TALK TALK:
A UNIFIED FREGEAN THEORY OF QUOTATION
Cappelen and Lepore[1] have laid down the gauntlet for opponents of Davidson’s paratactic analysis of indirect quotation[2] and their own unified Davidsonian account of quotation more generally. Rather than merely taking potshots at the Davidsonian picture, they have challenged their interlocutors to provide an alternative general account of reported speech and to compare it to the unified Davidsonian account with respect to simplicity and elegance. I mean to take up this challenge.
I: Indirect Quotation – Theories and Principles
There are three central principles that seem to be widely endorsed by philosophers engaged in the indirect quotation literature: semantic innocence, compositionality, and opacity. Consider the following example:
1. Mary said that Hesperus is roughly spherical.
According to the principle of Semantic Innocence, the meanings[3] of expressions do not vary systematically with the type of sentential context in which they occur.[4] So, for example, the meaning of ‘Hesperus’ in,
2. Hesperus is roughly spherical,
would be the same as its meaning in (1). The force of semantic innocence is to rule out theories which invoke Fregean reference shift[5] in quotational and other propositional attitude contexts.
According to Compositionality, the meaning of a sentence (or utterance) is wholly determined by the meanings of its subsentential constituents and its grammatical structure. An implication of this principle is that the product of replacing a component of a sentence with an expression having the same meaning is a sentence which means the same as the original (assuming, of course, that the grammatical structure remains unchanged). So if, for example, ‘the evening star’ had the same meaning as ‘Hesperus’ (at least in quotational contexts), compositionality implies that
3. Mary said that the Evening Star is roughly spherical,
would have the same meaning as (1).
According to the principle of Opacity, co-extensive expressions are not in general substitutable salva veritate in propositional attitude contexts. This implies that the product of replacing a component of a sentence with an expression having the same extension often will be a sentence which differs in meaning from the original.[6] As a result, according to the principle of opacity,
4. Mary said that Phosphorus is roughly spherical,
may well differ in meaning from (1).
What is interesting to note is how difficult it is to develop an account of indirect quotation, and propositional attitude reports more generally, that satisfies all three of these principles. In fact, three of the most prominent approaches to indirect quotation – direct reference, Davidsonian parataxis, and Fregean reference shift – satisfy only two of the principles, differing in which of the three they reject.[7]
The Direct Reference view combines a compositional account of the meanings of indirect quotations with a semantically innocent account of the contribution of complement sentences and their constituents.[8] Paradigmatically, direct reference theorists take ‘said’ and other propositional attitude verbs to be two-place predicates satisfied by pairs consisting of speakers and propositions.[9] We might, slightly more formally, represent the logical form of the schematic sentence ‘A said that p’ as
5. Said (A, that p).
The proposition named by the ‘that’-clause (‘that p’) just is the proposition expressed by the complement sentence (‘p’) in ordinary extensional contexts, thus securing innocence. And given that the proposition expressed by the complement sentence is determined by the meanings of its subsentential components, if it has any, compositionality is secured. But, as should be obvious, this picture violates opacity. Since, for example, ‘Hesperus’ and ‘Phosphorus’ are coreferential, (2) and
6. Phosphorus is roughly spherical,
express the same proposition. As a result, according to the direct reference theory, (1) and (4) contain coreferential ‘that’-clauses, thereby ensuring that they have the same meaning.
The Davidsonian Paratactic view satisfies innocence and opacity at the cost of compositionality. On this picture, indirect quotations – single sentences at the level of surface grammar – are two distinct sentences at the level of logical form. One sentence expresses the proposition that a relation obtains between a speaker and an utterance picked out demonstratively; the other is used to produce an utterance that can be picked out demonstratively by the first. The logical form of ‘A said that p’ on the paratactic account is as follows:
7. Said (A, that ). P.
Since the demonstrated utterance is not produced by the person to whom speech is being attributed, Davidsonians typically give a further analysis of indirect quotation in terms of a “samesaying” relation that (putatively) holds between the utterance produced by the speech reporter and a prior utterance of the person to whom speech is attributed. The Davidsonian analysis of our schematic sentence would be
8. $u(Said ( A, u) & Samesays (u, that)). P,
where the existential quantifier ranges over token utterances. On this view the complement sentence occurs as a separate utterance from the speech attribution in an ordinary extensional context. As a result, semantic innocence is trivially satisfied. And since, for example, an utterance of (2) could stand in the samesays relation to an utterance of Paul’s while an utterance of (6) does not, opacity is satisfied as well.[10] But since the meaning of the speech attribution is not, on this view, a function of the meaning of the complement sentence, compositionality fails to be satisfied.[11]
The Fregean Reference Shift approach – the one that I favour – satisfies compositionality and opacity at the cost of innocence. Paradigmatically, Fregeans take ‘said’ to be a two-place predicate satisfied by pairs consisting of speakers and Fregean thoughts (that is, the sense expressed by a sentence). We might represent the logical form of ‘A said that p’ as
9. Said (A, s(‘p’)),
where ‘s(‘p’)’ names the sense expressed by ‘p’ in extensional contexts. It is worth noting that, broadly construed, the reference shift approach to indirect discourse (may but) need not invoke Fregean senses. What is important is that the meanings of complement sentences differ systematically in quotation contexts from their meanings in extensional contexts. Nonetheless, the meanings of complement sentences are compositional (as are the meanings of the larger speech reports of which they are a part): they are wholly determined by the meanings (in quotation contexts) of their constituents and their grammatical structure. Moreover, opacity is secured by to the (putative) fact that coextensive expressions need not express the have the same meaning in quotation contexts.
