Uriah Kriegel

Uriah Kriegel (University of Arizona), “Real Narrow Content."  The paper can be found here.

Comments

[Brad Weslake]

I think this is a very elegant paper. I have two questions about your Objection Six.

Firstly, since this objection can also presumably be raised against a response-dependent theory of colour, shouldn't you have a nice set of stock replies to draw on from that literature?

Secondly, it seems to me that the reason why your narrow content properties are non-twin-earthable is that they are response-dependent. But in your second, preferred reply to Objection Six, you advocate modifying the account by identifying the narrow content properties not with the response-dependent properties themselves, but with their macro-level categorical bases. Doesn't this undermine the whole account by making them twin-earthable again?

[uriah]

Thanks a lot of this, Brad. My first inclination is to say that you’re right on both scores.

Regarding the first point, you’re probably right that there should be something usable
for me in the literature on color (though I don’t know about “loads” – not everybody working on color cares about phenomenology), which unfortunately is less vivid in my memory than would be ideal.

Regarding the second point, my first reaction is to think that you’re probably right and it was a mistake for me to offer the macro-basis response to Objection Six. I think the other response I give, in terms of the distinction between representing a dispositional property and representing it as a dispositional property, is plausible enough to carry all by itself the burden of responding to the objection. There are reasons why I wanted to have another response, but I’ll keep them secret for now…

There might nonetheless be a way for me to hold on to the macro-basis response. It seems to me not entirely implausible that macro-bases are somehow defined in terms of the dispositions they are bases for. On the view I have in mind, what makes a macro-base property F the property it is just is the fact that it grounds dispositions D1,…Dn. If so, then given that the dispositions associated with watery stuff are non-twin-earthable, the macro-bases must be non-twin-earthable as well. This line of reasoning depends of course on a substantive claim about macro-bases – namely, that their identity conditions are given by the dispositions they ground – but it doesn’t strike me on the face of it as all that implausible.

[Benj]

Hi Uriah, very cool paper. I had a question about the motivation. You worry about 2D that "two-dimensional semantics does not clarify how the elements making up the mental analog of a full sentence get hooked up with the corresponding elements making up the state of affairs constituting the truth condition". Wouldn't the 2Dist say that 'a is F' has a 1-truth condition and a 2-truth condition, the former of which is determined compositionally by the 1-intentions of 'a' and 'F', while the latter is determined compositionally by the 2-intentions of 'a' and 'F'?

[David Chalmers]

A bit more on the same point. The "not real content" objection to narrow content that I am familiar with says that narrow content is not content as it is not truth-conditional. But as Benj says, narrow content as understood on the 2D story is certainly truth-conditional.

Your paper seems to raise a slightly different "not real content" objection, namely that 2D narrow content doesn't yield appropriate cognitive contact between narrow content and objects and properties. Now the obvious understanding of this contact is that narrow content serves as a mode of presentation of those objects and properties, roughly as sense is a mode of presentation of reference. Perhaps the objection is that real content should in fact be object-involving, not just mode-of-presentation-involving. But that seems too strong a requirement for content -- otherwise Fregean sense will also be dismissed as "not real content", which seems absurd.

From the substance of your proposal, it seems that the extra feature that you want to claim is that your contents are property-involving. I think the same point arises here. But in any case, the difference with 2D accounts looks pretty minor. Primary intensions are already very close to properties. Strictly speaking they are relations (to subjects and times), because they are functions over centered worlds rather than worlds simpliciter. But relations are still worldly in the relevant sense. And further, the primary intension for a specific subject and time can immediately yield a relational property rather than a mere relation -- i.e. the relational property of standing in the relevant relation to that subject and that time.

In fact it looks like your own proposal can be put into 2D form. The story about response-dependence can be seen as an account of the primary intension associated with certain mental items, and the story about their categorical bases can be seen as a story about extensions and secondary intensions. Where in effect, you fill in the subject (and presumably the time) so that the primary intension yields a relational property and not just a relation.

Incidentally one downside of seeing content as involving the relational property rather than just the relation is that the content is now not "narrow" in the traditional sense of being shared between duplicates. This worry seems to apply to your account. On your account, Oscar and his intrinsic duplicate Twin Oscar will have different contents associated with their expressions 'water': the property of causing appropriate responses in Oscar and the property of causing appropriate responses in Twin Oscar, respectively. Maybe it is narrow in some other sense (e.g. not dependent on the environment), but this is a weaker sense, and one could worry that the resulting object-dependence in the content will lose some of the explanatory advantages of narrow content in the pure sense.

