Today in the PhD course on Herman Cappelen and Josh Dever’s The Inessential Indexical we talked about an objection posed by Manolo García-Carpintero to the impersonal-action rationalizations that are discussed in chapter 3 of the book (section 3.3). Herman called this objection Reply 1*. Here I would like to propose a line of argument against Reply 1* that Herman and Josh (henceforth C&D) didn’t explore during the session. I will focus my attention on Nora’s scenario (pp. 36-37).

Here is C&D’s impersonal rationalization of Herman’s action of closing the door in Nora’s scenario:

Impersonal Action Rationalization 2

Belief (B1): Nora is in danger.

Desire (D1): Nora not be hurt.

Belief (B2): If the door is closed, Nora will be safe.

Action: Herman closes the door.

And here is Reply 1* (as I understand it): a full folk-psychological explanation of Herman’s action will attribute Herman a number of indexical beliefs, such B3, B4, and B5.

• B3: I am Herman

• B4Nora is here

• B5: The door is there

Without the attribution of indexical beliefs like B3 – B5, a rationalization of Herman’s action is incomplete.  One way to show that this is so is to consider a possible scenario in which Herman has the non-indexical states B1, D1 and B2 but lacks B3 – B5. In this situation, Herman wouldn’t close the door. Moreover, his not-closing-the-door behavior wouldn’t be irrational in the envisaged scenario. It would be perfectly rational for him to have the relevant non-indexical states while not performing the action of closing the door. Thus, the attribution of appropriate indexical mental states is necessary in order to get a proper rationalization of Herman’s action.

I think the intuition behind Reply 1* is compelling. Intuitively, some background indexical beliefs are necessary in order to fully explain Herman’s action in the original scenario. However, it is important to note that a variety of background non-indexical states also seems necessary to fully explain Herman’s action. Consider, for example, beliefs B6 – B8.

• B6: An ordinary door can be closed by a normal human-size agent

• B7Closing Nora’s door will not put Nora in a greater danger

• B8: Superman is not going to kill anyone who tries to close a door

By mimicking Reply 1*, one can argue that the attribution of non-indexical beliefs such as B6 – B8 is necessary in order to get a proper rationalization of Herman’s action. Here is a copy-pasted version of the argument: without the attribution of B6 – B8, a rationalization of Herman’s action is incomplete. Consider a scenario in which Herman has the non-indexical states B1, D1B2, has the indexical beliefs B3 – B5, but lacks B6 – B8. To make the scenario more vivid, suppose that Herman believes that closing the door will put Nora in a greater danger. In this situation, Herman wouldn’t close the door. Moreover, his not-closing-the-door behavior wouldn’t be irrational. It would be perfectly rational for him to have all the attributed mental states while not performing the action of closing the door.

What this line of argument seems to suggest is that the incompleteness of the original rationalization of Herman’s action doesn’t reveal anything special about indexicality. Many background states must be attributed to Herman in order to have a complete (folk-)psychological account of his action. Some of these states are indexical, others are not. Reply 1*, one might conclude, doesn’t show that there is an interesting psychological difference between indexical and non-indexical mental states. If this is right, essential indexicality can be resisted even if one finds Reply 1* compelling.

The advocate of essential indexicality will probably think that indexical beliefs such as B3 – B5 and non-indexical beliefs such as B6 – B8 are not on a par. The former, she may suggest, must be included in a proper rationalization of Herman’s action, whereas the latter belong to a background of unstated ceteris paribus conditions. But how does one draw the line between background and non-background mental state attributions? Wouldn’t be easier to save one step and say that B3 – B5 are also background mental states that are understood to be possessed by Herman ceteris paribus? I do not deny that the essential-indexicality theorist might find a satisfactory way to a draw principled distinction between the two kinds of mental states at stake. But I think it is fair to say that at this point of the discussion she has the burden of proof.

End of the original post

Comment by Josh Dever (July 4)

Thanks, David. This is helpful and an interesting line of thought.

I think a defender of IIC, and Reply 1*, might want to push a bit on the question of whether B6-B8 are really necessary in the way that (supposedly) B3-B5 are necessary. (You suggest trying to break the analogy by suggesting that B6-B8, unlike B3-B5, get swept into the background ceteris paribus conditions. That’s one option, but that’s already to concede that there is a need for the beliefs specified in B6-B8, while relocating that need in the shape of the explanatory scheme. I’m exploring instead the idea that they might not be needed at all.)