II: A Cure for Onto-Phobia
Recently, Fregean theories of propositional attitude contexts seem to have fallen into disfavour. The direct reference theory is the reigning paradigm for belief contexts, while the Davidsonian paratactic theory dominates discussion of (direct and indirect) quotation contexts.[12] One reason for this is, of course, the devastating criticism to which Kripke subjected descriptivism in Naming and Necessity.[13] But since (i) the reference shift strategy does not require that one endorse descriptivism and (ii) it is far from clear that Kripke’s arguments apply to expressions occurring within propositional attitude contexts, I am going to ignore such considerations. Another reason, which I do want to address, is the fear that Fregeanism brings with it a commitment to the existence of intensional entities – mysterious “creatures of darkness.”[14] Ludwig and Ray, for example, go so far as to make it a desideratum of an adequate theory of quotation (and other opaque) contexts that it not quantify over such entities.[15]
What I want do here is to argue that the fear of the intensional implications of Fregeanism is, in a sense, ill-founded. The reason is that intensional entities can be viewed as the products of the commitment to two distinct and orthogonal theses: that there is an intensional typing schema for utterances (and belief states, etc.), and that types are abstract entities. Let me elaborate. What one does by means of indirect quotation is report actions and, in particular, speech acts. And in reporting actions, while one does typically name the agent, one only rarely names the (token) action. Instead, what one does is describe the type of action the agent performed.[16] When reporting speech acts, part of what one does is specify the type of illocutionary act that has been performed; while our focus here is on acts of assertion, many of the same issues arise for direct and indirect quotations of questions, commands, promises, etc. In addition, one specifies the type of utterance the agent (or speaker) has produced in the performance of this illocutionary act. Exactly how one goes about typing a speaker’s utterance will depend both on what one knows about the utterance, as well as the conversational purposes for which the speech report is being made. And it is worth noting that, in speech reports, utterances are typically not typed in terms of the intrinsic features of the noises (or inscriptions) produced – the semantic properties of such noises, and even their standing as words, are relational.
In indirect quotation, utterances are typed intensionally. This is, of course, not very informative: to be typed intensionally is just to be typed in terms of substitutability in indirect quotation (and other propositional attitude) contexts. One thing that I think can be said with a fair degree of confidence is that intensional typing is at bottom epistemological. The reasoning underlying this suggestion is that (i) failures of substitutivity in indirect quotation contexts are of a piece with similar failures in belief contexts and (ii) the explanation of this latter phenomenon is a lack of knowledge of certain facts on the part of the subjects of belief attribution. This will be developed more fully in the discussion of my positive view below.
As I have been casting things, what a commitment to intensional entities amounts to is a commitment to intensional types. And what I want to suggest is that what sorts of entities this commits one to depends upon one’s theory of types. If one holds that types are abstract Platonic entities standing outside of space-time, then one will hold the same view of intensional entities. But one might instead hold that types are immanent universals, or classes of abstract or concrete particulars, or what have you. The central point is that a Fregean approach to indirect quotation need bring with it no new ontological commitments beyond those stemming from one’s background metaphysical views.[17] Of course, one might object to the existence of types altogether, however they are construed. But short of some such more radical thesis, the sort of Fregeanism I am appealing to here is ontologically neutral. It is worth noting that this is something of a departure from Frege’s own account. For Frege, senses were objective, abstract entities grasped in thought, and (capable of being) expressed by means of utterances, rather than types of utterances, as I would have it. Where Frege typed utterances in terms which sense they stood in the expressing relation to, I am suggesting that we eliminate this second relatum altogether and identify intensional entities with the types themselves. This enables us to deploy the reference shift approach to opacity without either retreating to quotationalist accounts of indirect quotation or facing ontological disaster.
III: Neither Guilty nor Innocent
Another reason Fregean theories are unpopular these days is their rather blatant failure to preserve semantic innocence. As above, one of the defining characteristics of Fregeanism is that there is a systematic difference between the meanings of sentences when they occur in extensional contexts and when they occur in propositional attitude contexts. What I want to do in this section is to present some reasons for being suspicious of innocence. My claims here, however, are meant to be more suggestive than decisive.
The first thing to note is that there are (at least) two distinct uses of quotation sentences – let’s call them “attributive” and “justificatory” uses. In my view, a plausible case can be made for thinking that, in justificatory uses, complement sentences occur innocently. But I want to suggest that (i) insofar as we want to view quotation as of a piece with propositional attitude contexts more generally, our primary concern is with attributive uses, and (ii) it is far from clear that both uses deserve uniform treatment.
Consider the occurrence of the sentence,
10. Peter is the most talented member of our trio
in the following pair of dialogues:
Dialogue I:
Mary: Why does Peter deserve a higher percentage of the profits
than you or I do?
Paul: Peter is the most talented member of our trio.
Mary: Why should I believe that?
Paul: Bob Dylan said so.
Dialogue II:
Mary: Why does Peter deserve a higher percentage of the profits
than you or I do?
Paul: Bob Dylan said that Peter is the most talented member of our trio
In Dialogue I, (10) occurs in an extensional context in which it is used to make an assertion (regarding, in particular, the relative talent of Peter). The point of invoking Bob Dylan’s utterance is entirely justificatory – it is an appeal to authority meant to provide evidence of the truth of the earlier assertion. Dialogue II contains an example of what I mean by a justificatory use of an indirect quotation sentence.[18] It is, in effect, an abbreviated or succinct version of Dialogue I. By means of his use of the single indirect quotation sentence, Paul communicates what it took two sentences to convey in the longer dialogue – that Peter is the most talented of the three and that Bob Dylan has given testimony to this effect. Now given the coincidence of conveyed information and the extensional occurrence of (10) in Dialogue I, I might well concede that (10) occurs innocently within the quotational frame in Dialogue II. Moreover, Paul’s quasi-demonstrative reference to his utterance of (10) in Dialogue I might even yield grist for the Davidsonian mill when it comes to justificatory uses of indirect quotation.
But consider now an attributive use of the very same indirect quotation sentence in the following dialogue:
Dialogue III:
Mary: Why do you think Bob Dylan is going to invite Peter to
perform on his summer concert tour and not you or me?
