¹ One might even be tempted to say that here the distinction between the two breaks down, or perhaps more accurately that the distinction was, historically, not one that was yet clear-cut. Not surprisingly, one can find suggestions of the importance of grammar in Plato; in addition to the obvious Cratylos, see Theaetetus 183a and the ‘Seventh letter’. T. Benfey argues for the idea that Plato is suggesting the idea of a universal language scheme in ‘Über die Aufgabe des platonischen Dialogs Kratylos’, Abhandlungen Göttingen, Philologische-Historische Klasse 12 (Göttingen, 1866), pp. 189–330. Sextus Empiricus gives a long discussion of grammar in Against the Professors [Mαφεμακους], Chapter III. Such examples can be easily multiplied from the history of Classical, Medieval, and Renaissance philosophy. See also Frede, Michael, ‘The origins of traditional grammar’, in Essays on Ancient Philosophy (Minneapolis: University of Minnesota Press, 1987), pp. 338–59, which emphasizes the Stoic contribution to this history.Google Scholar

² Russell, Bertrand, The Principles of Mathematics (London: Allen & Unwin, 1903), p. 42.Google Scholar

³ I have tried to suggest why this is in ‘Why doesn't Kant care about natural language?’, Dialogue, 40 (2001), pp. 25–51.CrossRefGoogle Scholar

⁴ One might recall that at A78=B103 Kant originally wrote that ‘Synthesis …is a blind but indispensable function of the soul’, but changed ‘soul’ in his own copy (Nachträge XLI) to ‘understanding’. This is probably how this claim from the Logik Wiener should be read, as well. It should also be noted that Kant never develops his remark here about a ‘transcendental grammar’, qua grammar, for reasons I have presented in some detail in ‘Why doesn't Kant care about natural language?’

⁵ As Béatrice Longuenesse makes the point, for Kant ‘no judgment [as psychological activity] can take place without linguistic expression.’ Kant and the Capacity to judge., trans. Wolfe, Charles T. (Princeton, NJ: Princeton University Press, 1998), p. 100Google Scholar n. 47. See also Brandt, Reinhardt, Die Urteilstafel (Hamburg: Meiner Verlag, 1991) p. 77 n. 54; ‘For Kant there is no thought that would not be grasped [gefaβt] linguistically.’Google Scholar

⁶ This naive characterization is put forth to emphasize the similarities in the procedures shared by linguists and logicians, rather than a precise taxonomy of the subdisciplines of linguistics, and in particular is designed to focus on the universal features (that Kant sees) shared by grammar and logic. To be sure, a linguist treats morphological and syntactic issues as involving separate but related questions, while the logician would treat both as syntactic, in contrast to semantics. (Not infrequently, the close tie between morphology and syntax in linguistics is expressed by the hybrid ‘morpho-syntax’.) I also do not take up the important differences the linguist might describe in terms of ‘competence’ and ‘performance’, nor those complications introduced by Chomsky's conception of ‘degrees of grammaticality’.

⁷ Kant scholars have generated their own cottage industry in discussing the metaphysical deduction, including debates over what its argument is, how it proceeds, and even where it is located. Understanding the claim at A79 = B105 may well be the key to understanding the metaphysical deduction, as indicated by the amount of attention it has received, from such classic texts as Paton's Kant's Metaphysic of Experience (London: Macmillan, 1951Google Scholar) and Reich's, The Completeness of Kant's Table of Judgments (Berlin: Richard Schoetz, 1932Google Scholar), on to Allison's, Kant's Transcendental Idealism (New Haven, CT: Yale University Press, 1981)Google Scholar, Guyer's, Kant and the Claims of Knowledge (Cambridge: Cambridge University Press, 1987Google Scholar) and Pippin's, Kant's Theory of Form (New Haven, CT: Yale University Press, 1982Google Scholar), up to recent treatments in Longuenesse's Kant and the Capacity to Judge (Princeton, NJ: Princeton University Press, 1998Google Scholar) and Greenberg's, Kant's Theory of A Priori Knowledge (University Park, PA: Pennsylvania State University, 2001)Google Scholar; dozens or more books, and scores of papers, could be added to this list. Greenberg offers a particularly ambitious - and complex - defence of the claim that the functions referred to in this passage ‘make their own contribution to the content of knowledge in general’. As Greenberg notes, many might regard this as a ‘surprising thesis’ (p. 138), a surprise that seems merited, given the stark conflict it seems to generate with Kant's own characterization of general logic. Taking up the issues involved here, in the detail required and with respect for the rigour of the arguments mentioned above, would take this paper too far afield from its focus on Kant's remarks on grammar and language as informing the analogy he exploits in his discussion of general logic and transcendental analytic.

