A DIRECT REALIST AND DEVELOPMENTAL ACCOUNT
* * *
* * *
1.2 Motivations for foundationalism
1.4 The structure of the essay
1.5 What this account of foundationalism will look like
Foundationalist accounts of justification claim that justified propositions are divided into two classes. Some propositions are independent, not deriving their justification from other propositions. The rest are dependent, deriving their justification ultimately from the propositions that do not depend upon other propositions for support. The former are, accordingly, typically called "basic" propositions; these serve as the foundation for the rest. Relations of justification are ultimately one-directional, tracing their way back to the foundational, basic propositions. Questions of justification thus have a stopping point, eventually reaching the point where all one can state in response to a justificatory query is some proposition such as, "It's self-evident," or "Look at this test tube," or "Listen!" Foundationalist accounts of justification, then, make two claims about the structure of justification: it is hierarchical and it is terminal.
Antifoundationalists, by contrast, claim there are no basic propositions. All justified propositions are justified in relation to others. Every proposition depends upon others. Relations of justification are multi-directional, and no proposition is ultimately more fundamental than any other, which implies that there are no terminal points in a set of justified propositions.
During the 20th century defenders of foundationalism have been in the small minority. There are some signs, however, that this situation may be changing. Revivals of old and offerings of new versions of foundationalism have begun to appear in the literature, though nonfoundationalists continue to be in the majority. In this essay I present and defend a version of foundationalism seldom discussed.
1.2 Motivations for foundationalism
Why have philosophers traditionally thought that knowledge must have a foundation? A common negative argument begins by noting that some of us (at least) justifiably believe some propositions, and then asking what the phenomenon of justification implies for the structure of our set of beliefs. Initially, the standard response is that two conditions must be met for a given proposition p to be justified: there must be another proposition, q, and q must make p probable or certain. We in turn set the same conditions to the justifying proposition q and ask what justifies it. And to whatever proposition is then indicated, we raise the same question. In the end we shall have to choose between four options. Either the process of justification halts or it doesn't. If it doesn't, then there are two possibilities: (1) the justifying process continues, in principle, on to infinity; (2) the justifying process eventually comes full circle and one ends up citing as justification some of the very propositions one started out attempting to justify. On the other hand, if the justification process does eventually halt, then again there are two possibilities: (3) at some point we discover some propositions that all the rest are held to be justified on the basis of, but which are themselves arbitrarily held, i.e., held without any form of justification whatever; (4) the process eventually arrives at some basic, non-arbitrarily held propositions on the basis of which all the rest are justified. The foundational propositions specified in (4) could either be themselves "self-evident" or "self-justifying," or be "directly justified" (e.g., justified by a transition from non-propositional sensory/perceptual states to propositional belief that does not depend upon other propositions). Crucial to option (4) is the claim that the justification of some propositions — the basic ones — does not depend upon other propositions. Some justification is nonpropositional.
To motivate foundationalism, we form a disjunction of the four options, eliminate as unsatisfactory those involving circles, infinite regresses or arbitrary starting points, and then conclude that option (4) must be true.
It is not possible to hold an infinite number of propositions, let alone an infinite number of inferentially linked propositions, so no advocacy of option (1) can demand that an individual actually possess an infinity of believed propositions. Instead, on this view, as long as each proposition is justifiable by its successor in the sequence, the proposition in question can be considered justified. One has only to trot out as much justification as the situation demands. The major problem with this account, according to foundationalists, is that it yields only conditional justification: p follows from q and so is justified if q is; q follows from r and so is justified if r is; and so on. We never get any actually justified beliefs, merely a long conditional proposition. This does not sit well with foundationalists, so option (1) is rejected.
Circular reasoning proves nothing, we learn in introductory logic classes. A circular set of propositions is a floating, detached body of beliefs, and it is hard or impossible to see how such a set could have any connection to reality. If the purpose of justification is to help us achieve truth, and truth depends on the way reality is, then justification requires input from reality. A closed system, which is what our circle is a metaphor for, has no such input. Hence, option (2) is rejected.
If arbitrarily believed propositions are allowed into a justificatory structure, then any
proposition whatsoever can be justified, thus collapsing the distinction between justified and unjustified propositions. Thus, option (3) is out.
This leaves option (4). To adopt (4) is to say that the existence of the phenomenon of justification requires that there be basic propositions that serve as the foundation upon which the entire structure of knowledge is built. This is the foundationalist position.
