A prominent set of moderate rationalists hold, roughly, three theses: (i) that at least some justification is a priori in the sense that some beliefs are justified independently of experiential evidence; (ii) that a priori justification is defeasible on new evidence; and (iii) that whatever non-experiential evidence underwrites (or explains or constitutes) a priori justification is an ineliminable element of what George Bealer (Reference Bealer, Wagner and Warner1993) calls our “standard justificatory procedure.” Our standard justificatory procedure is comprised of what we generally take to constitute prima facie legitimate (or basic) sources of evidence, including sense data, testimony, and memory. The ineliminable element defended by rationalists has been identified variously as “intuitive induction” (Chisholm Reference Chisholm1977), “reason” (Plantinga Reference Plantinga1993), “intuition” (Bealer Reference Bealer, Wagner and Warner1993, Reference Bealer2000; Sosa Reference Sosa2007a, Reference Sosa2007b), and “rational insight” (BonJour Reference BonJour1998). Even though these accounts differ to some degree, I take it that all proponents are attempting to capture the same epistemic state, and therefore, I will use the term “intuition” to refer to whatever epistemic state moderate rationalists are attempting to distinguish from experiential evidence.
In contrast to this moderate rationalist view of intuition, epistemologists who characterize their projects as “naturalized” resist characterizing intuition as a priori; they either reject treating intuition as a legitimate source of evidence, or they reject the idea that it is a basic source of evidence, or they oppose its characterization as non-experiential.Footnote 1 Though naturalized epistemologists are primarily interested in specifying and developing (and, in some cases, collapsing) the relationship between epistemology and other sciences, particularly psychology, these projects often entail either a rejection of a priori justificationFootnote 2 or include a normative prescription to remake a priori justification in the naturalist's image – in some cases, while retaining the qualifying phrase “a priori.”Footnote 3
A popular response to naturalistic rejections of moderate rationalism is to attempt to show that such a rejection leads to an unfortunate dilemma.Footnote 4 According to these rationalists, naturalists hold two incompatible claims: (a) that experience is the only legitimate source of evidence, and (b) that generalizing principles of inference (i.e., principles that allow one to infer claims that include content beyond what is directly perceived) constitute grounds sufficient for justifying beliefs derived from arguments employing them. Though these dilemma arguments vary to some degree, the general thrust is that the conjunction of (a) and (b) leads to something like the following problem: either naturalists are implicitly committed to some sources of evidence in addition to experience (by accepting generalizing principles of inference), in which case, they are not actually naturalists (and thus, any argument attempting to exclude putatively non-experiential evidence using these general principles of inference is self-defeating); or naturalists explicitly reject non-experiential evidence, in which case, apparently, they must also give up generalizing principles of inference (thus, either begging the question against rationalism by offering arguments against the a priori, or implicitly committing to a pernicious form of skepticism).Footnote 5
Those who argue that the second horn is question-begging (Siegel Reference Siegel1984) claim that it is not obvious that the methods of the empirical sciences have anything to say about the legitimacy of non-experiential sources of evidence. Those who argue that it leads to skepticism (Bealer Reference Bealer, Wagner and Warner1993; BonJour Reference BonJour1998) claim that consistent naturalists are denied the conceptual machinery necessary for believing much of anything.
Albert Casullo (Reference Casullo2000, Reference Casullo2003) and Shane Oakley (Reference Oakley, Shaffer and Veber2011) respond to this challenge by arguing that even if dilemma arguments are effective against naturalized epistemology, they are also effective against moderate rationalist accounts of a priori justification. To show this, each offers a parallel dilemma argument against rational intuition, and concludes that this defuses any advantage rationalists claim to have over naturalists. I will argue that successful parallel dilemma arguments require ascribing illicit assumptions to the rationalists who offer dilemma arguments, assumptions to which they are not committed, either expressly or implicitly in the structure of their dilemma arguments. I will then argue that the force of dilemma arguments derive from a conception of empirical evidence that naturalists do, by and large, accept. Because of this, I argue that dilemma arguments constitute a serious, though perhaps not insurmountable, challenge to naturalized epistemology. I conclude that, although there are strategies for avoiding the implications of dilemma arguments, naturalists must, nonetheless, take dilemma arguments seriously and pursue one of these strategies.