An obvious point to start with: we need to distinguish between not having a belief, and having the negation of that belief. Certainly if someone has the negation of B8, the belief that Superman will kill anyone who tried to close the door, then the presence of that belief would block rationalization of the action. So we need at least the absence of ~B8. The crucial question then is whether the mere absence is enough, or whether we need the replacement belief B8 as well.

(Side note: I think in general we need to be careful with the thought experiments being used in these considerations. When we try to consider someone who is acting absent the self-identity beliefs — without, for example, belief B3 — one easy way to envision the scenario is to imagine someone who is positively identity confused. So we can imagine Herman believing that he is Josh, and so not believing that he is Herman. But then we don’t know if changes in the imagined behavioral profile are consequences of the positive misidentification belief, or just of the absence of the identification belief. Even if we’re careful not to impute a misidentification belief (as Perry is in the Lingens case), the scenarios change a lot more than just the presence/absence switch on the identification belief. The scenarios always involve someone who is messed up in some way, and when we imaginatively develop such a scenario (think about the closest worlds in which there is someone like that), we may be imputing additional features that do the actual explanatory work in the changed behavioral profile.)

I think there’s a reasonable grip to the thought that one doesn’t need to believe that Superman won’t kill you for undertaking an action, in order reasonably to undertake it. Surely in normal cases the question of whether Superman will kill you for acting *just doesn’t come up*, and plays no role in your acting. If that’s right, then one doesn’t need B8.

(Another side note: it does feel like there’s an important difference here between not believing ~B8, and having the epistemically modalized belief that Superman might or might not kill you for acting. To have that modalized thought is to raise the issue of Superman’s intervention, and once the issue is raised, there’s some temptation to think that it then plays a role in determination of action in a way that it didn’t when it was simply not under consideration. This may mean that we really need models for deliberation that are tied into some kind of broadly dynamic semantics, in which contexts are keeping track of issues under consideration, and modals are partitioning or expanding the space of worlds.)

It’s clear that if we do allow that B8 is needed for action, we’re quickly going to place very high cognitive demands on acting. For any P you believe, we can consider the belief that it is not the case that if P, then dire consequences will result from undertaking action X. Then we’d end up needing to impose all of those beliefs, which means one additional “harmlessness” belief for everything you believe. (Plus a regress, as those beliefs create a need for further harmlessness beliefs.) It’s very hard to accept that rational action comes with such a high cognitive entry cost.

(Maybe we just have these beliefs tacitly? One line of thought will then have it that our very acting is the evidence that we have them tacitly — that a background functionalism will impute the beliefs to us via our acting. This is a line of concern in the de se case as well — that IIC can be defended by saying that anyone who acts ipso facto has the de se belief, since all that’s meant by a de se belief is one that is necessary for action. There’s a lot to say here, but broadly speaking I think it’s hard to see how an *interesting* version of IIC comes out of this approach.)

Of course, that there’s reason to reject the demands for B6-B8 isn’t enough to help out the defender of IIC. They also need that those reasons don’t extend as well to B3-B5. I won’t try to make that positive case here, but I do think that many people will be more gripped by the apparent positive need for B3-B5 than for the apparent positive need for B6-B8, and that it’s less obvious how a kind of over-expansiveness worry arises with the specific category of de se beliefs.

Comment by Manuel García-Carpintero (July 10)


Thanks for this, David – and to Josh for the very helpful reply. I’d like to take advantage of your post to try and explain better than I did during the course my take on Herman & Josh’s discussion of the agency argument.

I think of the agency-based arguments for the essential role of first-personal thoughts (the famous thought-experiments involving characters such as Perry the Messy Shopper, Lingens the amnesiac lost at Stanford Main Library, Castañeda amnesiac soldier reading his biography, etc) as modest, but nonetheless compelling causal experiments.

Imagine, for illustration, a situation in which there is a standing ball, B, which starts moving after being hit by another, A. We believe B’s movement to have been caused by A’s hit. I take this to be an ordinary application of Folk Physics, underwritten by (a folk version of) Mill’s Method of Difference. We compare the many imagined/perceived situations in which neither A’s hit nor B’s movement occurs, with the situation in which both the hit and the movement occur, and conclude that the hit makes the difference, thereby constituting the relevant causal factor of the movement: without it, the latter would not have occurred.