At this point the big question for foundationalism is, What is the nature of these basic propositions? Where do they come from? They are said to be justified, not by means of other propositions, but rather by being "self-justifying" or "self-evident" or "directly justified" — what does this mean, and how is it possible? If a given proposition is a "basic" proposition, is it infallibly justified and forever immune from revision? Will it always remain a basic proposition? Questions of this sort form the primary set of problems for the foundationalist to address. The onus is on the foundationalist to provide a positive account of such propositions and to defend it against the many critiques purporting to show its impossibility.
There is a further question for foundationalism: How do the basic propositions in turn confer justification upon other, nonbasic propositions? This, I think, is by far the easier of the two sorts of question to offer answers to and is therefore not where most of the controversy lies. Once one is in the propositional realm, so to speak, what we know of logic makes it seem much simpler to show how justification can be transmitted from proposition to proposition. Even so, a foundationalist cannot offer an account of proposition-to-proposition justification in isolation. One's approach to this issue depends on one's account of the nature of basic propositions, since whether one emphasizes inductive, deductive, or some other sort of connections depends on how much one sees the basic propositions as giving one. More on this to come.
I will be focusing almost exclusively in this essay on the question of where the basic justified propositions come from. The question of how further justified propositions can be derived from basic propositions will not be dealt with except in passing and in those cases in which such an account is crucial for seeing how a particular version of foundationalism proceeds.
In addressing the problem of nonpropositional justification, it is typical for foundationalists to appeal to sensation, perception, or experience as the source of such basic propositions, to argue that sensation, perception or experience provides direct, non-propositional justification for the basic propositions. Everyday experience suggests why this is a popular move. If I want to justify my propositional belief that I have indeed once again locked my keys in my car, I look in the window, see the keys dangling from the ignition slot, and unhappily consider myself justified in my belief. Nonpropositional perception, on the face of it, has justified a proposition.
But, critics have wondered, how can there be any such thing as non-propositional justification? "Does it follow from the sense-impressions that I get," asks Wittgenstein, "that there is a chair over there? — How can a proposition follow from sense-impressions?" Richard Rorty, agreeing with Wilfrid Sellars, who agrees with Wittgenstein, says no such thing exists. Laurence Bonjour, agreeing with all of them, argues that the notion is "more than a little paradoxical," and so on. What are their objections and how can they be met? If the basic propositions are based on perception and justified by it, then must we hold that perception is propositional? Not all versions of foundationalism do — but then, how could a non-propositional percept "justify" a propositional perceptual judgment?
In addition to local concerns about avoiding justificatory circles, infinite regresses and arbitrary starting points, there are three global epistemological considerations that have traditionally motivated versions of foundationalism: preserving objectivity, universality, and certainty in human knowledge. A few words on each.
Objectivity in knowledge means the object, and not the knowing subject, sets the terms. One's focus as a knower is radically outward; the subject is essentially passive. And since our senses are our root points of contact with reality, a premium is placed on sensory evidence and on showing how the rest of our knowledge stems from it. This is of course the empiricist commitment, which is why foundationalism has often been seen as best preserving any commitment to empiricism. Knowledge of reality comes from the senses, and objectivity requires preserving the justificatory connection to the senses throughout the structure of one's knowledge.
For knowledge to be universal the same standards for what counts as knowledge must hold for every person. This will naturally include our standards for justification. The
opposite claim is that of relativism: standards for justification and knowledge are individual- or culture-specific. In contrast to relativism, foundationalism is motivated in part by wanting to discover a universal methodological procedure for determining what's justified and what's not. Add to this a bedrock level of justification, an ultimate evidential court of appeal, and relativism has no place to stand.
Skepticism has been at the forefront of epistemology (and philosophy more generally) since Descartes, and the desire to avoid it has motivated all versions of foundationalism since. In Descartes's case, this motivation was explicit: find a secure base, something indubitable, something certain, so we can say we know things about reality. Descartes thought the skeptic could be beaten decisively; he is in the minority. Chisholm speaks for the majority when he claims that skepticism is not refutable: "we can deal with the problem only by begging the question," by assuming we do know some things and proceeding from there. This, he grants, is an arbitrary starting point, but the skeptical option he thinks is equally arbitrary; and if our choice is between two arbitrary postures, why not go with the one that best squares with our pre-analytic intuitions? Chisholm is more pessimistic than I am about our chances of disposing of skepticism, but I think he is right that if foundationalism turns out to be incoherent, we may have to embrace skepticism.
