Propositions and higher-order attitude attributions

Abstract

An important objection to sententialist theories of attitude reports is that they cannot accommodate the principle that one cannot know that someone believes that p without knowing what it is that he believes. This paper argues that a parallel problem arises for propositionalist accounts that has gone largely unnoticed, and that, furthermore, the usual resources for the propositionalist do not afford an adequate solution. While non-standard solutions are available for the propositionalist, it turns out that there are parallel solutions that are available for the sententialist. Since the difficulties raised seem to show that the mechanism by which sentential complements serve to inform us about attitudes and about sentence meaning does not depend on their referring to propositions, this casts doubt on whether talk of propositions should retain a significant theoretical role in the enterprise of understanding thought, language and communication.

Keywords::

• propositions
• sententialism
• propositional attitude reports
• semantics
• modes of presentation

Notes

^1. I draw on the account in (Ludwig 1998). I assume attitude reports are relational, but this is also a presupposition of the puzzle that I want to explore.

^2. For convenience I focus on ‘that’-clauses. With some more circumlocution the discussion can be extended to other sentential complements which encode sentences without exhibiting them fully in surface structure.

^3. See (Schein 2012; Ludwig 2010) for overviews.

^4. This point has been urged also by (Burge 1978; Higginbotham 1991, 2006; Seymour 1992).

^5. See (Ludwig 1998) for a list with replies; for a different approach, see (Higginbotham 2006).

^6. I have in mind the equivalence relation among utterances that propositionalists have in mind when they are willing to say that the utterances express the same propositions. This does not signal or presuppose commitment to propositions.

^7. Cf. Kripke 2013, 258–261. Kripke says it must be ‘revelatory,’ showing what the referent is, but the cash value in this case is that grasp of the mode of presentation must suffice for grasp of what it presents.

^8. Perhaps it is exactly acquaintance that provides what the Fregean needs. Here is a suggestion by Kripke:

My suggestion … is that Frege, like Russell, has a doctrine of direct acquaintance. Every time we determine a referent, we are introspectively acquainted with how the referent is determined, and that is the corresponding sense. And our introspective acquaintance with this sense gives us a way of determining it, and of referring to it, and this is the indirect sense. (Kripke 2011, 271)

The idea is that if we think that p, which on Frege's view refers to a truth value, we are acquainted with how we determine it. This ‘how we determine it’ is to be identified with a mode of presentation of the truth value, i.e., the thought expressed by ‘p’, i.e., its sense. Being acquainted with how we determine the truth value of the thought that p ‘gives us a way of determining it’ in turn, which is a mode of presentation of the thought (proposition) itself.

It is not quite clear how we are to think about acquaintance giving us a way of determining what we are acquainted with. I think the idea is that acquaintance itself is a kind of thinking about an object, and so something of the sort that plays the right role. I will assume this is the intent. Then either acquaintance with the proposition is mediated by a mode of presentation or it is not. Suppose it is mediated by a mode of presentation (and put aside the worry that this undercuts the idea that it is acquaintance that is involved). Then the mode of presentation must be sufficient for grasp of its object or it will fail to meet the needs of the Fregean. But if the relevant notion of acquaintance involves a mode of presentation that must be sufficient for the grasp of its object, appealing to it does not explain how a mode of presentation could be sufficient to grasp a thought, but instead presupposes it. Calling it acquaintance doesn't help. Suppose then that it is not. Then, even if it suffices for grasp of the proposition, it does not in fact provide a sense that suffices for grasp of its object, for its object is not thought about via a mode of presentation at all. One could say: but can't we say that the sense is given by the phrase (relativized to speaker and time) ‘the sense that I am now acquainted with’? Even so: grasp of that does not suffice for grasp of the thought one is acquainted with: acquaintance (whatever that is) is what does the trick. (For I could grasp the sense of that expression in someone's mouth without grasping the thought it refers to.) So we have still not found any account of a mode of presentation grasp of which suffices for grasp of its object.

