Thursday, July 25, 2013

Epistemology and Kinds of Knowledge

This is a prolegomena to a discussion of the status of Mises’s human action axiom and praxeology. Mises is usually thought to have argued that the human action axiom had a “synthetic a priori” status. Whether that is credible is highly questionable.

But before a serious discussion of this issue, some comments about epistemology are necessary. Hence this post.

“Epistemology” is the branch of philosophy concerned with the theory of knowledge. How do know truths? What kind of knowledge and truths are there?

Though it is a simplification, in essence there are four positions held since the late 18th century on the kinds of knowledge:
(1) The radical view of Willard Van Orman Quine that the analytic–synthetic distinction is a dogmatic myth;

(2) the empiricist view, associated with the logical positivists, that there are only two relevant distinctions:
(i) analytic/a priori, and
(ii) synthetic/a posteriori truth and propositions.
(3) the view of Immanuel Kant that recognises three distinctions:
(i) analytic a priori,
(ii) synthetic a posteriori, and
(iii) the synthetic a priori;
(4) the view of Saul Kripke that, as well as the analytic and synthetic, there are additional kinds of knowledge:
(i) contingent a priori;
(ii) necessary a posteriori.
In addition, the basic conceptual distinctions in the nature of propositions and truth can be given, as follows:
(1) analytic versus synthetic;

(2) a priori versus a posteriori; and

(3) necessary versus contingent.
The analytic versus synthetic distinction, formulated properly by Immanuel Kant, is usually taken to refer to propositions (or statements).

A statement that is true solely by virtue of the meaning or definition of the terms used is analytic, and it has logically necessary truth. This means that a denial of such a proposition would entail self-contradiction.

An older definition (used by Kant) is that a proposition whose predicate expresses ideas conceptually already contained (explicitly or implicitly) in the subject is analytic, such as the following propositions:
(1) All sad widows are sad.

(2) all red coloured things are red.

(3) all bachelors are unmarried.

(4) All (normal) bipeds are two-footed.

(5) All triangles have three sides.
In propositions (1) and (2) above, you can see that the idea expressed in the predicate is already explicitly verbally contained in the subject (or, strictly speaking, in the noun phrases that function as the subject). In propositions (3), (4), and (5), the predicate expresses an idea implicitly contained in the subject, which is to say that the subject already carries that idea in its implicit definition, as understood in a language or community of speakers.

Analytic propositions are true a priori (or prior to or without experience, experimentation, or empirical evidence) and are necessarily true (and their negation is impossible).

Such propositions are called analytic a priori.

Many such propositions can be regarded as tautologies and are non-informative, in the sense that they do not by themselves add new information to our knowledge of the world, because the knowledge in the predicate was already contained in the subject or deducible from it. Another way of saying this is that many are often trivial truths.

All propositions that are not true by virtue of the terms used are synthetic, and their negation is logically possible. Another way of saying this is that the predicate does not express an idea contained in the subject:
(1) Some widows are sad.

(2) Some swans are black.

(3) the earth’s sun contains hydrogen and helium.
Here the predicate is not obviously contained in the subject of the propositions, and the propositions are true a posteriori, or by means of experience, experimentation, or empirical evidence (although often this empirical verification has already been done by people before us). They have contingent, not necessary truth. Such propositions are called synthetic a posteriori.

But there is one more type of proposition proposed by philosophers and logicians, as we can see here (listed after the other two I have described above):
(1) analytic a priori;

(2) synthetic a posteriori, and

(3) synthetic a priori.
This is (3): the synthetic a priori proposition. Nevertheless, the existence of synthetic a priori propositions is debated.

Some proposed examples can be given:
(1) nothing is both green and red all over.

(2) all events are caused.
Are these, as Kant thought, synthetic a priori, in the sense that they are necessarily true but not analytic (that is, not vacuous)?

