Jackson's two-dimensionalism

An excellent conference on Concepts and Conceptual Analysis here at the ANU has just finished.  Lots of interesting talks -- David Braddon-Mitchell on fine-graining two-dimensionalism (into three-dimensionalism!) to handle a priori equivalent but cognitively distinct concepts, Laura Schroeter and John Bigelow on Against Apriori Reductions (arguing against Chalmers and Jackson 2001, by appealing to an "improv" model of concepts), and Anna Wierzbicka on her remarkable project of conceptually analyzing all linguistic expressions into 65 or so conceptual primitives that can be found in all languages.  Also I gave a talk on a different approach to primitive concepts (no paper yet, but here's a Powerpoint version).

Here I'll say a few words about Frank Jackson's talk, which set out his version of two-dimensionalism and made clear a number of differences with my version.  Frank complained that people sometimes assume that he believes everything I believe (except about consciousness), so it's useful to clarify the differences.

A first, relatively predictable difference is that Frank doesn't like versions of two-dimensionalism that conceive of the two dimensions as operating over two different modal spaces (epistemic possibilities and metaphysical possibilities, say) -- he wants a single space of worlds, with at most a difference in centering.

A second difference is that Frank holds that while linguistic expressions have two-dimensional content, mental states do not -- they only have primary intensions (A-intensions), not secondary intensions (C-intensions).  I think this is tied to Frank's endorsement of a "map" theory of belief, on which there aren't really token beliefs, and certainly nothing like sentences in a language of thought (it may also be tied to his view of the special role of linguistic conventions in determining content).  I'm more sympathetic with token beliefs than Frank is, but I think that even if belief states were holistic map-like states, one could still associate these with individuals and properties represented, and so with entities like secondary intensions, in a way that's useful for many explanatory purposes (e.g., for tracking certain sorts of agreement and disagreement across believers, and certain sorts of communication). The alternative view has some odd results: e.g. no relevant mental content will be shared by two people who believe I am hungry and He is hungry of the same individual, although intuitively these people are agreeing and in some sense believe "the same thing".

The third difference, and the most surprising to me (well, the second would have been just as surprising, but I knew about it already) was Frank's suggestion that there aren't really two different "true at a world" relations between sentences and worlds, but just one.  Now one might think that the existence of two different relations here is a sine qua non for two-dimensionalism (at least for versions with a single space of worlds).  But what Frank meant by this was that the "true at a world considered as actual" relation (i.e. the primary intension or A-intension relation) is derivative on the ordinary "true at a world" relation.  His view was something like the following: sentence S is true at world W considered as actual iff a related sentence S' is true at W simpliciter.  For example, when S is 'the actual F is phi' (and F is "one-dimensional", introducing no 2D variations of its own), S' is 'the F is phi'.  Frank suggested that this model generalizes: all "two-dimensional" expressions are equivalent to expressions of the form 'the actual F' for some one-dimensional F, and their two-dimensional evaluation can be defined in these terms.  (Aficionados will note the resemblance to Davies and Humberstone's framework, where 2D evaluation is defined entirely in terms of the behavior of 'actually').

This claim surprised me in a few ways.  First, the claim that all names, natural kind terms, and indexicals are equivalent to expressions of the form 'the actual F' is a strong and substantive assumption, one that I've always taken it that the two-dimensionalist isn't committed to.  The claim looks especially difficult in the case of indexicals such as 'I'.  Frank suggested that 'I' is equivalent to something like 'the actual utterer of this sentence', but then one has the further indexical 'this sentence' to worry about, and so on.  (In discussion afterwards Frank said he might want to have a somewhat different treatment for evaluation of indexicals.)  And for names and natural kind terms, the relevant F is at least not easy to find (especially given the requirement that F be one-dimensional).