Paul: Bob Dylan said that Peter is the most talented member of our trio
Despite my concession regarding the innocence of the complement clauses of quotation sentences in their justificatory uses, there are reasons to suspect that attributive uses ought to receive different treatment. As a result, arguments for innocence in the former case do not simply carry over to the latter. The first thing to note is that the central subject matter differs as between attributive and justificatory uses. In the former case, the subject matter is the person to whom speech is attributed – in Dialogue III, for example, the subject matter is Bob Dylan. In the latter case, however, the subject matter is (paradigmatically) the subject of the complement sentence – Peter, in Dialogue II. Moreover, in justificatory uses, the truth or falsity of the complement sentence is of central import. But in the case of attributive uses, the question of the complement’s truth-value need not even arise. In Dialogue III, whether or not Peter really is the most talented member of the trio is not even at issue. Finally, and most telling towards the case for separate treatment in attributive but not justificatory, uses of indirect quotation sentences, failure of substitutivity phenomena arise. Suppose that Paul and Mary know Peter to be an obnoxious jerk, and on this basis have nicknamed him ‘Jackass’. Moreover, suppose that Bob Dylan, who does not suffer fools gladly, is aware of neither the nickname nor the characteristics for which it is richly deserved. In Dialogue II, ‘Jackass’ could be substituted for ‘Peter’ salva veritate. Since how Bob Dylan conceives of Peter is not at issue in this dialogue, but the properties of Peter himself, any designative device which refers to him can be used in this context. In Dialogue III, in contrast, ‘Jackass’ could not be so substituted. Given Dylan’s ignorance of Peter’s character and corresponding nickname, the quotation sentence resulting from this substitution would mischaracterize Dylan’s speech act.[19]
Not only is the requirement of an innocent analysis of attributive uses of quotation sentences ungrounded in the innocence of their justificatory counterparts, such an analysis runs into difficulties in its own right. The problem is that, typically, speech reports occur in different contexts of utterance than does the speech of which they are reports. As a result, there will generally be systematic differences between what a sentence means when it occurs within a quotational frame and what it means when it occurs alone, in any given context of utterance. The kind of worry I have in mind can be illuminated by looking, first of all, at direct quotation. Consider the occurrences of the sentence
11. I am here now
in the following contexts of utterance:
Context I: Turcotte Hall,
Paul-a: I am here now.
Context II: Mary’s apartment,
Mary-a: I am here now.
Mary-b: Paul said, “I am here now.”
As should be obvious, the two
occurrences of (11) in Context II cannot
have the same meaning if the speech attribution to Paul is to be accurate.
After all, what it means in Mary’s first utterance (Mary-a) is that Mary was in
her apartment at 8 pm on July 15, whereas what Paul asserted in Context I was that he, Paul, was in
Turcotte Hall at 12 noon on July 10. The lesson is that, in the case of direct
quotation, there will at least be the following systematic difference in
meaning between sentences occurring in quotational frames and sentences
occurring alone: the meanings of the latter will be determined by features of
the actual context while the meanings of the latter will be determined by
features of the reported context.
Now, of course, this sort of example will not serve undercut the putative innocence of the complement clauses of indirect quotation sentences.[20] After all, Mary could use
12. He was there then.
to make an indirect report of Paul’s speech act. And accurate reporting does not require that the following two occurrences of (12) differ in meaning:
Context II: Mary’s apartment,
Mary-c: He was there then.
Mary-d: Paul said that he was there then.
But the speaker and the spatio-temporal location are not the only contextual features that play a role in determining the meaning of a sentence on an occasion of use. In addition, conversational presuppositions play a crucial role. Suppose that Paul and his conversants believe that Smith died as a result of his gunshot wound, while Mary and her conversants believe that, although Smith was shot, what killed him was poison.[21] Consider now the occurrences of
13. Smith’s murderer [whoever he is] is insane[22]
in the following contexts of utterance:
Context I: Presupposition – gunshot wounds killed Smith
Paul-b: Smith’s murderer [whoever he is] is insane.
Context II: Presupposition – poison killed Smith
Mary-e: Smith’s murderer [whoever he is] is insane
Mary-f: Paul said that Smith’s murderer [whoever he is] is insane
The denotation of Paul’s use of (13), in any given circumstances of evaluation, is the unique individual who shot Smith, if there is one. The denotation of (13) in Mary’s first utterance (Mary-e), in any given circumstances of evaluation, is, in contrast, the unique individual who poisoned Smith, if there is one. As a result, (13) will have to differ in meaning as between Mary’s two utterances if her speech report is to be accurate. An advocate of innocence might try to insist that correct speech attribution requires that Mary utilize
14. Smith’s shooter is insane,
rather than (13), but this seems to have little independent motivation.
The considerations raised here, even if they are right-headed, fall far short of showing that expressions refer to Fregean entities when they occur within quotational frames. A direct reference theorist could still maintain that sentences containing proper names and similar referring devices express Russellian propositions when they occur in such contexts, albeit distinct Russellian propositions from those express in extensional contexts. I do think, however, that the sorts of cases appealed to above put some pressure on the Davidsonian analysis. This issue will be taken up in section (VI) below.
IV: A Unified Fregean Theory of Quotation
Cappelen and Lepore have argued that a general theory of quotation should give a unified account of direct, indirect, and mixed quotation.[23] In direct quotation, the reporter quotes a speaker by means of mentioning the words the speaker used. An example would be the following:
15. Mary said, “Hesperus is roughly spherical.”
In indirect quotation, the speaker need not have used any of the words the reporter uses to quote her. An example would be,
1. Mary said that Hesperus is roughly spherical.
In mixed quotation, the reporter quotes a speaker by mentioning some of the words she used, as in,
16. Mary said that Hesperus is “roughly spherical.”
Moreover, they argue that direct and indirect quotation should receive distinct semantic treatment but that direct and mixed quotation, as well as indirect and mixed quotation, should receive overlapping semantic treatments.[24] What I am going to do in this section is develop a Fregean theory that satisfies Cappelen and Lepore’s desiderata.