Consider the proposition that “all events are caused.” Supposedly, this is synthetic a priori and necessarily true. But modern quantum mechanics would appear to cast doubt on that idea (in its conventional sense) since it is believed by some scientists that certain quantum mechanical events can be uncaused. At the very least, there is genuine dispute and, ultimately, the proposition that “all events are caused” was arguably really just a synthetic a posteriori proposition, and never had necessary truth.

The logical positivists denied the existence of synthetic a priori knowledge, and argued only two real distinctions are useful and valid:
(i) analytic/a priori, and

(ii) synthetic/a posteriori.
By contrast, Willard Van Orman Quine (1908–2000) attacked this view in his famous article “Two Dogmas of Empiricism” (1951; reprinted in Quine 1981), in which he argued that the analytic versus synthetic distinction does not exist. Although this paper was directed against the logical positivists, its thesis is equally damaging to apriorist rationalists, and in fact Quine was himself an empiricist.

Moreover, in Quine’s view, synthetic a priori do not exist either.

However, many think Quine’s critique is ultimately unconvincing (Grice and Strawson 1956), and that a strong case for the existence of analytic propositions can be made. I agree with this.

For example, Putnam argued that there is an analytic versus synthetic distinction but that it is ultimately a trivial one (Putnam 1962: 361). He also contended that there are a vast number of propositions that cannot be easily classified into either category (Putnam 1962: 364), and a better model is one with three classes:
(1) analytic
(2) synthetic
(3) various other kinds (Putnam 1962: 364).
Putnam (1962: 370) points out that modern science has revised the status of propositions which previous scientists or philosophers thought were a priori true as analytic statements, but which subsequently turned out to be synthetic statements which were false.

A fundamentally influential work was Saul Kripke’s Naming and Necessity (1980) (on Kripke, see Berger 2011; Sosa 2006; Burgess 2013; Hughes 2004). This advocated an even more complicated view. Kripke held that the notion of “necessity” as distinct from “contingency” must be metaphysical/ontological, whereas the “a priori” versus “a posteriori” distinction is epistemic (Scruton 1994: 162). The synthetic versus analytic distinction is semantic.

These concepts do not coincide, and Kripke presented arguments for additional types of knowledge, such as:
(i) the contingent a priori;

(ii) the necessary a posteriori.
Kripke also argued that identities are necessarily true, that names are “rigid designators,” and that some things have real essences (Scruton 1994: 164). This leads into the idea that “necessary a posteriori” truths exist: such as, for example, that the “morning star is the evening star” (namely, the planet Venus). On Kripke’s theory, that identity is necessary, but was discovered empirically: hence it is a “necessary a posteriori” proposition.

Another category is the “contingent a priori,” such as the idea that we know a priori that a metre rod is a meter long, but nevertheless “metre” could have been historically and scientifically defined at a different length (so it is therefore contingent).

But whether Kripke’s argument for “contingent a priori” propositions really succeeds is disputed (Scruton 1994: 165). At first glance, Kripke’s “contingent a priori” looks similar to Kant’s “synthetic a priori,” but on closer inspection there is an important difference. Kant’s “synthetic a priori” knowledge cannot be contingent, but Kripke’s “contingent a priori” can.

Though it is a subject for another post, the only epistemology capable of supporting Mises’s human action axiom, with its alleged “synthetic a priori” status, is really the Kantian theory of knowledge. But Kantian epistemology looks rather outdated in terms of the modern theory of knowledge.

Berger, Alan. (ed.). 2011. Saul Kripke. Cambridge University Press, New York and Cambridge.

Burgess, John P. 2013. Kripke. Polity, Cambridge.

Fodor, Jerry. “Water’s Water Everywhere” (review of Kripke: Names, Necessity and Identity by Christopher Hughes)

Grice, H. P. and P. F. Strawson. 1956. “In Defence of a Dogma,” Philosophical Review 65: 141–158.

Hughes, Christopher. 2004. Kripke: Names, Necessity and Identity. Clarendon Press, Oxford.