Further, it seems to me that we test hypotheses about whether N is equivalent to 'the actual F' in part by considering scenarios as actual ("what if the world turned out this way?"), and seeing whether the two expressions give equivalent results (as in e.g. the standard 2D construal of what's going on in Kripke's 'Godel' argument).  But this suggests that we have a grip on evaluation of expressions in worlds considered as actual that doesn't derive from our grip on associated descriptions.  Of course one could hold that an underlying description is tacitly guiding our evaluation.  But my view has always been that a two-dimensionalist needn't make this claim, and that we can evaluate expressions in worlds considered as actual simply by reasoning about those worlds (e.g. by what one should infer if one accepted the hypothesis that W is actual), whether or not there is an associated description.

What emerged from the discussion is that for Frank, two-dimensionalism is a sort of byproduct of a prior descriptivism. On his view there is 2D-independent reason to think that names and the like are equivalent to rigidified descriptions, and once we have this equivalence we can use it to ground a 2D analysis.  On my view, by contrast, things are closer to the other way around.  We have an independent grip on how to evaluate sentences in worlds considered as actual and counterfactual (grounded in our grip on certain epistemic and subjunctive modal notions), and this grounds a 2D analysis of content.  This analysis might then be used to ground a sort of quasi-descriptivism, or at least a framework that resembles descriptivism in some respects by delivering some of the results results that descriptivists wanted. But the descriptivism plays no essential role in grounding the two-dimensionalism, and some of the stronger claims of descriptivism need not be true (for example, I don't think that names are semantically equivalent to expressions of the form 'the actual F').  I think that in addition to being correct, this version is more effective dialectically, in that to presuppose descriptivism is to presuppose a view that a lot of people think is false. 

In any case, it's good to get clear on the alternatives.  The difference is in some respects analogous to that between Frege's and Russell's views in the philosophy of language.  As Kent Bach notes, although both Frege and Russell are sometimes described as descriptivists about names, only Russell is a true descriptivist who thinks that names are abbreviations of definite descriptions.  Frege thinks that names have non-extensional semantic values, senses, which are tied to epistemic role and which are in certain respects like the senses of descriptions.  One might say that Frank's two-dimensionalism is "Russellian" while mine is "Fregean".

Comments

[Jim]

Thanks for the report Dave. It's useful to have an official precis of where you and Frank diverge.

[wo]

Yes, very interesting comments.

Isn't another important difference between you and Frank Jackson that his (two kinds of) intensions are semantic values for public languages, constituted by linguistic conventions, whereas your intensions are defined epistemically for individual speakers, largely independent of a larger community and its conventions?

It seems to me that this is the main difference at least between your two-dimensionalism and that of Lewis (e.g. in Plurality, pp.40-50, where 1-intensions are called "truth-conditions" and 2-intensions are unnamed functions from worlds to truth values). And I always thought Jackson was very close to Lewis in this respect.

[djc]

Well, Frank certainly allows that the A-intension associated with an expression (especially a name) can vary between speakers in a community, so I don't know if there's a big difference here. (I'm not sure about Lewis, as I don't have my copy of Plurality handy.) Also, note that while I define primary intensions in terms of an expression's epistemic role for a speaker, it may be that for many expressions (though probably not for names) the epistemic role associated with competent use is itself determined by linguistic convention, so merely having a constitutive tie to individual epistemic role doesn't entail independence from linguistic convention.

Re your trackback comment (which I'll comment on here to avoid fragmentation): yes, it looks like your interpretation was closer to the mark than I thought at the time. After the talk, though, Frank resisted the claim that he was reducing A-intensions to C-intensions (saying that he thinks that if anything A-intensions are more fundamental): rather I think he sees both as grounded in a single "true at a world" relation (maybe defined only over "one-dimensional" expressions?). Certainly he doesn't think that the relevant descriptions have to be expressible by more basic expressions in the speaker's vocabulary -- the description (or perhaps better, the property/relation?) might just be tacitly grasped. Of course this raises questions about what constitutes this tacit grasp. If it's constituted by some sort of disposition to apply a term in specific cases, then the difference with my version won't be as great as it initially appeared to be. But maybe there is a stronger relation involved.

[Peli]

I hope this is not too much of a digression (in that case, simply ignore it, and receive my apologies) but I was never quite able to grasp Frank Jackson's new position on consciousness.

