigate, we learn what particular justification we have for the claim in question, and we learn more clearly what the general criteria of justification are. The justification is already implicitly there and simply needs to be brought out into the open. This is the method of particularism.
My procedure in this section will be to lay out Chisholm's program step by step and then to raise criticisms of it at the end of the section.
Putting particularism into practice, Chisholm follows in the tradition of Descartes and C.I. Lewis in holding that while justification is often a matter of degree, all justification must terminate in something certain. "If anything is probable for S, then something is certain for S." If everything were only probable, we would get either an indefinite regress or a circle, and thus no genuine probabilities. Hence, we need to find something certain.
Chisholm's candidates for indubitable certainties are first-person propositions about one's experiences. Experiences are "self-presenting properties." One undergoes them, and if one reflects upon one's experience, then the proposition that one is experiencing in a specific way is certain for one. Self-presenting properties are either intentional (thinking, hoping, fearing, intending) or sensible (the various ways in which we sense). The important point here is that sensations, according to Chisholm, are non-intentional. (More on this later.) Both species, however, are such that if one has them, and considers oneself having them, then it is certain for one that one has them. For example, if one is wondering about what to have for dessert, then it is entirely justifiable for one to believe one is wondering what to have for dessert. That one is wondering is thus certain and indubitable for one. This is an important point, for we have now introduced epistemic properties. In the case of self-presenting properties, the epistemic property of certainty supervenes on non-epistemic facts, i.e., one's conscious states.
In this way, self-presenting experiences are a source of certainty. One cannot be mistaken that one is having an experience, whether intentional or sensuous. One immediately apprehends it. But, and this is the important point, all that one is justified in believing one has access to immediately is the experience. What is certain is not that one is experiencing something out there in the world; that would be too grand a begging of the question against the skeptic, for it would assume a direct realist view of perception. For certainty, we must restrict ourselves to data about our own conscious states. Only our conscious states are self-presenting. "I seem to see a lime" is indubitable for me, for example, but I could go astray if I say "I see a lime." I don't know initially that I am seeing a lime; I know only that I am having an experience that is lime-seeing-ish. If we are to start with certainty, then we must start with subjectively qualified appearance propositions. Only these are "self-justifying." Self-presenting properties are the source for these self-justifying propositions, for while they "may mislead us about other things, they are not a source of error about themselves."
As with much of Chisholm's technical terminology, his use of the phrase "self-justifying proposition" is somewhat misleading since it is not in accord with common usage. The propositions in question do not themselves form their justification, contrary to what the phrase suggests. The presented experience and one's reflection upon it are necessary ingredients, for only in the context of their occurrence does a proposition become self-justified. Chi-sholm would perhaps be more comfortable with saying that the experiences generate justified propositions rather than that experiences transmit justification, since the latter implies that experiences themselves have justification. He doesn't want to say that self-presenting experiences have justification; they are what confer evidence, but they themselves are pre-evident. They are "prime movers" in this sense: they confer evidence without being themselves evident. This is what he means by holding the resulting proposition to be "self-evident." Chisholm is thus registering at least partial disagreement with the common antifoundationalist premise that justification is solely a propositional phenomenon.
But it is crucial for Chisholm's account that only propositions about first-person psychological experiences can be justified in this manner. For example, of the two statements "I am hungry" and "I see a key," "we may say that the first is self-justifying and the second is not." The first is internal, the second external. The general rule, then, is to formulate propositions about one's experiences non-relationally, i.e., without presupposing that there is an external object to which one's conscious state is related. That there is such an object will be a conclusion to be arrived at later. "To arrive at what is self-presenting in these [i.e., perceptual] cases, we must remove the reference to the external thing" — for example, "to the wine in 'This wine tastes sour to me'."
This was also
Descartes's procedure. One starts with
knowledge of one's own conscious states, this knowledge being certain and
incorrigible; and then one goes on to make inferences about an external world. Chisholm, therefore, is following the
broadly Cartesian representationalist pattern. The primary difference between Chisholm and Descartes is that
Chisholm places much more emphasis upon empirical sources for the initial
certain knowledge, while Descartes relies more heavily upon innate or a priori principles.
Thus far Chisholm has an account of how experience can generate justified propositions. He has, therefore, satisfied the weaker of foundationalism's claims: that it is possible to derive some justified propositions that do not depend upon other propositions for their justification. The stronger foundationalist claim — that all justified propositions are justified, at least in part, by basic propositions — has yet to be established. Yet the bulk of the justified propositions we are interested in (our common sense beliefs) are propositions about external reality, and not merely about our subjective psychological states. So we need to see how Chisholm thinks external world propositions can be justified on the basis of subjective propositions. And for this we need to investigate in greater detail Chisholm's views about the nature of perceptual evidence. For while Chisholm does not accept a representationalist account of perception, his views on the import of skepticism and of consciousness in general condition his method of proceeding in deriving justified propositions about the external world.
Sensory experiences are usually said to present both other things and themselves. The former claim is problematic, since it raises the question of whether direct realism is true. Due to the standard problems of illusions and hallucinations, Chisholm feels that one cannot assume at the outset that there exists an external object corresponding to one's experience. Therefore, when propositionalizing one's experience one cannot straightforwardly assert that one is perceiving an X, where X is an external object. The proposition must be qualified subjectively.
On the other hand, in order to subjectively qualify the proposition, traditional representationalist theories of perception typically introduce intermediary objects of perception, e.g. sense data. If, for example, one is hallucinating a pink mouse running up the wall, there is no real pink mouse corresponding to the hallucination. But the subject of the hallucination takes himself to be perceiving one. The representationalist analysis concludes that there must be a pink-mousish sense datum that is the actual object of one's experience.
Chisholm also rejects representationalist theories of perception. Sense-data theories raise a host of questions their advocates have not successfully been able to answer: "can sense-data exist unsensed? Can two persons experience numerically identical sense-data? Do sense-data have surfaces which aren't sensed? What are sense-data made of? Are they located?"
So neither direct realism nor representationalism is acceptable for Chisholm as a starting point for his foundationalism. His solution is to attempt to present his foundationalism entirely neutrally with respect to any theory of perception. The trick then is to formulate the basic propositions subjectively without adopting or even suggesting the adoption of a representationalist theory of perception. This is where his use of adverbial formulations enters.
For example, suppose one seems to see a kiwi fruit. Considering the fact that one's experience could be an illusion or hallucination, the proposition that is justified for one cannot be, "I see a kiwi fruit." Nor does Chisholm want to say that the justified proposition in this case is, "I see an apparent kiwi fruit," since that sort of talk invites sense data. The solution is to drop the talk of objects of perception altogether and instead to formulate the proposition as a report of one's experiencing: "I sense kiwi-fruitly" would be appropriate in this instance. What the proposition formulates is an activity of consciousness, with the adverbial ending indicating a modification of the manner of that activity.
Adverbialism thus attempts to solve the problems hallucinations or after-images pose for any theory of perception without having to suppose sense-data. Adverbialism shifts the focus away from what the apparent objects of experience are, to the subject and the subject's manner of experiencing. This maneuver thus leaves entirely open the question of the connection, if any, of the experiential state to the external world.
