One of these sentences is not like the others.
(1) Donna knows which road goes to San Jose.
(2) Donna knows the way to San Jose.
(3) Donna knows the mayor of San Jose.
The interesting question is, which one is it? If you know the answer, you can probably skip this post, but I think it’s a pretty interesting puzzle.
It’s a philosophical commonplace that the meaning of ‘knows’ in (4) and (5) is different.
(4) Vicky knows who won the 1992 election.
(5) Vicky knows Bill Clinton.
Somewhat tendentiously, we’ll call the relation denoted by ‘knows’ in (4) propositional knowledge, and the relation denoted by ‘knows’ in (5) objectual knowledge. We won’t discuss sentences like (6) here, though my view (following Stanley and Williamson) is that they too involve propositional knowledge.
(6) Vicky knows how to play the violin.
So (1) involves propositional knowledge, and (3), at least on its most natural reading, involves objectual knowledge. What does (2) involve? Jonathan Schaffer and I spent a fair chunk of the drive from Moscow to Seattle on Sunday discussing this, and this post is basically just a report of the conversation. For the most part the considerations in favour of propositional are mine and the considerations in favour of objectual are Jonathan’s, but to a large extent it was a joint effort. Since Jonathan doesn’t have a blog and I do, I get to write it up for now though!
This is potentially a really embarrassing admission, but I know of no work on constructions like (2). If there’s some big paper on knowledge-the (as I’ll call them for want of a better phrase) this might all be a waste of time. But it’s a fun exercise, and if there isn’t much on this it might even be original.
There are three considerations that tell in favour of a propositional interpretation of (2). None of them are decisive, but I think they make a reasonable case. It will be nice to have a few more cases to work with, just to check there’s nothing distinctive going on with the cases about ways. So we’ll often refer to (7) and (8) as well.
(7) Andy knows the time the bar closes.
(8) Sarah knows the answer to the first question on the history exam.
I find (7) a little marginal, but I think it’s ultimately OK. We’ll come back to the history exam a few times, so let’s note at the start its two questions.
Q1 (4 points). Who won the 1992 US Presidential election?
Q2 (6 points). Who won the 1996 US Presidential election?
On to the arguments.
_Substitution Failures_
Not all constructions with ‘knows the’ fall into the same category as (2), (7) and (8), which seem to form a natural kind. In particular, (3) is not like that. (3) is a clear case of objectual knowledge. And in these clear cases of objectual knowledge a certain kind of inference looks clearly valid. The inference from (3) and (9) to (10) is valid, as is the inference from (3) and (11) to (12).
(3) Donna knows the mayor of San Jose.
(9) Liz is the mayor of San Jose.
(10) Donna knows Liz.
(3) Donna knows the mayor of San Jose.
(11) The mayor of San Jose is the world’s most prominent epistemologist.
(12) Donna knows the world’s most prominent epistemologist.
It will be relevant later that (3) and (11) don’t entail (13).
(13) Donna knows the world’s most prominent epistemologist as such.
Some of the time I can hear (12) as having a reading on which it means (13). But its most natural reading is not like that, and on that natural reading it is entailed by (3) and (11). However the matching inferences involving the knowledge-the claims are all invalid.
(2) Donna knows the way to San Jose.
(14) The way to San Jose is the way to the epistemology conference.
(15) Donna knows the way to the epistemology conference.
(15) is not entailed by (2) and (14). If Donna doesn’t know the epistemology conference is in San Jose, knowing the way to San Jose won’t suffice for her to know the way to the epistemology conference. Similarly in each of the following triples the first two sentences don’t entail the third.
(7) Andy knows the time the bar closes.
(16) The time the bar closes is the time the last train leaves.
(17) Andy knows the time the last train leaves.
(8) Sarah knows the answer to the first question on the history exam.
(18) The answer to the first question on the history exam is the answer to the second question on the history exam.
(19) Sarah knows the answer to the second question on the history exam.