I. CASULLO'S DILEMMA ARGUMENT AGAINST RATIONALISM
BonJour offers what Casullo takes to be two dilemma arguments against naturalism.Footnote 6 Casullo's first parallel dilemma argument against rationalism is aimed at what he calls Laurence BonJour's “Sceptical Argument” against naturalism (Reference BonJour1998: 4). BonJour allows that at least some beliefs are justified by direct experience, particularly, beliefs about particular objects or events at specific places and times (presumably, we are to consider immediate experiences such as seeing a red ball roll off a table at a particular time). Of course, like most of us, naturalists do not restrict themselves to beliefs justified in this way; on the basis of directly experiential beliefs, they also infer “beliefs whose content goes beyond direct experience or observation: beliefs about the past, the future, and the unobserved aspects of the present” (4). For example, from our experience with the red ball, we might infer the existence of certain natural laws such as gravity that govern the movement of the ball off the table, that the ball could not be both blue all over and red all over at the same time, that the ball might have been blue rather than red, and, further, that if a blue ball were rolled toward the edge of the table, it, like the red, would fall off. BonJour argues:
[I]f the conclusions of inferences genuinely go beyond the content of direct experience, then it is impossible that those inferences could be entirely justified by appeal to that same experience. In this way, a priori justification may be seen to be essential if extremely severe forms of skepticism are to be avoided (Reference BonJour1998: 4).
Thus, BonJour poses a dilemma for naturalists: if naturalists do not accept that the above beliefs about the red ball are justified, they are stuck with a nasty form of skepticism. If they do accept that they are justified, they must accept the legitimacy of a priori justification.
In response, Casullo constructs a parallel argument to show that this argument “proves too much,” namely, the “very same argument can be marshaled against moderate rationalism”:
Assume that some beliefs are directly justified by [intuition].Footnote 7 Either some beliefs whose content goes beyond direct [intuition] are justified or skepticism is true. The justification of beliefs whose content goes beyond direct [intuition] requires principles of inference that are justified a posteriori: For if the conclusions of the inferences genuinely go beyond the content of direct [intuition], then it is impossible that those inferences could be entirely justified by appeal to that same [intuition] (2003: 101).
If he is right, naturalists and rationalists are in the same predicament: neither can justify the employment of their most basic evidence without begging important epistemic questions. If both appeal for justification to content unavailable through their most basic sources of evidence, both are equally subject to whatever criticisms attend the beliefs so justified.
This is Casullo's first parallel argument, but he doesn't find it particularly compelling. This is because he thinks the rationalist could avoid this conclusion by appealing to other rational intuitions to play a sort of corroborating role. This would allow them to indirectly justify, without vicious circularity, generalizations beyond particular instances of direct intuition:
The obvious reply is that a belief directly justified by some rational insight R, in conjunction with a belief in a general principle directly justified by some other rational insight R*, can indirectly justify a belief whose content goes beyond that of the directly justified beliefs (2003: 101–2).
Of course, this option also seems available to naturalists:
[Naturalists] can maintain that a belief directly justified by some experience E, in conjunction with a belief in a principle of inference directly justified by some experience E*, can indirectly justify a belief whose content goes beyond that of directly justified beliefs (2003: 102).
Because of this, Casullo thinks no rationalist would accept this option. If this is right, Casullo concludes, BonJour and others “must show that experience cannot directly justify principles of inference” (102).
Unfortunately, Casullo seems to have misread BonJour's original argument. BonJour explicitly states that the problem is that foundational empirical beliefs “are particular rather than general in their content and are confined to situations observable at specific and fairly narrowly delineated places and times” (4). According to moderate rationalists, intuitions aren't correspondingly restricted. If they are correct, Casullo's disjunction, “Either some beliefs whose content goes beyond direct [intuition] are justified or skepticism is true,” is false. The content of direct intuitions is general in a way that direct experiences are not.
