How People (Actually) Speak of the Colors

Imagine that you take a look at an ordinary CD and conclude that it is round. Now imagine that an alien being from another planet looks at that same CD and concludes that it is triangular. Would you say in such a case that (a) the alien being must be getting it wrong or (b) there is simply no fact of the matter?

Unsurprisingly, studies show that people say that there is a definite fact of the matter in a case like this. They say that the CD truly is round and that the alien being is wrong. 

Now suppose we ask exactly the same question in the domain of color. Suppose you conclude that a tomato is red, while the alien being concludes that the tomato is green. Now, ask yourself: Is the alien getting it wrong here, or might it be that there is just no truth of the matter?

Philosophers have often suggested that the folk view in such a case is that there must be some single objectively right answer. In fact, I recently attended a talk by a prominent metaethicist which started out by boldly announcing that people treat colors as objective and then went to ask the question as to whether people treat moral properties in the same way... 

But, as a new paper by Jonathan Cohen and Shaun Nichols shows, there is good reason to think that this traditional view is not quite right. Though people do give objectivist answers about questions of shape, approximately half of subjects actually give relationalist answers about color. That is,  approximately half of the subjects say that if the human and the alien disagree, then neither has to be wrong, since there is just no fact of the matter about unqualified claims like 'The tomato is red.' 

In my view, studies like this one show something striking about people's ordinary understanding of color but also point toward a deeper sort of question as to how people decide, in a whole variety of different domains, between more objectivist and more relationalist views. Definitely an exciting topic for further research. 

Comments

[David Chalmers]

An obvious worry is that the experimental results appear quite compatible with the view that (subjects hold that) colors are intrinsic. Many philosophers accept both (i) colors are intrinsic, and (ii) there can be inverted spectra without illusion, so that alien perceivers (with different patterns of color experience and judgment) will still have veridical experiences and true beliefs about colors. The results suggest that many subjects accept something like (ii). They don't seem to do much to suggest that the subjects reject (i). The worry is that the interpretation of the results in the paper conflates the two issues.

There are also lots of use/mention worries around here. Obviously, the worry about linguistic differences (in particular, differences in referent for color terms) in inversion cases isn't removed just by having some interaction. It would be nice to see the experiments run in formal mode too: e.g. saying that Harry says 'The ripe tomato is red', and without the somewhat prejudicial uses of 'so' and 'no fact of the matter about' -- neither of which are really apt if the different-referent hypothesis is correct. I suspect the results wouldn't be hugely different (because I suspect that subjects won't be hugely sensitive to these use-mention matters), but it would nevertheless help with the philosophical interpretation of the results.

[Justin Sytsma]

Fascinating paper – I really enjoyed reading it! And, of course, I like the basic argument you run and fuel with your research!!

I wonder if your results don’t overestimate the relationalism about colors of the folk, however. In particular, it strikes me that the way you phrase the questions serves to test the brand of non-relationalist view that you discuss in Section 1, but that a less radical version is possible. So, you write:

“Anti-relationalist views about color predict that, in such cases of perceptual variation with respect to color, just as in the case where you and I disagree in (say) the shape we ascribe to an object, at most one report can be correct in what it says about the object’s color.” (2)

It seems possible, however, for an anti-relationalist to accept that there is variation in people’s vision, including their ability to discriminate between different colors. This opens up the possibility that two people could disagree about the color of an object while both being correct – one person giving a more specific characterization of it that the other person is incapable of (and so gives a coarser characterization). At an extreme, a nice illustration of such a difference in visual ability is between a person who is red-green colorblind (call this person Andrew) and a person who is not (call this person Harry). Say that the colorblind person says that a tomato is green, while the non-colorblind person says that the tomato is red. In a sense I want to say that the colorblind person is mistaken, in a sense not: Andrew just can’t discriminate between the relevant colors; recognizing that, I have some inclination to take him to be expressing the judgment that it is red-green (or in this range of colors), which is true.

It strikes me that the background story that you give for your study is compatible with this; presumably the colorblind person learned English in the normal way and have native fluency (although they arguably should be understood as not meaning quite what the non-colorblind person means by some color words in some contexts, owing to the differences in their perceptual system). Applying the discussion of colorblindness to your scenario, I have some inclination to answer both 1 (“The tomato is red, so Harry is right and Andrew is wrong.”) and 3 (“There is no fact of the matter….”). In a sense 1 is true – Harry made the better color judgment – but I also hesitate to say that Andrew is wrong. I think I would probably lean toward answering 3 (focusing on terms like “absolutely”) if given this scenario out of the blue, despite leaning toward an anti-relationalist view about colors.