The basic picture I am going to defend is one according to which ‘said’ is a two-place predicate satisfied by pairs of speakers and utterance types.[25] Quoted complement sentences and ‘that’-clauses are names of utterance types on this analysis.[26] And since the types they name are (typically) distinct from their meanings in extensional contexts, the view counts as a version of the reference shift approach. In order to secure compositionality, I treat these expressions as complex names, picking out types of whole utterances in virtue of the types of utterance fragments denoted by their sub-sentential components. Moreover, what distinguishes the various kinds of quotation are differences in the typing schemata at issue.[27]
i: Direct Quotation
Let us begin with a discussion of direct discourse. As a first pass, a sentence of the form ‘A said, ‘p’’ would receive the following analysis
17. Said ( A, T(‘p’))
where ‘T(‘p’)’ denotes the type named by ‘p’. Given that, on the unified Fregean analysis, expressions name different types in direct and indirect discourse, a more perspicuous analysis would be the following:
18. Said ( A, Td(‘p’))
where ‘Tw(‘p’)’ denotes the type named by ‘p’ in direct quotation contexts. So far so good, but we have yet to capture the compositional structure of complement sentences in our notation. What I propose to do is to use curly braces – ‘{….}’ – as word-order indicators. This will enable us to represent (15) as,
19. Said (Mary, {Tw(‘Hesperus’), Tw(‘is’), Tw(‘roughly’), Tw(‘spherical’)}).
Since, in (ordinary) direct quotation contexts, what words name are word-types,[28] the type of utterance denoted by the second relatum in (19) includes all and only those consisting of the words ‘Hesperus’, ‘is’, ‘roughly’, and ‘spherical’ in that order. And the speech report will be true just in case Mary produced an utterance of that type with the requisite illocutionary force.
One wrinkle in this so far rather simple story arises as a result of ambiguity. Consider the sentence,
20. I saw you with my glasses.
The meaning of an utterance of this sentence depends not only on the speaker and the person to whom the utterance is directed, but also on its grammatical structure and the meaning of the word ‘glasses’. In general, the meanings of particular utterances are determined by packets of contextual parameters including such things as the speaker and the audience, the time and location of the utterance, the demonstrata and/or anaphoric antecedents of various referring expressions, the purpose of the talk exchange and prior “moves” within it, shared presuppositions, speaker intentions, etc. As a result, an utterance of (20) can be disambiguated by indicating the requisite contextual parameters.[29] Consider now,
21. Mary said, “I saw you with my glasses.”
The meaning of an utterance the whole sentence will of course be determined by features of the actual context in which it occurs – I will use the subscript ‘a’ as a label for this packet of contextual parameters. And one thing the actual parameters will determine are the contextual parameters in terms of which the complement sentence is to be interpreted.[30] One possibility involves interpreting the complement in terms of the actual contextual parameters. In the case at hand, this can be represented as,
22. Said (Mary, {Twa(‘I’), Tw(‘saw’), Twa(‘you’), Twa(‘with’), Twa(‘my’), Twa(‘glasses’)}a).
Note: the subscript outside the parentheses indicates that the grammatical structure is to be interpreted in terms of the actual contextual parameters and the subscript of each word-type name indicates that the meaning of each token of the named type is a function of the actual parameters.
Of course, except in jest, no one would utter (21) intending the complement to be so interpreted. After all, Mary is unlikely to have asserted (20) using ‘I’ to refer to the person who subsequently reported her utterance. Instead, a reporter would presumably utter (21) intending the complement to be interpreted in terms of features of the context in which reported utterance occurred. Using the subscript ‘r’ as a label for the contextual parameters of the reported context, we get,
23. Said (Mary, {Twr(‘I’), Twr(‘saw’), Twr(‘you’), Twr(‘with’), Twr(‘my’), Twr(‘glasses’)}r).
In some cases, however, the reporter might be ignorant of some or all of the (relevant) features of the reported context. Using the subscript ‘o’ to indicate that the interpretation has been left open,
24. Said (Mary, {Two(‘I’), Two(‘saw’), Two(‘you’), Two(‘with’), Two(‘my’), Two(‘glasses’)}o)
would correspond to a case in which the reporter is making a claim only about the word order in Mary’s utterance. And,
25. Said (Mary, {Twr(‘I’), Twr(‘saw’), Two(‘you’), Twr(‘with’), Twr(‘my’), Twr(‘glasses’)}o)
would correspond to a case in which the reporter makes no claim about the grammatical structure of Mary’s utterance nor to whom it was directed. In any event, a speech report will be true, on this view, just in case the subject of the report produced an utterance consisting of tokens the requisite word types in the requisite (temporal) order with the requisite meanings and with the requisite illocutionary force.
ii: Indirect Quotation
In indirect quotation, the reporter makes no claim about either the words the speaker actually used or word order. There is not even any presupposition that the speaker’s utterance was in the same language as the speech report. Instead, in indirect quotation, utterances are typed in terms of propositional structure (or, if you prefer, logical form) and speakers’ epistemological relations to the constituents of the propositions they express. I will take up each of these features in turn.
Consider the sentence,
2. Hesperus is roughly spherical,
Any (normal) utterance of (2) would express the proposition that the object Hesperus (=Venus) has the property of being roughly spherical. Using double “half-diamond” brackets – ‘<<…>>’ – as propositional structure indicators, we can represent the proposition expressed by such an utterance as,
26. <<Being roughly spherical; Hesperus>>[31]
More generally, we can represent the proposition expressed by a sentence of the (logical) form ‘Fa1a2…an’ as
27. <<F-ness; a1, a2, … , an>>
Truth-functionally complex and quantified propositions can be similarly represented by treating truth-functions (and other propositional operators) as relations between propositions and quantifiers as relations between properties.