To be fair, Casullo is trying to capture the subtlety of BonJour's case: BonJour's first argument refers primarily to conclusions inferred from foundational empirical beliefs, and highlights the problem of how naturalists could justifiedly arrive at conclusions with more content than found in the original empirical beliefs. To do so would require a principle of inference to bridge the gap. According to BonJour, principles of inference are “at least partially justified a priori” (4), and this motivates his dilemma: without a principle of inference, naturalists are left with an undesirable skepticism; alternatively, invoking an a priori principle of inference denudes naturalism of its restriction to non-empirical sources of evidence. Casullo suggests both naturalists and empiricists could avoid this dilemma by appeal to indirect justification. But this is not quite right. Assuming BonJour is right that principles of inference are justified, at least in part, a priori, Casullo is mistaken in thinking that “a principle of inference directly justified by some experience E*” could help justify “a belief whose content goes beyond that of directly justified beliefs.” No experience could directly justify a general principle.
We need not linger here, for Casullo responds to BonJour's second dilemma argument with a parallel argument he does find compelling. In what Casullo calls the “Generality Argument,” BonJour attempts to show why direct experience is no more plausible for justifying general principles of inference than for justifying general beliefs:
Could an argument of any sort be entirely justified on empirical grounds? It seems clear on reflection that the answer to this question is “no.” Any purely empirical ingredient can, after all, always be formulated as an additional empirical premise. When all such premises have been explicitly formulated, either the intended conclusion will be explicitly included among them or it will not. In the former case, no argument or inference is necessary, while in the latter case, the needed inference clearly goes beyond what can be derived entirely from experience (Reference BonJour1998: 5, footnote omitted).
The idea is that any direct experience will include content restricted to particular events at a particular time and place. It will include no content about future, alternative, or counterfactual events. Thus, any particular experience will constitute evidence about that event, time, and place, and only about that particular event, time, and place. If a conclusion is drawn about that particular event, time, and place, it will simply be a reiteration; no inference is needed. If a conclusion extrapolates to a future, alternative, or counterfactual event, no directly experienced premise will be sufficient to justify the conclusion. Casullo reformulates this argument as follows:
1. Experience is limited to particular objects.
2. No experience can directly justify a belief whose content goes beyond that of experience.
3. Principles of inference are general.
4. Therefore, experience cannot directly justify principles of inference (Reference Casullo2000: 2, emphasis his).
The most immediate problem, according to Casullo, is that whatever force the argument has against naturalism it has the same against rationalism:
Premise (2) of the argument appears to be a consequence of a more general epistemic principle:
(2*) No cognitive state can directly justify a belief whose content goes beyond that of the state.
Premise (3) is also general. Indirect a priori justification, as well as indirect empirical justification, requires general principles of inference. Hence, if rational insight is limited to particular objects, the Generality Argument also establishes that rational insight cannot directly justify principles of inference (Reference Casullo2003: 102).
Casullo is worried about the sort of content that direct intuitions could convey relative to direct experiences. If intuitions are also limited to particular mental states at a time, they are equally subject to BonJour's dilemma. Casullo doesn't explicitly formulate the parallel argument he takes (2*) to imply, but we may presume it would look something like this:
1’. Intuition is limited to particular objects.
2’. No intuition can directly justify a belief whose content goes beyond that of intuition.
3. Principles of inference are general.
4’. Therefore, intuition cannot directly justify principles of inference.
Casullo concludes that “if rational insight is limited to particular objects, the Generality Argument also establishes that rational insight cannot directly justify principles of inference” (102). If this is right, the rationalist can claim no advantage over the naturalist.
There are at least three serious problems with Casullo's case against BonJour. First, it is not obvious that either argument is valid. Each version is supposed to highlight a contrast between the “particular” nature of basic evidence and the “general” nature of principles of inference. But note that premises (1) and (1’) refer to particular “objects,” while premise (3) simply says that principles of inference are “general.” The conclusions would follow if principles of inference are “general objects.” But what might this mean? Perhaps principles of inference are ante rem universals, which may be infinitely instantiated and are therefore “general” in their instantiability. But even ante rem universals, if there are such entities, are individually delineated objects with empirically identifiable content. A single experience of “redness” is not general, even if redness could be infinitely instantiated. Even allowing, for the sake of argument, that we can make sense of a general object as a universal, it is not clear that BonJour intends to imply that principles of inference are general in the sense that universals are general. To be sure, whereas redness is contingently instantiated, principles of inference hold necessarily. More plausibly, we might regard the content or implications of principles of inference as general: their content applies generally; it is generalizable to all contexts. If this is right, Casullo's parallel argument fails because he equivocates on the meaning of generality.