For Chisholm, this maneuver falls out of his general non-relational view of consciousness, of which perception is a special case. In the case of perception, one should not think of an experiential act as necessarily relating one to an object, whether an object in the external world or a sense datum. One can think of something non-relationally, and one can experience non-relationally: "It just does appear white — and that is the end of the matter."
At the outset, then, there is no commitment to an independent X. One starts knowing only that one is in an experiential, adverbial state; in general, the pattern is "I am appeared to X-ly." Then the project is to discover that the experiential state implies the presence of an external object. Once attained, one's justified belief can be expressed in a proposition of the form "There is an X."
It is important to note that while Chisholm rejects representationalism in perception, he still is committed to a representationalist theory of justification. The same indirect pattern of deriving external world propositions from antecedently known subjective propositions is followed. This representationalist pattern is explored and criticized in Chapter 3.
Adverbialism is incomplete as an analysis of perception. Chisholm still needs an account of what the subject is perceiving in the normal case. But for Chisholm this is an issue to be dealt with separately. His foundationalism is intended to be neutral with respect to the question of the directness of perception.
The next question brings us to the heart of Chisholm's enterprise: How does one discover that experiential states imply the presence of external objects? What epistemic principles can bridge the gap between propositions about subjective sensory states and propositions about the external world? The needed principles are, logically enough, called "bridge principles."
Another angle on the same problem comes from Chisholm's above-mentioned distinction between intentional self-presenting properties and sensible self-presenting properties. Thinking, hoping, fearing, intending, and so on, are intentional. The ways in which we sense or are appeared to are sensible properties, yet propositions generated by these have been formulated adverbially to remove any sort of intentional object. So we need to bridge the gap between the two.
We are looking for criteria of the form: "So-and-so-tends to make it evident to S that he is appeared to by an F." Then, if the criteria obtain, S will be justified in believing he is appeared to by an F; and if he is appeared to by an F, then he will be justified in believing the proposition, "There is an F." This is the way that people proceed. "In the case of being appeared to, there is something, one's being appeared to in a certain way, that one interprets as a sign of some external fact." So the question is, Under what circumstances can one take one's experiences as a sign of a real object and not merely as an illusion or hallucination?
What makes this the heart of Chisholm's enterprise is that this is the standard exceedingly difficult problem representationalists since Descartes and Locke have had to grapple with. As a working example, take the transition from "I am appeared to mauvely" to "There is a mauve thing before me" as an example. The transition from the subjective proposition to the external-world proposition cannot be justified deductively, for all that follows deductively from propositions that are directly evident, i.e. propositions about one's subjective states, are further propositions about one's subjective states. Nor can enumerative induction justify the transition (supposing one had formulated a series of subjective propositions), for in an enumerative induction based solely on subjective propositions one has only a series of singular propositions about experiences and not any sort of proposition about an external world by which to assert a connection. So getting to the external world requires faith plus a "hypothetical induction." The conclusion "There is a mauve thing before me" must be viewed as the best explanation for the fact asserted by the self-justifying appearing expression — in this case, "I am appeared to mauvely." Belief in the external world is thus an inductive hypothesis, and realism requires a dose of Santayana's animal faith. One cannot establish conclusively that the hypothesis is true; for this reason, of all empirical propositions, only self-justifying propositions about subjective experience will have the highest degree of certainty; external world propositions derived by the faith/hypothetical induction method will have something much less.
The animal faith is not unlimited, however: Chisholm is careful to delineate the principles that yield justified propositions from hypothetical inductions from those that result from rather wilder leaps of faith. And since justification comes in degrees, the principles he offers attempt to reflect those degrees. If there are such things as justified propositions about the external world (believing that there are is the faith element), then the principles will catalog the steps that must obtain. It would be a misunderstanding to hold that the principles Chisholm offers are intended as a way of arguing one's way to an external world; rather they are intended as formulations of the principles that must obtain given that one already accepts that there are justified propositions about the external world. The G.E. Moorean faith is bedrock.
The principles are intended to capture degrees of justification descending along a continuum from "certainty" to "counterbalanced," where "counterbalanced" means that one is as justified in believing a given proposition as its negation and vice versa. Self-justifying propositions are of course certain for one. Being "obvious" is the next level down from certainty, followed successively by "evident," "beyond reasonable doubt," epistemically in the clear," "probable," and finally "counterbalanced."
In making the transition from certain subjective propositions to external world propositions, the first step up the hierarchy is to a proposition that is "probable" for one. "Probable" is the first positive degree of justification stronger than "counterbalanced," and is defined as follows:
D2 p is probable for S =Df S is more justified in believing p than in believing the negation of p.
The principles that capture this notion of minimal probability are his Material Principles 2 and 3:
MP2 Accepting h tends to make h probable
MP3 If S accepts h and if h is not disconfirmed by S's total evidence, then h is probable for S.
The claim here
is that a proposition about an appearing tends to make probable a proposition
about the external world. And if the
external world proposition is both accepted by the subject and the rest of the
subject's beliefs do not go against it, then the proposition in question has
been made probable for the subject.
Accepting an external world proposition under these circumstances gives
it prima facie probability.
The accepted proposition achieves a higher degree of justification under the following circumstance: If the denial of the proposition is not probable given all of the propositions that are probable for the subject, then the proposition is "epistemically in the clear" for the subject. This higher degree of justification Chisholm labels "epistemically in the clear," and it is defined as follows:
A proposition is said to be epistemically in the clear for a subject S provided only that S is not more justified in withholding that proposition than in believing it. This definition introduces the notion of "withholding," which is defined as follows: A person may be said to withhold a proposition h provided he does not believe h and does not believe the negation of h.
Putting all this together, believing a proposition h is epistemically in the clear for a subject S if: (a) S accepts h; (b) h is not disconfirmed by all of the evidence available to S; and (c) given all of the propositions that S accepts and that are not disconfirmed by all of the evidence available to S, not-h is either not accepted by S or disconfirmed by the propositions that are not disconfirmed by all of the evidence available to S. Then and only then is believing h more justified for S than believing neither h nor not-h, which is the degree of justification the concept of "epistemically in the clear" is intended to capture.
Chisholm continues in this vein. One more level up the hierarchy will suffice for our purposes, for then a fairly strong level of justification — "beyond a reasonable doubt" — will have been achieved for external world propositions, and we will have a clear enough picture of the general procedure Chisholm is using as a basis for evaluating it.
Three additional definitions are required for the principle that governs propositions that are justified "beyond a reasonable doubt." The first is that of "taking":
S takes there to be an F =Df (1) S is appeared — to;
it is evident to S that he is appeared — to; and
(3) S believes that there is only one thing that appears — to him and that that thing is F.
This definition requires the concept "evident," which is defined as follows:
p is evident for S =Df For every proposition q, believing p is at least as justified for S as is withholding q.
The concept in question also needs a definition.
p is beyond reasonable doubt for S =Df S is more justified in believing p than in withholding p.
