Skip to main content.
December 17th, 2003


Is this sentence ambiguous?

(1) Vegemite could have tasted icky.

(I assume ‘icky’ is unambiguous.)

I half think it has readings with each of the following truth-conditions.

(2) There is a world w such that Vegemite in w is disposed to cause icky-tasting reactions in normal observers in the actual world.
(3) There is a world w such that Vegemite in w is disposed to cause icky-tasting reactions in normal observers in w.

I think (2) is the most natural reading, but I half think (3) is a possible reading. Do you agree?

(Two caveats. First, the dispositional analysis of tastes here is really crude, but I don’t care as long as something like it is going to work. Whatever the right story is, we’ll still be able to ask this kind of question about whether there are ambiguities. Second, I’m leaning towards treating ‘normal’ as a MacFarlanesque relative intension predicate, so what’s normal is fixed by the context of evaluation, not the context of utterance. I think that doesn’t matter to the question of whether (1) is ambiguous, but I’m a little less certain of that than I am of the first caveat.)

Posted by Brian Weatherson in Uncategorized


This entry was posted on Wednesday, December 17th, 2003 at 7:56 pm and is filed under Uncategorized. You can follow any responses to this entry through the comments RSS 2.0 feed. Both comments and pings are currently closed.

10 Responses to “Ambiguity?”

  1. Geoff says:

    Yes, it is ambiguous. I don’t know which reading is more natural, though I’m inclined to agree with you. Certainly, if the subject in question is something that actually exists and doen’t taste icky, that’s the right disambiguation:

    (4) Ice cream could have tasted icky.

    That’s made true by a world where the only ice cream flavors are anchovy, sauerkraut, and the like, a reading that is in line with (2). It would be a stretch to say that (4) is made true by a world where chocolate ripple is disposed to cause the same reactions it actually does in us, but icky reactions in the residents of w.

    You could force the latter, (3)-ish reading by saying:

    (5) If people’s palates were different, ice cream could have tasted icky.

    But the much more natural thing to say instead of (5) is

    (6) Ice cream could have seemed icky.

    Maybe that’s why (2) seems more natural as a reading of (1). But still, since I don’t know what Vegemite tastes like, the ambiguity is much stronger than it is in (4).

  2. Alan E Brain says:

    I’m a Software Engineer by profession, so Logic and removal of ambiguity is my stock-in-trade.

    A useful tool (as adduced above) is to embed the sentence in a “Test Bed” to provide context.

    “Had the original recipe used Fermented Horse Manure instead of Yeast…” + S implies a world w similar enough to reality® that the characteristics of a normal-Vegemite-taster have not been changed.

    “Had it been invented by Dinosaurs…” + S implies an ambiguous world w which may or may not have changed the characteristics of normal-Vegemite-tasters.

    “If Humans had never evolved, and the dominant species been Intelligent Duck-Billed Platypusses…” + S implies a world where the characteristics of normal-Vegemite-tasters are different indeed.

    On its own, the sentence has no contextual reference, so a default value of “all other things being equal” may be a valid assumption – this depends on meta-context, are we talking in a concrete rather than hypothetical context.

    Whether the sentence is ambiguous or not can only be decided when more information is known. Given that the Probability of a concrete rather than hypothetical much-changed universe being the context is high, the P(~a) Probability-of-non-ambiguity is also high.

    All statements can be ambiguous given different contexts and meta-contexts ( Elegant Proof of this too large to contain in the margin ), so measurement of P(~a) is the only useful metric. At some high value of P(~a), it is most useful to speak as if P(~a) = 1.0.

    But what that value is depends on meta-meta-Context, dammit! In normal, everyday use, P(~a) is near enough to 1.0. When talking in a context like this, it’s only “probably unambiguous”.

  3. Thomas Dent says:

    How can Vegemite in an alternate world w cause icky-tasting reactions in observers (normal or otherwise) in the actual world?

    How can you or I taste anything in an alternate world?

    I would go for (3) all the way because (2) makes no sense. Hence, no ambiguity.

    Consider rephrasing the sentence as

    “It could have been that Vegemite tasted icky”

    I.e. we imagine an alternative world w where the following sentence is true: “Vegemite tastes icky.”

    Now, in this alternative world w, who else can be doing the tasting but observers in w?

    In what context could one utter the sentence (1)? Presumably, in offering an alternative history of sandwich spreads (for example). In place of the actual world where Vegemite is successful, we have an alternative world where Vegemite is unsuccessful because it tastes icky to the people in the alternative world.

    The reason why it tasted icky might have been because it was made up differently, or because people’s taste buds worked differently, or both. (Here there is an alternative – but not an ambiguity.) In the former case, (2) and (3) become more or less equivalent because the way anything tastes in w is the way it tastes in this world.

    We could even imagine a world where Vegemite is made the same, people’s tastes are the same, but sometimes an infiltrator from Bovril sneaked in a chemical that smelt strongly of rotten eggs. But this is no more an ambiguity than saying “I could have died last week”: you don’t specify how you died.

  4. Brian Weatherson says:

    I agree that if we think of another possible world as such it’s hard to think of us having taste reactions to things in that world. But I think we can make sense of counterfactuals like “Had Vegemite tasted like beer actually tastes,…” where we compare the counterfactual taste of Vegemite to the actual taste of beer.

    The controversial thing I’m assuming here is that we can individuate tastes independently of evaluating them. This is clearly true for other senses, but it isn’t clear I guess for taste.