Second, disambiguating the argument to resolve the equivocation doesn't help Casullo's case. Suppose we revise (1) and (1’) to read, “… includes content limited to particular events, times, and places” and premise (3) to read something like: “Principles of inference include content that applies to events, times, and places in general.” The argument is now valid, but not obviously sound. The central motive for distinguishing intuition from sensation is that the former seems to provide access to content that generalizes while the latter does not. As BonJour puts it, “[A]n apparent rational insight purports to be nothing less than a direct insight into the necessary character of reality” (107). The point that BonJour's dilemma highlights is that naturalists have no means of justifying appeals to beliefs that include content that generalizes since the content of direct sensory perception doesn't. In contrast, rationalists claim that intuitions do just that. Therefore, moderate rationalists may happily accept (2’) and simply regard intuitions as sufficient for providing access to general content.
Casullo anticipates but rejects this sort of move on the grounds that BonJour cannot sufficiently demonstrate how intuitions provide access to general content. He allows that, “[Moderate rationalists] can avoid the skeptical consequences of the Generality Argument only if they can sustain the claim that we apprehend abstract entities such as properties” (102). Unfortunately, he takes this requirement from a misreading of BonJour. BonJour writes:
A person apprehends or grasps, for example, the properties redness and greenness, and supposedly “sees” on the basis of this apprehension that they cannot be jointly instantiated. Such a picture clearly seems to presuppose that as a result of this apprehension or grasping, the properties of redness and greenness are themselves before the mind in a way that allows their natures and mutual incompatibility to be apparent. (Reference BonJour1998: 162)
Casullo focuses on the apprehension of properties as the relevant generalizing content of this description. But of course, it is no mystery to believe that we directly apprehend properties: every time we see that an object is round, large, red, or cold, we are “apprehending” properties. Surely, only a hardened nominalist would deny this. However, what is apprehended a priori, according to BonJour, is a feature of those properties that allows us to see their implications in other circumstances. In some cases, those features are relational: if this fork were 2” longer, it wouldn't fit in my drawer. In some cases, those features are modal: it is not possible that his fork is longer than it is.
Nevertheless, we can still make sense of Casullo's objection. He ultimately argues that BonJour's account of how we come to apprehend properties is unsatisfying. BonJour offers only “the bare outline” of how we might have access to mental content containing particular objects, and much less about how we might access content that generalizes. “Since he does not provide an account of how a thought can have as its content some property, he fails to provide any account … of the apprehension of properties. Hence, BonJour's moderate rationalism does not avoid the skeptical consequences of the Generality Argument” (104).
Notice that this criticism misses its mark. Even if Casullo is right that BonJour has no satisfactory account of this process, Casullo is conflating two different projects: a philosophical explanation of how intuition justifies general principles and a phenomenological seeming that intuitions justify general principles. BonJour may happily concede that he doesn't know how intuitions seem to include generalizing content, but since it is widely agreed that such content is not included in direct experience, a plausible prima facie alternative is that it is a priori. Casullo's challenge is analogous to someone who rejects the claim that an object falls when released over an open space on the grounds that no one can satisfactorily explain why it happens. The virtues of a hypothesis can be evaluated independently of our ability to answer every important question about it.
A third reason to reject Casullo's argument is that the most plausible reading of his (2*) involves an implausible interpretation of BonJour's argument. On one reading, a moderate rationalist may grant (2*) with equanimity: indeed, no cognitive state can directly justify a belief whose content goes beyond that of the state. But this isn't what BonJour claims; the content of the cognitive state we call “intuition” is general; it allows us to access, in BonJour's words, “the necessary character of reality.” And so, this reading has no bite against the rationalist; (2*) is true, but irrelevant to dilemma arguments against naturalism. Therefore, if (2*) is to do its work, Casullo must have an alternative reading in mind. For instance, he might intend (2*) to imply that no particular cognitive state can include general content. If so, then certainly no moderate rationalist could, in good conscience, accept it. If BonJour were implicitly assuming this reading in his description of direct experience as “particular,” then Casullo would rightly note the devastating parallel dilemma. Yet, there is no reason to think a moderate rationalist needs this reading of (2*) to construct a successful dilemma argument against naturalism. All the moderate rationalist needs show to achieve an epistemic advantage over naturalism is that an ostensibly a priori process plays an ineliminable role in any non-skeptical epistemic view. It is enough to read (2*) in the former, innocuous sense to do that, and thus, BonJour's dilemma argument withstands Casullo's criticism.