Finally, Material Principle 5:
If S takes there to be an F, and if it is epistemically in the clear for him that there is an F which he takes to be an F, then it is beyond reasonable doubt for S that he is perceiving something to be F.
The basic and apparently straightforward intuition behind this complex formulation is, according to Chisholm, that "The wise man will make use of whatever apparently probable presentations he encounters, if nothing presents itself that is contrary to that probability." Let's work this through step by step.
We suppose that a given external world proposition, "There is an F," is epistemically in the clear for S, as explained above. Now, if S also takes there to be an F — which requires that S be appeared to X-ly, that the proposition "I am appeared to X-ly" be evident for S (which means that for no other proposition would S be more justified in withholding belief in that proposition than believing "I am appeared to X-ly"), and S believes that only one thing appears X-ly to him and that the thing that appears X-ly is F — then Chisholm holds that it is beyond a reasonable doubt for S that he is in fact perceiving something to be F. And from that it follows in due course that the proposition "I perceive an F" or "There's an F" is justified for S. So for S an external world proposition has been justified beyond a reasonable doubt on the basis of subjectively qualified, self-justifying propositions about experience.
Chisholm's account is at least minimally foundationalist. There are such things as propositions that do not depend upon other propositions for their justification. And some other propositions (at least) depend for their justification upon those basic propositions. Thus, at least part of the structure of justified propositions is both hierarchical and terminal. Does it follow from this that the stronger foundationalist claim — that "every justified [empirical] statement, about what we think we know, is justified in part by some statement that justifies itself" — is true? As far as I can tell, Chisholm does not give an argument for this stronger claim.
The account is foundationalist, yet as we have seen the foundation consists in part of epistemological faith in the face of skeptical objections. If we're going to be reasonable, Chisholm believes, i.e. if we believe we're able to acquire knowledge and be consistent in our beliefs, then we have to accept his principles. The only fundamental alternative is skepticism, and that's the only fundamental argument that can be made for the principles. This doesn't undermine the skeptic, let alone prove the skeptic wrong. But that project is impossible; that is the "epistemic predicament." All we can do is affirm the central presupposition of the epistemological project:
I am justified in believing that I can improve and correct my system of beliefs. Of those beliefs that are about matters of interest or concern to me, I can eliminate the ones that are unjustified and add others that are justified; and I can replace less justified beliefs about those topics by beliefs that are more justified.
Let me now raise criticisms of Chisholm's account.
In the first place, his response to skepticism is certainly problematic. He grants that the skeptic cannot be beaten, which means he is unable to find any flaws in the skeptic's arguments. What, though, is the reasonable thing to do in the face of unanswerable arguments? Accept them. Instead Chisholm overthrows his rational judgment in favor of a faith he says cannot be rationally defended. This puts Chisholm in a paradoxical position. He claims to be giving an account of justification, yet nothing can be said to be justified if it rests on an act of faith. And if one's faith requires ignoring apparently compelling arguments, then any plausible sense of justification is undercut.
Chisholm is forced into this by his pessimism about the possibilities of overthrowing skepticism. Yet he does not beg the entire question against the skeptic, for the apparent force of skeptical objections also leads him to condition his methodology by making subjectively qualified appearance propositions the foundation. This in turn leads him to adopt the standard representationalist project of making justified belief in external world propositions indirect. But if it should turn out that skepticism is not unbeatable, and if representationalism must fail necessarily, then it follows that Chisholm's entire project is misguided. I think both of these antecedents are true; my investigation of them comprises Chapter 3
An equally serious problem, to my mind, is his admittedly arbitrary adoption of particularism in response to the problem of the criterion. Particularism as a method is certainly invaluable in many cases as a means of determining consciously and in retrospect how exactly one arrived at a given conclusion. Yet Chisholm presents the options of particularism and methodism as an exclusive disjunction. In presenting the options, the common premise is that only one or the other can have priority. Yet why can't both proceed simultaneously? The possibility that content and method go hand in hand, that neither has priority, is not explored. In Chapter 5 we will do so.
This point is not unrelated to the issue of Chisholm's representationalist methodology, so the issues discussed in Chapter 3 will bear upon it. The assumption of the problem is that as an adult one has a set of beliefs and a set of methodological principles, and that from the "inside" one must find some way to determine which are reliable. Chisholm recognizes that on these grounds there is no way to determine which set is reliable, and so he concludes that one simply has to arbitrarily choose one set over the other. The assumption is that we have no direct access to reality to ground or validate either the content of the beliefs or the methodological principles in question. That assumption is the representationalist premise, which I challenge in Chapter 3.
The above two points concern Chisholm's general positioning of himself philosophically. Let us turn now to five points of detail.
First: for Chisholm the proper foundational proposition is of the form "I am appeared to X-ly." The key concept here is that of "appearance." Grasping the concept "appearance," though, requires a prior grasp of the concept "reality." That is to say, before one can understand what it is for something to appear a given way one has to have grasped the sophisticated point that appearance and reality can sometimes diverge. This implies that in using appearance concepts as foundational, Chisholm is either smuggling in an implicit reference to reality or presupposing other concepts. This is a dilemma, for the former alternative is contrary to his representationalist methodology of working from the "inside" to the "outside," while the latter alternative means that no proposition of the form "I am appeared to X-ly" can be foundational, for understanding the proposition requires antecedent knowledge.
Chisholm is aware of this line of criticism, but he rejects the view that "appears" is parasitical upon "is." "Appears" language is not necessarily used as a hedge, i.e., as a way of indicating one's awareness of the fact that things may not be as they appear to be. It can be, instead, descriptive of ways of self-presenting appearing.
In the context of defending C.I. Lewis's foundationalism against the same charge, Roderick Firth adopts the same strategy that Chisholm does. Firth phrases the problem this way:
[I]f the coherence theory of concepts is correct and we cannot fully understand 'looks red' unless we possess the contrasting concept 'is red,' then it would seem that it is not logically possible to have the concept 'looks red' before we have the concept 'is red.' This is clearly a problem for the Lewisian approach: It is these expressive judgments [e.g., 'It seems to me as if I were seeing something red'], according to Lewis, that enable us to escape the coherence theory of justification; and if it should turn out that these judgments all make some covert reference to physical objects, then — depending on the kind of 'covert reference' — it may no longer be possible to make the epistemological distinction which Lewis requires.
Chisholm's program also requires the distinction. And if "appearance" concepts depend upon "reality" concepts, then appearing statements smuggle in a reference to reality, and then Lewisian/Chisholmian foundationalism cannot work.
Firth attempts to resolve the problem by noting that the "looks red/is red" distinction is the adult distinction between appearance and reality, and by claiming that children have prior to this a more primitive concept of "looks red" which is not dependent upon the adult concept of "is red." If this is true, then it will allow Lewis's program to preserve its required starting point of "it seems to me" without any physical object presuppositions being built into it.
The question then is: Is there a primitive "looks red" concept? In general, are appearance concepts prior? Is it possible, as Firth and Chisholm claim, for "appears" concepts to be used entirely independently of "is" concepts?