I wonder what would happen if you gave people a more graded choice of options? What I am thinking is that instead of essentially given participants the choice between Harry is totally right / Andrew totally wrong and neither Harry nor Andrew are right or wrong, you allowed people to say that Harry was more right than Andrew. Maybe using a Likert scale with your 1 at one end, your 2 at the other end and your 3 at the center point. I predict that if you did this on 7-point scale (your 1 = 1, your 3 = 4, your 2 = 7), you would get a lot of 2/3 answers and a corresponding drop in 4 answers.

[oldog]

...this will probably sound foolish, but i'm actually a bit unsure as to how *i* would answer the question about the shape of the cd. the question seems somewhat analogous to the following: i look at a standard schoolchild's ruler and say "it is a foot long." the alien looks at the ruler and says "it is 9 inches long." at first blush, it seems like one of us must be right and the other wrong. but of course, the question does not specify whether or not the alien is a reference frame moving at a high velocity in the lengthwise direction of the ruler. in other words, the question about the ruler, and, i believe, about the cd (and maybe the colors and moral properties too) does not contain enough information for the respondent to give a decent answer. perhaps this is linked to the above comments as well...

[Jonathan Cohen]

Dave:

I definitely agree with you that you need much more of a story to get from (ii) to the denial of (i). I think, however, that it would be asking an awful lot of ordinary Ss to get them to articulate such stories. So we didn't ask for that. In fact, the way I think of the paper is as doing something exceedingly weak (Shaun may differ here, so this is just me talking). Namely, I take the paper just to be demonstrating that there's some amount of evidence in the kinds of phenomenally-cum-ratiocinatively informed responses that some normal Ss give you that is compatible with (note: I didn't say 'entails') relational construals of the target properties, so it requires more than pre-theoretical phenomenal introspection cum ratiocination to decide that colors are not relational. We didn't aim to show that no such story could be given, but just that something substantive that goes beyond the untutored introspection (and hence that is bound to be philosophically contentious) has to be said at this point. And we, or at least I, took it to be worth saying this because it seems that many philosophers have thought simple, untutored phenomenal/ratiocinative introspection closes off that possibility from the outset. Think of this, if you like, as an invitation to those who have that view to provide the missing details....

About the use/mention issues: I agree that it's hard to control for these (didn't think to run things in the formal mode; that's a good idea), and I agree with your prediction that this is unlikely to change the results. For what it's worth, we did try to preempt an explanation turning just on linguistic differences by the background description stipulating "They learn English by reading books, and attain native fluency. Their use of English words is no more different from yours than that of other native speakers of English is from yours...."

Justin:
If I read you right, then your idea is that ordinary color terms -- eg 'red' are ambiguous between two different levels in the determinate/determinable heirarchy. On one reading (the one expressed by Harry's uses), 'red' applies to ripe tomatoes and English post-boxes but not ripe gooseberries or limes. On the other (expressed by Andrew's uses), 'red' applies to ripe tomatoes, English post-boxes, ripe gooseberries, and limes. Hence, you want to say that our choice 1 is correct on the first disambiguation of 'red' but that our choice 3 is correct on the second disambiguation.

This is an interesting interpretation (and one I hadn't thought of) because, if correct, this would mean that our subjects who answered 3 weren't (in at least one way) really as conceptually open to relationalism as we were thinking they were.

However, I'm not super-optimistic about this suggested interpretation of what our Ss were up to. One point is that people who DO fail to discriminate reds from greens typically themselves retain the two different color words 'red' and 'green' and take these to be more or less disjoint (even if they themselves can't pick out the extensions reliably). Since those people don't think that 'red' has a reading like your second disambiguation (and since they'd presumably be more likely to take the word to have such a reading than others), I doubt that our subjects took the word that way.