1.3 Foundationalism's competitors
As the project has traditionally been conceived, the above are the considerations motivating foundationalism. And so efforts have been focused on the two desiderata seen as required by a complete theory: on showing how basic propositions can be validated or justified or grounded in experiential states, and on showing how basic propositions can be used to derive other, justified propositions. The former, involving as it does sense-perception, has been conceived of as the more empirical wing of the project; while the latter, most often seen as involving methods or rules or principles of inference, has been conceived of as the a priori or analytic wing of the project.
Both projects have been subject to severe criticism. On the one hand, philosophers such as Sellars and Rorty argue that attempting to justify propositions on the basis of experience arises from a confusion of causal and justificatory accounts of knowledge, and so should be abandoned. And on the other hand, philosophers such as Quine argue that the whole notion of analytic truth does not make much sense and that we should, accordingly, dispense with it. The result of combining these two types of attack is that in the 20th century the traditional model of foundationalism has come to be viewed in most quarters as a dead theory.
In its place, many varieties on nonfoundationalism have arisen. Any theory of justification that rejects the claim that justification has a terminating hierarchical structure can be classified as nonfoundationalist. Justification is then typically held to be either solely a matter of networking propositions, of coherent integration, or, on more pragmatist accounts, of coherent integration plus considerations of workability; metaphorically, the discussion is in terms of webs or rebuilding rafts while at sea, rather than the skyscrapers or pyramids foundationalists speak of.
On such accounts what justifies any given belief is its cohering with a comprehensive set of other beliefs. The belief in question does not have to be itself infallible or indubitable, nor need it have been inferred validly from other infallible or indubitable beliefs. The system of beliefs as a whole is the unit of justification; any given belief is justified to the extent that it can be integrated with the rest. All beliefs are dependent upon all of the others for their justification; there is no linear or hierarchical ordering.
Obviously, the coherence theorist continues, we come to have a lot of beliefs due to perceptual experiences. Perception is a major contributing source of beliefs, but perception does not provide any justification for them. Perception is the cause of such beliefs, but it is never their justification. We must make a sharp distinction between causal and justificatory questions. Justification arises out of the mutual consistency of the given beliefs. The beliefs perception causes may be considered as "truth-candidates," and can be taken as having some initial plausibility. One tries (or should try) via inferences and explanations and various massagings of the data to come up with a maximally coherent system. Those items that cannot be integrated with the majority are written off. All of the remaining beliefs fit, in the sense that any one of them can be shown to cohere with the others: it may be an entailment of some, it may stand in an explanatory relation to others, and so on.
That is the general coherentist picture. At this point there are some options. One can be an individualist coherentist, arguing that the locus of a justified framework is within an individual mind. Or one could opt for a collectivized version of coherentism, as Lehrer does, arguing that the system of beliefs that is the unit of justification is larger than one's personal set; it includes others in society. Justification is a social product, on Lehrer's account, which requires asking of any belief whether it coheres with what others believe.
Whatever the unit of justification, the final claim for many coherence theorists is that the maximally coherent set of beliefs must be true. Compare being given 10,000 jigsaw puzzle pieces to put together. You work at it, and eventually find some that fit together. Eventually an overall picture emerges. However, you come across a few pieces that just don't fit, no matter what. These you throw out, leaving all the pieces that can fit together, fitting together, and those that can't, discarded. The puzzle has been solved. That must be the true picture. It must be the true picture, according to coherentists, since that is the best explanation for the system's coherence.
Coherentism is not foundationalism's only opponent on the current epistemological scene. Rounding out the line-up are two approaches that see foundationalism and coherentism as having in common some significant basic errors. One opponent is the cluster of closely related views variously named reliabilism, naturalized epistemology, and evolutionary epistemology. The common denominator of these views is the reduction of epistemology to the natural sciences. "Justification" and "knowledge" are not, on these accounts, concepts that have a subject matter independent of what the special sciences will tell us about our belief-forming processes. Traditional epistemology, on these accounts, is a product of attempting to answer skepticism without drawing upon the resources of the special sciences. As such, it has set itself up as an autonomous area of investigation. The various naturalized epistemologies reject this in favor of attempting to reject skepticism from within the context of what science teaches us. And here, as Quine puts it, "[t]here is some encouragement in Darwin." Evolutionary theory tells us that organisms that are not well-adapted die off; only well-adapted beings survive and reproduce. We humans have survived and reproduced, so we must be well-adapted. Since our primary survival mechanism is our belief-forming mechanism, it must be a well-adapted byproduct of natural selection — "well-adapted" in this context meaning that it predisposes us to believe truths.