These failures of entailment strike me as a strong consideration against the claim that (2) involves objectual knowledge. If (2) involves propositional knowledge, if it says roughly that Donna knows what the way to San Jose is, then the failures of entailment are just as we’d predict. This is I think the best argument for the propositional view of (2). But it isn’t the only view.
_As Such_
We noted above that (13) makes sense, and indeed is not entailed by (3) and (11).
(13) Donna knows the world’s most prominent epistemologist as such.
Actually (13) is ambiguous, though the ambiguity is not likely to arise in practice. It’s easier to see this with one of (13)’s contraries, namely (20).
(20) Donna knows the world’s most prominent epistemologist, though not as such.
On the most natural reading, the phrase ‘as such’ is anaphoric on the description ‘the world’s most prominent epistemologist’. But it also has a reading on which it is anaphoric just on the description ‘the world’. In that case it expresses something we might more explicitly say with (21).
(21) Donna knows the world’s most prominent epistemologist, though not as _the world’s_ most prominent epistemologist. She knows her as _Planet Earth’s_ most prominent epistemologist.
Obviously this isn’t the natural reading. (20) has a very natural reading on which ‘as such’ gets its reference (as it were) from the whole description. What’s interesting is that none of (2), (7) and (8) can be modified in this way.
(22) *Donna knows the way to San Jose, but not as such.
(23) *Andy knows the time the bar closes, but not as such.
(24) *Sarah knows the answer to the first question on the history exam, but not as such.
The only acceptable readings for these three sentences are where ‘as such’ binds onto descriptions (or names) smaller than the whole definite description after ‘the’. So the only reading of (23), for instance, is where Andy knows the time the bar closes under some description of the bar other than ‘the bar’.
Jonathan suggested that maybe we could get applications of ‘as such’ phrases in these cases when the alleged knower might not have the relevant context. So if Fido knows it is meal-time, so knows the time he is fed is now, but doesn’t have the concept of time, perhaps (25) is OK.
(25) Fido knows the time he is fed, but not as such.
I think that’s pretty marginal at best, and it still doesn’t explain why we can’t modify our original cases by an ‘as such’ phrase that binds onto the whole definite description.
Again, note that this is just what we’d expect if (2), (7) and (8) make propositional knowledge claims. If (2) requires Donna to know of a way that it is a way to San Jose, which is what we’d expect on the propositional reading, then it is contradictory to say she knows this but doesn’t know it as such. On the other hand, if it is just an objectual knowledge claim, it’s hard to see why Donna couldn’t be acquainted with a way, say as the way to the epistemology conference, but not know it is the way to San Jose.
(Disclaimer: I’m being very loose with talk about ‘as such’ phrases binding or being anaphoric here. I don’t really know how these phrases work, so I’m guessing a bit about their syntax and semantics. But I hope the distinction I’m drawing is clear, and that it will be statable even by someone who knows the right theory of ‘as such’ clauses.)
_Conjunctions_
All of the following sentences sound OK to me.
(25) Donna knows the way to San Jose, but not how long it will take to get there.
(26) Andy knows the time the bar closes, but not whether it’s possible to get the train home after it does.
(27) Sarah knows the answer to the first question on the history exam, but not how many points it is worth.
In each case we’ve got a pair of conjoined knowledge claims, with one occurrence of ‘knows’ and the second unambiguously being a propositional knowledge claim. So that’s a prima facie argument that the first conjunct is also a propositional knowledge claim. This argument is a little weaker than the previous two, because it seems to overgeneralise, as Jonathan pointed out.
(28) Andy knows the bar manager, and how to get free drinks off her.
(29) Andy knows the bar manager, and what she likes to drink.
(30) ?Andy knows the bar manager, and that she likes Red Bull and vodkas.