Why can't one just say: acquaintance suffices for grasp and is a mode of presentation! The foregoing argument just assumes that if a mode of presentation is involved, it must be something independent of acquaintance because indirect, but we should instead extend our notion of a mode of presentation to cover any way of thinking about an object, even thinking about an object directly. Then if there is a way of thinking about an object (acquaintance) that suffices for grasp of it, we have got what the Fregean, or, at any rate, the propositionalist needs. However, acquaintance is not suitable as a sense to be assigned to an expression in a public language. Obviously, assigning the relation (as a type) to an expression is no help, for to understand ‘x is acquainted with y’ is not ipso facto to be acquainted with y. And token relations of acquaintance between subjects and objects aren't suitable at all. One might give a rule: the sense of ‘that p’ as uttered by x at t is the token acquaintance relation obtaining between x at t and that p. But what would it be to ‘grasp that sense’ (if it is to suffice to grasp its object) except to stand in the relation? But then only x could in principle grasp the sense, and his interlocutors would be at a loss. The situation is not improved if one selects any other pair < y, t>. In addition, to come to stand in the relation, one would have to independently figure out what proposition was being referred to and how to grasp it. Appeal to acquaintance in this way does not engage with any mechanism the public language could use to put one in touch with the right objects.

^9. In chapter 8 of (Peacocke 2008), Peacocke offers what might be thought to be a way of satisfying the requirement in the idea of a canonical concept of a concept F (can(F)). The idea is that there are ways of referring to concepts that uniquely fix them, namely, by way of their individuative application conditions (The Leverage Account). The general idea is expressed in (*).

(*) For an arbitrary concept C to fall under can(F) is for the fundamental condition for something to fall under C to be the same as the fundamental condition for something to fall under the concept F. (291)

Here we imagine for particular cases that ‘the fundamental condition for something to fall under the concept F’ to be replaced by a statement of the condition. For example, if ‘F’ =  ‘red’, then: For an arbitrary concept C to fall under can(red) is for the fundamental condition for something to fall under C to be being red. To put it another way, can(red) =  the concept of being a concept the fundamental condition for falling under which is being red. Here the concept of red is deployed in the specification of the condition, and so in deploying the concept can(red) one deploys the concept of red, as in deploying the concept of a red ball one deploys the concept of red, though in the former unlike the latter the concept applies to the concept of red as well. The purpose of this is to describe a way of generating a hierarchy of concepts of concepts by starting with grasp of a first order concept, and tacit knowledge of (*). There is one more thing that we need to add to the story, for what we want is that one thinks about a concept, grasps it, and thinks about it as the one grasped. What is needed for this is that one know that if condition F is the fundamental condition for falling under a concept C, then ‘F’ expresses C, or the concept one deploys in thinking of the condition F is the concept C. Note that the total effect this is to prompt one to deploy the concept of red and to think of the concept that one there deploys. We are in effect given instructions of the following sort: think of the concept the deployment of which is required in thinking of the condition of being red. This is a clever idea. Does it supply what the Fregean needs?

What is the semantics for ‘the condition of being red’, for this gives us the sense attached to it. How more specifically does ‘being red’ pick out the right property or condition? It picks out the property attributed or the condition specified by ‘red’ in English. But we don't want that to be part of the specification of the concept can(red), because it would make it in part metalinguistic. Moreover, it is clear that using ‘red’ here as a way of specifying the condition or property is crucial for ensuring that the concept of red is deployed in deploying can(red). Here is a solution: We can say instead that ‘being red’ refers to that property an object is fundamentally required to have in order to fall under the concept red. Then we avoid the appeal to any metalinguistic element! But now we have another problem. We are now using in the specification of can(red) a term that refers to the concept red (and that is what can(red) is supposed to enable us to do). How does ‘the concept red’ pick out the concept red? We could say that it picks out the concept that ‘red’ expresses. But this gets a metalinguistic element back into the content. So we could say that it picks out that concept the fundamental condition for falling under which is being red. But this reintroduces the problem we started with. We could appeal to a mode of presentation of the concept red that suffices for its grasp and attach that to ‘the concept red’. But that was what the proposal was supposed to supply us with. The fundamental problem has not actually been avoided, but like the bump under the rug, it has been relocated. Perhaps there are other moves to be made here, but perhaps we do not need to make them. At least, that is what I will suggest in sections 6 and 7 (for the propositionalist) and in section 8 (for the sententialist). We can of course think about a thought that we are entertaining, and we do so by understanding the complement of an attitude report while understanding that it is to refer to what we thereby grasp or understand. In higher-order attitude reports, we need to say that when attributed using that-clauses for content positions, the person to whom we are attributing them grasps what it is that they refer to, or something equivalent, in thinking about them. (There is something similar to this in Peacocke's own suggestion about the semantics of higher-order attributions (see 307).) More details below.