As for the specific account in terms of response-dependence, it seems to me that this will give a plausible account of the content (corresponding to the primary intension) of response-dependent concepts. But on the face of it not all concepts are response-dependent, and one could worry that this account will give implausible results for non-response-dependent concepts. E.g. the concept expressed by 'phlogiston' is plausibly not response-dependent -- it doesn't pick out what causes phlogiston-appearances (if it did, it would pick out something to do with oxygen). So it looks like there will be cases where our phlogiston-beliefs are false but the narrow content you associate with them will be true, which seems an undesirable result. Something similar might apply to any other non-response-dependent concept. By contrast, the 2D framework has enough generality to avoid this sort of undesirable result.

Perhaps a middle ground is to see this as a specific account of the narrow content of some concepts, one that is broadly compatible with the 2D approach but with greater specificity. Of course the tradeoff for greater specificity is less generality, so there will be some concepts that it does not apply to. But nevertheless, the account could have considerable explanatory power within its domain of application.

[Justin Fisher]

I have two concerns, both of the form that Uriah's proposed sort of content won't turn out to be as narrow as he hopes.

Uriah says, roughly, that a "potential narrow-content-constituting property" is a disposition to produce certain cognitive/phenomenal responses in a given observer in normal circumstances. Uriah thinks that a given viewer can be related to certain of these properties just in virtue of her internal constitution, and hence that such properties could potentially constitute narrow content.

My concerns are (A) that the relevant notion of *normalcy* isn't narrow, and (B) that the relevant notion of *cognitive*response* isn't narrow. If these aren't narrow, then Uriah's proposed sort of content fails to be narrow, contrary to his advertisements.

(A)Normalcy - Whether H2O will normally cause 'water'-thoughts in me depends in part upon what sort of surroundings I'm normally in -- e.g., on whether there is normally ambient lighting that will reflect off the H2O in a way that will look watery to me. But what sorts of surroundings are normal for me is clearly a non-narrow property of mine.

Potential response: Maybe Uriah could say that it isn't H2O itself that is the Earthly realizer of the narrowly determined response-dependent property -- maybe it's H2O-in-an-Earthly-environment that realizes the response dependent property. (Effectively this response offloads what-counts-as-normal into the property.) However, this response fits poorly both with Uriah's thought that it is H2O simpliciter that plays the watery role, and with his hope that there would be a simple correspondence between narrow and wide content.

(B) Cognitive Responses - A fair number of people, including me and other teleofunctionalists, think that a state's counting as *cognitive* is itself not a narrow affair. E.g., we find it attractive to think that the *cognitive* states are roughly those that have been naturally selected to play an information processing role in helping creatures successfully navigate their environments (another clearly non-narrow property). Uriah is committed to opposing this view, and he needs there to be a fully internalist account of what properties count as 'cognitive'. I wish him luck.

[uriah]

Hey, Benj – thanks for this. You’re right, there’s a sense in which 2D semantics can provide for the internal structure of a truth-conditional content. Mind you, however, this structure will not map the individual components onto their real-world referents/denotata as we normally conceive of them. Thus, if I think that Jim is cool, you suggest (I think) that be mapped onto a function from centered worlds to particulars. But when I think that Jim is cool, what I think about is Jim, not a function. It is in this sense that 2D semantics is vulnerable to the accusation that the narrow content it assigns to is "not real content."

Dave and Justin: will write back soon - have to run for dinner...

[uriah]

Dave:

Thanks for catching a mistake in how I defined my narrow contents – I’ll get there in a second. But first I want to insist that 2D semantics is vulnerable to the “not real content” objection. (I should say: “insist that the *existing varieties* of 2D semantics are vulnerable” – you’re quite right that there is a general form of 2D semantics where many accounts, from Frege’s to this paper’s, are 2Dal.)