But remember the problem raised by such conflict cases is not unique to colors in particular. Are you prepared to say the same thing about our other cases of apparent conflict? Eg, that 'rectangular' is ambiguous between the more determinate shape had by the fronts of cereal boxes (expressed when Harry says the word) and the less determinate shape had by both cereal box faces and also dinner plates (expressed when Andrew says the word)? I guess I find that pretty implausible; and that makes me doubt that your characterization gets at how our subjects were reading the options we gave them.

What do you think?

[Jonathan Cohen]

oldog:

If you think that, then you are exactly endorsing relationalism about colors (which is what we were interested in inquiring about). If, as you say, length properties are constituted in terms of a relation to a frame of reference, then that amounts to a kind of relationalism about lengths. So, too, the relationalist about color claims that colors are constituted in terms of relations --- in this case, relations to subjects/viewing conditions. Our cases were intended to probe the extent to which Ss were conceptually open to the treatment of cases of apparently conflicting ascriptions that goes with the relationalist line about colors....

[Autumnal Harvest]

"Now, ask yourself: Is the alien getting it wrong here, or might it be that there is just no truth of the matter?"

My answer would be that either answer can be correct; it's an arbitrary choice of how we want to translate words. This alien presumably didn't come to Earth knowing English, and at some point must have learned definitions for "green" and "red." I might tell the alien that he's wrong, because I call things "red" if they, in white light, primarily reflect light in the range of 630-660 nm, and the bulk of the radiation from that tomato is in that range. By that definition, the alien is objectively wrong, and we can objectively demonstrate his wrongness (just as we can objectively demonstrate that the CD does not meet the human definition of "triangle").

The alien can reasonably counter (for the color, not for the shape) that he/she/it understood "red" and "green" not to be defined in terms of monochromatic wavelengths (and that neither the tomato, nor any of the other examples that they used to learn English colors were monochromatic anyway). They understand that by saying "the tomato is red," humans mean that the cones that are stimulated in their eye are roughly similar to those stimulated by other "canonical red objects," such as blood or roses. And the aliens might than show that they have a different eye structure such that the nerve endings stimulated in their eyes by the spectrum of light reflected by the tomato, as processed by their brain, more closely resembles that of grass or emeralds. We then just have to talk with the aliens, and decide if we want to define colors as (1) monochromatic ranges of the electromagnetic spectrum (2) human-processed categories of optical nerve stimulation or (3) sentient-being-processed categories of optical neve stimulation. All three are reasonable choices, but they're choices about whether we want to translate words, in this instance, in a relational or objective way, not a statement about what colors "really are" (whatever that means).

[Justin Sytsma]

Jonathan,

Thanks for the reply! I’m not sure that the terms need to be ambiguous; I was thinking more along the lines of there being a pragmatic reason in some cases to read the use of a color term as indicating a different level in the determinate/determinable hierarchy than it typically does. That is my inclination with colorblindness: Certainly when the colorblind person calls a red thing, “green,” he is making a mistake; at the same time, knowing that he is colorblind I recognize that he isn’t totally mistaken. One way to express the sense that he got something right is to say that his use of “green” is best understood as including red.

The shape case I find somewhat different, in part because I’m not familiar with the relevant type of shapeblindness (as it where). Pragmatically, I’m more inclined to say that Andrew is just wrong when he says it is a triangle. Another way to put that is that it isn’t clear to me how Andrew could get it wrong and yet my sense from the vignette is that we should take him to be being honest. Actually, that almost 31% of participants give answer three for shape is rather striking! One way to read this is that if we think that the person is making an honest perceptual judgment, then we are disinclined to think that the person is just wrong; rather, it seems that the person must be getting something right, if they are honestly making perceptual judgments. For the case of colors, colorblindness gives us a handle on how someone could honestly make the incorrect perceptual judgment… and in doing so shows how they are nonetheless getting something right in the judgment. With regard to shape, however, people are not familiar with a corresponding case (no triangle-circle shapeblindness), so somewhat fewer give Andrew the benefit of the doubt. But even in that case more than 30% do!

Anyway, this is clearly an empirical issue and I think rather readily tested. I’ll try to run a study asking the color and shape questions on a larger scale. If I am correct, then we should see a significant shift away from the relationalist answer when people are able to answer, in effect, that Andrew is wrong, but not totally wrong. I’ll let you know what I find.

[Jonathan Cohen]

Autumnal Harvest:

Couple of points.

