Ungrounded reason

Journal of Philosophy

Volumn 105, Issue 9, 2008, Pages 453-471

  Horwich, Paul ^a  

^a NEW YORK UNIVERSITY   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 57149135067     PISSN: 0022362X     EISSN: 19398549     Source Type: Journal    
DOI: 10.5840/jphil2008105928     Document Type: Article

References (49)

1. I will be using the terms 'rational', 'justified', 'reasonable', 'warranted', 'entitled', 'what ought to be believed', and so forth, more-or-less interchangeably - although, as we will see later, it is important to distinguish between epistemic obligations and epistemic permissions.
2. The fundamental epistemic norms governing a priori belief (if there is such a thing) are likely to be somewhat different from those governing a posteriori belief. Nonetheless, when it comes to the present foundational question of which phenomena are responsible for what appear to be the basic epistemically normative principles, the answer may very well be the same in the two cases. Indeed, each of the most commonly proposed answers purports to cover both domains. So, even if one happens to be especially concerned with the origins of a priori justification, it would be a mistake to assume that the appropriate investigation will be more restricted than the one that will be conducted here.
3. This sort of strategy was deployed by Hilbert, Poincaré, and the logical positivists. It has been advocated more recently (in a variety of different forms) by Paul Boghossian in "Analyticity Reconsidered," Noûs, XXX, 3 (1996): 360-91
4. in "How Are Objective Epistemic Reasons Possible," Philosophical Studies, CVI (2001): 340-80
5. and in "Blind Reasoning," Proceedings of the Aristotelian Society, Supplementary Volume, LXXVII, 1 (2003): 225-48;
6. by Christopher Peacocke in A Study of Concepts (Cambridge: MIT, 1992)
7. in "How Are A Priori Truths Possible?" European Journal of Philosophy, I (August 1993): 175-99
8. and in Truly Understood (New York: Oxford, 2008);
9. by Bob Hale and Crispin Wright in "Implicit Definition and the A Priori," in Boghossian and Peacocke, eds., New Essays on the A Priori (New York: Oxford, 2000), pp. 286-319;
10. and by Stephen Schiffer in "Non-Inferential A Priori Justification" (unpublished manuscript).
11. I am using the term, 'accept', in a technical sense, for the particular psychological relation between a person and a sentence that correlates with his believing the proposition expressed by the sentence. In light of this correlation, I will feel free to reformulate norms of belief as norms governing the acceptance of interpreted sentences.
12. For simplicity, the focus (in this particular section) is on epistemic principles that concern the rationality of unconditionally accepting certain interpreted sentences. But the discussion (with some obvious adjustments) is also intended to cover norms of inference, and externalist norms dictating the acceptance of certain sentences in certain specified conditions.
13. Notice that this way of demonstrating the rationality of an acceptance-practice cannot be completely general, since it depends on more basic epistemologically normative assumptions: namely, those that will need to be deployed in certifying our justification for accepting the premise and for making the inference from it.
14. We should distinguish two versions of this principle. One of them concerns idiolectical meanings, which vary from one person to another within a community - each one being fixed by an individual's own linguistic activity. It says that S's meaning-constituting practices are rational for him. The other version concerns communal meanings, which are assumed to be constituted by the common linguistic activity of the majority (or, of the relevant "experts"). It says that if S is a member of linguistic community, C, then C's meaning-constituting practices are rational for S - even if S personally does not engage in them. I will focus on the first version, since I believe that this is the one that proponents of the semantogenetic strategy have had in mind. But my criticisms will apply to the second version as well.
15. Any plausible explanation, in wholly non-normative terms, of the most basic epistemically normative facts would have to involve a reductive analysis of the concepts, RATIONAL, JUSTIFIED, OUGHT, and so forth. But, for familiar reasons - for example, Moore's "open question" argument, and considerations regarding the motivational power of normative belief - it seems unlikely that a naturalistic analysis of these normative notions is possible. Therefore, although some proponents of the semantogenetic view might be interpreted as suggesting that it offers a way of grounding epistemic norms in entirely non-normative facts about meaning, I think that a more charitable interpretation runs along the less ambitious lines that I consider in the text, whereby the aim is merely to explain what are in practice taken to be the basic norms in terms of a yet more basic one.
16. See, for example, Prior's "The Runabout Inference Ticket," Analysis, XXI (1960): 38-39
17. Dummett's Frege: Philosophy of Language (London: Duckworth, 1973)
18. and Boghossian's "Blind Reasoning."
19. The point is that, since a person cannot decide to perform every one of the alternative actions open to him, he cannot have every one of the beliefs that would motivate those decisions. For more on this, see my "Implicit Definition, Analytic Truth, and A Priori Knowledge," Noûs, XXXI (1997): 423-40.