Kripke, Saul A. 1980. Naming and Necessity (rev. edn.). Blackwell, Oxford.

Putnam, H. 1962. “The Analytic and the Synthetic,” Minnesota Studies in the Philosophy of Science 3: 358–397.

Quine, W. V. 1951. “Two Dogmas of Empiricism,” Philosophical Review 60: 20–43.

Quine, W. V. 1981. “Two Dogmas of Empiricism,” in From a Logical Point of View. Harvard University Press, Cambridge, MA. 20–46.

Schwartz, Stephen P. 2012. A Brief History of Analytic Philosophy: from Russell to Rawls. Wiley-Blackwell, Chichester.

Scruton, R. 1994. Modern Philosophy: An Introduction and Survey. Penguin Books, London.

Sosa, D. 2006. “Saul Kripke (1940– ),” in A. P. Martinich and D. Sosa (eds.), A Companion to Analytic Philosophy. Blackwell, Malden, Mass. and Oxford. 466–477.


  1. Missed a key figure there, LK. But you're not the first. Another repression from our Anglophone friends.

    1. Interesting post, and I am certainly keen to learn more about Johann Georg Hamann, given his close relationship to Kant and his role in translating Hume for German speakers.

      Some quick reading in Gwen Griffith Dickson's Johann Georg Hamann's Relational Metacriticism (Berlin, 1995) shows that Hamann was skeptical about synthetic a priori knowledge (p. 290).

      Dickson then quotes (n. 86, on p. 290) Irmgard-Maria Piske's Offenbarung, Sprache, Vernunft: zur Auseinandersetzung Hamanns mit Kant (Frankfurt am Main and New York, 1989, p. 162), to the effect that Hamann rejected synthetic a priori knowledge, because (in his view) it can't be grounded in "pure thinking," but must, like all empirical knowledge, be grounded in reality. Moreover, Hamann thought that apodictic certainty -- or epistemic necessity -- rests "only on a priori signs" (such as in mathematics).

      Now this is all a perfectly reasonable empiricist position.

      We get confirmation of this is on p. 291 when Dickson says that Hamann thought even geometry must be grounded in "the spatial, the empirical world of experience", and on p. 294 it is said that Hamann thought our knowledge of space and time arise "from our experience" and "senses".

      My reading is hardly anything more than a smattering of course, but it appears that the epistemological views of Hamann -- at least in his critique of Kant -- can be fitted into the empiricist category:

      Knowledge is either (1) a priori (and necessary only because it is analytic or mathematical), or (2) empirical (a posteriori).

      At any rate, that is the sense I get.

    2. This is a classic misreading of Hamann. It goes back to what I said about Berkeley in the original post. Hume just took over his position on particulars and universals and divorced it completely from his appeal to tradition, custom and belief. This is why people often think that Berkeley is an empiricist. Because he made an argument that, in retrospect (i.e. after Hume) looks empiricist. But his argument is very far from the empiricist argument which, remember, for Hume ends in skeptical nihilism (Berkeley was a Bishop!).

      The reading of Hamann above is similar. He is not an empiricist because, as he recognises, empiricism falls apart into skepticism ("especially at the final close of a critical century when empiricism on both sides, struck blind, makes its own nakedness daily more suspect and ridiculous"). He has an a priori and that a priori is language. As I said, language is dominated by custom. And so if Reason is dominated by language then Reason is dominated by custom.

      This is what some social constructivists, like Levi-Strauss and Lacan, argue too. Our language is generated through custom and is largely arbitrary and our Reason is then subject to this. Note that this is not Wittgenstein. For him we just need to purge language and make it conform to Reason. But these thinkers say that trying to do this will only end in a mess. (Something Wittgenstein, to his credit, realised in his later writings).

      So, no. This is not empiricism. This is what some might consider a conservative argument that customs and traditions are primary and Reason is secondary.