Firth's argument for this point begins by noting that a child will call things "red" whether the things are "really red" or only made to look that way (by, for example, the child's looking through red glass). "In fact," Firth continues,
at this stage the child says 'red' just in those circumstances in which we, as adults, could truthfully says [sic] "looks red to me now," so that it would not be unreasonable to assert that the child is using 'red' to express a primitive form of the concept "looks red."
This argument is completely inconclusive, for if the child also uses 'red' when adults use 'red' — as Firth recognizes — then it could as easily be that the child has a more primitive concept of "is red." And this would seem to be the more reasonable assumption, since it seems that children proceed as direct realists, always taking there to be real objects of experience. That one can make more limited claims of the purely phenomenologically descriptive sort seems to be an acquired skill, learned much later.
So while it is certainly true that adults have such a purely descriptive use of "appears," as Chisholm points out, the use of such a concept for a defense of foundationalism is not satisfactory, for it only pushes the initial criticism to a deeper level. A necessary component of the purely descriptive "appears" is that it eschews external object commitments; it limits the resulting description to the features of the experiential state. But this is not part of the child's understanding: he or she behaves as a direct realist. So in using "appears" only as a description of self-presenting experiences, Chisholm (or anyone in this context) is adopting the philosopher's context of knowledge: he has made the distinction between appearance and reality, between subject and object; he can focus upon the way things appear in abstraction from any consideration of the way things really are; and thus he can formulate a technical use of "appear" designed to capture experiencing from the subjective side. So we as philosophers certainly can speak of "appearances" without implying a direct actual or possible contrast with reality, just as Chisholm claims. But this technical noncontrastive use of "appears" presupposes the more general contrastive use of "appears."
If this is correct, then Chisholm's foundationalism is not neutral with respect to questions of realism, contrary to his hopes. And this means that propositions of the form "I am appeared to X-ly" cannot be foundational, since they presuppose prior distinctions.
A more general point can be made about this problem with Chisholm's approach to foundationalism: That it is not possible to do epistemology entirely neutrally. In the case of foundationalism, questions about the hierarchical structure of justified belief can be addressed in abstraction from questions about the nature of perception, but they cannot be addressed neutrally, i.e., without making either implicit or explicit commitments to some view on the nature of perception.
Second: I believe there is also a lacuna in Chisholm's account of how self-presenting experiences generate self-justifying propositions. An experience is a concrete phenomenon: it is determinate and particular — one experiences, for example, a particular dog of a particular size, shape, color, and so on. A proposition, by contrast, is abstract and (sometimes) universal — it utilizes concepts of entities and attributes that refer to more than one entity. In propositionalizing an experience, one uses abstractions in subsuming the experience to a universal type. This means that one is connecting this particular experience to other particular experiences. If so, then forming a proposition on a given occasion means doing more that simply summing up the experience; more is going on than an isolated statement. This presents two problems for Chisholm's account. The first is that he introduces a new level of conscious phenomena — concepts, abstractions — without an account of what validates their use, particularly in connection with their relation to given experiences. This lacuna could be closed if Chisholm presented a theory of abstraction, though this is not something he has done to date.
The lacuna, however, is in part a product of Chisholm's abovementioned allegiance to a non-relational view of consciousness. If each proposition is a self-contained phenomenon, if one can think of a given cognitive item without necessarily making connections to other cognitive items, then there is no need to present an account of how this particular use of the proposition "I am appeared to quadrilaterally" relates to other uses of the proposition.
I do not think that a given judgment about one's experience merely sums it up in isolation from the rest of one's experiences. In using abstractions, one is integrating the experience of the moment with previous similar experiences. I do not think that this integrative phenomenon entails any form of coherentism, as I will explore in Chapter 5. Yet I think Chisholm thinks it does, which I think explains his resistance to the notion that a theory of concept-formation is needed.
Third: another possible lacuna falls out of Chisholm's non-relationalism. Nelson Goodman's problem of imperfect community, though usually raised in the context of metaphysical bundle theories, also bears upon the issue of the abstractness of the resulting self-justified propositions and their relations to the experiential given. Suppose one is confronted with a red pyramid sitting on a blue cube. All of the following propositions will be self-justifying for one: "I am appeared to redly," I am appeared to bluely," "I am appeared to triangularly," "I am appeared to squarely." Some account is needed of why the propositions "I am appeared to red-triangularly" and "I am appeared to blue-squarely" will be justified while the propositions "I am appeared to red-squarely" and "I am appeared to blue-triangularly" will be unjustified. The latter four propositions are conceptually compound, and their derivation from conceptually atomic propositions or directly from experientially atomic states is not obvious.
Fourth: Chisholm's account formally divides into two sets of propositions: a set of definitions of terms indicating locations along the justification continuum, and a set of principles governing the use of those terms in particular situations. Let me raise one general criticism each of the offered definitions and of the principles.
On the "definitions" side, Chisholm's ostensible purpose is to provide an account of the concept of "justification." Yet most of the offered definitions of the degrees of justification ("probable," "beyond reasonable doubt," "evident," and so on) use the concept "justification" in their definition. This at the very least makes the account circular, which is a certainly a weakness in an account of epistemic justification. Perhaps the best response to the charge of circularity is that Chisholm sees this wing of his project as merely clarificatory. Just as we have the Moorean faith that we know some things, we have going into the project a pre-analytic sense for what justification is all about, and the project's goal is simply to make more precise our use of the term. Thus Chisholm's definitions are only paving the way for future investigations of what the root concept of justification really is. But if this is so, then Chisholm has left entirely open questions about whether our pre-analytic notions of justification are adequate, whether yours are the same as mine, and if not, whose are better.
Fifth: On the "principles" side, I think his MP2 and MP3 are symptomatic of a central problem with Chisholm's approach.
MP2 Accepting h tends to make h probable.
MP3 If S accepts h and if h is not disconfirmed by S's total evidence, then h is probable for S.
If MP2 read "Accepting h makes h probable," then it would clearly be invalid; it would be a form of "Believing makes it so," and thus be a form of the "epistemic conservatism" Chisholm is concerned to avoid. People come to accept propositions for all sorts of bizarre reasons and because of all sorts of strange causes, so the fact that someone accepts something by itself confers no probability. MP2, however, makes less of a claim by adding the "tends to" qualifier. What does it mean to say that accepting a proposition "tends to make" that proposition probable? "What is intended by the locution, 'e tends to make h probable,'" Chisholm states, "may be put somewhat loosely by saying, 'If e were the only relevant evidence you had, then you would also have some justification for accepting h.'" Now, MP2 claims that "Accepting h tends to make h probable". If we perform the relevant substitutions, the acceptance of h is the e in question, for the only relevant evidence available is the fact that one accepts h. And this is once again simply to say that the mere believing of a proposition confers probability upon it for the subject.
Clearly the import of MP2 is to confer upon accepted propositions some minimal level of probability based merely upon the fact that the proposition is accepted. But this then is a watered-down version of "Believing makes it so." One can only claim that accepting a proposition tends to make it probable if the mechanism by and context in which the proposition was accepted are such that they tend to make propositions probable. But if we know that people can and do accept propositions for utterly ridiculous reasons, then we cannot confer any degree of probability, however minimal, upon a proposition given only that it has been accepted. We have to know something about the mechanism by which the proposition is accepted and the context in which it is accepted — and this is what principles of justification should spell out for us.