Similarly, we compare the previous condition in which the Messy Shopper lacks the first-personal thought – even though he may well have other de dicto and de re thoughts about himself, to the effect that he is making a mess –, all other relevant folk-psychological conditions also obtaining (in particular, the pro-attitude to prevent sugar-spilling in supermarkets being present), with the situation in which he comes to have it, and conclude that the first-personal thought is a required causal factor. Here, I mean first-personal thought in the theoretically neutral way we accepted during the first session: one that we are disposed to express by using sentences including ‘I’ and its variants.

I take the thought-experiments to make plausible the general point that a first-personal thought is always causally needed to get a folk-psychologically complete rationalization (which I take to be a form of belief/desire causal explanation) of behavior, by applying Mill’s Method of Difference together with the causal-explanatory features of Folk Psychology. Of course, we run the relevant causal experiments in our imagination; but this is neither here nor there, as we do the same with physical causal thought – after all, this is the way Galileo did many of his celebrated “experiments”.

Hence, I do not take these arguments to be in the business of defending Herman & Josh’s IIC, because they crucially involve quasi-empirical elements: who knows what sort of rationalizations are needed to account for the behavior of angels and Gods? I do not have a clue, nor interest in engaging in unconstrained speculation about such matters. I am only interested in defending a restricted version of IIC, not considered by C&D (cp. IIC2, p. 40): As a matter of deep folk-psychological necessity, indexical beliefs are needed in order for human beings, constituted as they are, to act.

Before moving on, a comment about the disparaging remarks Herman made on Folk Psychology: of course, some unrestricted principles that psychologists studying Folk Physics take to be part of it have been conclusively refuted. But, firstly, they have been refuted by very compelling new evidence that, together with alternative physical theories, physicist have put forward. In fact, that combination of evidence and theory also refuted much more theoretically sophisticated theories, which were already previous alternatives to some of the general principles of Folk Physics, such as Newtonian mechanics. And, secondly, in the same way that adequately restricted claims in Newtonian mechanics survive the refutation, many ever so modestly articulated claims of Folk Physics also survive the challenges; among them, the claim that, in scenarios like the one imagined above, the hit by A was a causal factor of B’s movement.

Now, regarding the first point, as far as I can see, the most that C&D offer as a positive theoretical alternative to the folk-psychological explanatory role of first-personal thought is the “Action Inventory Model”. It is fair to say that this falls widely short in comparison with the challenging impact on Folk Physics of, say, early XXth century physics. As C&D admit, it is not even clear that their model is really an alternative; for it may well be that their category of “actionable contents” essentially involves first-personal thoughts – and in fact this is what I take the arguments we are discussing to establish. Secondly, even if the Action Inventory Model, understood in the strictly third-personal way C&D want, provides some adequate rationalizations for some science-fiction agents (perhaps Gods or angels), it is still open that the folk-psychological explanations of human actions involving first-personal/indexical thoughts that the arguments we are considering posit are still the perfectly adequate ones to be had at the “personal” explanatory level.

I move now to discuss David’s interesting argument. In line with the previous interpretation of the arguments for the essentiality of indexicals, I’d like to emphasize that I assume that thoughts such as B3-B5 are thoughts that ordinary subjects for which rationalizations such as IAR2 are given (Herman in this case) do in fact have, and they do in fact play a causal role in eliciting the behavior to be accounted for. Epistemologists distinguish two kinds of rationalizations of beliefs, propositional justification and doxastic justification. The former applies to beliefs a subject may or may not have; the latter to beliefs the subject has actually formed. A similar distinction is usually made with respect to practical rationalization, here between justificatory reasons and explanatory reasons. We are interested in the second variety, involving what in the epistemic case is called the “basing” relation. It should be clear from what I said before that I assume that such relations are in part causal (cf. the articles on the “Epistemic Basing Relation” and on “Reasons for Action: Justification vs Explanation” in the SEP).

