Ishani was discussing the following kind of example, and I wanted to get feedback from others who might have clearer intuitions than I.

Billy lives in a house that is painted red on the outside, but every room is painted bright yellow. To get a sense of how yellow, imagine every room looks like this

except without Dizzee Rascal, and with yellow floors. His house is yellow.

Suzy lives in a house that has a perfectly normal, white walls brown wooden floors interior, but every part of the outside is yellow. When driving down the street, people often look out and say That house is yellow.

No other house on the street has a yellow exterior or a yellow interior.

Question: How many yellow houses are there on the street?

My sense is that we can’t say two here. There is a context where we can say that Billy’s house is yellow. And there’s a context where we can say that Suzy’s house is yellow. But there’s no context where we can say that both houses are yellow. That strongly suggests that there’s something funny going on with the meaning of ‘yellow’, possibly something that requires a contextualist solution.

These kind of cases – swapping colour between interior and exterior – can be multiplied easily.

Jack’s car is painted white, but you can’t see that now because he’s been out rally driving, and it is covered with brown mud.

Jill’s car is painted brown, but you can’t see that now because it is literally covered with ice and snow, so if you look at it you only see white.

There’s a context where you can say Jack’s car is brown, and a context where you can say Jill’s car is brown, but not I think a context where you can say Both Jack’s car and Jill’s car are brown.

15 Replies to “Yellow”

  1. Brian, your claims about various sentences all sound right, but I’m not sure that I’d say we should become contextualists about yellow — maybe we should become contextualists about IS yellow. Sometimes, “the house is yellow” means “the exterior of the house is yellow”, and sometimes it means “the interior of the house is yellow”. “Yellow” itself always refers to the same color.

  2. what about this one – Jill has dyed her once brown hair blonde. On some occasions it would be seem correct to say that her hair is blonde, but on others it doesn’t. Not sure if we can say that her hair both is and isn’t blonde however.

  3. I completely agree with Jonathan that the cases you describe do not commit us to contextualism about “yellow”. But I’m not sure that the solution is to adopt contextualism about “is yellow”. I doubt that on one occasion, “that house is yellow” means that the the interior of some house is yellow, whereas on another occasion it means that the exterior of some house is yellow. I think both utterances mean the same thing: that a house has a certain property, that of being yellow. But a house can have that property in different ways (one by being yellow on the inside, one by being yellow on the outside). Why should we assume that the semantic contents of the utterances include that additional information? Perhaps the utterances mean exactly the same thing (setting aside the fact that we are talking about 2 different houses), and pragmatic factors are responsible for an utterance conveying that we are talking about the inside, rather than the outside.
    One should be careful to distinguish the proposition communicated by an utterance from the semantic content expressed by the utterance. Trying to account for everything that can be communicated by an utterance into one’s semantic theory of the terms involved seems to me both impossible and unnecessary.

  4. just in relation to jonathans suggestion, why no go contextualist about “house” instead? that seems the more obvious way to go. Sometimes “house” means the interior of a building, sometimes it means the exterior? same goes for the hair dying example – “jills hair” refers to different temporal parts of Jill on different occasions of utterance.

  5. Leo, I was agreeing with Brian that the linguistic data suggests that we might have to go contextualist about something; if there are just two different ways for the house to be yellow, then it seems like we should be perfectly comfortable saying that both houses are yellow — one in one way, and one in the other. (I have two cousins — one on my mother’s side, and one on my father’s.) I was taking it as given that this is unacceptable in this case; the fact that we’re unwilling to say that both houses are yellow indicates that it’s not just that there are two ways to be yellow.

    Rich, your interpretation of ‘house’ as ambiguous between ‘exterior of a building’ and ‘interior of a building’ strikes me as odd. Billy has a yellow interior and a red exterior; he doesn’t have a yellow house and a red house. Jill’s car has a cheap brown exterior and very, very expensive green leather upholstry on the interior. Jill’s car is brown and expensive, even though neither the interior nor the exterior is.

  6. So, after you paint Billy’s rooms blue, is there still a context in which you can say that his walls are yellow?

    If so, what if Billy’s house is made of glass (still painted red outside and yellow inside). Are there contexts in which it’s correct to say that (the outside of) the house is yellow (both before and after painting the interior walls blue)?

    Suppose Billy’s glass house painted yellow underneath blue on the inside has no paint on its exterior, but is instead covered with mud. Is there now a single context in which it’s correct to say that there are two yellow houses on the street?

    For what it’s worth, my inclination with respect to surface colors is that ordinarily we convey that it’s the outer visible surface of an object that bears a color. Further qualification, implicit or explicit, of which object is under discussion and which portion of that object is in question grounds correctness otherwise. So, I’m not inclined to think that there’s one context in which it’s correct to say that watermelons are green, another in which it’s correct to say that watermelons are red, and another in which it’s correct to say that watermelons are brown. Rather, it’s correct to say that watermelons are green (lemons are yellow, apples are red…), their insides are red, their seeds are brown, and their slices are green, red, and brown.