⁸ In his ‘Frege, Kant, and the logic in logicism’, The Philosophical Review, 111 (1) (2002), 25–65CrossRefGoogle Scholar, John MacFarlane takes up the issue of whether Kant would recognize Frege's Begriffschrift as a logic, or would adopt the position that Frege's conception of logic simply ‘changes the subject’. While the details go beyond the scope of this paper, it should be pointed out that the move from general to transcendental logic (analytic) results in a very different approach to the relationship between logic and mathematical content. As MacFarlane writes, ‘Thus logic, for Frege, cannot abstract from all semantic content: it must attend, at least, to the semantic contents of the logical expressions, which on Frege's view function semantically just like non-logical expressions. And precisely because it does not abstract from these contents, it can tell us something about the objective world of objects, concepts, and relations, and not just about the “forms of thought”’ (pp. 29–30).

⁹ Melnick, Arthur, Kant's Analogies of Experience (Chicago: University of Chicago Press, 1973), p. 40.Google Scholar

¹⁰ Svendsen, Lars Fr. Händler, Kant's Critical Hermeneutics: On Schematization and Interpretation (Oslo: Faculty of Arts Dissertation, 1999), p. 244 n. 170.Google Scholar

¹¹ A claim by Roger Bacon encapsulates much of Kant's thinking on logic, and its relation to language, that ‘in substance grammar is one and the same in all languages, although it allows variation in terms of accident[s]’ (Bacon, Roger, Oxford Greek Grammar, ed. Nolan, E. and Hirsch, S. A., (Cambridge: Cambridge University Press, 1902), p. 27Google Scholar [part II.1.2], dated by the editors as written around 1252; the original reads ‘grammatica una et eadem est secundum substanciam in omnibus linguis, licet accidentaliter varietur’).

Kant, that is, comes very close, in his characterization of both logic and grammar, to embracing a view strikingly similar to that put forth, in various ways, by those arguing from a different perspective for a ‘universal language’ or a ‘universal grammar’, although this connection has remained, as far as I can tell, almost entirely unexamined in the literature on Kant. The only exception to this with which I am familiar is Lars Fr. H. Svendsen, in his dissertation Kant's Critical Hermeneutics. Jere Paul Surber briefly considers the possibility in his Language and German Idealism: Fichte's Linguistic Philosophy (Amherst, NY: Humanity Books, 1996) p. 97Google Scholar, but quickly rejects it. Kant is entirely absent from the discussions in Umberto Eco, The Search for the Perfect Language (London: Blackwell, 1997) and Rossi, Paolo, Logic and the Art of Memory (Chicago: University of Chicago Press, 2000).Google Scholar Henry Allison considers a work relevant to this issue, Arthur Collier's Clavis Universalis, but focuses only on Collier's role in Kant's understanding of Berkeley's works; ‘Kant's critique of Berkeley’, Journal of the History of Philosophy, 11 (1973), pp. 43–63.Google Scholar

Kant rejects any natural language as a candidate for a model of a universal grammar, but focuses on the logical features that any such language must satisfy, and constitute a contribution that the judging subject makes a priori (albeit discovering it only on the basis of what Kant calls ‘transcendental reflection’). That contribution, articulated in terms of the employment of a priori concepts and rules, as delineated in the First Critique, would then constitute a universal grammar, or logic, of thought and reason. We see a typical feature of Kant's two-front strategy in action here: a fundamental challenge to the sceptic to make meaningful claims (including the capacity to make claims that assert sceptical views or conclusions) without presupposing (and satisfying) certain universal and necessary conditions, and a powerful critical weapon against the inevitable temptation to proceed cognitively beyond the legitimate boundaries of possible experience.