You say it’s too strong a requirement on real content that it be property-involving rather than just mode-of-presentation-involving, if only because that would make Fregean sense not a real content, which is absurd. My response is that none of this is absurd. The content of our thoughts is what we think about. Our water thoughts are about a certain kind of stuff, not a mode of presentation. When we think that water is wet, what is being thought by us is not the wetness of a mode of presentation, nor the wet presentation of a mode of presentation, nor anything in the vicinity. We rarely think about modes of presentation, so modes of presentation are rarely what is being thought by us, hence rarely the contents of our thoughts.

I guess the force of the “not real content” objection is in the fundamental idea that real content is that-which-is-being-thought-about. That idea sounds to me as solid as they come. The only absurd thing in this area is to deny it.

I suspect you might come back and say that this is an entirely verbal issue, that there’s a notion of content that’s like this (=meets this requirement) and another that’s like that (=doesn’t), and that semantic theory needs to posit both. One way to understand my paper is as denying the last bit. But there are deeper issues in the background (which we’ve discussed before in the context of your sympathy for pluralisms and my lack thereof). If our theory of content is an endeavor in the philosophy of linguistics and the philosophy of psychology, and our goal is to clarify the posits of those disciplines, then this kind of pluralism is warranted. But if our theory of content is motivated by the traditional question of the relation between mind and reality, then it’s not. There aren’t as many notions as we care to define of the relation in which we stand to the world when we perceive or think about the world.

On a different topic, you nicely show that the way I defined the contents I’m interested in, they come out wide, because they’re not shared by Twins. I need to revise my account. What I should do is remove reference to the subject in the definition of the content. Instead of saying that the relevant response-dependent properties are dispositions to elicit certain responses in certain subjects, I should say that they are dispositions to elicit certain response-in-certain-subjects. That’s the same as saying that they are dispositions to elicit subjective responses. Thus when the subjective responses in Oscar are type-identical to the subjective responses in Toscar, the disposition to elicit them is the same, and so the content constituted by that disposition is the same. The inter-subjective type-identification of responses would then have to advert to intra-cranial factors only, but that’s not a problem.

Finally, you claim that my account can only handle response-dependent concepts. It entails that the concept of phlogiston does not pick out a response-independent property, when in fact it does. Moreover, my account assigns the same narrow content to the concept of phlogiston as to the concept of oxygen, which is bad. I don’t get this objection. My claim is that the concept of phlogiston picks out both a RD property and a non-RD property. It narrowly picks out the former and widely picks out the latter. (Talk of narrow and wide picking out of properties is natural and necessary once narrow content is understood to be constituted by real-world properties.) So my account does allow for the picking out of different properties by the concepts of phlogiston and oxygen, and it does allow for the picking out of a non-RD property by that concept. Furthermore, even the property narrowly picked out by the two concepts may be different: although the perceptual responses the two are disposed to elicit are the same, but the cognitive responses they are disposed to elicit may not be.

[uriah]

Justin:

Thanks for the input. You’re quite right, in your second point, that the term “cognitive response” may be infelicitous. I should have clarified more. But we can get around that by taking a cognitive state, subtracting any alleged extra-cranial elements from it, and calling what’s left “cognitive response.” Someone might say that the term would be a misnomer, but that’s fine by me (which is not to say that I agree it is). Regarding the first point, I didn’t quite get it. The normalcy is stipulated to hold, however one construes it. The question is whether there could be two intrinsically indistinguishable subjects, placed in the stipulated conditions, that would have different “response-dependent contents” (let’s call them that). If not, these contents are narrow.

[David Chalmers]

On "not real content": OK, now it looks like the objection is that narrow content isn't referential content. Most proponents of narrow content agree that narrow content isn't referential content -- indeed, much of the very motivation for narrow content is tied to the need for a notion of content that isn't tied so closely to reference. (See e.g. my paper for this conference.) So any force to this objection must come from reasons for thinking that nonreferential content isn't real content, or equivalently, from reasons for thinking that the only real content is referential content. It would be interesting to hear those reasons. Of course I think that most of this is going to turn on quasi-terminological issues about what one stipulates to be the constraints on "real content".