  7. Jonathan (and anybody else),
    I still don’t see why one has to be a contextualist about anything in order to account for the linguistic data that Brian offers. Contextualism, at least as I understand it, is some kind of semantic theory. The linguistic data are judgments about the truth/appropriateness of what is conveyed by utterances. What is conveyed by utterances may be the result of a complex interplay between the semantic content of the utterances and various pragmatic factors. So it seems too quick to move directly from these linguistic judgments to a semantic theory like contextualism, without considering how pragmatic factors might play a role in determining what is conveyed. Earlier I suggested that what is semantically expressed by “that house is yellow” is just that it has the property of being yellow. You seem to think that this would have the consequence of making it perfectly acceptable to say “there are two yellow houses on the street”. This again seems too quick — one shouldn’t move directly from a semantic claim to a conclusion about what it would be okay to say, ignoring the complex ways in which pragmatics could get into the mix.

  8. So what are these “complex ways that pragmatics could get into the mix”, Leo? I think cases where you can’t go from ‘a is F’ and ‘b is F’ to ‘both a and b are F’ are pretty good cases of ambiguity, or context-sensitivity, or something very much in the neighbourhood. Are there any other explanations here that look plausible, or any other cases where this kind of collapse doesn’t work but it isn’t because of semantics? I’m assuming, admittedly without giving all my reasons in this post, that these kinds of syntactic facts tell us something about semantics, not just something about pragmatics. If this is wrong I’d like to see a reason why it is wrong, not just an assertion that it is, and preferably a case or two where the kind of inference I’m making goes awry.

    I think the idea that ‘house’ might be the source of context sensitivity is interesting, but I don’t think it can explain all the data. For one thing, I don’t see how we’d run a similar story about the cars. For another, note that if each of the houses has three bedrooms, then (1) and (2) are fine, but (3) is bad.

    (1) Billy has a yellow, three-bedroom house.
    (2) Suzy has a yellow, three-bedroom house.
    (3) Billy and Suzy each have a yellow, three-bedroom house.

    If the problems are with ‘house’ rather than ‘yellow’, then we are forced to say that both the exterior and the interior of the house have three bedrooms. That’s not absurd, but it is I think a little uncomfortable.

  9. I discuss cases like this in connection with the problem of change. Suppose you’re driving along a curvy section of Hwy.101. You can properly assert “The road is curvy”. Suppose that sometime later you are driving along a straight section of that same road. In that context you can properly assert “The road is straight”. Now imagine that you are flying high overhead and can consider the entire road at once. You might say “The road is both curvy and straight”. You might even say “The road is curvy, but it’s also straight” (where the connective is clearly sentential).

    Same story with color. Suppose there is a long wall, with red sections and blue sections. Standing in front of a blue section you can say “The wall is blue”. Standing in front of a red section you can say “The wall is red”. Stand back, look at the whole wall and you might say “The wall is red and blue” or “The wall is red and it’s also blue”.

    Same story along the temporal dimension. When the banana is unripe you can say “The banana is green”, when the banana is ripe you can say “The banana is yellow”, when you “stand back” and take the atemporal perspective on the world, you can say “The banana is both green and yellow.”

    This turns out to be an important point, since traditional ways of stating the problem of change include a premise like “Nothing can be both green and yellow” or “Nothing can be both straight and bent”. For reasons given above, you might think that these sentences express truths from the perspective of any particular time, but they express falsehoods from the atemporal perspective.

    As several people have pointed out here, there are different ways of making sense of the data. Going back to the orignal example, you might take ‘the road’ to denote different sections of of 101, relative to different contexts. But those who reject temporal parts will be suspicous of this general strategy, since you don’t have the differnt candidates for reference in the temporal case. There’s just the one banana.

  10. Brian,
    Some utterances seem to convey more content than just what is provided by the words in the utterance. Kent Bach has lots of examples like this in “Conversational Impliciture”:
    (1) Joe is not going to die (today).
    (2) Joe has finished (speaking).
    (3) Joe has nothing (appropriate) to wear.
    (4) Joe has eaten breakfast (today).
    The material in parentheses is not spoken but is implicitly conveyed by saying the words that are not in parentheses in the right context with the right intentions. Obviously different contexts and different intentions will yield different implicit material. Now, I don’t doubt that there are contextualist approaches that can explain how the implicit material is part of the semantic content of the utterance by supposing that terms in (1)-(4) are context-sensitive. But that’s not the only way to go. It’s at least plausible that what is semantically expressed by (1)-(4) is just the non-parenthetical material (which does not contain context-sensitive terms)and what is implicitly conveyed is largely a matter of what the speaker intends to convey in addition to what his words actually mean. So it seems at least plausible that the implicit material is added by pragmatic means.

    Now, (1)-(4) have the following counterparts:
    (1*) Bill is not going to die (ever).
    (2*) Bill has finished (the heart surgery).
    (3*) Bill has nothing (at all) to wear.
    (4*) Bill has eaten breakfast (in his life).
    Even if we can truthfully utter (1)-(4) in some contexts, and (1*)-(4*) in other contexts, we can’t appropriately utter (1+)-(4+), while continuing to implicitly convey what was conveyed by (1)-(4) and (1*)-(4*):
    (1+) Neither Joe nor Bill is going to die.
    (2+) Joe and Bill have finished.
    (3+) Neither Joe nor Bill have anything to wear.
    (4+) Joe and Bill have eaten breakfast.
    You don’t convey the same thing with (1+)-(4+) as you do with (1)-(4) and (1*)-(4*), because you don’t succeed in implicitly conveying the same information. These are cases where the collapse fails. But I don’t think we are forced into contextualist explanations of (1)-(4). So it doesn’t look like collapse failure is conclusive evidence for contextualism.

