The most fun seminar Ive been attending
this semester has been Jeff Kings seminar at Harvard on the
semantics/pragmatics distinction. (Hang on, isnt that the only seminar youve
been attending? – ed. Not at all, Ive also been attending my own seminar,
and normally Id think that would be the most fun seminar, because I get to
talk .) The main theme of the seminar has been a sustained
attack on theories that provide too small a role for semantics in a theory of
communication. (Some of the attack is presented in this
paper co-written with Jason Stanley.) These theories usually say, in one
way or another, that the explanation for the success of certain kinds of
communication is pragmatic not semantic. (They often go on to say other things
too, but thats the part that Im most interested in.) So, to provide a
representative sample, consider two stories about how (1) gets the intuitive
truth conditions that it has.
(1) If
Charlie drank ten beers and drove home, she broke the law.
Intuitively, (1) is true, because (1) is
true iff it is the case that if Charlie drank ten beers and drove home shortly
afterwards, she broke the law, and thats clearly true. How could (1) have those
truth conditions? Some theorists (including some time-slices of me) say that
the semantic content of (1) is just that if the conjunction (Charlie drank ten
beers Ù
Charlie drove home) is true then it is true that Charlie broke the law. The
intuition is explained by the truth of some more or less complicated pragmatic
theory, that somehow predicts that if Charlie drank ten beers and drove home
is normally only said if the events happened in that order, then (1) is
normally only said if Charlies drinking and driving in that order implies
that she broke the law. And of course theres a story in Grice about why Charlie
drank ten beers and drove home is normally only uttered if the events occurred
in that order, even if the ordering is not part of the truth conditions.
Jeff doesnt want to accept any of that. He
argues that the most plausible story about the semantics of (1) has the
intuitive truth conditions fall out as being the truth conditions. The first
point to note is that every sentence in English (and every other natural
language) is tensed, and the tenses are presumably part of the semantic
content. So Charlie drank ten beers has as its semantic content $t (t is in the
past) (Charlie drinks ten beers at t). Importantly, the quantifier here is
restricted. Whether Charlie drank ten beers at Bill Clintons second inaugural doesnt
really matter to the truth of an ordinary utterance of Charlie drank ten beers
unless for some reason we are talking about Clintons second inaugural.
Arguably (and
better philosophers than I have persuasively argued for this at length) every
sentence that isnt in the present tense literally expresses a proposition that
contains a quantifier over time. And this quantifier isnt present because of
some mysterious pragmatic process, its encoded in the verbs of the sentence,
just like most semantic content is encoded somewhere in surface structure. And
what goes for whole sentences goes for constituent sentences too, so to a first
approximation, the semantic content of (1) is (2).
(2) If $t1 (Past t1)(Charlie
drinks ten beers at t1) and $t2 (Past
t2)(Charlie drives home at t2) then
$t3 (Past
t3)(Charlie broke the law at t3).
This isnt much
help yet, but if we also hold (a) all three quantifiers here are restricted,
and (b) the restrictions are somehow co-ordinated, then we can have the
semantic content of (1) really be something like (3).
(3) If $t1 (Past t1)(Salient
t1)(Charlie drinks ten beers at t1) and $t2 (Past
t2)(t2 is shortly
after t1)(Charlie
drives home at t2) then
$t3 (Past
t3)(t3 = t2)(Charlie
broke the law at t3).
This is obviously
very rough, because as it stands weve got variables appearing outside the
scope of the quantifers that bind them, but at least this is a workable
suggestion for how (1)s truth conditions might match its intuitive truth
conditions. And to the extent that the argument for radical pragmatic theories was
premised on the assumption that there isnt even a workable suggestion for how (1)s
truth conditions might match its intuitive truth conditions, well those
arguments are looking fairly weak. (That would include some arguments Id
previously adopted. Oh well – you cant be right all the time.)
But not all the
examples of alleged separation between truth conditions and intuitive truth
conditions are handled with quite such ease.
(4) If Hannah insulted Joe and Joe resigned,
then Hannah is in trouble.
