Introduction
C) Finally, there is a risk that the kind of meta-linguistic primary intension we are discussing, as well as not allowing for a priori knowledge when this would seem desirable,
2.4. Saving the Ideal (-ized Rational Reflection)?
2.4.2. Internalism Strikes Back, in a Circle
If one is convinced that 2, which is central to two-dimensionalism, is untenable without 1, and one wants to save two-dimensionalism, one might perhaps think we have been too quick to give up 1; we could have explored a different strategy, i.e. using 2 to defend 1.
The move of re-interpreting Bert’s and A’s words as expressing a non-standard
concept is not unmotivated after all, it could be claimed. It is motivated exactly by the fact that they would express non-standard judgements on IRR. In particular, they have the same concepts TwinBert and B have, since they share judgements under IRR. In the case of Bert, this would involve claiming the belief he expresses saying ‘I have arthritis in my thigh’ is true after all. In the case of A, presumably the primary intension of his concept would be something like ‘the objects of sofish shape of my acquaintance’; this would pick up sofas in our world, and so, if he forms the belief that ‘sofas are not meant to be sat upon’, that belief would be false. However, it would not be a priori false, while we are assuming that the content standardly expressed by competent speakers with the same (syntactically individuated) words could be known a priori to be false.
The problem with this move is that it risks being a sort of stipulation. Of course it is perfectly legitimate for the two-dimensionalist to define primary intension in such a way to make it supervene on judgements under IRR. Then we also note that judgements under IRR plausibly supervene on internal states. The conclusion we should reach at this point is that primary intensions, if they exist, supervene on internal states of the subject. However, if there are cases in which two subjects intuitively differ in the content their concepts have, while their judgements under IRR would be the same, or, the other way round, their contents are intuitively the same, while their judgements under IRR would plausibly diverge, it is not a good reply that such cases are impossible by definition. The sort of content the two-
dimensionalist has defined might after all fail to have any psychological reality, or semantic role. Two-dimensionalism took its initial motivation from its ability to account for the (alleged) pre-theoretic judgement (or intuition) that in TwinEarth-like cases there is a level of content in common between the two subjects, which externalism cannot capture. It would therefore seem extremely awkward if in order to apply the two-dimensionalist theoretical machinery we had to deny, in a large number of other cases, our pre-theoretic judgements of sameness or difference in content. But this seems to be clearly the case at least for the envisaged re-interpretation of Bert’s beliefs; to claim he has a true belief about the disease he has at his thigh is just to ignore our pre-theoretic judgements, as it were. He has a false one, thinking it is arthritis. Moreover, to have the true belief he would need the concept TWARTHRITIS that is used in TwinBert’s community. And this applies to A as well; in order for his beliefs to b contingently false, he has to use B’s concept. But how did they acquire
these new concepts? Surely not by being taught. Nor did they ever intend in the slightest to create a new concept. They simply should have acquired the new concept by going wrong in the application of the old one; but this is quite incredible in these specific cases. Surely we can be inclined to say about a community of speakers that goes wrong for a long time and in a systematic way about the application of a concept that they created a new one; a new norm of application has taken over the old one. But how could Bert’s or A’s peripheral mistakes have
such an effect? And what is the advantage of attributing to them such new concepts? These questions seem to have no answer, unless the stipulation that primary intensions supervene on IRR is invoked. But appealing to that stipulation would beg the question here, since primary intensions, defined that way, might not figure at all in the best explanation of Bert’s or A’s
mental states and linguistic behaviour.
In addition to the imaginary cases discussed so far, consider the following real case of theoretical disagreement46: Most epistemologists hold that knowledge entails belief; it is not possible for a subject to know that P if the subject does not believe that P. However, A minority of epistemologists deny this claim. It is surely implausible to charge these people with linguistic deficiency about ‘knowledge’ or ‘belief’. There is clearly a disagreement here; the theorists defending the minority view therefore do not have a deviant concept, otherwise there could be no disagreement, but just talking past each other. Again, to stipulate that in this kind of cases the deviant theorist must have created a new concept is unmotivated. But the two-dimensionalist does not need to say that; she might claim that on IRR the theorists who
46
The example of the disagreement between McGee and defenders of modus ponens considered before would serve equally well.
are wrong, let us suppose they are the ones denying the entailment between knowledge and belief, would recognize their error. If this is so, the disagreement is real, because they are wrong on the primary intension of their own concept, for lack of adequate reflection. The case would be analogous to one of two mathematicians disagreeing about the truth of some complex mathematical statement.