One, it was stipulated in the description of the cases that the alien DID come to earth knowing English at a level of native fluency and that his use of English words is no more different from the subject's than that of other native speakers of English is from the subject's.

Two, I take the first point to more or less entail that the alien DIDN'T learn definitions for color terms -- because having done so would make it the case that the alien's use of those English words is much more unlike the subject's than that of other native English speakers. (For we don't learn word definitions in vastly most cases outside of formal mathematics; occasionally we do learn incomplete and semantically inessential paraphrases from dictionaries, etc, but I would hazard that roughly no one learns color words by such paraphrases -- let alone by appeal to definitions.)

Three, putting the above qualifications aside, I think your main point can be put this way: there are (at least!) three separate properties, including some relationalist ones and some non-relationalist ones, that are candidates for being expressed by an English color word such as 'red'. That seems fair enough. But if so, then it seems like a legitimate question to ask which, if any, of the candidates considered is in fact expressed by the English word. And one prima facie interesting way of asking that question is to set up cases of perceptual variation such as those we used, using English words like 'red' and 'green', and seeing whether native English speakers respond to those cases using words like 'red' and 'green' in ways that suggest those words are picking out relational properties (in which case they would be prepared to accept multiple variants), or non-relational properties (in which case they would most likely not be prepared to accept multiple variants).

Of course, it is a possible outcome that there's no determinate fact of the matter about which candidate is expressed by our color words. (Maybe this is one way of putting the core of your suggestion?) If so, then I would expect that the responses would be pretty randomly distributed, as cooperative Ss would be reduced to guessing. But we didn't see this.

Anyway, that's how it seems to me (again, not speaking for Shaun, whose ideas are, I'm sure, much smarter and better thought out than my own). Does that seem right to you?

Justin:
I think your suggestion, then, is that 'red', etc, are not lexically ambiguous at the level of their semantics, but that some kind of pragmatic accommodation kicks in when we hear them uttered by people with known visual-system defects --- kinda like what Kripke says we do with descriptions such as "the man in the corner drinking the martini" that we know don't semantically apply to the speaker-intended referent ("Speaker's Referent and Semantic Referent").

While I think this is a definite improvement over the semantic ambiguity story I was (wrongly, I now see) attributing to you before, I'm still doubtful about it because (like you!) my intuitions are so different in the shape variation cases and others like it. I concede that we aren't familiar with an analogous kind of discriminatory failure for shapes, but I don't think it needs to be named and familiar for us to consider the case of someone who systematically disagrees with all of our shape term applications.

I would have thought that, if your story is the right one, then we don't need to know anything about -- or even know about the existence of -- the mechanisms underlying the different discriminations in the alien guy. All you need to know is that he gets the property applications reliably wrong, strictly speaking; and then you go on to do the pragmatic accommodation business because you are such a nice, charitable, kind of guy. If we don't do this for the shape case, as both your and my intuitions suggest, then I take this to cast some doubt on the general story.

One other point. Although we didn't make this explicit in the vignette we told Ss (because we weren't considering your alternative interpretation at the time), I, at least, was thinking that the different variants coming from the different Ss were at the very same level of determinacy. So it is not that Andrew was willing to apply 'green' to a superset of the things to which Harry applied 'green'; rather, it was more like a reversal, so that the extension of Harry's 'red' is exactly the extension of Andrew's 'green' AND the extension of Harry's 'green' is exactly the extension of Andrew's 'red'. What this means is that if there's a pragmatic accommodation to be carried out wrt Andrew's 'green', the right way to accommodate wouldn't be by going to the superset/less determinate property, but by going to a completely different/equally determinate property that, as it happens, has a perfectly good and salient lexical expression in English. Here's an empirical conjecture: subjects are less willing (but not unable) to do this last kind of accommodation.

If you do run such a followup study, I'd be interested to see if it turns out differently depending on whether extension Andrew is responsive to is a superset of Harry's extension (as you were conceiving the case) or disjoint from Harry's extension (as I was conceiving the case). I bet it does....