Note that reading Casullo's parallel argument according to the second interpretation of (2*) seems to presuppose that proponents of a source of evidence must be able to make a positive case for that source. We could read Casullo's response colloquially as: No, we cannot justify our general principles, but neither can you, so you are no better off than we. If this is what Casullo intends by (2*), it implies that he conceives BonJour's challenge as a meta-justificatory challenge: a challenge to naturalists to justify their basic sources of evidence. If that were the accurate reading of BonJour, Casullo's response would hit its mark. If, however, BonJour is not making (or need not be interpreted as making) a meta-justificatory demand, but simply arguing that naturalists have no non-skeptical way to exclude non-experiential sources of evidence, Casullo's parallel argument is not apropos. This point will help illuminate the following discussion of Oakley's parallel argument against Harvey Siegel.
II. OAKLEY'S PARALLEL ARGUMENT AGAINST RATIONALISM
Shane Oakely offers a parallel argument against dilemma arguments that he takes to be “similar, if not identical” to Casullo's (Reference Casullo2000) (171, n. 26). Oakley is responding to the dilemma argument against naturalism offered by Harvey Siegel (Reference Siegel1980, Reference Siegel1984),Footnote 8 which, though aimed primarily at Paul Roth's (Reference Roth1980, Reference Roth1983) defense of Quine's naturalization project, constitutes an interesting challenge to any epistemic theory that rejects non-empirical sources of evidence.
Siegel understands Roth's argument to be that Quine's critique of analyticity combined with the Quinean/Duhemian thesis (that the sentences of a theory have meaning and constitute evidence only as an interrelated set) imply naturalized epistemology, that is, that there are no extrascientific (a priori) sources of evidence. In response, Seigel poses a dilemma for the naturalized epistemologist. On one hand, the Quinean/Duhemian thesis needs to be justified, and even Roth apparently admits that a purely scientific justification is implausible:
It is worth re-emphasizing that in one respect the naturalized epistemologist's position is self-defeating. For it seeks to justify naturalized epistemology in precisely the way in which, according to it, justification cannot be had. The Duhemian thesis cannot lead to the rejection of “old-fashioned” justification, for it must itself be justified, as I argued earlier and Roth admits, in the old-fashioned extrascientific way. (Reference Siegel1984: 675)
On the other hand, to claim that science warrants naturalized epistemology is question-begging, since it is the legitimacy of extrascientific sources of evidence that is at issue (675–6).Footnote 9 If this is right, Siegel concludes, epistemic naturalism is untenable.Footnote 10
Oakley begins his response to Siegel's dilemma by defining epistemic naturalism as the normative thesis that “principles of evidence and epistemic justification should be seen to be the principles that are employed in or are discoverable by the methods of science” (160), Oakely then reconstructs Siegel's argument as follows:Footnote 11
5. If there is an independent justification of naturalism, then it cannot be justification in terms of the naturalistic thesis, i.e., naturalism is false.
6. If there is no independent justification for naturalism, then any defense of the thesis is viciously circular.Footnote 12
7. Either there is an independent justification for naturalism or there is not.
8. Therefore, either naturalism is false or any defense of naturalism is viciously circular (Reference Oakley, Shaffer and Veber2011: 164).
Oakley raises an interesting problem for premise (6), constructing an argument showing that it is possible that naturalism is non-viciously self-justifying. I will offer some brief comments on this argument in section IV, but for now, Oakley argues that even assuming premise (6) is true and that the argument is sound, its reasoning can be applied, mutatis mutandis, against rationalism:
5*. If there is an independent justification for rationalism, then rationalism is self-defeating, and hence false.
6*. If there is no independent justification for rationalism, then any defense of the thesis will be viciously circular.
7*. Either there is an independent justification for rationalism or there is not.