I see such naturalized epistemologies as a subset of the set of representationalist theories of knowledge, which will be discussed in greater detail in Chapter 3. Some representationalists attempt to answer skepticism by searching for an external guarantor for our cognitive faculties: In naturalized epistemologies, natural selection plays roughly the same role a benevolent God plays for Descartes's representationalism. As such I will discuss these theories in Chapter 3 in the broader context of rejecting any form of representationalism.
The other approach is that of the antirealists, for which Rorty has been the most prominent spokesman. What foundationalism and coherentism share in common, according to Rorty, is a commitment to the Kantian project. The view so-named gives epistemology a central seat in the halls of knowledge, for its task is to give us a universal and ahistorical context into which every other inquiry can be placed. Epistemology tries to find necessary starting points and an absolute framework within which to construct the edifice of human knowledge. The Kantian thesis, according to Rorty, is thus that philosophy is a foundational discipline; this is precisely the view that informed peoples everywhere need to reject.
Foundationalism so conceived certainly implies an account of justification, one that Rorty thinks the coherentists have critiqued correctly. Foundationalists strive for correspondist accounts of justification; to reject this is to follow the lead of Sellars and Quine in recognizing "that nothing counts as justification unless by reference to what we already accept, and that there is no way to get outside our beliefs and our language so as to find some test other than coherence." This is the standard idealist/coherentist point that our beliefs and language are obstacles to our knowing reality as it really is, so thus far the antirealists have no argument with the coherentists.
It is on the subject of truth that Rorty emerges as far more radical than the coherentists. In seeking a universal truth coherentism reveals itself as merely another version of analytic philosophy, as still in the Kantian tradition of looking for all-embracing principles of human knowledge. The radical point for Rorty is that not only are there no system-external justificatory connections (the coherentists got that right), there also are no system-external truth connections (both foundationalism and coherentism are wrong about this). There is no one true system. Coherentists abandon correspondence accounts of justification, yet generally want to maintain correspondence accounts of truth. The coherentists are thus hesitating before the inevitable entire collapse of the foundationalist project, trying to maintain at least a shred of a connection to an external reality. The coherentists still want to be realists, disagreeing with the foundationalists only about overall strategy.
Instead, following the leads of Quine and Sellars, Rorty believes that we should accept the view that the conceptual scheme of our knowledge is entirely a social product and that all standards for what counts as justification or for what we want to call true arise only within the scheme and cannot be directed to the scheme itself. Truth and justification are, in the end, socially subjective; there is no neutral, independent, or foundational benchmark for justification and truth. What is true or justified depends upon the set of beliefs held by one's social community. Nor is there any necessary homogeneity across social communities; here Kant was wrong in positing categories universal to the species. Each community has its own contingent starting points and evolution; as a result, to borrow MacIntyre's wording, "each of these sets of beliefs and ways of life will have internal to it its own specific modes of rational justification in key areas and its own correspondingly specific warrants for claims to truth."
Actually, Rorty feels, it would be best to drop altogether all talk about truth. This is a rather difficult point to put without falling into paradox, as he notes, since in making it he has to "avoid hinting that this suggestion gets something right, that my sort of philosophy corresponds to the way things really are." That would be to fall back into foundationalism.
To say that we should drop the idea of truth as out there waiting to be discovered is not to say that we have discovered that, out there, there is no truth. It is to say that our purposes would be served best by ceasing to see truth as a deep matter, as a topic of philosophical interest, or "true" as a term which repays "analysis." "The nature of truth" is an unprofitable topic.
So there is nothing for epistemology to do — no deep analysis of "justification" or "truth" or "knowledge." These terms are not success words signifying universal types of connections between thought and an external reality; they have only the particular significances that particular communities give them. We must get rid of language that speaks of language as mirroring the world:
we have to understand speech not only as not the externalizing of inner representations but not as representation at all. We have to drop the notion of correspondence for sentences as well as for thoughts, and see sentences as connected with other sentences rather than with the world.