Now one suggestion might be to give the following analysis to (1):
28. Said (Mary, <<Being roughly spherical; Hesperus>>),
where, ‘<<Being roughly spherical; Hesperus>>’ names not a proposition but a type of utterance.[32] The trouble, of course, is that this would afoul of opacity. After all, since ‘<<Being roughly spherical; Phosphorus>>’ names the same type of utterance as does ‘<<Being roughly spherical; Hesperus>>’, this would imply that (4) means the same as (1). A better suggestion, then, would be to analyze (1) as,
29. Said (Mary, << Te(‘is roughly spherical’); Te(‘Hesperus’) >>)
where ‘Te(‘p’)’ denotes the type named by ‘p’ in indirect quotation contexts. The speech report would be true, on this analysis, just in case Mary uttered, with assertoric force, a sentence having subject-predicate (logical) form, while standing in the requisite epistemological relationships to the constituents of the proposition she expressed.[33]
The challenge here is to offer an adequate account of epistemological typing. Given my attempt to recast Fregean senses as intensional types in section II above, it would be perverse to characterize such types in terms of relations to intensional entities. An alternative might be to type utterances in terms of the direct (i.e., unmediated by intensional entities) cognitive relations of individual speakers to the objects of thought and talk. The trouble with this suggestion is that there are strong reasons to believe if typed in this way, utterances of distinct speakers could not be of the same epistemological type.[34] And this would impede the theoretical accommodation of group speech reports, such as
30. Everyone who went to the concert said that Keith Richards was stoned,
as well as
raising difficulties for an account of reporter
ability to denote the requisite types.
In order to avoid these worries, a typing schema for utterances that socializes cognitive relations to objects of thought and talk needs to be developed. Let me briefly sketch how this might be done. Individual thinker/ speakers stand in various sorts of cognitive relations to things in the world. These include conceptual as well as perceptual (and other causal) relations.[35] Moreover, individual speakers have a repertoire of referring expressions they use talk about the things to which they stand in specific cognitive relations.[36] In order to communicate with one another, however, speakers need to presuppose that, although they stand in distinct subjective cognitive relations to things in the world, they are so related to the same objective things. As a result, they need also presuppose that the corresponding referring expressions are co-referential. Suppose, for example, that John stands in a perceptual relation to an object he refers to as ‘the jolly red coated stranger’ and ‘that [pointing] guy’ and Mary, John’s telephone conversant, stands in a conceptual relation to an object she refers to as ‘Santa’ and ‘Pere Noel’. In order to communicate, they will have to presuppose that they stand in distinct cognitive relations to the same object and that their referring expressions pick out the same guy. There is, of course, no guarantee that their presupposition is true. Nonetheless, modulo this presupposition, John’s uses of ‘the jolly red-coated stranger’ and ‘that [pointing] guy’ and Mary’s uses of ‘Santa’ and ‘Pere Noel’ are of the same epistemological type. Given that the packets of contextual parameters discussed in section VI-i above include such presuppositions, we can make this explicit by representing (1) as,
31. Said (Mary, << Ter(‘is roughly spherical’); Ter(‘Hesperus’) >>),
where, as above, the subscript ‘r’ is used as a label for parameters of the reported context. This picture does need to be developed a lot more fully, but the sketch given here will do for present purposes.
iii: Mixed Quotation
What remains to be done, in order to meet the challenge of Cappelen and Lepore, is to give an account of the semantics of mixed quotation that overlaps with the treatment of both direct and indirect quotation. And this is not hard to do. What one reports via mixed quotation is the logical form of the speaker’s utterance, as well as the epistemic type of some of the constituents of the utterance and the (ordered) word-type of the remainder of the constituents. Consider again,
16. Mary said that Hesperus is “roughly spherical.”
What is being claimed here, on my view, is that Mary produced an utterance of the epistemic type denoted by ‘<< Ter(‘is roughly spherical’); Ter(‘Hesperus’) >>’. But moreover, in so doing, she used the expression ‘roughly spherical’ to pick out the property she was attributing to Venus. This can be represented by replacing ‘Te(‘is roughly spherical’)’ in (29) with ‘{Tw(‘roughly’), Tw(‘spherical’)}’, yielding,
32. Said (Mary, <<{Tw(‘roughly’), Tw(‘spherical’)}; Ter(‘Hesperus’) >>).
As above, we can include context-indicating subscripts to disambiguate the ordered word-types, if necessary.
There is, in addition, another kind of mixed quotation that a unified account of quotation ought to accommodate. Suppose Mary says,
33. He is a schmuck.
Someone could report Mary’s speech act by saying
34. Mary said, “He is a schmuck.”
But unless the reporter’s audience knows about whom Mary was talking, (34) will leave them ignorant of this potentially important feature of Mary’s utterance. To alleviate such difficulties, a reporter will often replace a referring expression used by a speaker with one more amenable to the reporter’s audience, while indicating in some way that this was not the expression used by the speaker. For example, someone might report Mary’s utterance of (33) by saying,
35. Mary said, “[Joe] is a schmuck”
instead of (34).