  5. Daniel Elstein says:

    I was just wondering why there isn’t another possible meaning:

    (4) There is a world w such that Vegemite in the actual world is disposed to cause icky-tasting reactions in normal observers in w.

    Take another example:

    (A) Vegemite could have been my favourite food.

    That seems in one way a simpler case, in that we don’t have to worry about a dispositional account of ‘icky’, but if we do give a dispositional account of ‘icky’ it should give (1) more or less the same ambiguities as (A).

    For me, the most natural reading of (A), assuming that the possibility that I haven’t actually tried Vegemite is eliminated by the context, is that Vegemite could have been the same but I could have been sufficiently different that I liked it. The response: ‘Yes, if it had been like Nutella’ would usually be a joke. The reason is that the reading required for that answer would make it either trivially true or hard to assess for truth. It would be trivially true if the claim is just that there is some world where Vegemite (or its counterpart) is differently constituted; presumably we want it to be a ‘close’ world, where some kind of contextual factors determine how close it needs to be. So it will then only be true if I know that, say, that a small change to the production process would make Vegemite taste great (to me). But I won’t have evidence for this on most occasions when I contemplate making claim (A).

    In contrast, a conditional like ‘If I had been given Vegemite as a child, it would have been my favourite food,’ might well be assertible, depending on evidence that is quite likely to be at hand (do most people given Vegemite as children…?).

    So I think that in my case (A) there is an ambiguity, but it is generally resolved in the (4) direction by Gricean rules. Does this transfer into the interpretation of (1)? Naively we would think that so long as we were prepared to countenance a dispositional theory of taste it would do. But that won’t be clear if we’re both dispositionalists and projectivists, because we might think that the projection of a ‘real’ property will make the property basis more rigid. I can’t believe it makes it completely rigid, because the same kind of Gricean thoughts still seem to have weight. I tend to think that (2) is more natural on syntactic grounds, but (4) makes more sense if we want to interpret utterances of (1) as sensible. I don’t see any chance that (3) could be the intended meaning, but if (2) and (4) are possible interpretations, then I suppose we have to admit (3) as well.

    As for the wider issue on the independence of individuation from evaluation, I wouldn’t be too worried, because I’m sure that the following conditional is both true and well-supported:
    ‘If I had liked liquorice then I’d have liked fennel too.’

  6. Jeremy Pierce says:

    I agree that (2) is the more natural reading. I’m not even sure it’s ambiguous. My sense is that it’s semantically ambiguous but that pragmatics does the work to get us to hear it as (2), but my only reason for thinking that is that you can paraphrase (3) with a fairly natural-sounding English sentence:

    (3’) It could have been that we found Vegemite icky-tasting.

    I can’t think of a similar sentence for (2) or even anything close to it. There’s got to be some Gricean kind of rule for sentences that could be ambiguous where one of the readings could easily have been said but the other one couldn’t have been.

  7. DJC says:

    I hear (3) more naturally than I hear (2). It seems correct to say that if our taste receptors had been different in a certain way, then vegemite would have tasted icky. From this it seems to follow that vegemite might have tasted icky. And this requires a reading more along the lines of (3) than (2) (though one arguably needs to stipulate that the relevant counterfactual tasters are counterfactual versions of ourselves). Likewise, I’m inclined to say that if there had been no tasters, then vegemite wouldn’t have tasted any way at all. That also suggests (3) rather than (2).

    I’d say the same sort of thing for “looks red”. It seems reasonable to say that “looks red” goes with the way our sensory apparatus is set up in the relevant counterfactual world, even if “red” does not. If, like most philosophers, one thinks it isn’t necessary that red things look red, one will naturally be inclined to say that “could have looked red” goes with something like (2), even if “could have been red” goes with something like (1). Same, I take it, for “could have tasted icky”. Perhaps there’s a case that “could have been icky” goes with (2) rather than (3); this isn’t obvious, but it’s not obviously false, either.

  8. Matt Weiner says:

    I can’t think of a similar sentence for (2) or even anything close to it.

    “Vegimite could have had an icky taste”?

    As for sentences that could be ambiguous where one of the readings could easily have been said but the other one couldn’t have been
    I think a good example is
    (13) Fred might have bet on red.

    This is ambiguous as to whether “might” has scope over the past tense. The reading on which “might” has narrow scope can be forced by saying
    (14) Fred may have bet on red.
    [In my idiolect.] But there’s no way AFAICT to force the reading on which “might” has wide scope—still I find no pressure to interpret (13) so that “might” does have wide scope. So the Gricean approach doesn’t seem to work here.

    (Of course Brian has written a whole paper about epistemic modals, and may have already addressed this. Certainly he might have. See what I mean?)

  9. DJC says:

    Typo in my message: the sentence numbers in the middle sentence of the second paragraph are wrong. It should read:

    If, like most philosophers, one thinks it isn’t necessary that red things look red, one will naturally be inclined to say that “could have looked red” goes with something like (3), even if “could have been red” goes with something like (2).

  10. dan fineman says:

    Forgive me, but why is the concept of possible worlds here the governing concept? The language is clear, if modal. That this could be narrated differently does not entail any ontological consequence any more than if I say “X could have been Y”: in this algebraic expression it’s clear that the statement makes no claims outside its logical consistency. If I hear, “Bill and Ted went to the store. He’s such a buyer.” There’s ambiguity because this is not a WFF, not because of something in the world(s). The problem of what and how referring expressions do what they do is interesting but not one of ambiguity (for me, not even the ambiguity of sense and reference).