In Chapter 4 of his (2008), Robert Brandom presents an argument to show how our modal thought and thought about counterfactuals is legitimate and indispensable. I find myself in substantial agreement with much of the reasoning, but there are a few things worth discussing (I hope!). I discuss a few issues regarding the argument: First, how bad is the problem it is supposed to solve? Second, how much does the argument prove? Third, can a thoroughgoing pragmatist theory, cashing out modality in terms of commitments, get off the ground? And finally, is there a problematic circularity in the Kant-Sellars thesis? As things stand right now, the situation regarding modals is quite unsatisfying. On the one hand, modal (and counterfactual) expressions are regularly used in the explanation of other concepts, and in the development of for instance scientific theories. On the other hand, existing theories of modality and of our knowledge of possibility and necessity (including counterfactuals) leave a lot to be desired. Thus Kit Fine and Christopher Peacocke have voiced their dissatisfaction in the following ways: It is an oddity of current thinking about modality that it has been heavily influenced, one might even say dominated, by two extreme and highly implausible views. The first of these, associated with the name of Quine, is that modal notions are lacking in sense. … The second of these two views, associated with the name of David Lewis, is that the possible and the actual are on an ontological par. Other possible worlds and their inhabitants are just as real as the actual world and its inhabitants; and there is no difference between them in regard either to the degree or to the kind of reality that they possess. (Fine 2005, p. 1) And Peacoke, in his treatment of problems concerning the knowledge of modals: The metaphysics of necessity seemingly [has] … this distinction: that there is practically no philosophical view of the matter so extraordinary that it has not been endorsed by someone or other. (Peacocke 1999, p. 119) Much of the problem with our knowledge of modals has its origins in an empiricist conception of what the basis for knowledge and understanding amounts to. How could we, by observing contingent facts, come to know that something must be the case, or that if something were the case, then something else would be the case? Observation seems only to tell us what is in fact the case, with no counterfactual frills, as it were. William Whewell presented this problem, with its obvious roots in Hume, in a forceful way. He said that experience: can observe and record what has happened; but she cannot find, in any case, or in any accumulation of cases, any reason for what must happen ... To learn a proposition by experience, and to see it to be necessarily true, are two altogether distinct processes of thought ... If anyone does not clearly comprehend this distinction of necessary and contingent truths, he will not be able to go along with us in our researches into the foundations of human knowledge; nor indeed, to pursue with success any speculation on the subject. (Whewell1840, pp. 59-61)
If our grip on the world exclusively depends on what our observations can tell us, our access to knowledge of necessity and possibility is endangered, at least if we can make good the suppressed empiricist premiss here: that there is such a thing as pure observational knowledge of the world that is untainted by modal knowledge. In his (2008), Brandom presents an argument designed to show that all our knowledge, observational or not, is shot through with modal knowledge.1 This argument has acknowledged Kantian and Sellarsian roots, but Brandom spells it out in greater detail than his predecessors. If Brandom’s argument works, empiricist worries are misplaced, or alternatively, we might say that the empiricist is powerless to meet these worries on her own terms, but that a better view of empirical knowledge will leave room for our knowledge of modal and counterfactual facts. If we know anything at all, then we know at least some modals. So Whewell’s problem was never something we should have worried about, since it rests on an incorrect conception of observation. Brandom’s argument is interesting, because if it works, we get a new kind of support for appeals to knowledge of counterfactuals, different from attempts to provide a direct rationalist support for such knowledge. Brandom’s central argument for his claim about knowledge of modals is in five steps, with both a preliminary and a more considered conclusion. It is mainly set out in chapter 4 of Brandom (2008). The argument starts with what Brandom calls the Kant-Sellars thesis about modality, which is that mastery of ordinary empirical vocabulary requires that we already know how to use modal vocabulary (2008, pp. 96ff, see also p. 115). Grasping any claim, modal or not, already presupposes grasping some counterfactual or modal claim. This kind of claim is not only supported in Kantian circles; we can find a recent endorsement of something much like it in Timothy Williamson:
In practice, the only way for us to be cognitively equipped to deal with the actual is by being cognitively equipped to deal with a wide variety of contingencies, most of them counterfactual. (Williamson 2007, p. 137) So understanding counterfactuals is a necessary precondition of understanding anything at all about the world. Brandom proceeds to develop this idea more systematically, in an argument in five steps. First, the argument in outline: 1. 2. 3.
Observationality: Every discursive practice must have some vocabulary that can be used observationally (Brandom 2008, p. 106).
Goodness of material inference: Those who engage in discursive practices must distinguish in practice between materially good and materially bad inferences (ibid.) Non-monotonicity: Material inference is in general non-monotonic. It is defeasible, and its defeasibility cannot be cancelled by some exhaustive spelling out of the possible defeaters (ibid.).