As Jeff and Jason
note, (4) seems to express the proposition that if Hannah insulted Joe and Joe
resigned as a result of Hannah’s insult, then Hannah is in trouble. The
suggestions above about using restricted quantifiers over times wont help
here, because they wont get the causal link between Hannahs (possible) insult
and Joes (possible) resignation into the proposition. So what can our heroes
do? They start by taking a rather sensible approach: when in trouble, ask What
Would Bob Stalnaker Do?
As Robert Stalnaker has argued, indicative conditionals normally
exploit a similarity relation that counts only worlds compatible with the
mutually accepted background assumptions as the most similar worlds for
purposes of semantic evaluation.
An indicative conditional is true if and
only if the consequent is true in every one of the most relevantly similar
worlds in which the antecedent is true. (King and Stanley, 48)
Well, Im not
sure thats exactly what Stalnaker said, for reasons that shall become apparent
presently. Anyway, applying this theory to (4) we get the following
conclusions.
Fortunately, however, there is no reason to give a non-semantic account
of the intuitive readings of (4). The relevant reading of (4) is simply
predicted by the semantics for indicative conditionals that we have endorsed.
In a context in which the speaker has in mind a causal relationship between
Hannah’s insulting of Joe and Joe’s resignation, all relevantly similar worlds
in the speaker’s context set in which Hannah insulted Joe and Joe resigned,
will ipso facto be ones in which Joe’s resignation is due to Hannah’s insult.
The speaker’s context set is what is epistemically open to her. This may include
worlds in which the conjunction holds, and there is no causal relationship between
the conjuncts. But given that she has a causal relationship saliently in mind,
such worlds will not be the most relevantly similar worlds in the context set.
So, if she has a causal relation in mind between the two events, that is just
to say that the similarity relation for indicative conditionals will select
those worlds in which there is a causal relationship between the conjuncts of
the antecedent as the most similar worlds to the world of utterance in which
the antecedent is true. So, the causal reading of (4) is predicted by the
simple semantics for the indicative conditional that we have adopted above.
(King and Stanley, 53, numbering adjusted.)
Imagine
that all the following circumstances obtain:
(5) Jeff and Jason are right about the
semantics of indicative conditionals;
(6) Hannah recently insulted Joe;
(7) Shortly after that, Joe resigned
(8) Joes resignation was not due to
Hannahs insult
(in fact it was because he just realised
he always wanted to be a lumberjack)
(9) Hannah is not in trouble.
(10) Someone uttered (4) knowing (6)
and (7), but not (8).
In
those circumstances, I think the utterance of (4) may well be true. All the
epistemically open scenarios in which (6) and (7) are true are ones in which
Hannah is in trouble. And according to Jeff and Jason, if the antecedent of (4)
is true iff (6) and (7) are true. So all (epistemically) nearby worlds in which
the antecedent is true are worlds in which the conclusion is true, so the
utterance of (4) is true.
But,
per hypothesis, the actual world is also a world in which (6) and (7) are true,
and hence the antecedent of (4) is true. And the actual world is a world where
the consequent of (4) is false. So the actual world is a world where the
premises of the following argument are true and the conclusion false.
If
Hannah insulted Joe and Joe resigned, then Hannah is in trouble.
Hannah insulted Joe and Joe resigned.
So, Hannah is in trouble.
So modus
ponens is not a valid argument form. Something may have
gone awry. Theres two problems here, both of them potentially serious. First, on
the formal semantics Stalnaker adopts for the indicative conditional, modus
ponens is valid, yet Jeff and Jason claim to just be implementing Stalnaker,
and theyve ended up rejecting modus ponens. Either Stalnakers got his own
theory wrong, or Jeff and Jason have got him wrong.
Secondly,
THEYRE REJECTING MODUS PONENS. Isnt this something that should be a serious
issue? I mean, its at least somewhat surprising. Not as surprising as, say, the
fact that Rocky
VI is going to get made. But surprising. Reading through Jeff and Jasons
papers, and certainly listening to Jeff, one gets the impression that the
forces theyve lined up against present views that are seriously flawed in some
way or other. I do hope that rejecting modus ponens is not the only
alternative to these positions.