MP2 is Chisholm's first positive link between subjective propositions and external world propositions. The propositions that people find themselves accepting are external world propositions, and Chisholm is searching for a way to justify them. Yet the fact that people accept realism does not make or even tend to make the propositions involved probable.
MP3 claims that a genuinely positive degree of justification is conferred upon a proposition given two conditions: it is accepted and none of the evidence available to the subject disconfirms it. The first condition alone is invalid, for the reasons given above. The second alone is also problematic, for it claims that a lack of evidence against a given proposition translates into a positive degree of probability for the proposition. The fact that none of the evidence available to me disconfirms the proposition that one of the moons of Jupiter is inhabited by eight-celled organisms that eat frozen methane does not make the proposition probable for me. Argumentum ad ignorantiam is a fallacy at any level. The only remaining possibility is that the two conditions together can yield probability. This does not seem plausible; two fallacies do not a valid inference make.
This concludes my presentation and initial criticism of Chisholm's foundationalism. Let us now turn to Moser's account.
2.4 Moser's foundationalism
Paul Moser's 1989 Knowledge and Evidence is a synthesis and expansion of material presented in his many journal essays published in the 1980s. As the title suggests, he is offering a general account of the conditions for knowledge, of which his account of the conditions for epistemic justification forms only one part. That one part is the major part, however, being dealt with at greatest length in the book's central chapters. Since my concern is only with Moser's foundationalist account of justification, we can pass over the details of his account of the nature of belief and truth, his remarks on practical rationality and on the nature of knowledge in general, in order to focus on his account of justification. I believe Moser's version of foundationalism to be the most sophisticated of its type, so it is well worth investigating.
To begin, let us highlight the major features of Moser's foundationalist account.
Moser accepts the traditional definition of knowledge as justified true belief. In the context of this minimal definition he presents his analysis of what is required for the justification condition to be met. If a proposition is to be counted as knowledge, then it must be justified. So, what makes a proposition justified for someone? Moser actually starts by asking a simpler question: What makes a proposition justifiable for someone? First we must distinguish a proposition's being justifiable for S and S's being justified in believing a proposition. There is a difference between justification as justifiability and justifiedness, the former requiring weaker and the latter stronger conditions. If we can lay out the lesser requirements for justifiability, then we can focus our efforts on what must be added for justification.
Moser first connects the concept of justification to the concept of probability. We speak of propositions as being more or less justified, indicating a connection between justification and probability. So we can recast the question: What makes a proposition probable?
Moser rejects all accounts of probability that involve propositions' being self-probable, propositions' depending for their probability upon an infinitely regressing chain of probable propositions, or propositions' being made probable by a circular chain of propositions. The most common reasons for rejecting these have been well-rehearsed in the literature, and Moser is not concerned to repeat those reasons. His primary objection to all such accounts is that — even supposing they could meet all the commonly raised objections — they are much too narrow to be general accounts of probability. All such versions make probability solely a matter of relations between propositions, and Moser is concerned that they leave out of the picture any sort of nonpropositional probability-makers. The reason for his focus on nonpropositional probability-makers is two-fold. First, most foundationalist accounts of justification attempt to ground justification in some nonpropositional psychological states, such as sensation or perception, and Moser places himself in this tradition. Second, Moser takes the failure of all entirely propositional accounts of probability as indicating that a complete account of probability will have to involve an account of nonpropositional probability-makers.
Since Moser's foundationalism attempts to ground justification in experience, his candidates for nonpropositional probability-makers are subjective experiential states. Seeming to sense and seeming to perceive are such states. (Moser includes both sensation and perception in his discussion, but for brevity I will shorten "sensation and perception" to "perception.") Perceptual states are taken by Moser as subjective because they are states of the subject and do not necessarily involve the objective existence of the apparent object of sensation or perception. (This manner of describing the status of sensory/perceptual states will be the subject of one of my major criticisms of Moser's foundationalism.) Such experiential states are also nonconceptual states, since being in such a state does not involve the application of a concept. Concepts, on Moser's view, are classificatory structures or items that a subject can impose upon his sensory or perceptual experiences. One way of making the distinction between sensuous/perceptual states and conceptual states is to note that the subject is psychologically passive in the former in comparison with the latter. Moser may not want to say that perception is an entirely passive process, but it does not involve active manipulation or classification of what is presented. Conceptual states involve precisely such active manipulation and classification.
Perception and conception are distinct modes of awareness. Even if one does conceptualize one's perceptual experience, Moser notes, this does not alter one's perceptual experience. There is interaction between the two sorts of states, since conception draws at least some of its material from perception and since conception can direct where one focuses one's perceptual mechanisms, but the two are modularized states.
Taking sensory/perceptual states as nonconceptual, a further distinction must be made. It is possible, Moser notes, for one to perceive something without noticing that one is doing so. One can, for example, be reading a book and not notice at all the constant rumble of road traffic outside one's window. And then, a moment later, one can attend to the sounds of the traffic. Both of these are perceptual states, and yet the distinction between the two is crucial for Moser's account of nonpropositional probability making. Moser takes an internalist position, which requires that one be aware of whatever is justifying evidence for a proposition if that proposition is to be justified for one. In the case of a subject who never noticed the rumble of traffic, the fact that his ears were stimulated by the sounds would not be evidence for him that trucks and cars are driving past his window. For perceptual awareness to make a proposition probable, the subject must attend to it. We want to have more going on in justification of basic propositions than just passive stimulation of sensory receptors, and yet we don't want that extra justificatory activity to be conceptual.
Moser uses the concept "attention-attraction" to designate the psychological relationship of a subject who is alert to his perceptual states. The subject has his attention drawn to some feature of his experience. Once his attention is so drawn, that experiential item can play a justificatory role. This is distinguished from attention-focusing. Attention-focusing is a more active process on the part of the subject; the subject initiates and maintains the focus. The difference between attention-attraction and attention-focusing is, for example, the difference between one's psychological state that results from a nearby car unexpectedly backfiring and one's psychological state when one stares at some feature of one's visual field. One is attending to some feature of one's perceptual field in each case, but one is psychologically active in the latter case in a way one is not in the former.
Perceptual attention-attracted states are at the base of Moser's justificatory structure. They justify propositions that stand in a certain relationship to them. In one sense, the relationship between a proposition such as "That's an orangutan" and a perceptual state such as seeming to see an orangutan seems unproblematic. As James Van Cleve puts it, "if I am having the experience of seeing a tree but not that of seeing an elephant, am I not (other things being equal) more justified in believing in a tree than an elephant?" Yet it is just such justifying relationships that nonfoundationalists have traditionally denied, and it is against such denials that Moser presents his new (at least in this context) account of the nature of that justifying or probable-making relationship.
Moser's proposal is to make explanatory relationships the source of the justification for a proposition given a perceptual state. Moser reverses the usual procedure of cashing out explanation in terms of making-probable by explaining making-probable (i.e., justifiable) in terms of explanation. The central thesis here is that propositions attain some minimal probability simply in virtue of having explanatory power relative to some experiential state.