Justification: Many of a subject’s beliefs could only be justified by exhibiting them as conclusions of material implication. A believer is “epistemically responsible” insofar as she acknowledges a commitment to being able to justify many, if not most, of her beliefs (p. 108).
Epistemic responsibility: To count as a discursive practitioner, one must be at least minimally epistemically responsible (p. 108).
These five steps yield a preliminary conclusion, the updating problem: “Every change of belief ... is potentially relevant to the justification of every prior belief” (p. 108). I observe changes in the world around me all the time, and if every such change is potentially relevant for any one of my prior beliefs, I am in trouble. How are we supposed to be able to hold on to the right set of beliefs, and update successfully? Brandom argues that the only solution to the updating problem is that people who use a vocabulary already from the outset must have an idea of the “counterfactual robustness” for their material inferences. Two speakers, who on the surface agree on a factual claim, may turn out to not be in agreement, if they turn out to disagree about virtually every counterfactual related to the factual claim. In such cases, we can start to wonder if they even agree about the 1 Arguments of this form are also found in Brandom (1994), see for instance pp. 633-636. In Price (2007), Price considers a different line of interpretation of Brandom’s project, concentrating upon the genealogical side, the side that is interested in finding out how our modal thought develops, not what justifies it. I am not averse to such a reading, but have chosen to concentrate on understanding Brandom’s reasoning as an attempt to find a justification. basic, factual sentence. Agreement about factuals requires some kind of underlying agreement about counterfactuals.
The next step is that such counterfactuals can be used to introduce modal locutions, in the way Ryle suggests his (1950): understand “If p were true, q would be true” as being equivalent with “It is not possible that p and not q”. Then we can take our use of counterfactuals to account for our knowledge of modal truths. Since we are discursive creatures, we can be granted such knowledge, and hence we have provided a transcendental argument for modal knowledge: modal knowledge is needed for non-modal knowledge, and since we have such knowledge, we should realize that empiricist worries about modal knowledge are without basis. This picture of how our modal knowledge should best be understood requires further grounding. As it stands, it is little more than a sketch (so my account is a sketch of a sketch), and some of the details may turn out to be problematic. How severe is the Updating Problem? How much epistemic responsibility is required? What kinds of counterfactual are relevant for understanding a given factual statement? How do we know the counterfactuals? How much disagreement about counterfactuals can we tolerate? But I think that the general thrust of Brandom’s argument should be clear enough from my brief account, and it is this general strategy for grounding our knowledge of counterfactuals that I will focus on in my talk.
Some of the steps in the reasoning above are not controversial. For instance, (1) appears to be little more than a truism, whereas some other steps may be more problematic. The claim about non-monotonicity seems to me to be substantially correct, but I guess there are people out there, still trying to come up with completely monotonic patterns of reasoning for observational matters. But even if we grant all steps above, a central difficulty remains. What backing does the five-step argument give to our ability to use counterfactuals and modals? From an empiricist point of view, there is something almost miraculous about our supposed knowledge of counterfactuals: how could we know these, when there is no basis in observation for them? There may still be room for the empiricist to wonder.
First, how bad is the Updating Problem? There is at least a way of understanding this problem that is a good description of the scientist’s predicament: we simply don’t know in advance which new observations that are relevant for the standing of the beliefs we happen to hold right now. This goes with the non-monotonicity Brandom rightly stresses: if non-monotonicity means that we cannot spell out in advance what kinds of defeaters there will be for our beliefs, then we will be stuck with the Updating Problem, no matter what. Appeals to counterfactual robustness won’t help us here. It is part of scientific reasoning that new evidence, unanticipated or arriving from some completely unexpected source, can make us change our firmly held beliefs. So even if the Updating Problem maybe is mitigated by Brandom’s suggested way out, we will never be completely rid of it. For the second set of questions, it seems that Brandom’s argument gives us a weak conclusion – we are only given a guarantee that we, as responsible subjects, must make use of some counterfactuals, but nothing much is said about which counterfactuals, and it is not shown that we actually will know any counterfactuals, or which counterfactuals that are true. Perhaps all the five-step argument gives us is a weaker conclusion: in order to count as knowing facts, we must be using counterfactuals – but there is no real check on our ability to use them; it has not been shown that we must be able to use them correctly, as long as there is some agreement between speakers. The argument will by itself not give us that much check on our uses of counterfactuals. It can perhaps give us a backing for a conclusion that is slightly weaker, but not without bite: in order to count as communicating, we must hold each other responsible to roughly the same counterfactuals. Again, there is one way to understand this that even the staunchest empiricist can go along with. Rough agreement in counterfactuals is just a special case of rough agreement in beliefs, and will so not by itself have to be disallowed by the empiricist, even if she of course will have grave doubts concerning the legitimacy of such beliefs. So we have two readings of the conclusion of the argument: Weak reading: we must be in agreement on some counterfactuals, if we are to see each other as epistemically responsible and hence discursive creatures.