The sentences are a social product, part of a historical sequence which shapes them and gives them meaning. Sentences point to other sentences, which point to other sentences, and so on ad infinitum back into the contingent history of one's social community — and never to a sentence-independent reality.
Epistemology then withers away and is supplanted by a historically sensitive awareness of what has been passed down through the generations of one's social community. Our only real alternative is "to accept our inheritance from, and our conversation with, our fellow humans as our only source of guidance." This will give us the sense that "what matters is our loyalty to other human beings clinging together against the dark, not our hope of getting things right."
Neither this chapter nor this essay is a polemic against antifoundationalist views. But it is useful — for my purposes in this essay — to have at least a sense of what the alternatives are, in order to have a context for some of the criticisms that will be raised against foundationalism by advocates of the alternatives.
1.4 The structure of the essay
In this essay I provide a positive account of foundationalism. To this end the essay addresses three questions: What are the current foundationalist alternatives? What can and has been said against them? What is my position?
First is a presentation of the prominent contemporary versions of foundationalism. Roderick Chisholm is obviously the major figure here, so I will give an analysis of his account. Paul Moser has recently written a number of articles and two books defending a different, though related, version of foundationalism. Investigating the particular strengths and weaknesses of these accounts and isolating their major premises makes up Chapter 2.
I believe Chisholm and Moser offer two variations of one basic approach to foundationalism — what I call representationalist foundationalism. Chapter 3 is devoted to substantiating this claim, arguing that representationalist foundationalism cannot help but fail, and showing how the problems that traditionally motivate representationalism can be handled on direct realist grounds.
Next, in Chapter 4, comes a presentation of the major contemporary critiques of foundationalism — the cluster of arguments that comprise the attack on the given. However, virtually all of the critiques presuppose that representationalist foundationalism is equivalent to foundationalism simpliciter. This I think is a mistake; if I am correct, then this also implies that their critiques of foundationalism are not as general as they hope. Contemporary antifoundationalists and most foundationalists share certain key premises; by the end of Chapter 4 I hope to have made clear what they are and to have indicated what alternatives there may be to them.
Drawing upon the results of Chapters 3 and 4, I present the option I favor in Chapter 5. I will not argue for foundationalism by arguing against coherentism and antirealism. Rejecting false alternatives is fine as far as it goes, but it doesn't go very far. For any account of foundationalism to be satisfactory, what is needed is a positive account of how it is supposed to work, so this will be my primary goal. The negative work of getting rid of the unsatisfactory options may clear the field of errors, but it doesn't teach us anything about why the right position is the right position. So my emphasis is on giving a constructive account of foundationalism. This naturally involves defusing some antifoundationalist critiques, and outlining, comparing and contrasting different competing versions of foundationalism, before getting on with the hard work of presenting positively the version of foundationalism I think works.
1.5 What this account of foundationalism will look like
I argue that cognitively we start with perception, and then form concepts by abstraction from perceptual evidence; the latter is what makes possible our propositional beliefs as well as our grasp of general methodological principles; and the latter of these is what makes possible our inferences from our basic propositional beliefs. This amounts to a thoroughgoing empiricist foundationalism: perception gives us all of our raw material for knowledge, including not only the material from which we form our simple concepts such as "dog," "green," and "running" but also the methodological principles by which we organize, integrate, and analyze our beliefs. So a complete defense of foundationalism needs accounts of four things: (1) what perception is, (2) how the process of concept-formation works, (3) how basic propositional beliefs are arrived at and can be said to be justified on the basis of perceptual states, and (4) how other propositions can justifiably be derived from the basic propositions.
That is a lot. To help keep the project manageable I will devote less space to item (2). It would require, first of all, a thorough discussion of the strengths and weaknesses of realism and nominalism, neither of which is, I think, acceptable. It would also require something I am as yet unable to provide: an account of concept-formation that works. This is because for all of the centuries of discussion of concepts, I think this is still largely terra nova. However, I will outline the general desiderata of a theory of concepts as far as a defense of foundationalism requires, and respond to criticisms of foundationalism based upon incompatible theories of concepts (e.g., Sellars's coherence theory of concepts).
That leaves items (1), (3) and (4), of which I think (1) and (3) are the central and controversial issues. A few words on each.