Possibly (30) is fine, though I hear it as a little awkward. But the problem is that (28) and (29) seem OK, which suggests that it’s possible to have reduced conjunctions involving both objectual and propositional knowledge claims. Now one possibility is to run with the argument and say this shows there is one overarching category of knowledge, of which propositional and objectual knowledge are sub-types. But I’m not prepared to draw that radical a conclusion, in particular because it would leave us a lot of work to do to explain why these conjunctions so often sound bad. So I’ll just say that this argument for the propositional interpretation of knowledge-the claims is weaker than the first two.
_The Counter-Argument_
So why not just conclude that knowledge-the claims are propositional knowledge claims? Well, because syntactically they look a whole lot like objectual knowledge claims, and it’s hard to tell a plausible story about how the syntax could be so misleading. The problem is we want (8), say, to express the proposition (31).
(8) Sarah knows the answer to the first question on the history exam.
(31) Sarah knows *what* the answer to the first question on the history exam *is*.
But what justification could we have for thinking that somehow those words either side of the DD get deleted in moving between the LF and surface syntax? (If we even believe in things like LFs.) Of course we can go ahead and posit a rule that permits their deletion. Jonathan called this rule wh-deletion, though in the spirit of generative semantics I’d rather call it Josh. The problem is that we don’t have much evidence to believe Josh exists, and some reason to believe that it doesn’t. We can’t, for instance, use (32) to express (33), as you’d think Josh would allow.
(32) *Sarah wonders the answer to the first question on the history exam.
(33) Sarah wonders *what* the answer to the first question on the history exam *is*.
As Jonathan points out, if (8) is really an objectual knowledge claim, we get a simple explanation of the badness of (32), since ‘wonder’ doesn’t have an objectual form.
(34) *Andy wonders the bar manager.
On the other hand, perhaps Josh would let us explain the uses of the a sentences below to express the b propositions.
(35) a. Bo learned French.
b. Bo learned *how to speak* French.
(36) a. Bo learned the guitar.
b. Bo learned *how to play* the guitar.
But these aren’t even obviously instances of Josh, and in any case if we were going to lean on such cases we’d have to have a story about why we can’t do the same thing in (37) and (38)
(37) a. ?Bo learned lasagne.
b. Bo learned *how to cook* lasangne.
(38) a. Bo learned the skateboard
b. Bo learned *how to sell* the skateboard.
This doesn’t look to be helping. There really is little reason to believe Josh is a rule.
We could go a quite different direction and claim that (8) expresses (31) not by some kind of deletion, but by (free) pragmatic enrichment. So perhaps the literal, or minimal, content of (8) is an objectual knowledge claim, but we can use it to say something like (31). There are, I think, four problems with this approach.
First, there’s no such thing as (free) pragmatic enrichment, as “King and Stanley argue”:http://www-philosophy.ucdavis.edu/phildept/document/sem-prag.htm. (Disclaimer: I know there are huge issues here, and I’m being more than a little flippant in just taking sides without giving reasons. But I don’t need King and Stanley’s arguments to argue against *this* application of pragmatic enrichment, as we’ll soon see.)
Second, we’d still need an explanation of why we can’t use (32) to express (33). This is related to King and Stanley’s point that enrichment theories often overgenerate explanations.
Third, if ‘knows’ really is ambiguous, this would be a very odd style of enrichment, and quite contrary I’d think to the spirit of similar pragmatic proposals. Even if we believe in free enrichment, it is very plausible that this is something that happens *after* ambiguity resolution, and the effect is never to overturn that resolution.
Fourth, it’s not clear how this proposal would explain the sytactic data in the _as such_ and _conjunctions_ point, even if it did make some progress towards explaining the data about permissible inferences in the first section.
So I don’t see either a plausible deletion story or a plausible enrichment story that explains how (8) can be used to express (31). Nevertheless, I think there are very good arguments that (8) does express (31). So there’s some work to be done here either in showing that the arguments for a propositional interpretation of (8) are wrong, or in explaining how their conclusions _could_ be true.