[Autumnal Harvest]

Jonathan,

I find your second point, that "the alien DIDN'T learn definitions for color terms" difficult to understand. Surely at some point she must have learned, impliclty or explicitly, a definition for "red." Otherwise why would it even occur to her to make that sound? As you say, few people learn words by explicit dictionary definitions (although aliens might need to), but if she hasn't learned an explicit one, she must have deduced some implicit definition by discering that humans describe fire trucks, cherries, and blood as red (e.g. "things that stimulate the eyes in the manner of fire trucks, cherries, and blood, are red). She has to have some rough definition of "red" in mind when she uses that word, and that definition has to correspond to at least a plausible human definition of "red," or else she is wrong because she is simply mistranslating. (i.e. she cannot correctly use the word "red" to communicate to English-speaking humans that an object is heavy, for that is not a reasonable definition of the Enlish word "red").

I agree with your summary of my views, and I agree that this is an illustrative way of investigating an interesting question. It's a rather neat experiment (although I haven't read the full paper yet). However, I disagree that my claim that there's no determinate fact of the matter entails that the responses should be random. I'm saying that how to define "red" under these circumstances is a choice, akin to choices about how to define "chair." If you ask people whether stools are chairs, the English word "chair" is not defined precisely enough for there to be a clearly correct answer, but people will still have real (differing) ideas on the matter, not randomly distributed guesses.

[Bob Kruk]

What is so surprising about this? A triangle is defined in a way that color is not. Triangles have definite fixed properties. Color does not. Instead, it is learned through association with various sets of concrete objects. As that actual set will vary from person to person and culture to culture (Native Americans did not originally have apples), we can expect different notions of red. Unlike the triangle which has no room for notional flexibility (3 sides always equals 3) red occupies a range across a spectrum that in different languages is broader or narrower depending on the number of basic words for color.

[Jonathan Cohen]

Autumnal Harvest:

I think our disagreement is at least largely verbal. I agree that in order to get the extension right (etc) the alien has to have some kind of semantic representation of the expression. But this leaves it very much open what the nature of that semantic representation is. By 'definition', I meant something that specifies the kind of necessary and sufficient conditions thing you get in math class. My point was that vanishingly few of our expressions can be given non-trivial definitions in this sense, from which it follows that we don't learn (gain mastery of) those expressions by learning necessary and sufficient conditions for them. In your response you seem open to the semantic representation's being far weaker than what I meant by 'definition'. Cool.

So where we are, I think, is that we agree that the alien has to have some semantic representation for the lexical item 'red', and that the subject of our experiment likewise needs to have some semantic representation for the lexical item 'red'. If so, then our hope was to probe the content of that semantic representation in certain constrained ways....

Let me say a word about what I think the random responses we didn't find would show if we had found them. Of course, I agree that one CAN stipulate a definition for 'chair' or 'red' in lots of different ways. Our experiment didn't make any such stipulations for 'red', 'square', or the other names for the properties we were probing. So -- without our having made any stipulative intervention -- either the semantic representations our Ss have in mind determinately pick out one from among the candidate properties (some relational, some not), or they don't. If they do, and if our experiment is put together decently (a big caveat, I know!), then questions using the terms in question just might reasonably be thought to be answered in a way consistent with how one or more of the candidate properties reacts in cases of apparent representational conflict. If not, then each subject we ask could only possibly answer our questions by, in effect, making largely unconstrained stipulations about the referents of the terms on the spot; but, since there's nothing to constrain different subjects to stipulate in the same way (specifically, we are now supposing that the semantics of the words don't impose such constraints), one should expect that Ss will not react in any consistent way. I.e., their responses should be more random than they appear to be.

The analogy with 'chair' that you suggest seems to me like a good one. Assuming (plausibly) that you are right that the semantics for 'chair' requires determinately neither that a stool counts nor that a stool fails to count, what people have by virtue of their mastery of the semantics of English leaves it open whether they'll count a stool as in the extension of 'chair'. If you ask, 'is this [pointing at stool] a chair?', subjects have to stipulate on the spot; and if nothing (eg nothing pragmatic) constrains their stipulations to line up with one another, then what you should expect is more or less random variation in the answers you get. (I'm not disagreeing that people will be prepared to justify their answers; only I'm saying that what's going on in their heads, whatever they say the think is going on, bottoms out in a stipulative choice, in the imagined scenario.)

I think we are probably in agreement about much of this given what you say. Am I right?

Bob Kruk:
What is surprising about this? Maybe nothing. It is potentially surprising if you think that (i) triangularity is a non-relational property with a fixed (geometric) essence, and (ii) red is a non-relational property with a fixed essence of some other kind.