[O]ne's subjective nonconceptual contents can make a proposition, P, evidentially probable to some extent for one in virtue of those contents' being explained for one by P in the sense that P is an essential part of an explanation for one of why those contents exist, or, equivalently, why those contents occur as they do.
If a proposition can explain to some degree one's subjective perceptual states, then it is to that degree probable for one. And since justification is a matter of being made probable, the proposition is thus justifiable for one to that degree. This is not yet to say that the proposition is justified for one or even that the proposition is a better explanation for one than any or all other possible propositions. The proposition has so far acquired only some minimal probability — but this is a big step, for we now have connected propositional probability to nonpropositional perceptual states. Now the project is merely to show how to "jack up" the degree of probability of a proposition to the point that it is justifiable for one to believe it, and then to add whatever it takes to get one to the point where one is in fact justified in believing it. This will involve, as the next two steps, stating what an explanation is and explaining what it is for a given proposition to be a better explanation than another.
Explanation is explained in terms of making understandable. A proposition explains when it makes the thing being explained understandable to some extent.
a proposition makes certain subjective contents' occurring as they do understandable to some extent if and only if anyone who assents to that proposition as a direct result of directly experiencing those contents will thereby understand why those contents occur.
The subjective perceptual state is taken as something requiring explanation. A proposition explains the state to some degree if it can make the subject understand why he is in the state he is in. A proposition capable of conferring such understanding is thereby minimally probable, or justifiable to some extent, for the subject. So Moser has connected justification to making-probable, grounded making-probable in explanatory relations between perceptual states and propositions, and cashed out explanation in terms of making understandable. Next we need criteria for better and worse explanations, in order to be able to isolate those propositions it is most justifiable for a subject to believe. We will of course have to have notions of better and worse explanations that do not depend on the notion of probability in order to avoid circularity, since probability is being cashed out in terms of explanation.
To clarify the notions of better and worse explanations, let us work through two related cases: why ordinary external world propositions are better explanations for one's perceptual states, and how Moser uses his explanatory account to defeat skepticism.
First, ordinary physical-object propositions. Suppose Mr. Doxas seems to smell an apple pie. His attention is attracted to the scent-experience. None of Mr. Doxas's other beliefs or other features of his experiential state lead him to doubt the proposition, i.e., there are no underminers. Then Mr. Doxas's sensory state is best explained by the physical object proposition that there is an apple pie around. Accordingly, that ordinary, external-world object proposition is justifiable for him.
So, the proposition "There's an A," where A is some posited physical object, is justifiable for S if
(i) S's attention is attracted by an apparent perception of an A,
(ii) the proposition "There's an A" can make understandable why S seems to perceive an A, and
(iii) nothing in S's experience suggests to S that no A is present or that the explanatory relation of (ii) does not hold.
In such cases, the proposition "There's an A" is the best explanation of S's experience, and that is all there is to justifiability. The subject has an experience. Given that experience, some propositions — those with explanatory power relative to those experiences — will have some probability given the experience, but one of those propositions will better explain the experiences and so will be more justifiable for the subject than the rest. In the normal perceptual case, the best explanatory proposition will be an ordinary external-world proposition. Thus, ordinary external-world propositions come to be justifiable for a subject.
This is not yet to say that the subject is actually now justified in believing the proposition; that requires more to be laid out shortly. But what we have so far is enough to see how Moser takes his account as answering the skeptic.
About a given apparent perceptual experience you may have, skeptics will offer any number of explanatory propositions. It could be, for example, that you are dreaming or hallucinating. Or you could be a brain in a vat. Or an evil demon could be playing a little joke on you. Or it could be that you are perceiving a physical object that exists independently of you. Given all of these possibilities, how could it be justifiable for you to believe any one of them and not the others? Moser's point is that your experience all by itself gives you something to go on in evaluating the merits of these proposed explanatory hypotheses. To grasp the general pattern, compare the evil demon proposition with the physical object proposition. Both have some explanatory power given the experience, but not equally so because the evil demon proposition makes reference to items not contained in any sense in the given experience. It posits gratuitous entities, while the physical object proposition does not. As such, given the experience, the evil demon conjecture has less going for it and so is not as good an explanation. Thus the physical object proposition is a better explanation and so more justifiable.
Better explanations do not posit gratuitous entities. A gratuitous entity is one not represented in any manner whatsoever in the item to be explained. Any skeptical explanations that arbitrarily posit entities are thus explanatorily weaker. Moser does not take the arbitrariness of skeptical objections to rule them entirely out of court, but he does claim that it is legitimate to set aside as weaker the skeptical explanations when compared to the explanations provided by ordinary object propositions. Better explanations for one's perceptual experiences are ordinarily provided by physical-object propositions, and so the skeptic is simply wrong to conclude that it is unjustifiable ever to believe physical-object propositions. If one proposition is clearly a better explanation than another, then it is more justifiable to believe it.
So far we have an account of what it is for a proposition to be justifiable for a subject. The traditional definition of knowledge that Moser wishes to uphold requires that the proposition be justified for the subject and not merely justifiable. So we need to seek out what in addition is required in getting from justifiable propositions to justified propositions. There are three conditions.
First, a minimal condition: the proposition must be more probable, given the evidence, than its denial. This rules out one's being justified in believing a proposition when one has only weak explanatory hypotheses available.
Second, the subject has to consider and understand the proposition.
Third, since justifiability does not require that the subject has associated the proposition with its supporting evidence (the subject only has to have the evidence for justifiability), and since Moser wants to add the internalist requirement of the subject's grasping in some sense the relation between the evidence and the proposition it justifies, justification requires justifiability plus the subject's having associated the proposition with its evidence. Some sort of awareness of the connection between the two is required. This is where things get tricky.
If Moser were an externalist, then his account of justification could end with the second condition added to justifiability. The subject would have an experience, one proposition would be a better explanation for that experience, the subject would consider and understand that proposition and thus be justified in believing it. But as an internalist Moser wants something more: the subject must grasp the significance of the nonpropositional experiential evidence relative to the proposition it justifies, the subject must be aware of the connection between that evidence and that proposition, the subject must believe the proposition because it stands in a certain relation to the evidence — the point can be stated a number of ways. This requirement raises a difficult question, though: What exactly is involved in this grasping of the relation between evidence and proposition? So far this notion is vague.
Before exploring this as yet vague notion and the possible criticisms of Moser's position that arise from it, let us end this expository section by summarizing Moser's account of justification, highlighting its foundationalist features and contrasting its distinctive features with those of other prominent foundationalist positions.
Moser's position is foundationalist because all justification stems from a starting point: all justification (at least, all empirical justification) is traced to nonpropositional, perceptual experience. Not all justification is a matter of relations between propositions, contrary to the claims of coherence theories of justification, for one species of justification is a matter of relations between nonpropositional experiential contents and propositions. Perception can give rise to justified propositions if those propositions have strong explanatory power relative to the perceptual experiences. Then, given such basic, justified propositions, other propositions can be justified by means of deductive and other inductive inferences. Justification, then, on Moser's account, exhibits the two-tiered structure typical of foundationalism.