8*. Therefore, either rationalism is false or any defense of rationalism is viciously circular (Reference Oakley, Shaffer and Veber2011: 171).
If Oakely is right, then, as with Casullo's parallel argument, naturalists are in no worse an epistemic predicament than rationalists.
Before we evaluate Oakley's parallel dilemma, it will be important to identify what about his argument Oakley takes to be practically identical with Casullo's. BonJour's dilemma and Casullo's parallel dilemma seem to be about sources of justification for generalizing principles, whereas Siegel's dilemma and Oakley's parallel seem to be about meta-justifications for restricting what counts as a basic source of evidence, that is, for excluding intuition as a basic source of evidence. If we interpret Siegel's argument as highlighting a problem with the Quine/Duhem thesis itself, Oakley's response would be analogous to Casullo's. This is, however, implausible because Seigel isn't concerned that the Duhemian thesis is itself an a priori source of justification (which BonJour does claim for generalizing principles).
Siegel and Roth seem to admit, however, that in order to justify the Duhemian thesis, one would need to appeal to at least some extrascientific source of evidence. The best that Roth can do is claim that empirical evidence confirms the application of the Duhemian thesis (a move that Siegel calls question-begging). This appeal to a principle to exclude sources of evidence may have affinities with our second interpretation of Casullo's (2*). Recall that Casullo's last bid against BonJour was to invoke meta-principle (2*) as a means of formulating a parallel dilemma argument for rationalism. Interpreted as a claim about content, we saw that this principle is trivial and innocuous against BonJour's dilemma. But interpreted as a claim about generalizability, it has skeptical implications for both naturalists' and rationalists' basic sources of evidence. If both naturalists and rationalists must justify their basic sources of evidence, they both face Siegel's dilemma: either they appeal to non-basic sources, in which case they are self-defeating, or they appeal to other basic sources, in which case they are circular. This seems to be the most plausible sense in which Oakley's parallel argument is “similar, if not identical” to Casullo's.
All of this suggests that the philosophically interesting parallel responses to dilemma arguments are those that presuppose rationalists are demanding that naturalists justify their basic sources of evidence – that is, that they are invoking a meta-justificatory demand. Since naturalists cannot justify their use of generalizing principles, they are compelled, on pain of skepticism, to abandon naturalism and accept some a priori justifications. The meta-justificatory parallel response is that, rationalists face the same dilemma when they attempt to justify generalizing principles, and therefore, dilemma arguments are trouble for everyone. But if rationalists are not making a meta-justificatory demand, if they are simply demanding that naturalists justify their rejection of intuition as a basic source of evidence, then meta-justificatory parallel arguments are inadequate to their task.
If this is right, the same objection that undermines Casullo's parallel argument also undermines Oakley's. If epistemologists are under a general obligation to justify their most basic sources of evidence (sensory, memory, intuitive, etc.) and neither can satisfactorily do so, then it is clear that highlighting the naturalist's inability to offers such a justification will confer no advantage to the rationalist; the rationalist is in no better epistemic position than the naturalist.
However, if epistemologists are content to regard certain sets of evidence as basic (irrespective of whether these sets are justified), then debates about which sources are legitimate take a more obvious form. For instance, we may ask why we are justified in regarding probability as a legitimate source of evidence while we are not justified in so regarding Tarot card readings. To make this sort of determination, we appeal to more basic sources of evidence to show that the former is reliable (in some acceptable epistemic sense) whereas the latter is not.
Now, an interesting question arises when a group of epistemologists attempts to reject a source of evidence that is widely regarded as basic, such as intuition. Of course, no one denies the legitimacy of general principles of inference or the seeming necessity of, say, the law of identity. However, when naturalists claim to restrict themselves solely to empirical sources of evidence, the question of how these basic reasoning processes are justified becomes contentious. This is where dilemma arguments against naturalism garner their force.