Of course, this does not mean that for a rationalization to explain/doxastically justify a behavior/belief, the subject must have consciously gone through the supporting states. For my present purposes, I think it suffices to note that it would be very easy to intuitively show that agents in such cases have the relevant first-personal thoughts, and that they play a causal role in their behavior. For instance, in the Herman case, we can imagine dialogues such as these in reaction to the impersonal explanation C&D provide, IAR2: “But how could Herman have closed the door, if it was not within his sight/if he was paralyzed – no, you are wrong, Herman had in fact a good visual perspective of the door from where he was/had already recovered from his paralysis and now he can walk normally”. These dialogues also suggest the other element in the full argument, which as I said I take to be an application of Mill’s Method of Difference: without the first-personal thoughts, the subject would not have acted the way he did. Note also that, in contrast with what Herman seemed to assume, for the argument we need not defend anything as strong as that visual experiences are themselves first-personal, etc. We only need to show that first-personal thoughts, in the pre-theoretical sense we are assuming, appear to be crucially involved in the explanation.

Now, what about David’s attempt at discrediting the argument with his “bad company” considerations? My reply would be essentially the one suggested by Josh, with a nod to recent debates surrounding dogmatism in epistemology. To put it bluntly, that some reason would undermine/defeat a rationalization does not entail that the rationalization must include a belief to the effect that the potentially undermining reason does not obtain. In the previous remarks, I have tried to add to what Josh says my take on why the status of B3-B5 is very different.


Thank you very much for the discussion yesterday, it was very useful. And sorry I was a bit thick at the end (and maybe before too ☺!).

So my main claim was that that contextualism straighworardly respects facts about the existence of disagreements, as these are sometimes ultimately constitute by contrasting non-doxastic attitudes. But that it does face a prima facie difficulty in accounting for facts about the expressibility of (existent) disagreements in ordinary disputes in the domains in question.

These I suggested to state as the presumption of contradiction:

In any ordinary, non-defective conversation it is common ground among the participants that utterances of (say) ‘Homer is funny’ and ‘Homer is not funny’ would contradict each other.

And I claimed that the presupposition of commonality approach I submitted is in good standing vis-à-vis accounting for these.

Now, as was stressed in discussion, it is quite compatible with a conversation presupposing commonality that (say) senses of humor be not actually shared (accepting is not believing), and this may well be known by participants. So it is compatible that a conversation being not defective that utterances of (say) ‘Homer is funny’ and ‘Homer is not funny’ need not contradict each other—and that this may be happening ever so often. And this may well be known (if the view is correct and known). But then it becomes a mystery why would it be common ground that things are otherwise.

Talking with Jose after the session, I think this is the last of the worries Elia was expressing. But if this is so, I think the response should be the one Jose was advancing on my behalf. Being part of the common ground does not require that people would believe that utterances of (say) ‘Homer is funny’ and ‘Homer is not funny’ would contradict each other, just that they would accept so. And this seems fine, if they are accepting that humor is shared. Is it not?


Lately I was agonizing over the phenomenon of retraction in MacFarlane’s relativistic framework. As widely known, for MacFarlane utterance-truth is relative both to a context of use and a context of assessment, the latter providing the values for the relevant parameters of the circumstances of evaluation of a given utterance. Also, as widely known (and accepted), there is a distinction between the truth of an assertion and its correctness – or, as MacFarlane calls it, its “accuracy” (I will use this term in what follows) – a distinction that he wholeheartedly accepts. However, given the connection between truth and accuracy, and the fact that truth for him is relative, accuracy will in MacFarlane’s framework be relative too: an assertion is accurate (as assessed from a context of assessment) if its content (as determined in the context of use) is true relative to the context of assessment from which it is assessed. Now, one of the phenomena that MacFarlane has argued gives relativism an advantage over contextualism (both of the indexical and the non-indexical sort) is retraction. Retraction is understood by MacFarlane to involve the acknowledgement that the content of a previous assertion is false (relative to the current context of assessment) usually accompanied by a formal act of “taking it back”. However, as he is keen to stress, retracting an assertion doesn’t mean that the speaker was “at fault” when she made it – that is, the assertion was accurate, albeit (relative to the current context of assessment) false.

This all seems nice. What I don’t understand, however, is how MacFarlane could make the last claim, given the other things he endorses. In other words, I see a tension between the claim 2 and 3, assuming 1:

1. Truth-relativism is correct: the content of assertions is true relative both to contexts of use and to contexts of assessment.
2. Like truth, accuracy is relativized to contexts of assessment: an assertion is accurate if its content is true relative to the current context of assessment.
3. Retraction is the acknowledgment that the content of a previous assertion is false (relative to the current context of assessment), but retracting a previous assertion doesn’t mean that the asserter was “at fault” – i.e., her previous assertion was accurate.

