[hist-analytic] A Priori/A Posteriori -- Revisited
Roger Bishop Jones
rbj at rbjones.com
Wed Nov 18 16:46:14 EST 2009
On Tuesday 17 November 2009 01:00:05 Jlsperanza at aol.com wrote:
> --- I am also reminded of Dummett, and Jones may like to elaborate on that!
Don't understand this one, did I mention him somewhere?
> For the intuitionists, a proof (or justification) is something that _TAKES
> time_: it's a step by step process. So, I guess they would go on to say
> that all proof (or justification) is _a posteriori_. Yes, we have had
> rounds of discussion discussing the a posteriori of mathematical truths,
> but the Dummettians take it _pretty_ seriously.
This is rather a nitpick, but you do seem simply to take orderings as if they
were necessarily temporal and this ain't so, and prior and posterior are
applicable in the context of any ordering temporal or not. Strictly speaking
the propositions in a proof are partially ordered by the transitive closure of
the relationship "A is a premise from which B is directly inferred (in the
proof)". This is not in itself a temporal order, though the ordering might be
presented temporally. It might be spatially presented on a page, or it might
be just something in the memory of a computer, neither spatial nor temporal.
I'm not familiar with the Dummett connection, I am very ignorant about
> -- I wonder if by merely analysing the 'proposition' one should determine,
> by the mere lights of one's intellect -- cfr. Enlightenment -- where the
> proposition 'p' requires an a priori or a posteriori justification. I would
> think so, but examples do not come out easily.
Before Kripke (at least, in Carnap!) this story was fairly easy, because the
analytic/synthetic distinction coincided with the other two, so to determine
the epistemic status you have only to decide whether the sentence is analytic
Nowadays you have to reject Kripke's analysis of the relevant concepts before
you can work in this way.
> "Computers can't think"
> strikes me as 'analytic' (true or false) rather than in need of a
> posteriori justification. On the other hand, Noel Coward was possibly
> being jocular when he wrote in his re-write of Cole Porter's "Let's do it"
> "Probably we'll live to see machines do it" (Let's do it, let's fall
> in love).
Those are both hard cases. But I'm pleased to see that you assume a
connection between analyticity and a priority.
> -- If I understand Jones aright, he is saying that Kripke makes a
> distinction (between a priori/a posteriori) that conflates with the
> 'necessary/contingent', if not the 'analytic/synthetic'. Wasn't Kripke's
> idea that while the apriori/aposteriori distinction is epistemic (or
> doxastic), the necessary/contingent is 'metaphysical', or ontic, and the
> analytic/synthetic logical? God knows!
The "essential" difference, between Kripke and myself and Carnap is that Kripke
held analytic, necessary and a priori all to be distinct, whereas we hold them
to be at least coextensional. I don't know how he described the
analytic/synthetic, I would call it semantic, and I would observe that in this
matter semantics and metaphysics are intricately intertwined.
Kripke cannot separate the two and be considered to be using the same concepts
as Carnap. However, our recent conversations have been on the epistemic
connection, and in this Kripke's distinctive conceptions are completely
> Back to propositions, I was amused (in a good way) by Jones's pragmatism.
> Surely he doesn't need to _specify_ what a 'proposition' is, and he is
> ready to have the notion as 'language specific' and 'contextual' in
The context in question here. is any context necessary to disambiguate a
sentence, so that "the proposition" which it expresses is definite, bearing in
mind that the one constraint I offered for the notion of proposition was that
the proposition should embody the determinate truth conditions somehow (though
it might be more than that).
So the context is used to identify the proposition, and the proposition is
itself insensitive to (this kind of) context.
I think elsewhere the word context might have been used with reference to the
"possible world" relative to which a propositions truth value is determined.
It remains of the essence of the proposition that its truth value (unless it
be necessary or contradictory) will vary in different possible worlds (or
situations if you like, or state descriptions).
It might be worth pointing out the differences in attitude towards propositions
which I take (depending on "context"!)
The talk of propositions here is in the context of a discussion of Carnap
against Kripke, and the doctrines in question (Carnap's at least, not
Kripke's) relate primarily to formal languages rather than to natural
languages. These are languages which have been designed
by some person (or committee!) and which have been given a
formal semantics. Now with the methods I advocate the notion of proposition
would feature in such a semantics and its details would be chosen to reflect
the desired semantics. However, one very general approach would be to have a
proposition as a truth valuation (map from possible worlds to truth values)
and the only bit of that likely to change from one language to the next would
be the domain, i.e. the notion of possible world. So the language specific bit
of proposition is the metaphysics you want to embed into the language.
On this conception metaphysics is a part of a formal semantics and this
explains how it is that a semantic concept like analyticity turns out to be
the same as a "metaphysical" concept like necessity.