So, what are we to make of Oakley's response to Siegel? The most vexing problem is that, as stated, it is a straw man. As we saw from his definition, Oakley interprets “naturalism” as a thesis about the sources of evidence that naturalists accept. Siegel, however, defines naturalism as a thesis about the sources of evidence naturalists reject, namely, extrascientific, or, for our purposes, a priori sources, in particular, intuition. If this is right, then Oakley's parallel argument is vastly off the mark – rationalists do not reject any sources of evidence naturalists find compelling. The question is: On what grounds can anyone rule out the a priori? The problem is that naturalists apparently cannot rule it out by appealing only to empirical evidence. Rationalists do not face this problem because their position does not entail rejecting any (historically) widely accepted sources of justification.
III. CLARIFYING THE ASSUMPTIONS BEHIND DILEMMA ARGUMENTS
Our discussion of Casullo's and Oakley's parallel dilemma arguments allows us to highlight two distinct assumptions in tension in interpretations of dilemma arguments. First, someone might read a dilemma argument about naturalism as implicitly demanding a meta-justification of naturalism's basic sources of evidence. Such a demand is predicated on something like the following principle:
Independent Justification Principle (IJP): Proponents of a theory of justification must be able to provide non-circular reasons for all of the evidential sources included in their standard justificatory procedure.
This sort of principle confers an obligation on rationalists and naturalists alike to account for the reliability of their basic sources of evidence. Many rationalists appreciate the desirability of IJP, but disagree agree widely as to its plausibility. BonJour (Reference BonJour1998) constructs a sort of negative defense of rationalism by first offering an exposition of its implications and strengths and then attempting to defuse arguments against it. He even acknowledges that any more direct attempt at justification would fall foul of a dilemma argument (143). Ultimately, he rejects the plausibility of this demand for “metajustification”; its demands are either implausibly strong or question-begging (144–5).Footnote 13 Given that he explicitly rejects the demand for meta-justification for the set of prima facie evidence acceptable to rationalists, it would be question-begging for him to argue that naturalists cannot do the same for empirical evidence. Thus, no charitable critique of BonJour's dilemma argument should interpret his dilemma argument as presupposing IJP.
Siegel, on the other hand, seems to accept the demand for meta-justification, though he argues that it doesn't affect his dilemma argument. He writes: “Roth is right to point out that the criteria of evaluation offered by epistemology themselves need to be justified. We do not want our epistemology to be illicitly self-validating ... [T]he question of the actual justificatory force, of justificatory practices and of criteria of evaluation of those practices, is always open. ” (669). Nevertheless, this worry is irrelevant to his dilemma argument because the question is not whether any epistemic processes are valid – the rationalist and naturalist agree that many are. Siegel's dilemma argument attempts to show that the naturalist's preferred epistemic processes are insufficient to justify a rejection of the rationalist's. “The question to be asked is: Is the abandonment of the traditional project appropriate? Should we give it up?” (671).Footnote 14
A second reading of dilemma arguments is that they simply demand that proponents of a theory of justification show how their basic sources of evidence sufficiently exclude certain cognitive states as legitimate evidence. This demand is predicated on something like this principle:
Evidential Restriction Principle (ERP): Proponents of a theory of justification must be able to provide non-circular reasons for restrictions on their critical apparatus, that is, to offer reasons sufficient for rejecting a piece or source of evidence as legitimate.
This sort of principle confers an obligation only on the epistemologist who is rejecting a putative source of evidence as legitimate. For example, the rationalist may incur this demand when she rejects Tarot card reading as a legitimate source of evidence. To do so, she might highlight the inconsistency in conclusions that Tarot card readings have relative to evidence that both she and the Tarot card proponent share (including empirical evidence and generalizing principles of inference, not least of which would be a restriction on inconsistency). Then she might appeal to the much greater scope of applicability of these agreed-upon sources of evidence to establish that they are more basic than Tarot card readings (notably, that they are the sort of sources necessary for confirming Tarot card readings in the first place).
The same demand faces the naturalist who rejects intuition as an a priori source of evidence. To do so, he must appeal only to those sources of evidence that both rationalists and naturalists accept. The difficulty, as we have seen, is showing how purely empirical cognitive states can reject principles that include content that applies generally without invoking principles with content that applies generally.
The central point is this: If dilemma arguments presuppose something like IJP, then Casullo and Oakley are right that dilemma arguments against naturalism fail. But if dilemma arguments presuppose only something like ERP, parallel arguments are straw men. With this distinction in place, it seems clear that moderate rationalists need nothing like IJP to make an ostensibly powerful case. If this is right, what we need is a dilemma argument that explicitly proscribes this sort of ambiguity in interpretation.