It sounds, from your comment, that you accept (i) and deny (ii). Wonderful! Then you should be open (at least one phenomenological grounds!) to the kind of relationalist story about colors that I (but not Shaun!) am trying to sell. Have I got a bridge (or at least a book) for you! On the other hand, it seems to me from a glance through the literature that many people have thought your combination of positions here is wrong, obviously wrong, and obviously wrong on grounds that are easily within the grasp of anyone who has ever looked at a ripe tomato. Our paper is designed to defend the view to which I think you're sympathetic from some of these considerations brought against in the literature. So we are on your side here -- you should be our friend!

[Brandon N Towl]

Another thing to consider, along the lines that Kruk has mentioned:

There are more constraints on shape than on color, meaning that there are more ways to ascertain an item's shape than its color. This might lead to different views of the two kinds of properties.

Let me flesh this out with an example. When we see a triangular CD case, we see a shape (presumably). But there are other ways to ascertain the case's shape than just by sight. For example, if the case truly is triangular, it should be able to fit in an appropriately sized triangular hole (but not, on the other hand, a flat slot). When stacked with other triangular things of the same size, it should form a stack with straight sides. It should balance much more easily when stood on a side than when stood on a vertex. Etc.

Shape is something that determines how many interactions between an object and the environment will go. I would suspect (admittedly, just speculation) that color is not a property that makes so many determinations. Color is something we see and react to, but not much else. Thus it is much easier to be a relativist about color, since there are few "independent" tests for color outside our seeing it.

But here would be an interesting follow-up. Suppose that we *could* come up with such tests. For example, maybe different colors of light transmit heat differently, or conduct electricity slightly differently. Now inform subjects about the existence of such tests (they could be fictional, even) and how people have used them to learn about the colors of objects. Then ask them about the same essence/relativism questions. I would wager that subjects given information about the other "properties" of color would then lean much more towards the real essence/fact-of-the-matter end.

[Jonathan Cohen]

Hi Brandon:

Thanks for the input -- you bring out an important disanalogy between the target properties. (And one that others have sometimes suggested in the literature --- see, eg, Byrne and Hilbert's 2003 BBS piece, section 3.4.)

I guess I agree that if colors did interact in a consistent and lawful way with the environment --- specifically, with the non-cognitive bits of the environment --- then this might give us a way of resisting the support lent to relationalism about color by considerations about perceptual variation (considerations we were trying to test in relatively naive Ss in the paper under discussion).

(I don't think this support would be weakened were it merely the case that sentient beings had an additional perceptual modality for representing colors -- for one could raise the same sorts of questions about variation in the imagined new modality, and in any case it's not clear why we would be justified in taking the representations of the new modality to be correct and authoritative over the representations of vision rather than the other way around.)

On the other hand, suppose, as you are prepared to accept, that colors don't interact with the non-cognitive environment in the ways you have in mind. Then the question is what you can learn about colors by asking subjects about what would happen if colors were different in these specific ways. And here I'm doubtful that we will learn what we hoped to learn.

Here's why. If you stipulate that color does a bunch of stuff that it in fact doesn't do --- stuff that is best explained by taking colors to have hidden essences, and then ask subjects whether they think colors have hidden essences, then your stipulation effectively begs the question in favor of the hidden essence view. It's just not clear (short of begging the question) that you're asking about color anymore, as opposed to the closely related property shmolor (which is just like color, except that it rigidly has the hidden essence of whatever it is that actually materially constitutes colors in our world).

Here's a way to make that point clear. Take any property that you think clearly lacks a hidden essence -- my favorite example is humorous, but feel free to substitute something else if you don't like my example. It does seem that, just as you say about red, humorous is limited in the range of its environmental reactions -- pretty much we perceive it and respond to it, and not much else, as you say.
Now let's (as you suggest for color) stipulate that there are some newly invented environmental interactions for the property that don't run through the reactions of sentient beings. Say, I invent a humor detection machine that has a green light that goes on when and only when you tell it a joke that is humorous. Stipulate that the machine works successfully. And suppose I tell naive Ss about my new machine, and then I ask them the same sort of essence/relationalism questions. My prediction is that they're going to say there's a hidden essence of humorous. (After all,
Ss are trying to be cooperative with your setup of the case, and it's hard to see how there could be the kind of invented device stipulated as part of the case unless the property has a hidden essence that could be detected by the new machine.)