On the new definition of your contents: I take it the relevant properties will be those of normally causing the relevant responses in creatures with property B, where B is an intrinsic property (one shared by Oscar and Twin Oscar). To understand the proposal, we need to know how B is specified. If the idea is that B is a detailed specification of brain state, then it seems pretty odd that this will be part of the content even though Oscar and Twin Oscar may have no idea about their brain states. If B is something less specific, it would be good to know what. Also, it now looks like the property in question will be possessed equally by H2O and by XYZ (assuming that Twin Earth is a distant planet in our universe). So this now raises a version of your original worry: how does Oscar's content connect him to H2O as opposed to XYZ? In the paper, one idea was that H2O was the stuff that has the relevant relational property, but now it looks like H2O and XYZ will have the property equally.

On response-dependence: I didn't mean that the concepts of phlogiston and of oxygen would have the same content. I just meant that insofar as the narrow content of 'phlogiston' is something like the property of causing phlogiston-esque responses, then this property will be instantiated by something to do with oxygen (e.g. the absence of oxygen). In any case it will presumably be instantiated by something, since something will normally cause these responses. But presumably that thing is not phlogiston, as phlogiston doesn't exist. So it looks like even though the belief we express by saying 'There is phlogiston' is false, its narrow content will be true on your account. This looks like an undesirable consequence. One could avoid the consequence by saying that the account applies only to response-dependent concepts, and that the concept of phlogiston isn't a response-dependent concept.

[uriah]

On “not real content”: Yes, the objection is that 2D semantics’ existing varieties’ narrow content is not referential, and the point of my paper is to develop a notion of referential narrow content (as I state in the 5th paragraph of my paper, the one that mentions 2D semantics). You ask what the reason is to think that non-referential content is not real content. Like I said, the reason is that content needs to be that-which-is-being-thought-about. Someone might ask for reasons to believe *that*, but then again someone might also ask why believe that bachelors are unmarried.

On the new definition of narrow content: No, the relevant properties are not “those of normally causing the relevant responses in creatures with property B.” Rather, they those of normally causing the relevant responses – period. And of course these properties are going to be instantiated equally by water and twater – that’s why the contents they constitutes are shared by Oscar and Toscar. The original worry doesn’t arise, because the original worry is not “how does Oscar’s content connect him to H2O as opposed to XYZ?,” but rather “how does Oscar’s content connect him to a real-world property?”

On response-dependence: I now see what the objection is, and see that it’s a good one. It does follow from my proposal that the narrow content of “there is phlogiston” is true. I also agree that it looks initially like an undesirable consequence. But there may be an upside to it. It shows that my proposal can provide referential narrow content even for empty terms. Once an account of content pulls that off, it is bound to be saddled with true narrow contents for existentials with empty terms. What is perhaps most important is to make sure that the wide content of “there is phlogiston” turns out false – which is not obvious the way I have the account working right now, but should be possible “to arrange.”

[Justin Fisher]

Uriah seemed to miss the point of my worry (A) above, that normalcy isn't really narrow enough to satisfy Uriah's needs. So here's another shot.

Imagine two subjects: Rene Descartes and Twin-Rene. Rene lives on Earth, where prevalent conditions are such that the presence of H2O normally causes the relevant sort of cognitive response in him - let's say it causes 'the watery response'.

Twin-Rene is intrinsically just like Rene, but lives in the Evil Demon's world. The Evil Demon has arranged Twin-Rene's surroundings in such a way that the presence of H2O never causes the watery response in Twin-Rene. (Perhaps the Evil Demon has even perversely arranged things such that the presence of H2O normally causes Twin-Rene to think that water is definitely *not* present.)

Now it seems that the following claims are true:

(1) H2O normally causes the watery response in Rene.
(2) H2O does not normally cause the watery response in Twin-Rene.

If these claims are both true, and Rene and Twin-Rene are intrinsic duplicates, then it must be that whether a particular kind of thing *normally* causes the watery response in a given subject depends on factors that are *extrinsic* to that subject and that kind of thing -- e.g., upon what sort of surroundings it is normal for that subject to encounter that kind of thing in. Hence, it seems that normalcy isn't narrow enough to play a cenral role in an analysis of narrow content.

(This is related to the exhange with Dave over whether *references*to*particular*subjects* are legitimately narrow. For a claim about what normally causes what to be well defined, you need somehow to specify which possible background circumstances would count as 'normal'. One natural way to do this would be indirectly, by way of a direct specification of a particular subject who dwells within the sort of background cirucmstances you want to talk about. But insofar as specifying a particular subject is problematic for Uriah, this strategy for legitimizing appeals to normalcy will also be problematic. And I don't see any plausible alternatives.)