From what I gather, he rejects Type B materialism in such a way that no change of position based on "practical" epistemological issues such as the Paradox Of Phenomenal Judgement (that is, that it's very, very unlikely for a phenomenal judgement to be correct if we accept epiphenomenalism) can make him accept it.
Does that mean he became a Type A materialist?

That seems almost impossible for someone who used to be a dualist. Perhaps only possible (psychologically and intellectually possible for a person, not philosophically possible) if one accepts some kind of very, very strong "Cognitive Illusion" view.

[djc]

Yes, Frank is now a type-A materialist, embracing Lewis-style analytic functionalism and the ability reply to the knowledge argument. He thinks that representationalism (and diaphanousness) make this conclusion easier to accept, as the properties that Mary seems to "learn" about are merely represented properties rather than instantiated properties. But a "cognitive illusion" is definitely involved -- hence the title of his paper "Mind and Illusion".

N.B. The excellent new collection There's Something About Mary: Essays on Phenomenal Consciousness and Frank Jackson's Knowledge Argument (MIT Press, edited by Ludlow, Nagasawa, and Stoljar) collects all of his papers on the knowledge argument as well as a lot of papers by others.

[Matt]

What's the difference between a primary intension and an A-intension? Is it just terminology?

[wo]

Interesting. Did he say whether the domain of the single "true at a world" relation consists of centered or un-centered worlds?

If the description isn't a real linguistic description, but something like a function from (centered) worlds to individuals, I don't see how talk about sentences S and S' and dropping "actually" can be applied. In fact, doesn't rigidification go with a loss of information? The property of being the teacher of Aristotle and the property of being the author of Nicomachean Ethics, when rigidified, both become the property of being Aristotle, the constant function from worlds to Aristotle. So there is no way back from the rigidified function to its de-rigidification.

[djc]

Matt: To a first approximation it's just terminology (Jackson and I may have slightly different views of the intensions, but the difference in names doesn't really mark the difference in views). There's a long history of multiple names for two-dimensional semantic values. The antecedents are Kaplan's "character" and Stalnaker's "diagonal proposition", but one can't use these terms here as they stand for subtly different sorts of entity. In my 1993 Ph.D. thesis I used "prior" and "posterior" intension, but I wasn't wild about these terms. For my 1996 book I thought a lot about the best terminology, and even polled a number of colleagues about options. A significant majority voted for "primary" and "secondary" -- including Frank Jackson! But for his 1998 book he introduced the new terms "A-intension" and "C-intension", I think on the grounds that they were more mnemonic. I've also occasionally used "epistemic" and "subjunctive" intension when specificity about the foundational notions is required.

Not long before he died, I received a letter from David Lewis urging me to stay with "primary" and "secondary" on the grounds that these terms were by far the most likely to catch on. I've come to think that David was right about this, and this seems to be the way that the literature is going. So I'll probably mostly use "primary" and "secondary" in future.

[djc]

Wo: Frank didn't say what was in the domain, but I presume it was uncentered worlds. Maybe one could handle indexical cases by putting an indexical-laden expression E into the form 'the entity that actually bears F to me' (where F is a one-dimensional expression for an appropriate relation), and then saying that 'E is phi' is true at a centered world W considered as actual when 'the entity that bears F to S is phi' is true (simpliciter) at the uncentered world W' corresponding to W, where S is a one-dimensional expression that when evaluated at W', picks out the individual at the center of W. If 'now' or other indexicals are involved, one can generalize this pattern in the obvious way.

If there's no actual linguistic expression F available, presumably one will do things in terms of the underlying relation. The idea might be something like: we have antecedent reason to believe that for all names N, there is a relation R such that the name functions to rigidly designate whatever actual entity bears R to the speaker. Then one can say that 'N is phi' is true at centered world W considered as actual iff (in the uncentered world corresponding to W) the entity that bears R to the individual at the center of W has the property denoted by 'phi'. Strictly speaking this doesn't appeal to a second sentence S', but there's still a broad sense in which this proposal defines truth at a world considered as actual in terms of truth at a world simpliciter. I don't think that the fact that rigidification involves a loss of information makes a difference here.