The foundation will be perceptual states. I defend a direct realist account of perception, and I take perception — not sensation — to be our primary mode of awareness. Perception gives us direct awareness of a world of three-dimensional entities integrated over space and across time with the usual constancies of size, shape, and so on, available to the normal adult perceiver. Here I rely upon David Kelley's defense of direct realism, as well as J.J. Gibson's and the ecological approaches to perception he inspired. Perception is a non-propositional form of awareness; that is, it is not conceptual or subject/ predicate in form. Nor is perceptual awareness constructed by inferential means on the basis of discrete sensations. Perception is direct and cognitive, but not propositional or inferentially constructed.
This is currently a minority position on several counts. Direct realism in perception has never been popular since Descartes and Locke. Further, the view that perception is inferential, i.e., that perceptual states are conclusions reached via conscious or subconscious inferences from immediate sensory data, is currently almost universally believed. And further still, the view that a cognitive state must be propositional is very widely held, especially by foundationalism's opponents.
Next comes item (3). Once perceptual direct realism is established, the next question is how propositional beliefs such as "This is the doggie," and "Here is Mama" justifiably arise from perceptual experience. In the formation of a basic perceptual judgment — a proposition identifying some feature of one's perceptual field — a transition occurs from a non-propositional state of awareness to a propositional one. The crucial feature of my defense will be to challenge the common premise that all justification is propositional. It is on the basis of this premise that foundationalism's opponents rule out as impossible any such phenomenon as non-propositional justification.
Why shouldn't we accept the premise that justification is solely a propositional relation? The basic foundationalist idea I defend is that justification requires awareness of evidence for a proposition. This places no a priori restrictions on what form the evidence must take. Since perception is a form of awareness of reality, it puts one in the position of being aware of evidence — the facts of reality are the ultimate evidence. So the fact that perceptual awareness is different in form from propositional awareness is not the fundamental justificatory issue; it simply means that justification must be of two species, as foundationalists have claimed all along.
In contrast to the usual versions of foundationalism offered this century, e.g., those of Chisholm and Moser, the version of foundationalism I defend will not follow the representationalist pattern those versions do. Older versions of foundationalism were based explicitly upon representationalist theories of perception. This meant that the basic propositions had to be reports of one's subjective states and that ordinary external world propositions were derived from them. Newer versions generally try to describe perception neutrally for epistemological purposes, i.e., without making any commitments about whether perception is direct or not. But since perceptual states have as yet no ontological status, one can justifiably derive external propositions from them only by indirect means, which is the representationalist pattern. One can either make the basics subjective propositions from which external world propositions are inferred, or make external world propositions justified indirectly by virtue of explanatory relations they stand in to the perceptual states which as yet have no status. These two options are taken by Chisholm and Moser, respectively. In each case one works from the "inside" to the "outside."
I do not think one can properly do epistemology in a vacuum, as the epistemologically neutral approach to perception attempts. And, having defended direct realism up front, there is no need to require that one argue one's way to an external world on the basis of perceptual evidence. If consciousness is essentially relational, as direct realism claims it is, then one simply starts out in direct contact with reality and builds from there.
Worries about skepticism play a much greater role in virtually every version of foundationalism since Descartes than they do in the version I will be defending. It is due to the place of honor that skepticism has been given by contemporary foundationalists that they feel constrained to adopt representationalist methodologies. I do not think skepticism is something that needs to be hedged against, or that it is something that implies that one must proceed cautiously for fear of begging questions. I think it can be met head on and set aside as wrong in principle at the outset.
Our last item, item (4), is the issue of how the nonbasic, derived propositions are related to the basic propositions. I suggested above that this is not really the central or controversial issue of the debate, so I will not devote much space to it either. However there is an interesting, related issue that bears upon the entire range of foundationalist/antifoundationalist issues. One way to raise the issue is to ask whether new foundations can be added as one matures intellectually. Can some perceptual experiences one has at age 30, for example, become foundational? Is it possible still to have access to a perceptual given as one grows cognitively and to integrate this new datum into the base of one's structure of knowledge? If so, the structure of our beliefs might, to speak metaphorically, take the shape of a pyramid with an ever-expanding base. Or is it the case that only a special set of propositions could be foundational because, for example, once one has a given network of concepts all perception becomes theory-laden and unsuitable for the status of being foundational? If so, then the structure of our beliefs might, again to speak metaphorically, instead resemble an inverted pyramid. The problem would then be finding enough of the right sort of beliefs to place at the base.