    I would treat your cases in the same way:
    (5) That house is yellow (all over the outside).
    (6) That house is yellow (all over the inside).
    What is semantically expressed is just that the house has the property of being yellow. The implicit material is pragmatically tacked on. Collapse failure occurs because you can’t convey the same implicit information you do with (5) and (6) by saying “Those two houses are yellow”.

    This may be all wrong — I would have stuck to my vague pronouncements about “complex pragmatic factors” if you hadn’t called me on it. But don’t accounts like this at least have to be considered before moving from the linguistic data to a semantic theory like contextualism?

  11. A few quick points.

    First, in all of these cases I’d be perfectly happy to go with a contextualist reading, and take the failures of conjunction to be evidence for it.

    Second, there seem to be cases where collapse is possible but different things are conveyed. Imagine it is common knowledge that I always have cereal for breakfast, and that Ishani always has eggs. Then you might think that (7) and (8) communicate what is inside the brackets, even when only what is outside the brackets is uttered.

    (7) Brian ate (cereal for) breakfast.
    (8) Ishani ate (eggs for) breakfast.

    But of course we can collapse these into

    (9) Brian and Ishani ate breakfast.

    The point is that when different things are conveyed, the conjunction still seems like a well-formed, and true, sentence, which it notably doesn’t in the house and car cases.

    Third, in all your cases the conjunction seems true, at least under its most natural interpretation. If John has finished what he was doing, and Bill has finished what he was doing, then we can truly say that John and Bill have finished. Again, this is a pretty striking difference from the yellow case.

  12. As I see it, there is a prima facie case for taking ‘house’ to be context sensitive and a prima facie case for taking ‘yellow’ as such not to be so. Yet, as Brian points out, it seems that not all cases are covered by the context sensitivity of the term standing for the object that instantiates the property, e.g. the cars case. So maybe it is ‘is yellow’ that we should take as context sensitive too.

    But here is a way to accommodate all these data. We may say that ‘is yellow’ is to be analyzed as ‘appears yellow in normal conditions’. The context sensitivity of ‘normal conditions’ can explain the context sensitivity of ‘yellow’ via the context sensitivity of e.g. ‘house’.

    For instance, in the case of the house, if normal conditions are those that put the observer in the perspective of looking at the house from a certain distance, then ‘house’ in that case is the exterior of it and Suzy’s house is yellow, Billy’s is not.

    In the case of the cars covered by mud or snow, if, for instance, normal conditions are those in which the car is clean, covered only by its original dye, then ‘car’ in this case is the car as not covered by mud or snow, and ‘is white’ or ‘is brown’ is applicable to it as a function of how it looks as originally dyed.

    If normal conditions are just those of the observer being at a certain distance from the car, ‘car’ is the car together with the mud or the snow that covers it. That is, mud or snow are part of the car, and the muddy car is brown, the car covered with snow is white.

  13. It’s all a question of definitions. What do you mean by “house is yellow”? You can break this up into an inquiry about the definition of “house”, “is”, and “yellow”, but that’s more work than is necessary for the specific question.

    I think it’s important to understand that, with questions like these, there isn’t a fundamental definition of “yellow” or of “house is yellow” floating around in the aether waiting to be discovered. We’re using the medium of language, which has evolved to be functional but not necessarily precise. Thus, when we stumble across a block like the one you pointed out, Brian, it’s more often a linguistic imprecision and nothing more. Once you isolate that, it’s entirely a question of preference – do you wish to define “house is yellow” in such a way so as to mean that the outside of the house is yellow exclusively? The inside, exclusively? Or in such a way as to include both? From a very quick examination, there doesn’t seem to be a good practical reason to pick one over the other two, so I would conclude that the question has no answer as stated, but that it is easy to pose an answerable form of it.

  14. Here’s a reasonable scenario where we can answer “2”.

    Say I want to determine the various ways that something can be yellow, then count the particular instances that are yellow in one or more of those ways. I would express this as, “count the houses on this street that are yellow in some respect”. Within this context, I can correctly say, “two of the houses are yellow”.

    Similarly with the car example. Generally we are referring to the painted color, sometimes the interior color, or the apparent color, but not usually all at the same time. If we want to do that, we make our intention clear by saying something like, “both cars are brown in some respect”.

    The breakfast example has no mutual exclusion implication, so that conjunction raises no potential problems.

  15. Yes, that is a reasonable definition to adopt. One could also adopt the definition that a house is yellow if more than 50% of its external surface area is yellow, thus resulting in there being one “yellow house” on the street.

    It becomes a good bit less interesting, though, once you see it’s just a matter of definition.

Comments are closed.