In an earlier draft of this paper, I discussed at some length this set of issues, particularly in light of the school of ‘Modistae’; yet, as a reviewer of that draft correctly noted, our current understanding of Kant's familiarity with this school has not been systematically explored, and any such connections are conjectural. Consequently, I have eliminated those remarks as too speculative.

¹² Kant knew some Greek, although it is not clear how much; the general scholarly consensus is that he had virtually no ability to read English; his Latin was excellent; and he was quite at home in reading French.

¹³ Harris's text was, however, known to Herder and Hamann as early as 1768 (see Cassirer, E., The Philosophy of Symbolic Forms, vol. I, p. 145 (New Haven, CT: Yale University Press, 1953))Google Scholar, and thus some of his views may have been transmitted to Kant indirectly. Warda's standard catalogue of Kant's library (Immanuel Kants Biicher (Berlin: Martin Breslauer, 1922)) doesn't indicate any copies of the kinds of texts on language discussed here, but is not a wholly reliable source in determining what texts Kant had access to or had read.Google Scholar

¹⁴ Svendsen, Kant's Critical Hermeneutics, p. 264.

¹⁵ Carboncini, S. and Finster, R., ‘Das Begriffspaar Kanon-Organon’, Archiv fur Begriffsgeschichte, 26 (1982), p. 29.Google Scholar Informally, one might think of general logic as a ‘filter’, through which any judgement must pass in order to be evaluated as true or false; at that stage, a syntactically ill-formed judgement would be eliminated as a candidate. Transcendental analytic would function as a second filter, in terms of evaluating a judgement of possible experience as true or false. In this sense, then, general logic is the logic of thinking in general; transcendental logic is the logic of thinking about objects. (Without arguing for it here, I believe that the strictures of the categorical imperative in Kant's practical philosophy function in an analogous fashion.) The fundamental point here is to see that, while a given judgement may well ‘pass’ through such a preliminary process, that leaves us quite far from determining whether that judgement is in fact true (or false), and that Kant's conception of logic seeks to establish precisely the former result. Similarly, a moral judgement that does not violate the categorical imperative is not ipso facto the thing to do. In short, we should not expect a set of rules - whether of general logic, transcendental logic or practical philosophy - to establish more than it is in a position to do qua a set of rules. I owe this way of formulating this aspect of the relation between the theoretical and practical philosophy to Daniel Farrell. It might be thought that this is a very rough and preliminary version of the ‘criterial’ account - specifically in terms of the moral law - developed by Kerstein, Samuel in his Kant's Search for the Supreme Principle of Morality (Cambridge: Cambridge University Press, 2002).Google Scholar As an anonymous reader of an earlier draft rightly points out, if general logic simpliciter lacks semantic content {pace Greenberg, above), it becomes difficult to determine the relationship between judgements about the ‘thing in itself and general logic, particularly if such judgements are said to possess content. The short way with this is to regard judgements of practical reason as subject to the laws of general logic and to distinguish them from judgements of possible experience in terms of the domain over which they range. The longer way is to take up the issue of the role the thing in itself plays in the first Critique, and whether (and how) one distinguishes that role from the role it plays in Kant's discussion of the relationship between freedom and the categorical imperative. Riidiger Grimm, for one, has claimed that it could be ‘persuasively argued’ that ‘the concept of the thing-in-itself is really quite dispensable to Kant's project in the Kritik’; Nietzsche's Theory of Knowledge (Berlin: de Gruyter, 1977), p. 56. This seems to be Kant's point at A30=B45, ‘The true correlate of sensibility, the thing in itself, is not known, and cannot be known, through these representations [of sensibility]; and in experience no question is ever asked in regard to if (my emphasis). Obviously enough, these issues go beyond the scope of the present discussion.

¹⁶ Keeping in mind, of course, that thought - one might say here ‘propositional content’ - is expressed by means of judgement, and that judgement itself must be expressed linguistically.