Suppose one retracts an assertion whose content is false (at the current context of assessment), but was true when made. Given 1, I don’t see how MacFarlane could claim both 2 and 3: because the connection between truth and accuracy, any assertion that is retracted is therefore inaccurate (relative to the current context of assessment).

What am I missing?


At yesterday’s session of the LOGOS Colloqium, Juan Comesaña argued against inductivisim, incompatible with what he took to be three also very plausible basic principles. Like others in discussion, I didn’t see clearly what was the intuitive motivation for one of them, Ruling Out Principle (details below). But in any case it seemed to me that the other two rule it out, thus independently of the issue regarding Inductivisim. The particular counterexample I submitted didn’t quite work, as I was using a necessary truth, but the point remains, or so it seems to me. At least on the antiskeptical assumption that one is ever justified in believing anything. Let A be one such contingency. So:

1. One is justified in believing that A.

Now consider the (contingent, known) consequence (A v (A v ¬E)), where E is one’s total basic evidence. By Closure,

2. One is justified in believing that (A v (A v ¬E)).

By Only Basic Evidence Justifies we then have

3. E justifies (A v (A v ¬E)).

If I understand the relevant notion of a way for something to be true, A’s being true and E’s being false are two ways for (A v (A v ¬E)) to be true, E rules out one but not all of these. Thus

4. E only rules out one way for (A v (A v ¬E)) to be true


If E only rules out one way for P to be true, E doesn’t justify P.

i.e., the Ruling Out Principle.

Am I wrong?

6 com

Regarding the discussion as to whether it is a condition for acceptance of T2 that T2 explains WHY T1 was successful, when it was.

1) I think we should distinguish strong and week explanations of this kind.

strong: Rel Mech explains why Class Mech was successful when it was, namely, because, roughly, the laws used in successful predictions of CM coincide with the laws used in predictions of RM for medium bodies and low speeds

weak: Helioc explains why Geoc was successful when it was, e.g. in predicting the apparent trajectories of the planets. How? well, SUPPOSING that the world is like Helio says, then GIVEN the theoretical claims of Helio we can infer that they have these and those true empirical consequences. But this is a very weak sense I think totally irrelevant at all for the realism debate

There may be cases in between, maybe Phlogiston

2) Some of you seemed to defend that Darwin explains WHY Nat Theol was successful, when it was. It is clear that the issue is NOT whether darwin explained WHAT Nat Theol (claimed to) explain. But WHY, i.e. whether if the world is like what darwin says, then Nat Theol succeeded in some “pre(retro)dictions”. I think that in this sense, this case is at the very bottom of the strong/weak divide in 1)
Is there any concrete example of biol phenomenon to which the strong sense (or at least stronger than the weakest one) in 1) applies in this case?

3 com

At yesterday’s LOGOS Colloqium, Anandi presented her thoughts on semantic zombies. As I understood it, the main element in the considerations had to do with the fact that, for a number of expressions, whatever in the thoughts, practices, and experiences of language-users could determine meaning, it does not determinate one in particular out of several admissible, equally natural candidate referents. Now according to the view of vagueness as semantic indecision, this is precisely what constitutes vagueness (or, anyway, the familiar kind of semantic indeterminacy, if one wishes to reserve ‘vagueness’ for the particular case of this associated with, say, sorites-susceptibility and the like). The worry I had was that then the candidate “semantic truths” S that allegedly are not apriori scrutable from the physical (plus the “that’s all” clause) would be either not (super-)true, corresponding to particular precisifications, or else such that the claim that they are not apriori scrutable cannot be grounded on the “inscrutability arguments”. What do people think?

5 com

At the last LOGOS Seminar, Delia introduced her notion of underarticulation, to be used in characterizing the phenomenon semantic underdetermination while trying to avoid a domestication policy of reducing it to more familiar phenomena, notably “indexicality”, broadly conceived. Her definition was something like this (I don’t have my notes in Idaho):

S is underarticulated iff there is a context C such that some other sentence S’ that “articulates” S would have the same truth-conditions in it.

where the notion of a sentence “articulating” another is left somehow intuitive, but so that ‘I’m sitting’ is articulated, in the relevant sense, by ‘The utterer of that sentence is sitting’.