The unified Fregean analysis can handle such sentences by means to a minor emendation to the account of direct quotation. As with the account of direct discourse, utterances are structurally-typed in terms of word order. But unlike the account of direct quotation, not all of the constituents of utterances are typed as words. Instead, some of the constituents are typed semantically, in terms of their meanings. A sentence like (35) can be represented on this account as,
36. Said (Mary, {Tta(‘Joe’), Tw(‘is’), Tw(‘a’), Tw(‘schmuck’)})
The superscript ‘t’ (for ‘transparent’) indicates that word tokens are to be typed in terms of their (extensional meanings), while the subscript ‘a’ indicates that the meaning (of ‘Joe’) in question is relative to the actual contextual parameters. On this analysis, (35) is true just in case Mary produced (in the following order and with the requisite force) an utterance consisting of a word which means what ‘Joe’ means (in the actual context of utterance), a token of the word ‘is’, a token of ‘a’, and a token of ‘schmuck’.
iv: Transparency and Inference
There are (at least) two distinct uses of indirect quotation sentences: opaque uses and transparent use. In the former cases, how speakers conceive of the objects of thought and talk is at issue, and co-referential expressions are not substitutable salva veritate within the complement clause of the quotation sentences. In the latter cases, in contrast, how speakers conceive of the objects of thought and talk is not at issue, and co-referential expressions are substitutable in complement sentences. The justificatory uses of quotation sentences discussed in section (III) above would fall into this latter category.
Thus far, we have considered only opaque uses of indirect quotation, but an account of transparent uses is easy to provide. As with opaque the opaque case, utterances are structurally-typed in terms of logical form, but the constituents of utterances are typed semantically, in terms of their meanings. So, for example, a transparent use of (2) would get the following analysis:
37. Said (Mary, << Tt(‘is roughly spherical’); Tt (‘Hesperus’) >>).
As above, worries about ambiguity can be handled by including subscripts for the relevant contextual parameters. On this analysis, (2) is true only if Mary produced a token of the type consisting of utterances expressing the proposition that Hesperus is roughly spherical. It is worth noting that ontological economy can be achieved here by identifying propositions with the corresponding utterance-types themselves. For type-nominalists, this would involve taking propositions to be (certain) sets of utterances.
Inference relations among quotation sentences, on the unified Fregean account, are ground in relations among the utterance-types named by their complement clauses. For simplicity of exposition, I am going to assume utterance-types are just sets of utterances. This will enable me to treat relations between types as set relations. The basic idea is that if an utterance-type T1 is a subset of a of type T2, then the inference from ‘Said (A, T1)’ to ‘Said (A, T2)’ will be valid. After all, if A produced an utterance that is a member of T1, and T1 is a subset of T2, then A produced an utterance that is a member of T2. For example, if the type named by ‘{Tw(‘Hesperus’), Tw(‘is’), Tw(‘roughly’), Tw(‘spherical’)}’ is a subset of the type named by ‘<< Te(‘is roughly spherical’); Te(‘Hesperus’) >>’, then inference from (15) to (1) (interpreted opaquely) will be valid. Strictly speaking, since the type named by the complement clause of any given quotation sentence varies with contextual parameters, we should only talk about inferential relations among utterances of quotation sentences rather than among the sentences themselves.
Quantified quotation sentences, such as,
38. There is something of which Mary said that it is roughly spherical,
can be handled using an analogue of Church’s denotation predicate.[37] The guiding idea here is that (38) can be inferred from (1) – which, recall, is represented as
30. Said (Mary, << Te(‘is roughly spherical’); Te(‘Hesperus’) >>) –
all expressions tokens of the type denoted by ‘Te(‘Hesperus’)’ pick out the same object. Using ‘xRy’ to abbreviate ‘x refers to y’, we can represent (38) as
39. ($x)($T)("eÎT eRx & Said (Mary, << Te(‘is roughly spherical’); T>>)),
where the x-quantifier ranges over material objects, the T-quantifier ranges over expression types, and the e-quantifier ranges over expression tokens. Only a minor modification of the analysis of quantifying in is required in order to accommodate de re quotation sentences, such as
40. Mary said of Hesperus that it is roughly spherical.
(40) can be represented as
41. ($x)($T)(x = Hesperus & "eÎT eRx & Said (Mary, << Te(‘is roughly spherical’); T>>)),
or, more simply, as
42. ($T)("eÎT eRHesperus & Said (Mary, << Te(‘is roughly spherical’); T>>)),
where the quantifiers are to be understood as before. A similar treatment can be given to quantified and de re direct quotation sentences, such as
43. There is something of which Mary said, “[It] is roughly spherical”
and
44. Mary said of Hesperus, “[It] is roughly spherical.”
V: Two Objections and One Reply
Before going on to compare the unified Fregean theory with the unified Davidsonian account, there are a couple objections to the former view that might be raised. Because the same response to both objections can be offered on behalf of the unified Fregean theory, I will present both of them before replying. The first objection comes from the article in which Davidson first defended his paratactic analysis. In this paper, Davidson argued as follows:
“What stands in the way in Frege’s case is that every referring expression has an infinite number of entities it may refer to, depending on context, and there is no rule that gives the reference in more complex contexts on the basis of the reference in simpler ones.”[38]
The basic idea underlying Davidson’s objection is that since propositional attitude constructions can (in principle) be iterated without limit, the reference shift strategy implies that every expression of the language can have a limitless number of referents. And, in Davidson’s view, this runs afoul of the need for a finite compositional truth theory for English (and other natural languages).
Although not focused specifically on a Fregean account of quotation, Heal raises considerations that might be thought to count against the view on offer here. She says,
“…in speaking a language, (a) we exercise a skill in producing and responding to complex structured items; (b) there are principles governing the construction and properties of these things; (c) we show that we are in some sense aware of these principles and properties by our skillful performances; but (d) we cannot state these principles or specify the properties explicitly.”[39]
In light of these facts, Heal concludes,
“Given that you have a know-how of producing utterances and appreciating them, but only limited know-that about exactly what you are producing or responding to, the sensible thing is to indicate the character of the original remark by producing one like it in as many of the significance-relevant aspects as you can deal with or are important to you.”[40]
The upshot, according to Heal, is that speech reports should be understood to involve the attribution to the utterances of speakers indexically picked out properties of complement sentences.