¹⁷ As Manley Thompson observes, there is an important distinction to be drawn here, relative to ‘thinking a contradiction’. ‘We think a contradiction only when we think it as such, as thought that cancels itself. In thinking a contradiction without thinking it as such, we fail to think anything at all - we are illogical.’ ‘On a priori truth’, Journal of Philosophy, 78 (1981), p. 471Google Scholar n. 8. For an extended discussion of Kant's notion of contradiction, from a very different perspective, see Michael Wolff, ‘Der Begriff des Widerspruches in der “Kritik der reinen Vernunft”’, in Tuschling, B. (ed.), Probletne der “Kritik der reinen Vernunft” (Berlin: de Gruyter, 1984), pp. 178–202.Google Scholar

¹⁸ This claim rests on Kant's contention that we can determine in an analytic judgement whether or not the predicate is ‘contained’ in the subject. Beck, L. W., in ‘Can Kant's synthetic judgements be made analytic?’ (in, Wolff, R. P. (ed.), Kant: A Collection of Critical Essays (New York: Doubleday, 1967), pp. 3–22Google Scholar) helpfully distinguishes between a ‘logical’ and a ‘phenomenological’ criterion for determining the analyticity of a judgement; only the former seems able to withstand the consistent criticism, first systematically treated by Eberhard, that the analytic-synthetic distinction is variable, dependent upon how much one knows at a given time. The phenomenological criterion does seem vulnerable to this charge; as Beck recognizes, ‘Kant was not free from a psychologizing introspective tendency in his decisions on what is analytic and what is synthetic’ (p. 9).

¹⁹ Junguis, J., Logica Hamburgensis, ed. and trans, (into German) Meyer, R. W. (Hamburg: Augustin, 1957), p. 1; the original reads ‘Logica est ars mentis nostrae operationes dirigens ad verum à falso discernendum.’ Kant would challenge the notion that a scientific logic ultimately is an ‘ars’.Google Scholar

²⁰ In ‘Zum Wahrheitsbegriff bei Kant’ (Kant-Studien, 60 (1969), pp. 166–82)Google Scholar, Prauss gives a detailed analysis of A57ff=B82ff, beginning with the ambiguity of Kant's statement (translated by Kemp Smith as ‘For if a question is absurd in itself … ‘) ‘Denn, wenn die Frage an sich ungereimt ist ….’ The ambiguity is whether this statement should be taken as (in Prauss's reformulation) ‘die Frage als solche ist ungereimt’ or ‘die Frage ist eigentlich (an sich) ungereimt’; that is, whether the question in itself is absurd, or the question is absurd in itself. Prauss argues that the former paints a history of philosophy in which all philosophers - before, after and including Kant - are posing a question that as such is absurd, and thus suggests the latter reading, in which the question is absurd as posed by, or to, general logic. It should also be noted that Kemp Smith translates here ‘die Frage’ as ‘a question’; the correct ‘the question’ shows more clearly what Kant is getting at, and would seem to support Prauss's interpretation. Guyer and Wood, and Pluhar, in their translations of the Critique, both translate the claim as ‘the question’.

²¹ Logick hath it's name from λογας ratio, because it's an art which teacheth Reason and Discourse. Ther is a twofold Logick, One Natural, which is nothing but natural reason, or of which every man (which is not an Idiot) doth in some measure partake: The other Artificial, which is the perfection of the Natural, nam Ars perficit naturam: this belongeth only to Schollars.’ A Brief English Tract of Logick (Oxford: L. Lichfield, 1677).Google Scholar

²² I take it as obvious that one who had zero beliefs would not qualify as an agent; while, technically, one who had exactly one belief would not be in jeopardy of being inconsistent, it is not clear that such a being would qualify as an ‘agent’ (except, perhaps, in some metaphorical or idealized Kierkegaardian sense). Indeed, the more interesting constraint may well come from the other direction: Christopher Cherniak has pointed out that, even for a physically idealized computer, the ‘supercycle’ of which is approximately 2.9x10–23 seconds, to use a truth-table to evaluate a set of 138 distinct beliefs could not be done ‘during the entire history of the universe’; Minimal Rationality (Cambridge, MA: MIT Press, 1986), p. 143Google ScholarPubMed, n. 3. As Cherniak observes, a set of 138 atomic beliefs might well be too ‘small-minded’ to be possessed by ‘anything we could call a full-fledged agent’; p. 94.

²³ Korner, S., Kant (Middlesex: Penguin, 1955), p. 26.Google Scholar Körner never really elaborates on his qualification of ‘in a sense’ relative to traditional logic.