Delia herself remarked that, as alluded in the example, indexical sentences can be underarticulated, in her sense. As observed by Genoveva, underarticulation in Delia’s sense, seems to be quite a general phenomenon, provided that, for instance, ‘indeed I’m sitting’ or ‘gee, I’m sitting’ would have the same truth-conditions as ‘I’m sitting’. That, as stressed by both Genoveva and Max, may give rise to legitimate worries in connection with the prospects of the notion vis-à-vis the characterization of what (allegedly) is peculiar to semantic underdetermination.

Be it as it were, Manolo M and Stephan offered cases sustaining the claim that, even if quite general, underarticulation was not a completely general phenomenon:

(a) This is five words long.
(b) This is articulated enough.

On reflection, I don’t see why the following wouldn’t be articulations of them, in Delia’s sense:

(a’) The relevant utterance of (a) is five words long.

(b’) The relevant utterance of (b) is articulated enough.

Any thoughts?

2 com

I will try to articulate better the objection I was making at the Q&A part.

In order to characterize Bert’s situation, vindicating in so doing realism about subjectivity and rejecting universality, Giovanni wants a fact that is only “otherpersonally” the case, while it is not the case “at me” said from Bert’s perspective, which is why he cannot know “it”. The asymmetry between what is the case “at one” vs. “otherpersonally” is supposed to be understood along the lines of the asymmetry between what is the case presently vs. at other times, in a seriously-tensed metaphysics. Now, my problem is that, on the basis of Giovanni’s examples or others I can think of, I cannot understand how the operators work so as to establish what Giovanni wants about Bert’s case.

Giovanni characterizes a fact of the intended sort as: there is no instance of PR. Now, this can be understood in two ways. I will successively express my difficulties with both readings.

(i) It can be understood as a temporal fact, that there (presently) is no instance of PR. But, to the extent that I have any grip on the operators “at one/me” vs. “otherpersonally”, this does not distinguish Bert from me, say: it is also not presently the case “at me” (let us assume) that there is no instance of PR, while it is “otherpersonally” the case that there is PR (it is the case “at” other subjects). But I assume that any knowledge argument essentially purports to distinguish Bert (and Mary) from me: it is irrelevant that Bert and me share the contingent fact that no instance of PR is going on in us, the point is that I am in a position to know facts involving PR, while Bert is not. (Incidentally, it is also irrelevant for the argument whether instances of PR are going on in other subjects; for we can run the argument with respect to in fact never-instantiated-Humean-shades-of-blue.)

(ii) Alternatively, we can think of the fact that there is PR not as a temporal one, but as a general, atemporal one: “at Bert” it is not just the case that, at the relevant moment, there is no instance of PR; it is rather that, “at him”, there is no PR (the property, not just its instances, somehow does not even exist for him). The problem with this is that I find obscure what the relevant notion of variable existence for properties is, and in particular what it means that a property exist or not “at a subject”. What sense does it make to say that “at me” properties with which I am not acquainted (perhaps theoretical properties to grasp which I lack the required cognitive equipment) do not exist? We can say, metaphorically, that they do not exist for me; but I guess this just means what I said, that I cannot conceive of them. More in general, I fail to see how the operators “at one”/”otherpersonally” can meaningfully operate on general facts, such as whether a property exists, or whether it is the property typically instantiated under such-and-such circumstances.


Today at the LOGOS Seminar Giovanni presented a very interesting paper. As the discussion was so rich, I didn’t get to ask my second question. Here it comes. I wasn’t clear I completely followed the dialectics. Suppose I genuinely exemplify phenomenal greyness, whereas you guys merely other-personally exemplify it. Suppose however that God exemplifies every phenomenal property. Thus (Universality) is false and (Realism about Subjectivity) is true. Suppose however that I am Bert. Doesn’t the same problem arise?

8 com

Towards the end of our RG there were some points raised concerning underdetermination arguments and compositionality. I was a bit baffled and I tried to clarify a few things for myself.

First, I think it is a bit misleading that Szabó calls the argument against compositionality that he presents, the underdetermination argument. This is a different argument than what is usually called such in the literature. I found this a bit confusing.