These worries can be alleviated, however, by appealing to Kaplan’s well-known distinction between character and content.[41] Meaning in the sense of character is a property of expression types; meaning is the sense of content is a property of utterances. In a nutshell, the character of an expression, together with feature of the context in which it is uttered, determines the meaning (or content) of an utterance of it. Now the unified Fregean theory is a theory of the contents of utterances of quotation sentences, and not a theory of the character of the sentence types themselves. As a result, insofar as the “indexical predication” that Heal invokes operates at the level of character and not content, it is compatible with the unified Fregean theory.
Let me elaborate. As I would have it, the content of complement of a quotation sentence, on any given occasion of use, is an utterance-type. The character, then, of a complement sentence, when it occurs within a quotational frame, is a function from contextual parameters to utterance-types. Nothing prohibits us from specifying the character as (the function whose value is) the epistemological (or other) type of the complement itself, or, the type consisting of utterances epistemologically (or otherwise) similar to the complement, in either case understood relative to some intended set of contextual parameters. The self-referential quality of this specification is no more problematic than the specification of the character of ‘I’ as (the function whose value) the person who produced the said occurrence of ‘I’.
Davidson’s worry can also be alleviated by focusing on the indexical characters of complement sentences when they occur within quotational frames. The fact that complement sentences can have an unlimited number of referents (or contents), on the unified Fregean theory, is no more puzzling than is the fact that ‘I’ can have a (potentially) limitless number of referents. Moreover, given that the contents of complement sentences depend systematically on features of the contexts in which they are used – the intended typing schema, and the context relative to which the complement sentence is intended to be understood – this unlimited number of referents poses difficulties for the linguistic competence of finite beings with respect to languages containing quotational constructions.
VI: Simplicity and Elegance
What remains to be done is to face Cappelen and Lepore’s challenge head on, and compare the unified Fregean view with their own unified Davidsonian picture of quotation. According to the unified Davidsonian theory, a sentence of indirect discourse such as (1) would get the following analysis:
45. $u(Said ( Mary , u) & Samesays (u, that)). Hesperus is roughly spherical.
A sentence of direct discourse such as (15) would be analyzed as,
46. $u(Said ( Mary, u) & ST (u, these)). Hesperus is roughly spherical,
where ‘ST’
means same-tokens.[42] And,
finally, the unified Davidsonian account gives the following analysis to
sentences of mixed quotation such as (16):
47. $u(Said ( Mary, u) & Samesays (u, that) & ST (u, these)). Hesperus is roughly spherical,
where the first demonstrative picks out the whole utterance of (2), whereas the second picks out only that part of the utterance consisting of the words ‘roughly spherical’.
Recall, for purposes of comparison, that the unified Fregean theory gives
29. Said (Mary, << Te(‘is roughly spherical’); Te(‘Hesperus’) >>),
as an analysis of (1),
19. Said (Mary, {Tw(‘Hesperus’), Tw(‘is’), Tw(‘roughly’), Tw(‘spherical’)}),
as an analysis of (15), and
32. Said (Mary, <<{Tw(‘roughly’), Tw(‘spherical’)}; Ter(‘Hesperus’) >>),
as an analysis of (16).
At a first glance, there seems to be little to choose between the two accounts. Both are relatively simple and powerful, handling a wide variety of cases. Moreover, they seem to yield the same truth value in the vast majority of cases. This should be more or less unsurprising, however, because, semantically at least, the fundamental difference between the views is that where one invokes two (primary) types of utterances, the other invokes two similarity relations between utterances. And although this may make a difference at the level of fundamental ontology (but see section II), it is unlikely to make much difference to our classifications of utterances. Each view does face a cluster of difficulties characteristic of the principle governing propositional attitude contexts (see section I above) it violates. The unified Davidsonian theory faces a number of well-known difficulties as a result of its failure to satisfy compositionality.[43] And the unified Fregean theory faces a similar cluster of problems as a result of its failure to satisfy innocence.[44] As a result, any decision to choose one of these theories over the other should be based in large part on which of innocence and compositionality one holds most dear.
That being said, there are, in my view, a couple of reasons to prefer the unified Fregean theory to the alternative. The first has to do with de re quotation sentences. Following Hornsby, we can offer the following Davidsonian analysis of a de re quotation sentence such as (40):
48. $u(Of (Hesperus, u) & Said (Mary , u) & Samesays (u, that)). x is roughly spherical,
where ‘Of’ denotes a new primitive relation and the demonstrative denotes an utterance of an open rather than a closed sentence. [45] The first worry with this analysis is that allows utterances of open and closed sentences respectively to stand in the samesays relations to one another.[46] Hornsby suggests that this need not cause us much alarm:
“This new found looseness in what SAMESAYS with a given utterance is in practice always compensated by further predications of the utterance; one must tell what it was of.” [47]
The compensation that Hornsby identifies, however, does not cover the damages. Suppose, for example, that in some context C,
1. Mary said that Hesperus is roughly spherical,
and
40. Mary said of Hesperus that it is roughly spherical
are true, but ,
4. Mary said that Phosphorus is roughly spherical,
is false. In order for this to hold on the Davidsonian picture, Mary needs to have produced an utterance which stands in the samesays relation to the utterances (in C) of both,
2. Hesperus is roughly spherical,
and
49. x is roughly spherical,
but not
50. Phosphorus is roughly spherical.
The trouble is that if we assume, as we should, that the samesays relation is symmetrical and transitive, this implies that (2) and (49) stand in the samesays relation. And if (2) and (49) are so related, then there seems to be no reason to deny that (50) and (49) are so as well.[48] Given symmetry and transitivity yet again, this implies that (1) and (40) are true, (4) must be as well. A natural reply would be to introduce a new primitive relation that holds between utterances of open and closed sentences. While this would solve the problem at hand, it would do so at a great cost of elegance, especially when compared to the unified Fregean analysis, which has no trouble reconciling the truth of (1) and (40) with the falsity of (4).