Carnap has a different explanation, which comes from taking necessity as "true
in all possible worlds" and analyticity (aka logical truth) as (following the
tractatus) "tautological" i.e. true for all state descriptions. (of course
you have to abandon Wittgenstein's insistence on the logical independence of
atomic propositions for this to work).
If one wants to talk about propositions expressed by natural languages the
situation is much more murky, for two reasons.
Firstly if you take a proposition to be the meaning of a sentence (in
context), then its not so easy to figure out what meanings are.
Secondly, if you want proposition to mean what it means in natural language
then that is something else again (in that case there are diverse uses to
grapple with). Those are problems which don't interest me.
On other hand, though in the account above in relation to formal languages,
propositions are just conveniences in presenting the semantics, there are
interesting issues which arise if you start to worry about the identity
conditions, and these I think set you off in the direction of something I might
call metaphysics. We can devise any number of different
formal languages in which the truths of elementary arithmetic
are expressible. According to the above notion of proposition,
what we might informally think of as being the same
arithmetic proposition is likely to be a different abstract entity in each of
these languages. At the same time, though for many purposes taking a
proposition simply to be the truth conditions will suffice, this has the
opposite disadvantage, all necessary propositions turn out to be identical.
So if you want a notion of proposition which is not just good enough to
express the semantics of some chosen formal language, but is good enough to
reflect the possibility that the same proposition can be expressed in many
different languages, and to ensure that propositions are not identified just
because they have the same empirical content (truth conditions), then it gets
complicated. That's a bit of "metaphysics", I don't know how
important it is, but most of my talk about propositions belongs to semantics
> I was reminded of a similar 'pragmatist' (but he'd call it
> transcendental) approach to 'propositions' by Grice in "Prejudices and
> Predilections" (aka "Reply to Richards").
> Drawing on conversations with Geo. Myro (the Russian emigre from Ukrania
> that Grice befriended since he settled in Berkeley in 1967 -- Myro had
> studied in Oxford, but I'm unaware if they had met back then), Grice
> mentions that
> what a proposition is
> may well depend on the theoretical role it may play in different
> A proposition, I hold, is what a theory of "propositional
> attitudes" needs.
Yes, though that's not the only reason we need them.
I would say however, that much of this is more to do with convenience than
Its not so much that one could absolutely not do without them, but that it
would be much more complicated to deal with such topics while eschewing a
This kind of pragmatism, which is in the spirit of Carnap whose "internal
questions" are the results of pragmatic choices of ontology in the design of
languages, I would urge upon Steve B., who constantly worries about supposed
metaphysical problems arising from a usage which need not be construed as
involving a metaphysical commitment (especially when its Carnap and we know
that everything which sounds like the kind of metaphysics which he proscribed
really isn't going to be that kind of metaphysics, and who invented the
internal/external distinction just so that he could talk as if he had a
metaphysics but not actually take that extra (content-less) step).
> We need propositions to have the propositional attitudes
> (so-called, Grice prefered, 'psychological attitude') hooked onto
"propositional attitude" is better.
Its what its about that makes it interesting here, not the kind of thing that
is about it, and we are really talking about propositional attributes which
are not psychological at all. so maybe "attitude" too strong.
Necessity and a priority, then, are propositional attributes?
Incidentally one of the main points historically about these things is that
they are places where some kinds of "transparency" fail, e.g. referential,
and so you can't substitute synonyms willy-nilly in propositions of which
someones attitude is asserted. These failures of substitution give us clues
about propositional inequalities (the proscribed kinds of substitution give
different propositions, that's why they can't be permitted). So they give us
information about some kinds of information which there must be in
propositions over and above truth conditions (though possibly only about what
must be in the propositions expressed in this particular natural language, not
necessary yielding metaphysical truths about propositions.)
> ---- believes that ----
> (For a psychological attitude like 'belief' holds between the 'arguments'
> of the believer and what is believed -- and beliefs and psychological
> attitudes are, contra Quine, compositional and relational, no?)
I don't know what that means.
> A metaphysician may need propositions for other reasons, i.e. to fulfil
> other theoretical metiers. Myro's point was that there is NOT just one
> answer as to what a proposition is, but many (or none).
But perhaps some are better than others.
> --- Grice played for years -- as evidenced in Reply to Richards -- with the
> idea of a COMPLEX (or propositional complex) as being more basic than
> proposition! A propositional complex (I think I've seen the same idiom in
> writings by Peacocke) is just the schematism of the _content_ of a belief,
> say, into its minimal components (the belief that SOME cat, Tibbles, is on
> some old rug in SOME kitchen, say -- rather than talk, in abstracto and
> out of context, of the proposition that the cat sat on the mat).
But propositions are closed under conjunction, so isn't a propositional
complex only doing what might as well be done with a complex proposition?
More information about the hist-analytic