Here is a reconstruction of what Szabó calls the underdetermination argument against compositionality (page 265 last paragraph). (I talk of utterance content and not speech act content in order to simplify things).

The sentence (S) ¨The leaves are green¨ is uttered in two different contexts. In one is evaluated as false, in the other as true. The two utterances of the same sentence have different contents in the two contexts (assuming that the circumstances of evaluation are the same, which is not uncontroversial).  But, the constituents of the utterances have the same contents in both contexts. That is, the constituents ¨leaves¨, ¨green¨, ¨the¨, ¨are¨ have the same contents in both utterances of S in the two contexts. If this is so, the contents of the two utterances of S are not compositionally determined out of the contents of the utterances’ constituents. This is the sort of argument that shows a failure of compositionality: two utterances with different contents but with the same syntactic structure and whose constituents have the same content.

The argument can be put along the following lines:

1) the content of utterance u of S in C is different than the content of utterance v of S in C’: Cu ≠ Cv

2) The contents of the constituents of the two utterances, u and v, are the same: C(green)v = C(green)u, C(leaves)v = C(leaves)u, and so on for all the constituents of the two utterances of S.

3) the two utterances have the same structure.

Szabo´s solution to this argument was to show that the difference in the contents of the two utterances u and v is due to, and can be traced to, differences in the contents of (some) its constituents. The utterance u of S has a different content than utterance v of S because some of the constituents of u have a different content than some of the constituents of v. The content of ¨are green¨ in the utterance u is different than the content of ¨are green¨ in the utterance v. If so, compositionality is preserved and the argument against it defused.   Szabó´s way of defusing the argument is to show that (2) doesn´t hold. (I think what’s really important is not the particular way in which Szabó does this, as much as the template he offers: to show that there is no failure of compositionality just show that (2) doesn’t hold and back up your claim with intuitions regarding the difference in contents of u and v’s constituents).

Now, two points were raised: (a) that (someone like) Travis can reject (2) and still run an underdetermination argument against compositionality and (b) that rejection of (2) and acceptance of Szabó’s way of defusing the argument is compatible with Travis’ contextualism.

As far as I can see (b) is correct but that’s not the case with (a).

If one wants to make an argument against compositionality one must accept (2) (in fact, must accept (1), (2) and (3)). Without (2) there is no argument against compositionality. To show that compositionality fails one needs to show that there are cases where our intuitions match the situation described in (1) through (3). That is, we must have the intuitions that two utterances u in C and v in C’ of S have different contents. We must also have the intuition that the constituents of the two utterances have identical contents (that is, any of the expressions uttered in u and v has the same contents in both utterances). And finally, we must have the intuition that u and v have the same structure. Finding an expression that satisfies all three intuitions is finding an example of failure of compositionality.  As long as one wants to run an argument against compositionality (of content) one cannot give up (2).

On the other hand accepting Szabó’s way of defusing the argument (by rejecting (2)) is compatible with contextualism (i.e. the thesis that the contents of u in C and v in C’ of S are different and nothing in the lexical meaning of S determines the difference in contents). Even if denying (2), Travis (& Travis-like people) can run the following argument:

1’) two utterances u and v of the same sentence S in two contexts C and C´ have different contents.

2’) The lexical meanings of S’s constituents are identical in both of its utterances in the two contexts.

3’) the two utterances share the same structure.

Therefore the contents of the utterances u and v of S are underdetermined by the lexical meanings of S´s constituents. Or, if the lexical meanings of constituents compose into the lexical meaning of S, one can reformulate as: the contents of utterances u and v of S (in C and C’ respectively) are underdetermined by the lexical meaning of the sentence S.

This argument is an argument about the relation between the lexically encoded meaning (contextually insensitive meaning) and content (utterance content).It supposedly shows that the lexical meaning of a sentence fails to determine the (truth conditional) contents of the utterances of the sentence. So, it should be distinguished from what Szabó calls the underdetermination argument against compositionality which is an argument about the compositionality of content. (notice that (2) and (2’) are different)

All in all, I see no reason why (b) should be false: one can accept that contents of utterances of S are compositionally determined out of the contents of their parts and accept that the contents of utterances are nevertheless underdetermined by  the lexically encoded meaning of S.



tag cloud