However, it’s not clear what guarantees that in typical cases of theoretical disagreement the theorists who are going wrong would change their minds on IRR any more than the subject in the sofa case. The evidence that we have so far indicates that they believe what they believe stably, on reflection; e.g. they believe it is not a priori impossible for a subject to know something she does not believe, and in fact they believe, on reflection, such a possibility to be actual. Of course, we might insist that, as in the mathematical case, the theorist would, given unlimited cognitive resources, come to the correct conclusion. At this point, though, I think we have lost our grip on the notion of IRR. Clearly, we do not have access to this notion except through the notion of what is a priori possible for a subject47. What is a priori possible for a subject should be determined by the primary intensions of her concepts; and the primary intensions are determined by what is possible under IRR. We are moving, it seems, in a circle of definitions; this might not be itself a problem, but it becomes a problem if we are supposed to motivate the counter-intuitive claim that there is no disagreement in some cases (the arthritis case, the sofa case, the latter case of theoretical disagreement) by appeal to what conclusion the subject would reach on IRR. There are plausible cases of disagreement about what is possible, e.g. about the possibility of knowledge without belief, which would not plausibly disappear on IRR. To stipulate this situation is impossible for some kind of content only makes it implausible that this is the kind of content that our concepts and words possess. To reiterate, two-dimensionalism gets some initial plausibility from its alleged ability to do a better job than other theories in capturing judgements about content. If it ends up forcing our judgements to be discounted in favour of some general principle, which is required by the two-dimensionalist’s theoretical commitments, then there is little to be said in its favour.
47
Chalmers is explicit about that; see e.g. Chalmers [2006b] section 3.3, and [2006a] sections 3.3. and 3.9; in the latter work, Chalmers makes the circle smaller by mostly avoiding talk of IRR and connecting primary intentions and apriority directly.
Conclusion
We began this chapter by describing a certain picture of conceptual analysis, on which our judgements on philosophical thought experiments teach us about our concepts, in virtue of their being (in some sense) epistemically analytic. However, we saw that there are good reasons to think there are not, or not nearly enough, epistemically analytic truths to make that picture attractive. I then considered epistemic two-dimensionalism as a strategy to overcome such difficulties. Two-dimensionalism also incorporates the view that there is a fundamental level of content supervening on the subject’s internal states. However, I have considered some thought experiments, mostly the ones presented in Burge [1979] and [1986], which are largely neglected in the literature on two-dimensionalism, and have found the two-dimensionalist defense of the idea that there is such a kind of content unsatisfying in the light of these thought experiments. In particular, I considered the response to the thought experiments provided in Chalmers [2002], and I found it implausible on independent grounds. I have then considered the possibility of conceding that point and still holding to a modified version of two-dimensionalism, involving a distinction between partial and full understanding. However, it seems this would require deep modifications of the view, and the resulting theory seems unstable, and unmotivated. Finally, I considered the possibility of going back to defending primary content by appealing to its definition in terms of judgements under idealized rational reflection, but I’ve found no reasons to think that content of such a kind is what human beings use in thinking or talking.
My conclusion so far is that the idea of conceptual analysis as I described it at the outset is also to be abandoned. Of course, this does not mean that we should stop considering imaginary cases. On the contrary, my arguments here rested on consideration of some such case, and I will say something positive about the epistemology of our judgements on thought experiments in chapter 4. Before doing that, however, I will discuss in the next chapter the consequences of the conclusion of this and the previous chapter on the notion of apriority. One might still think that, although not based on intuition, and not epistemically analytic, our judgments on philosophical thought experiments are still ‘exceptional’ in that they are a priori, in the sense, roughly of being independent of experience. I will argue in the next chapter that there is no useful way to specify the notion of independence from experience, thereby concluding my case against philosophical exceptionalism.