20. Of course, it is perfectly possible to come to believe (either rationally or irrationally) that the English are pretentious. And one might then proceed to infer in accord with the pair of rules under consideration - defining 'pom' as 'English'. But, in that case, the meaning-constituting disposition - the basic source of the word's overall use - would be to intersubstitute those two terms, not to follow the above pair of inference rules. Those rules could be what provides 'pom' with its meaning only if they were followed underived - which is impossible, given the present meanings of 'English' and 'pretentious'.
21. If the norm is taken to concern communal meanings (see footnote 8) then there is a little more scope for a response to this objection. For the norm would genuinely regulate anyone whose basic uses of a word do not match the practice of the majority (which is what is assumed to fix the word's meaning). The difficulty is still not entirely removed, however. For we surely feel that epistemic norms apply to everyone - not only to those whose understanding of language is deficient.
22. But see footnote 17 for further discussion.
23. One might attempt to resist the force of this observation by reference to social externalism - namely, that an individual member of a linguistic community may be credited with meaning what the majority (or the "experts") mean, even though his own usage diverges somewhat from theirs (which constitutes that communal meaning). But such a response is in fact beside the point. For even someone with a perfect understanding of some theoretical term might nonetheless explicitly dissociate himself from the predictive-explanatory practice for the sake of which the term was introduced. To put it another way: if there were a community - causally disconnected from ours - in which 'T (g)' was entertained and universally rejected, 'g' could nonetheless have the same meaning there as it does here, where 'T (g)' is universally accepted. Some philosophers (notably Timothy Williamson in chapter 4 of his The Philosophy of Philosophy (Malden, MA: Blackwell, 2007)) take the moral of the "dissenting expert" to be that there are no sentences whose acceptance is a necessary condition of our meaning what we do by them. And, for similar reasons, one might think that no acceptance-practice with a word (no matter how complex and tentative) is essential to its meaning. From this point of view, the semantogenetic approach is an obvious nonstarter.
24. The idea of some such two-way factorization originated in the works of Bertrand Russell, The Analysis of Matter (London: Allen and Unwin, 1927)
25. Rudolf Carnap, Der Logische Aufbau der Welt (Berlin: Schlachtensee Weltkreis, 1928)
26. Frank Ramsey, "Theories" (1929)
27. reprinted in his Foundations, D.H. Mellor, ed. (New York: Routledge and Kegan Paul, 1978)
28. and David Lewis, "How to Define Theoretical Terms," this JOURNAL, LVII, 13 (July 9, 1970): 427-66.
29. According to their way of implementing this idea, the two factors are tantamount to (i) accepting '∃ΦT(Φ)', and (ii) accepting '∃ΦT(Φ)→T(f)'. But my own tentative characterization of the two commitments diverges from this proposal in various respects. First, I avoid objectual second-order quantification (and the problem of what the variables would range over). Second, I aim for an account that will cover practices (such as logical reasoning) that do not consist simply in the acceptance of a certain conjunction of sentences. And third, I have come to think (post Reflections on Meaning (New York: Oxford, 2005)) that in order to mean what we do by 'f' it will not suffice that we are prepared (in suppositional contexts) to accept 'T(f)' given 'T(g)', or given 'T(h)', and so forth (which corresponds to factor (ii) in the above traditional model). We must, in addition, have a disposition to infer 'T(f)' from 'fs exist'.
30. If I am right in thinking that conditional commitment (b) is involved in fixing the meanings of the logical constants, then elimination rules - such as 'p and q/ :.p' and 'p, If p then q/:.q' - must be followed by anyone who perfectly understands the eliminated terms. So one might hope to argue that, after all, at least some of our basic logical norms can be explained in terms of meaning-constitution. But it would remain to overcome the difficulty discussed in section in. Intuitively, we are obliged to follow these rules (given what we mean); but how can an obligation to do something stem from actually doing it?! Now, one might respond that this intuition is mistaken - that in fact the normative status of the rules, given their meaning-constituting potential, is merely that it is legitimate for us to follow them. But then the question arises as to why we should be obliged to obey them. And (as mentioned in footnote 14) the answer would require invoking something like the following norm: If someone legitimately follows rule R(w), and if that is meaning-constituting, then he is epistemically obliged to obey R(w). But this seems ad hoc. Rules do not generally engender such obligations.
31. Boghossian argues (in "Blind Reasoning," p. 247) that the meanings of logical terms cannot be given by a conditionalized version of logical theory, since its formulation (like any other conditionalized theory-formulation) would itself require logical terms. But one might respond that, since meaning-constituting rules of use are followed implicitly, there is no need for us to have a prior understanding of any formulation of them; a certain practice will do; and I have just been indicating what form, in the case of logical terms, that practice might take.
32. In section II it was suggested (apparently in defense of the semantogenetic view) that there may be no irrational meaning-constituting practices. But we now see that this defense contains the seeds of the present objection - namely, that the practices necessary (as well as sufficient) for meaning what we do are too weak to be of epistemological interest.