IV. THE POWER OF DILEMMA ARGUMENTS AGAINST NATURALISM
With the distinction between IJP and ERP in mind, and by disambiguating dilemma arguments in favor of ERP, we can construct the following dilemma argument against naturalism:
9. Historically and in actual practice, philosophers and scientists employ both experiential evidence and intuitions in their set of prima facie evidence.
10. Naturalism entails the claim that intuitions should: (i) be rejected as a legitimate source of evidence; (ii) be considered an insignificant or irrelevant source of evidence; or (iii) be justifiably regarded as an empirical source of evidence.
11. Attempts to justify (i), (ii), or (iii) must appeal to evidence either within the naturalist's preferred set of prima facie evidence or outside of it.
12. If the appeal is made outside of the naturalist's set, then any conclusion rejecting intuitions is self-defeating.
13. If the appeal is from within the naturalist's set, then any conclusion rejecting intuitions is either viciously question-begging or entails skepticism.
14. Therefore, any justification for naturalism is self-defeating, question-begging, or entails skepticism.
Is this argument sound? Oakley argues that premise 13 is not obviously true by constructing a formal argument for the possibility that naturalism, and any other epistemic theory for that matter, is non-viciously self-justifying (167–71). There is, however, a severe problem with such an argument. It is surely right that circular arguments are valid; insofar as the rule of reiteration is valid, any claim is “self-justifying.” The more difficult question is whether the circularity is vicious, and this is an informal consideration. Moreover, this informal consideration presupposes a general principle of evidence – a type of evidence the naturalist must already be able to accommodate within the limits of experience alone. Thus, any argument to avoid 13 by showing that naturalism is non-viciously circular is also question-begging.
Let us assume, then, that the argument is sound. Is epistemic naturalism untenable? Not quite. There are at least two responses available. First, the naturalist might reject both IJP and ERP. If naturalists resist the attempt to justify both their rejection of intuitions and their basic sources of evidence, they may simply stipulate that all basic sources of evidence are legitimate and claim that this is consistent with their naturalistic view. On this approach, the problems generated by dilemma arguments will not arise.Footnote 15 This seems to be the primary motive for some naturalists to characterize their projects as a “direction” (Sellars Reference Sellars1922), “program” (Forrest Reference Forrest1996), or “strategy” (Giere Reference Giere1999) as opposed to a “philosophical thesis.”Footnote 16
On the other hand, a naturalist could attempt to show that every significant type of evidence that rationalists regard as non-experiential is plausibly included in a naturalistic framework, especially general principles of inference. This would require two steps. In the first, naturalists would need to show, using only uncontroversial experiential evidence, that a certain set of general principles of inference are actually unique forms of empirical evidence. This would require identifying a coherent set of conditions for identifying empirical evidence and showing that general principles of inference meet these conditions. Hilary Kornblith (Reference Kornblith1993) makes an admirable effort in this regard for induction. Of course, even if successful, this first step wouldn't sufficiently rule out the a priori – it would only restrict what rationalists could plausibly regard as non-experiential evidence. To complete the rejection of the a priori, naturalists would then need to show that this set of basic sources of evidence effectively proscribes any remaining putatively a priori sources of justification. This strategy amounts to a denial of premise 13. Whether any such attempts are successful is beyond the scope of this paper. The point is that neither this strategy, nor the previous, undermines the force of dilemma arguments (as Casullo's and Oakley's arguments would if they were successful); indeed, they presuppose the need to take them seriously.
In conclusion, we have seen that neither Casullo's nor Oakley's parallel dilemma arguments against BonJour and Siegel sufficiently undermine their force against naturalism. Disambiguating Casullo's equivocation, we see that his parallel argument is not actually parallel to BonJour's. Further, Casullo's and Oakley's meta-justificatory interpretation of dilemma arguments are straw men since dilemma arguments against naturalism more plausibly presuppose ERP rather than IJP. And despite the fact that naturalists have strategies available for responding to these dilemmas, dilemma arguments are not illicitly question-begging and constitute an important point of contact for debates concerning the legitimacy of a priori justification.