Moser's account is not typical in all respects. No certainty is required at the basis of the justificatory structure, in contrast to the foundationalisms of Descartes, Lewis, Chisholm, and others. All that is required is that the proposition be a decisively better explanation than its possible competitors. Nothing in the system requires or makes use of self-justifying propositions, in contrast to one way of interpreting Chisholm. Subjects also don't have to start with explicitly subjectively qualified propositions such as "I seem to perceive an X," as the subjects of Chisholm's & Lewis's systems do. Such propositions can be incorporated into the Moserian account, but subjects can also start with propositions of the form "There is an X"; and given the fact that most people proceed as direct realists, cutting out intermediate subjective propositions and proceeding directly to external-world propositions allows the system to be psychologically more realistic than versions having only propositions of the form "I seem to perceive an X" as the basis. And it seems to allow for a tidier disposal of skepticism, in the process allowing young cognizers and those who have never taken philosophy courses to have justified beliefs without ever having worried about Cartesian demons and being brains in vats.
Let us return to the question of the nature of the subject's grasping of the relationship between the experience and the proposition justified by it, for this is a natural lead in to several possible criticisms of Moser's approach.
First let us combine the conditions for justifiability with those for justification to work out an illustrative example. Suppose our subject, Ms. Credo, is facing a blue book in normal light with her eyes open. In stages, then, what happens is as follows:
1st: Ms. Credo experiences blue-bookly.
2nd: Ms. Credo's attention is attracted to the blue-bookish experience.
3rd: Ms. Credo conceptualizes her experience with the appropriate appearance qualifiers.
4th: She asks herself, Why am I experiencing blue-bookly?
5th: She generates at least one explanatory hypothesis in propositional form. In this case the proposition will most likely be something like "There's a blue book."
6th: Ms. Credo understands the proposition(s).
7th: She evaluates the explanatory merits of the proposition(s).
8th: She believes the best one, in this case most likely "There's a blue book."
This doesn't seem to be what Moser wants. For one thing, unless Ms. Credo is a high-powered epistemologist, it is psychologically unrealistic to expect that she will go through all these stages. That is just not the way people function, as Moser and many others are well aware. This sort of construction would raise a host of hard questions. What is the source of the necessary concepts and qualifications required by the third stage? Does she already have to know or believe or be aware in some form of the fact that sometimes there are hallucinations and illusions and sometimes there are veridical perceptions, and that the former two are epistemically dangerous but not the latter? How, in the fourth stage, does Ms. Credo come to conceive of her experience as something requiring explanation? What explains her grasp, in the seventh stage, of the criteria for better and worse explanations? The problem, generally, is that this construction requires that cognizers have an entire epistemology at their disposal prior to having a single justified belief. It overintellectualizes the process, and unless one is a nativist in a big way this is not an appealing option.
The flip-side of the problem with this construction is that if a theory insists that cognizers actually go through all of these stages prior to having actually justified beliefs, then it runs the risk of being, in Moser's words, inexcusably unkind to children and other possibly intelligent animal species by implying that they have no justified beliefs. If it is unrealistic to expect adults to go through all eight stages, then it is wildly so to do so for children and some animal species. For Moser's purposes, this reconstruction won't do, so some less demanding criteria are required.
The major problem seems to be with the seventh condition, involving Ms. Credo's evaluating the explanatory merits of the proposition(s) in question. There is a tension here, for on the one hand Moser's internalism leads one to expect that subjects must be aware of the significance of the evidence to the proposition, which, given his account of justifiability, seems to imply that the subject must have some background grasp of what counts as a better or worse explanation; while on the other hand, Moser positively does not want to have his account presuppose any form of background knowledge on the part of the subject, because doing so would make his account circular and also cause his view to be unkind to young cognizers. Propositions are to be justified solely on the basis of experience, so no background knowledge can enter the picture. But if explanatory relations are the crux of justification, then internalism requires that the subject have a handle on explanatory relations, and it is hard to see how this "handle" could be anything other than a form of conceptual knowledge. The tension, then, is between Moser's internalism and his explanatory account of justification.
Moser is aware of this tension and takes care to deny explicitly that the "handle" is any form of conceptual belief or knowledge. Subjects, he notes, do not have to have the concept of a probability-maker for probability-makers to be so for them. In other words, a subject doesn't have to have a belief that experiential contents X make probable proposition P in order for X to make P probable for that subject. The grasping, however we cash it out, must be entirely nondoxastic; it will be de re and not de dicto.
Perhaps an account of concept-formation could help fill the gap. One line of criticism of Moser's account is that it leaves unexplained subjects' abilities to come up with propositions, which are necessarily conceptual. He has provided an account of the relationship between percepts and propositions, but propositions are structured series of individual concepts. Where do the necessary concepts come from? What is their connection to perceptual experience? What legitimates their use? Moser is not the only foundationalist to whom these questions can be addressed, since most foundationalists do not provide an account of concept-formation or seem to think one is needed in a theory of justification. Accounts of concept-formation try to account for how subjects abstract from and universalize their determinate and particular perceptual experiences, and whatever is involved in the process must involve some nondoxastic, nonconceptual form of sorting and integrating perceptual contents. If Moser were to explore further this apparent lacuna in his account, it could clarify his notion of nondoxastic awareness of the relationship between experiential evidence and propositions.
Another alternative open to Moser, in order to go straight through both horns of the dilemma, is to posit the nondoxastic awareness as an irreducible primary. Subjects can grasp connections between their nonpropositional experiences and propositions, and that's that. So instead of having to describe in detail what sort of grasping it is and what it depends upon, one can only indicate the philosophical pressures and considerations that lead one to posit it, describe it in general (and negative) terms — it's nondoxastic, nonconceptual, nonpropositional — end of story.
Or perhaps Moser could get rid of the problem by adopting a position we might call "externalist inter-nalism." Externalism in general gives criteria too weak for justification, according to Moser. At best externalism gives criteria for justifiability. If the subject experiences X-ly, understands the proposition "There is an X," there are no underminers of that proposition in the subject's experiences or beliefs, and the subject's believing the proposition resulted from his experiencing X-ly via some nondeviant causal chain — then according to externalism the proposition is justified for that subject. This does not require that the connection between experience and proposition be grasped by the subject; for all the subject knows, the connection between the two could be entirely fortuitous; and so internalism wants something more: the entire justificatory story has to be within the subject's cognitive grasp — including the explanatory relationships, on Moser's account. What if we said the following? The explanatory relation has to be present to the subject's awareness, but we don't have to demand that the subject attend to it. All we have to require is that the subject has access to it in principle. In principle the connection would be accessible, so pure externalism is avoided, yet we satisfy internalist requirements while avoiding the pressures of making the awareness of the connection a doxastic state. So the subject needn't be aware of the full explanatory significance of the proposition relative to his experiences and can still be justified in believing the proposition. This squares with one of Moser's statements to the effect that the subject does not have to be aware of the fulfilling of a condition for justification (in this case, the explanatory relationship is strong); the subject just has to be aware of the evidence. The fact that the proposition explains the evidence is accessible to the subject, should the subject come to attend to it, so it is not in principle external to the subject's awareness.