33. There is a nice question (put to me by Andreas Kemmerling) as to why the meaning-constituting practice for 'bachelor' does not relate to 'The bachelors are the unmarried men' in a way that exactly parallels the conditionalized factor that I suggested is appropriate for scientific terms, for arithmetic, and for logic. I think the answer resides in the fact that, whereas the latter terms are introduced for the sake of being able to engage in certain corresponding substantive practices, 'bachelor' and other explicitly defined words are introduced merely for the sake of abbreviation and convenience.
34. And see also my discussion of it in Reflections of Meaning, chapter 5, pp. 163-65.
35. For further critical discussion of the relationship between epistemic rationality and meaning-constitution, see Joshua Schechter and David Enoch's "Meaning and Justification: The Case of Modus Ponens," Noûs, XL, 4 (2006): 687-715;
36. and their (this time by Enoch and Schechter) "How Are Basic Belief-Forming Methods Justified," Philosophy and Phenomenological Research, LXXVI, 3 (2008): 547-79.
37. See also my Reflections on Meaning, chapter 5. The main conclusions of that chapter are the same as those reached here. But note the following differences: (1) I am focused here on the idea of a basic semantogenetic ur-norm, rather than the idea of grounding the correctness of epistemic norms in non-normative semantic phenomena; and (2) I am now inclined (as sketched in footnote 16) to depart from the traditional way of articulating our conditionalized meaning-constituting commitments.
38. See Jerrold Katz's Realistic Rationalism (Cambridge: MIT, 1998);
39. Laurence Bonjour's In Defense of Pure Reason (New York: Cambridge 1998);
40. and George Bealer, "A Theory of the A Priori," Pacific Philosophical Quarterly, LXXXI (2000): 1-30. Note that the facts detectable through intuition might be held to include normative facts - for example, that whatever seems red should tentatively be assumed to be red.
41. See Alvin Goldman's Epistemology and Cognition (Cambridge: Harvard, 1986);
42. Marshall Swain's Reasons and Knowledge (Ithaca: Cornell, 1981);
43. and Ernest Sosa, "Beyond Internal Foundations to External Virtues," in Epistemic Justification, Bonjour and Sosa, eds. (Malden, MA: Blackwell, 2003), pp. 97-171.
44. Consider a correct, empirical, law-like regularity, 'All As are Bs', that would be reasonable to believe only on the basis of inductive support. And imagine someone who believes it despite having no such support. That belief is irrational even though it is true. Therefore, even though his method of arriving at the belief that something is B - namely by inferring it from the thing being A - is perfectly reliable, both the method and its epistemic products are irrational. As for rational but relatively unreliable belief-forming methods, surely our actual inductive practice falls in this category. It is far from 100% reliable. We very often make rational but false predictions. But if we knew the laws of nature, we would not make these mistakes. Thus, an inductive practice that happened (luckily) to build in the truth of the actual laws would be much more reliable than ours.
45. For detailed discussion, see my "The Value of Truth, " Noûs, XL, 2 (2006): 347-60.
46. See Robert Brandom's Making It Explicit (Cambridge: Harvard, 1994);
47. and Mark Lance and John Hawthorne's The Grammar of Meaning (New York: Cambridge, 1997). Brandom's formulation of the idea is that we have adopted certain implicit normative attitudes (manifested in our practices of acceptance, argument, and criticism) and that these commitments institute epistemic obligations. This strategy is often called "constructivism": the facts of how we ought to regulate our beliefs are said to be constructed, or constituted, from the facts of how we (implicitly) think we ought to regulate them.
48. Let me articulate this objection to Brandom's constructivist position in terms that are closer to his own. There may well be moral reasons (of cooperation) for a person within a linguistic community to operate with words in such a way as to manifest certain implicit normative attitudes. And these may indeed engender certain objective normative facts. But it is hard to see how any epistemically normative facts can be engendered ("instituted") in this way. We have been given no account of how our taking some pattern of reasoning (say, modus ponens) to be epistemically rational could conceivably make it so.
49. Thus we need not be moved by the second paragraph of the following passage from Boghossian's "Blind Reasoning" (p. 239): You want to know which inference patterns are permitted to be blind [that is, underived, (PH)]? These ones: Modus Ponens, Non-Contradiction, and a few others. Do not ask why it is precisely these inference patterns that are sanctioned. There is no deep answer to that question; there is just the list. What makes this brand of deflationary answer unsatisfactory is that it is hard to believe that the property of being warrant-transferring is simply a primitive property that an inference pattern either has or fails to have. Surely, if an inference pattern is warrant transferring there must be some property by virtue of which it is warrant transferring. And our question is: What, in the most basic cases, in which reflectively available support is not possible, could that property be? My response is that it is for certain identifiable and questionable reasons (just mentioned) that someone might find the deflationary "nothing-but-a- list" answer so hard to believe.