Current defenders of foundationalism seem to feel a pressure to preserve an access to a perceptual given that is completely independent of any concepts or background knowledge. If one is looking around, for example, one wants to be able to do so without any past concepts automatically springing to mind and labeling things; that, it is feared, would prejudice or distort the perception in some way. The idea is that once one has concepts, one has to worry about still perceptually being able to get at things as they really are. Concepts and language are seen as obstacles, as getting in the way of direct perception, as constituting and in a sense distorting the object of our perception. We become victims of our theories and conceptual schemes. This is a very Kantian model, of course, and the impression I get from current foundationalists is that to avoid it and to retain access to a perceptual given, they would like to completely modularize each percept and concept. Each item of knowledge or belief is a completely self-contained unit, unconnected to any other item until an external connection is made explicitly. The connections between items of knowledge or belief are made in a way analogous to stacking bricks to make a house. Placing one brick on another does not alter in any way the nature of each individual brick, yet each brick lends support to the others. This results in a hierarchy of connected "bricks." Yet, in principle, each item in the structure could be detached without destroying its meaning or cognitive significance. Thus we have one (as yet metaphorical) account of the integrated structure of our knowledge.
Antifoundationalists, on the other hand, argue that the integrating of each item of knowledge with others is mutually influencing. Each alters the others, in part coming to constitute the others' cognitive significances. Once integrated, one can no longer sever any item and expect it to retain its meaningfulness.
The debate between the foundationalists and the antifoundationalists on this issue arises from an apparent tension between the hierarchical and contextual nature of the structure of our knowledge. Sellars notes that these two dimensions are those which "epistemologists have sought to capture by the concepts of the given on the one hand, and of coherence on the other." Current foundationalists strongly emphasize the hierarchical aspect: some beliefs are more basic that others, and those others are dependent for their justification upon the more basic ones. But they worry about the contextual nature of the structure of our knowledge, seeming to think that if the items of our knowledge become too integrated with each other, then they will become mutually dependent and the structured hierarchy will be lost. So to preserve the hierarchy, they downplay the integration by making each cognitive item a distinct "atom." Antifoundationalists, noting the same tension between context and hierarchy, emphasize the context, coherence and mutually supporting nature of our knowledge and thus push to eliminate hierarchy and the need for foundations that involves.
There seems to me to be something radically wrong with this debate. The root premise shared by both sides seems to be that context and hierarchy are at odds with each other, and yet I don't see any reason why they should be. So Chapter 5 also investigates this issue in order to see whether the two can be reconciled. It seems right to say that our conceptual knowledge is contextual, that no one item stands completely self-contained. On the other hand, since I am defending foundationalism, I must think there is a hierarchy in the structure of justification, that knowledge has to start somewhere, that there are relations of one-way justificatory dependence between items of knowledge, and that new perceptual data can become integrated in a foundational role into a preexisting body of beliefs. So the question is whether the contextual and the hierarchical aspects of our structure of beliefs can be made compatible, thus making possible a developmental version of foundationalism.
tinct version of foundationalism.
probable or certain; there could be a set.
founded" (para. 253).
sy, as we shall see in Chapters 2 and 5.
 Sosa (1980b, p. 18).
ods. This is the more common version of foundationalism.
alist versions of foundationalism.
 Wittgenstein (1972, Part I, Section 486).
 Bonjour (1985, p. 30).
justifying basic statements. The latter notion he rejects.
recently Quine (1953/1961) is a clear counterexample.
 Chisholm (1982, p. 75).
problem. His solution will be outlined in Chapter Two.
there is a "rock-bottom."
tension between its interconnected pieces.
a broadly rationalist version of foundationalism.
 The term is Rescher's (1974, p. 703).
 Lehrer (1987, pp. 87-107).
thus allowing us to choose between them only arbitrarily.
 Quoted in Kornblith (1985, p. 4).
version of reductivism.
Edition, p. 7 [A viii]).
accept Rorty's characterization.
 E.g. Bonjour (1985, p. 4).
as impossible in principle" (Passmore, [1985, p. 118]).
 MacIntyre (1985, p. 8).
makes his suggestions.
 Rorty (1979, pp. 371-372).
 Rorty (1980, in Moser & vander Nat [1987, p. 215]).
plett (1987), and Haack (1990).
be explored in Chapter 5.
suspect it will founder" (1976a, p. 185).
 Sellars (1981, p. 3).