^10. In his ‘Postscript to “Belief DeRe”’ (Burge 2007, 65–81), Tyler Burge says that a thought about a thought, e.g., the thought that snow is white is true, specifies the thought ‘in the that-clause way,’ where this is de re and ‘the de re reference feeds directly off immediate understanding of representational contents, the res’ (70). Of this case, he says that ‘the representational thought contents that carry out the dere reference are completely conceptualized’ and so ‘there is a striking relation to a re that goes beyond merely conceiving of it or forming a concept that represents it’ (70). Burge goes on to say: ‘That-clause forms of representation in thought are individual concepts. They are complex structure- and content-specifying concepts when they name whole representational thought contents. … Mastery of such an individual concept, of either sort, requires comprehending the representational content that the individual concept names’ (70–71). There is, however, something not completely conceptual about the relation to the thought content because it involves ‘comprehending the re, not merely conceiving of it’ (71), and it is for this reason that Burge says that it is a de re and not a de dicto thought. And, he says, ‘This form of dere representation is possible only for res that are themselves representational contents’ (71). Thus, the account appears to hold that there is a mode of presentation (or way of conceiving) of an object that has representational content, namely, a thought content, that suffices for its grasp. Does this help us to understand how a way of conceiving or presenting a thought content could suffice for grasp of it? It does not. It merely describes the idea that we have been trying to make sense of: a way of presenting a proposition that suffices for grasp of it, that is, a mode of presentation of a proposition p grasp of which suffices for the grasp of p (an individual concept whose mastery requires comprehending the content it names). This does not respond to the independence argument, and it does not engage with the fact the mechanism by which that-clauses secure their referents go through our understanding the sentences that appear in them. One could, as I have noted, insist that since the theory requires such individual concepts, they exist, but this is an ad hoc defense of the theory, and as there is an alternative account, we should feel no pressure to adopt it.

^11. It might be objected that ‘that p’ is syntactically complex while ‘Bob’ is not. But, first, we can assign a complex sense to a simple expression, and, second, we could, in any case, introduce a complex expression whose components are not the words in that appear in ‘p’.

^12. Davidson remarked, ‘If we could recover our pre-Fregean semantic innocence, I think it would seem to us plainly incredible that the words “The Earth moves”, uttered after the words “Galileo said that”, mean anything different … than is their wont when they come in other environments’ (Davidson, 2001, 108). I think this is right, but I think we can say something stronger: if we did not understand those words in their usual sense, what Galileo said would remain opaque to us.

^13. Higginbotham (2006, 110–112) offers a response on behalf of the sententialist to Schiffer's objection. If I understand it correctly, it is that the relevant matching-in-content relation the sententialist needs can be construed so that ‘that Galileo believed that the earth moves’ does not stand in it to, in Higginbotham's phrase, its target truth conditions, as given by the analysis. This would amount, I believe, to rejecting premise 4 in the argument. Perhaps the suggestion I make in the text is a version of what Higginbotham has in mind, for it likewise rejects premise 4. However, it works by treating the position of the complement as sensitive not only to the referent but also the term used to refer to it, and Higginbotham's suggestion appears to be that it is the relation between the referents alone that does the work. Higginbotham does not elaborate, however, and it remains unclear to me how he intends the relation and the relata to be understood so that the right result is obtained.

Additional information

Notes on contributors

Kirk Ludwig

Kirk Ludwig is Professor of Philosophy at Indiana University Bloomington. He has published numerous papers on the philosophy of language, mind, action and epistemology. He is the editor of Donald Davidson (CUP 2003), co-author with Ernie Lepore of Donald Davidson: Meaning, Truth, Language and Reality (OUP 2005) and Donald Davidson's Truth-theoretic Semantics (OUP 2007), and co-editor with Ernie Lepore of the Wiley-Blackwell Companion to Donald Davidson (2013).