[uriah]

Thanks for elaborating, Justin. I still don't get it though. In Evil Demon's world, watery responses are caused not by H2O, but by something else - perhaps aluminum, perhaps Evil Demon's mental states, perhaps something different yet. But what *bears* the disposition to normally cause watery responses doesn't affect narrow content on my account. The narrow content is constituted by the *disposition itself*. In the present case, the disposition itself is the narrow content of both Rene's and Twin-Rene's water thoughts, so this is not a case in which two intrinsically indistinguishable subjects have different narrow contents. (I think that's what the case was proposed as, though I might be wrong.)

[David Chalmers]

Real content: well, if 'content is that-which-is-being-thought-about' is like 'bachelors are unmarried' for you, then you're using 'content' roughly as I use 'referential content'. So it looks like the objection is grounded in terminological usage, after all. Certainly, I'm happy to agree that 2D narrow content is not real referential content. But like many others, I don't take "all content is referential content" to be an analytic triviality (if it were, then anti-Fregean arguments about content would be a lot easier than people have made them out to be!). On a usage where this is not a triviality, then the claim that 2D narrow content is not "real content" requires substantive argument.

Definition: You'd said that you'd defines the contents as dispositions to exhibit certain responses-in-certain-subjects. I took it that "certain subjects" was to be cashed out by some narrowly type-identified class of subjects, hence B. But OK, if you don't want B, then I take it the property will just be that of dispositions to cause the relevant responses in some subjects or other (in conditions that are normal for those subjects, presumably -- I think this would address Justin's worry). Now the worry is that almost everything will have that property, at least if merely possible subjects count. If only actual subjects count, then an awful lot of things may still have the relevant property, at least if we live in an infinite world.

Response-dependence: well, if the narrow content of one's belief that there is phlogiston comes out true, even though the belief is false, then narrow content becomes much more detached from the ordinary explanatory roles of content than it was previously (and than it is, say, on the standard 2D account). One might have thought it was a platitude that a belief is true iff its content is true, but you'll have to deny this platitude for your narrow contents. You're also now going to be up to your ears in a strong content pluralism, which will weaken your objections on the first point above.

[uriah]

Definition: Interesting point. If we restrict to actual subjects, then many things will have the disposition to elicit the relevant responses, but I don't think many things will have the disposition to elicit the relevant responses *in normal conditions*.

Response-dependence: You make a good case for the undesirability of the consequence you pointed out in the previous comment. But I wonder if this consequence is really peculiar to my own account of narrow content. Let's see what happens with narrow content of existential beliefs on Twin-Earth, now thought of as a non-actual world. Suppose a Tearthling believes that there is water. The narrow truth condition of her belief is that there is watery stuff, so the narrow content comes out true. Isn't this something that every account of narrow content in some sense needs to be wedded to? What am I missing? How does this work on your account?

Real content: we're going a bit in circles on this, so I'll let it go.

[David Chalmers]

Definitions: well, if the world is infinite and there are infinitely many communities, then there's some reason to think that for almost any kind and any response, there will be some subject for whom it is normal for that kind to cause that response. If so, almost any kind will have the relevant dispositional property.

Twin Earth: Sure, on the standard 2D account, the content of Twin Oscar's belief expressed by his saying 'There is water' is true (of his situation, i.e. of the Twin Earth centered world). But likewise, his belief is true. When a belief is false, as with that expressed by one of us saying 'There is phlogiston', the 2D narrow content will be false. So on this account, a subject's belief is true iff its narrow content is true (of that subject's situation). But on your account, the truth-values of a subject's belief and of its narrow content (evaluated at their situation) can come apart.

[uriah]

Definition: If there are infinitely many subjects, then yes, I probably need to introduce a restriction (your B). I don't think it follos that B "enters the content" (which was your original worry). When I think that the chair is heavy, I represent the chair to have a certain relational property, where the other relatum is earth's gravity, but earth's gravity is not part of my thought's content.

Twin earth: Right. I got a problem with existential belief that not every account of narrow content. But now i'm thinking: every account of referential narrow content would have it. Maybe I can say that the truth value of existential beliefs is determined by their wide content, but that of other beliefs by their narrow content. Not a lovely consequence, but perhaps not awful.