If I'm right, then following your recipe should make people respond to the essence/relationalism question in a way indicating they take humorous to have a hidden essence. Should we believe them? I suggest NO! We initially chose humorous as an example of a property we thought (before messing around with Ss' heads) on the basis of our best, broadly empirical theories of the world, does not have a hidden essence. All the experiment shows is that we have doctored up the (uncontroversially contrary to fact) stipulations to get hidden-essence-compatible answers out of Ss.

I think the same thing is going on with your imagined experiment involving red. Viz., it's true that the exactly analogous stipulations will get Ss to respond in the hidden-essence-compatible way, but this should be understood as a result of what the (uncontroversially contrary to fact) stipulations built in to the description of the case, not as telling us something about the metaphysics of red.

Anyway, that's how it seems to me. Do you disagree?

[Autumnal Harvest]

Jonathan,

Yes, I think our disagreements are primarily (perhaps entirely?) semantic.

[Brandon N Towl]

What you say sounds right to me. I agree that the experiment would be "loading the deck", so to speak.

But even considering the experiment has given us a "recipe" for distinguishing properties with essences versus those that don't (or ones that we can be relativists about). It seems that multiple and varied interaction with the non-sentient environment leads us (with good reasons) to be realists about a property and/or to posit a hidden essence. Lack of such interaction at least leaves us open to be anti-realists/relativists. That is a theory worth developing.

So I guess my post wasn't so much of a challenge to what you've said, but more in the category of the other "are we surprised?" posts. Then again, having a good explanation to hand for a result that is not surprising can, itself, still be interesting (and/or surprising).

Thanks for taking the time to respond!

[Paul Torek]

I'd like to chime in with oldog on the suggestion that the geometric-disagreement question might not contain enough information for the respondent to give a decent answer. The respondent might, of course, think that it does, and that can be interesting enough in its own right. But that doesn't mean the experiment is truly getting at the difference between relationalism and objectivism.

For example, subjects might be familiar with Einstein's physics but take it as implicitly given that the alien and the earthling are in the same reference frame. In that case, the subjects will claim that one of them must be mistaken even though they count, at least in Jonathan's book, as relationalists about shape.

[davey z]

What I would like to comment on is a point I believe Autumnal Harvest started, but I feel was not fully satisfied in the disccusion.

Before I begin, let me direct your attention to a short video clip:
http://www.youtube.com/watch?v=27XEjKbgLCk
the reason this joke is funny is because prince akeem, who is used to the convention of having a pleasant response after greeting people pleasantly (in this case with good morning), and who is learning cultural mores in america, thinks that when the neighbor yells, "fuck you!", he is actually being pleasant. taking this as a cue and a newly-acquired pleasant greeting, he responds back with "fuck you too!" in a sincerely pleasant way.

this, i feel, is oddly relevant to the methodology of the study in how you phrased that "They [the aliens] learn English by reading books, and attain native fluency. Their use of English words is no more different from yours than that of other native speakers of English is from yours..."

the issue with this, is that if these aliens have "different perceptual systems from ours" (7), then all that which we label as "red" should be "green" for them. now, how do you suppose they become aware of the fact that there is a discrepancy in what we refer to as red and what they (would they understand that what they see is green) refer to as green?

just like prince akeem, it seems impossible that your aliens could know that red is red and green is green for humans if their perceptual systems were different.

It seems like, no matter which way you put it, they can only learn that green things for them are what English speakers refer to as "red" (and if it is the case that you phrase it as Justin Systma does with gooseberries and limes, than the issue still lies in the colorblinded-ness). You cannot cross that phenomenological divide to access and inform aliens that red is green. During the process of the aliens learning English, the only way to learn what the names of colors are are to point out instances in which the color exists. By pointing to the fire truck driving by and saying, "that is a red fire truck", regardless of whether the alien sees the qualia of red or green, he or she is still learning that the referent is named "red".

Now, as to whether this affected experimental results, this is up for question. It surely boggled me while going over the study. I propose that simply taking on the voice of an third-person omniscient narrator could improve your methodology, among other things.

Otherwise, great paper, and a very interesting topic. I also read your Color Relationalist Manifesto and found it very compelling.