Strong reading: we must know roughly the same counterfactuals, if we are to see each other as epistemically responsible and hence discursive creatures.
The weak reading appears to follow from the five steps and the Updating Problem, but does not by itself give us knowledge of counterfactuals. The stronger reading would do just that, but I cannot see how the five steps would give us knowledge. If Brandom’s five-step argument is to give us the strong, and desired, conclusion, we need something extra: something showing that the use of a particular modal or counterfactual statement is justified, that there is some way to distinguish correct from incorrect use. The bare assertion that our knowledge of facts must rest on our using counterfactuals in some way or other, does not by itself show enough, even if it shows that we must agree (roughly) in our uses of counterfactuals.
But how might we go about to show this? There is a dilemma for Brandom’s position here: the transcendental argument doesn’t show that our use of counterfactuals is correct, so something more is needed. But if we beef up the transcendental argument with some other argument that shows the correctness of the use of counterfactuals, then no transcendental argument is needed: then we have a direct argument for our use of counterfactuals. Such “beefing up” might for instance be some kind of traditional appeal to intuition of necessary truths. But few put much faith in such intuition, and if we were to have such faith, there would be no need for the five-step argument: in that case we would just have direct modal knowledge. So the ambitious argument to show the indispensability of modal knowledge for observational knowledge is left dangling. There is in fact another indirect way of reasoning, that might appeal to Brandom. This is saying that we do have knowledge in general, and some of this knowledge is observational. Then we can argue that since there is no such thing as a purely observational layer of knowledge, we must in fact be granted modal knowledge as well. Perhaps this extra reasoning can be made to work, but it still seems vulnerable to the observation that it by itself doesn’t give us knowledge of any specific modals. So the argument as it stands is still incomplete.
The third issue I want to raise concerns Brandom’s phenomenalism about modality.2 The connections between necessity and our commitments appears to be problematic. The claim is that necessity is in a way a product of our commitments (for reasons of space, I will skip the fancy machinery of Between Saying and Doing; not because it’s uninteresting, but because I think the points I will be bringing up are unaffected by the exact set-up of the claim). Modality is as it were arising out of incompatibility of commitments (see Brandom 2008, ch. 5 and Appendices). But it seems that even if we spell out the intricate and interesting details of this story much better than this, some problems remain.
We can for instance all mistakenly see ourselves as committed to a certain claim, whereas it in fact turns out that we are not – we had missed some subtle aspect of the reasoning. Or things can be the other way round: we think we are not committed to something that in fact is a necessary consequence of other things we are committed to. It took a while until people managed to come up with alternatives to Euclidean geometry. The parallel postulate was not necessary, but everyone was holding themselves and everyone else committed to it for a long time. Intuitionists think that some people are mistakenly committed to something that in fact is false. People working within the confines of classical logic think that dialetheists are wrongly seeing themselves as non-committed to the law of non-contradiction.
Perhaps we can try to counter such observations by saying that this doesn’t really matter: various communities can be determined by their different commitments. So instead of a whole group being wrong about a given commitment, we might see the group as existing as the result of the commitment. But this doesn’t really work, since determination is itself a modal notion (if A is determined by B, the truth of A is necessitated by B), so commitment, understood in this modified way, cannot be used to explain modality.
A related problem remains. Consider a purported necessary truth p. In a sense, consistent with the phenomenalism about deontological status, p is a necessary truth because we are committed to treat it like a necessary truth. Are we then committed to treat this commitment as creating a necessary truth? Neither alternative is good. If the answer is yes, then we need an extra commitment to make our commitment into a commitment that gives us necessary truths. If the answer is no, then it is unclear what the first commitment accomplished. No doubt this general sketch of an argument can and should be tidied up in several respects, not least to do better justice to the intricacies of Brandom’s developed theory, but I think a workshop on Brandom’s analytical pragmatism would be the right setting to at least start examining the issues.
A final point I would like to consider is probably pretty minor, but it seems to indicate something of interest. This is that there may be a kind of lingering circularity in the Kant-Sellars thesis itself. It is not clear from the wording, but the gist of the thesis must be counterfactual in nature: if we didn’t have counterfactual knowledge, we would not have observational knowledge at all. So someone who doubts that we know counterfactuals will not be swayed by a thesis, formulated in a way that assumes that we do know counterfactuals. Why would this still be a minor point – isn’t circularity always bad? Well, in this case, maybe not. It is not always unfair to assume an argued version of what you think is the most basic means of acquiring knowledge. The alleged circularity here is perhaps unavoidable but harmless – if our basic ways of knowing something really must integrate knowledge of counterfactuals, then it is no great surprise that the statement of such knowledge in itself involves counterfactuals, just as we have to lean on some kind of logic when explaining logic. We have to start somewhere, and starting in the middle of our ability to know stuff, just like any naturalist would, seems like a good place to start.