Yet this may be giving more to externalism than Moser wants, since it could be pointed out that from the standpoint of the subject, the connection between evidence and proposition could still be fortuitous. In addition, if Moser concludes the justification need only be accessible to the subject in principle, then he is concluding that the subject has access in principle to the fact that his perceptual experiences are subjective states and that they are items requiring explanation. But these, I do not think, are accessible in any significant way to the average subject. These are very sophisticated philosophical conclusions. (Witness introductory philosophy students.)
Further, such an externalist internalism does not square with Moser's more strongly internalist statements to the effect that the connection has to be more than available in principle, when he says that the subject must associate the proposition and the evidence in the appropriate manner. In his central section discussing the transition from justifiability to justifiedness, he characterizes justified propositions as those which are justifiable and for which the subject has related the evidence E and the proposition P in a certain way; the certain way is characterized by two conditions:
(i) the subject has a de re awareness of E's supporting P, and
(ii) the subject is in a dispositional state such that if he were to consider the evidence for P, he would focus on E.
Condition (i) states the stronger internalist requirement. Since the justificatory support in question is P's explaining E, this means the subject has a nondoxastic awareness of explanatory relations.
If this is the right way to interpret Moser, then we are still left in need of a clarification of what a nondoxastic grasping of explanatory relations is like. This project sounds difficult. Explanation seems a conceptual phenomenon in the most easily understood cases: relating one conceptually identified phenomenon to a broader context. This presupposes a why-question, which is another conceptual phenomenon. Explanation thus seems more a coherence phenomenon, and so seems out of place at the basis of a foundationalist account. Moser of course denies that explanation generally is either a conceptual or a coherence phenomenon. But unless it is made clear what this awareness is like, we can't evaluate this aspect of his account. And Moser is not altogether clear on this. We get some idea of what it is not supposed to be: no knowledge is involved, no beliefs are involved, no epistemic elements in general on pain of a regress. On the positive side, we know it does involve some notion of evidential support, but that is all. Moser's account is poised on the borderline between conditions for justifiability and justification, and it seems that if he makes the conditions stronger, he must smuggle in conceptual background understanding; but if he leaves the conditions weaker, no conceptual background understanding is presupposed but we end up with only justifiable beliefs; and if he adds a nonconceptual grasp, we seem to be left in the dark as to what it is. None of these options are satisfactory, given Moser's stated purposes.
The above questions arise from Moser's attempt to combine internalism with an account of nonpropositional justification in terms of explanation. As such, they focus on features specific to Moser's account. Yet I do not take these to be the most important problems. What I take as most problematic features are not those which distinguish him from other foundationalists, but rather the same general approach to defending foundationalism that he shares with virtually all other foundationalists this century. It is in the context of identifying this shared approach and criticizing it that I will continue my critique of Moser's foundationalism.
* * *
[1958, p. 187]; emphasis added).
 The first phrase is Scruton's (1981, p. 37), the second Windelband's (1901, pp. 276-ff and 390-ff.).
 It is significant that the subtitle of Descartes's second Meditation is "Concerning the Nature of the Human Mind, and
how it is more easily known than the Body."
 Kelley (1986, p. 184) makes this point.
problem of interaction for dualism.
take subjective propositions as the foundation.
 Quine (1969).
of science, are clearly antifoundationalist.
133-134, and p. 19).
it to me in discussion.
 According to Chisholm (1982, p. 69; 1988, p. 232).
 Chisholm (1989, pp. 72-73).
 Quoted in Chisholm (1982, pp. 186-187).
 For example, Moser (1989, p. 11).
 Chisholm (1989, pp. 1-4; also 1982, p. 75).
 Chisholm (1988, p. 232).
probability, must themselves be certainties."
 See Chisholm's Material Principle 1 (1989, p. 62).
 Chisholm (1989, pp. 19-20).
parallel case for perceptual appearing.
(1989, p. 19)].
to be mistaken about the given in some cases.
 Van Cleve (1985, p. 100) is clear on this point.
epistemically 'move' itself?" (1978, p. 5).
or subjective subject matter" (1982, p. 155).
 Chisholm (1989, p. 22).
the same general approach "Cartesian empiricists."
 See also Moser (1985b, p. 14).
Chapter 5 of this essay.
 Chisholm (1982, p. 144).
 Chisholm (1989, p. 43).
 Chisholm (1989, p. 67).
in the next section.
 Chisholm (1989, p. 9).
 Chisholm (1989, p. 10).
 Both from Chisholm (1989, p. 63).
makes it so."
(1989, p. 64).
 Chisholm (1989), p. 16 and p. 8, respectively.
 Chisholm (1989), pp. 41, 11, and 11, respectively.
 Chisholm (1989, p. 65).
 The words are Carneades's, as quoted in Chisholm (1989, p. 66).
(1977, pp. 82-84).
 Chisholm (1989, pp. 72-74).
 Chisholm (1989, p. 73).
its very nature a faith cannot be rationally defended.
 Sosa (1980, p. 561) thus suggests that perhaps we can reject the dichotomy.
 Chisholm (1989, p. 21).
proaches. Chapter 3 argues this point.
 Firth (1964, p. 547).
(1981, p. 11). I agree.
 Chisholm (1982, pp. 141-145).
ics." This issue is discussed in detail in Chapter 4.
 Chisholm grants that the phrase "at least as justified as" is taken as undefined (1989, p. 12).
 Chisholm (1989, p. 63).
 Chisholm (1989, p. 55).
 The minimal level Chisholm has in mind is prima facie probability (1989, p. 63).
Chapter Six of his (1989a).
 Moser (1989a, pp. 52-63).
 Moser goes into the details of the shortcomings of entirely propositional accounts in his (1988a).
 Moser (1989a, p. 7).
 Moser (1989a, p. 80).
 See for example, Moser (1988b, p. 193).
 Moser (1989a, pp. 81-ff).
 Van Cleve (1985, p. 95).
explicit in Lewis' writings."
 Moser (1989a, pp. 91-92).
 Moser (1989a, p. 93).
 Moser (1990, p. 132); also see his (1988a, p. 242).
 See, for example, Moser (1990, pp. 131-132).
posit is gratuitous. (Moser 1989a, pp. 162-163.)
 Moser (1989a, pp. 97-98; for details see pp. 158-165). Also see (1988f, pp. 138-ff) and (1990).
 Moser (1989a, p. 126).
 Moser (1990, p. 132).
 Moser (1989a, p. 145); Alston (1976a, p. 178); Pollock (1986, pp. 175-177).
 Moser (1989a, p. 158).
 Moser (1989a, pp. 106-107).
 This issue is dealt with at greater length in Chapter 5 of this essay.
 Alston (1989, pp. 233-234), generates an infinite regress from any sort of doxastic requirement added at this point.
 Moser (1989a, p. 111).
 Moser (1989a, p. 157).
 Moser (1989a, p. 141).
 Moser (1989a, p. 142).