²⁴ Kneale, William and Kneale, Martha, The Development of Logic (Oxford: Clarendon Press, 1962), p. 358.Google Scholar Similarly, Hegel relies on such a view when he writes: ‘Now if logic has not undergone any change since Aristotle - and in fact, judging by modern compendiums of logic the changes frequently consist mainly in omissions - then surely the conclusion which should be drawn is that it is all the more in need of a total reconstruction.’ Hegel, G.W. F., Hegel's Science of Logic, trans. Miller, A. V (London: Allen & Unwin, 1969), p. 51.Google Scholar (It is worth noting that, at least relative to syllogistic, Hegel writes that Aristotle's account was done ‘with such sureness and accuracy that, in essentials, there has been no need to add anything.’ The Encyclopedia Logic, translated by T. F. Geraets, W. A. Suchting and H. S. Harris (Indianapolis: Hackett, 1991), §183z (p. 260).) Peirce observes that ‘we are to remember that, according to Kant, nothing worth mention had been contributed to logic since Aristotle’; Elements of Logic, in Collected Papers of Charles Sanders Peirce, vol. II, ed. Hartshorne, C. and Weiss, P. (Cambridge, MA: Harvard University Press, 1931)Google Scholar, §39; P. F. Strawson notes that Kant believed ‘without question’ in the ‘finality’ of Aristotelian logic; The Bounds of Sense (London: Methuen, 1966), p. 23Google Scholar; E. W Beth clearly has this standard interpretation in mind when he writes that ‘Aristotle's logical work, far from being a closed system, already contained the germs of future development’; Beth, E. W., ‘Hundred years of symbolic logic’, Dialectica, 1 (1947), p. 332.Google Scholar

²⁵ Thompson, ‘On a priori truth’, p. 472. The ‘neo-Tractarian’ conception of logic here is that of Tractatus 5.4731: ‘What makes logic a priori is t he impossibility of illogical thought.’

²⁶ As one example, note Michael Friedman's suggestion that we take Kant's conception of logic as amenable to ‘at most, monadic quantification theory plus identity’ in his Kant and the Exact Sciences (Cambridge, MA: Harvard University Press, 1992), p. 63Google Scholar n. 9, while rejecting Thompson's claim that ‘The general logic required by Kant's transcendental logic is thus at least first-order logic plus identity’, a claim Thompson subsequently revised in arguing that predicating in terms of ‘Fx’ (in contrast to Aristotelian ‘S-P’ judgement-forms) would be a general logic developed from Kant's transcendental logic. See M. Thompson, ‘Singular Terms and Intuitions’, p. 334, and the postscript added in Posy, C., (ed.) Kant's Philosophy of Mathematics: Modern Essays (Dordrecht: Kluwer, 1992).Google Scholar The point here is that, if Kant's general logic is as ‘flexible’ as Friedman and Thompson (and others) indicate, then it makes little sense to restrict Kant's conception of general logic to Aristotelian syllogistic.

²⁷ A formal system is ‘weakly complete’ if and only if all logically true well-formed formulas are theorems of the system; a formal system is ‘strongly complete’ if no contradiction results from the addition to the system of an independent axiom.

²⁸ ‘[O]ur time's view of logic, as formal logic, or as symbolic logic, was completely foreign to philosophy until the beginning of the nineteenth century’; Tonelli, Giorgio, Kant's Critique of Pure Reason within the Tradition of Modem Logic, ed. Chandler, David (Hildesheim: Olms, 1994), pp. 2–3.Google Scholar

²⁹ This, again, is a fundamental aspect of what I earlier characterized, following Thompson, as a ‘neo-Tractarian’ conception of logic.

³⁰ Strawson, P.F., ‘Sensibility, understanding and the doctrine of synthesis: Comments on Henrich and Guyer’, in Förster, E. (ed.), Kant's Transcendental Deductions (Palo Alto, CA: Stanford University Press, 1989), p. 71.Google Scholar One might compare a still more surprising source - Quine - who notes that Kant's conception of logic excludes any notion of set theory, and suggests that we reserve the very word ‘logic’ for the kind of elementary logic Kant employs. See Quine, W V, ‘Carnap and logical truth ‘, in The Ways of Paradox (New York: Random House, 1966), pp. 103–4.Google Scholar