One final reason to prefer the unified Fregean theory over the unified Davidsonian account concerns quotation containing unknown words and sounds. If someone, say Nichola, utters
51.
she can be correctly quoted by
52. Nichola
said that
as Cappelen and Lepore point out.[49] And they offer the following analysis of (52):
53. $u(Said (Nichola, u) & Samesays (u, that) & ST (u,
these)).
where
‘that’ refers to the utterance of ‘
54. Nichola
said that “
As it stands, the unified
Davidsonian account would have to take the analysis of (54) to be given by
(53), where ‘these’ is understood to refer to both ‘philtosopher’ and ‘
55. $u(Said (Nichola, u) & Samesays (u, that) & ST (u,
these) & Same-sounds u, this)).
where
‘that’ refers to the utterance of ‘
56. Said (Nichola, <<Tph(‘philtosopher’); Tw(‘Alice’)>>)
(where the superscript ‘ph’ indicates that the typing in question is phonetic) considerations of simplicity and elegance count again in its favour and against the unified Davidsonian view.
VII: Conclusion
By developing a unified account of quotation and challenging their opponents to do the same, Cappelen and Lepore have raised the bar in the quotation literature. But, if I am right, they have raised the bar without winning the day. The Davidsonian approach to quotation has always struck as a paradigm of philosophical ingenuity, but I have never, even for a moment, been tempted by it. As Burge put it,
“…our grip on the syntax of the relevant discourse is too fundamental and firm for us to be persuaded that we have been misled regarding how to … divide among the sentences. At any rate, I think our grip on sentence division is much firmer than our grip on the semantics of ‘intensional’ contexts.”[52]
Peter Alward
Department of Philtosophy
Peter.alward@uleth.ca
REFERENCES
Alward, Peter. 2000. “Simple and Sophisticated “Naïve” Semantics.” Dialogue, 39: 101-121.
Alward, Peter. 2003. “Fregecide.” Dialogue, 42: 275-290.
Alward, Peter. Manuscript.
“Between the Lines of Age: Reflections on the Metaphysics of Words.” http://people.uleth.ca/~peter.alward/papers/words.htm.
Burge, Tyler. 1986. “On Davidson’s
‘Saying That’.” In Truth
and Interpretation: Perspectives on the Philosophy of Donald Davidson, Ernest
LePore (ed), 190-208.
Cappelen, Herman and Lepore, Ernie. 1997. “Varieties of Quotation.” Mind 106: 429-450.
Cappelen, Herman
and Lepore, Ernie. 1999. “Semantics for Quotation.” In Philosophy and Linguistics, Kumiko Murasugi, (ed), 209-211.
Church, Alonzo. 1951. “A
Formulation of the Logic of Sense and Denotation.” In Structure, Method, and Meaning: Essays in Honor of Henry M. Sheffer, P. Henle, H. Kallen, and S. Langer (eds), 3-24.
Clapp, Leonard. 1995. “How to be Direct and Innocent: A Criticism of Crimmins and Perry’s Theory of Attitude Ascriptions.” Linguistics and Philosophy 18,5: 529-65.
Crimmins,
Mark. 1992. Talk About Belief. MIT Press/Bradford Books.
Cresswell, Max. 1985. Structured
Meanings,
Davidson, Donald. 1968. “On Saying That.” Synthese 19: 130-146.
Frege, Gottlob. 1997. “On Sense and Reference.” In The Frege Reader, M. Beaney, (ed), 151-171. Blackwell.
Heal, Jane. 2001. “On Speaking Thus: The Semantics of Indirect Discourse.” The Philosophical Quarterly 51: 433-454.
Hornsby, Jennifer. 1977. “Saying Of.” Analysis 37: 177-185.
Kaplan, David. 1971. ‘Quantifying In.’ In Reference and Modality, L. Linsky,(ed).
Kaplan, David. 1980.
“Demonstratives.” In Themes from Kaplan,
Joseph Almog, John Perry, and Howard Wettstein (eds).
Kaplan, David. 1990. “Words.” Proceedings of the Aristotelian Society: Supplement, v. 64: 93-119.
Kripke, Saul. 1980. Naming
and Necessity. Basil Blackwell.
Ludwig, Kirk and Ray, Greg. 1998. “Semantics for Opaque Contexts.” In Philosophical Perspectives 12, Mind,
Language and Ontology, James E. Tomberlin (ed), 141-66.
Pietroski, Paul. 1999. “Compositional Quotation (without
Parataxis).” In Philosophy and
Linguistics, Kumiko Murasugi,
(ed), 245-258.
Quine, W. V. O. 2001. “Quantifiers
and Propositional Attitudes.” In The
Philosophy of Language, 4th
Edition, A.P. Martinich, (ed),
355-360.
Richard, Mark. 1990. Propositional Attitudes.
Salmon,
Nathan. 1986. Frege’s Puzzle.
Soames, Scott. 1989. “Direct Reference and
Propositional Attitudes.” In Themes from
Kaplan, Joseph Almog, John Perry, and Howard Wettstein (eds).
Stainton, Robert. 1999. “Remarks on the Syntax
and Semantics of Mixed Quotation.” In Philosophy and Linguistics, Kumiko Murasugi, (ed), 259-278.
[1] Cappelen and Lepore (1997) p. 448.
[2] Davidson (1968).
[3] Meaning here should be understood as the proposition expressed (or propositional content ) in the case of full sentences and the contribution to the propositional content in the case of sub-sentential expressions. Strictly speaking, meaning in this sense is a property of utterances rather than sentences.
[4] Cappelen and Lepore (1999) p. 215.
[5] Frege, 1997.
[6] I am assuming, of course, that the meaning (= propositional expressed by) of a sentence determines its truth value (relative circumstances of evaluation).
[7] I am, of course, ignoring the “hidden indexical” view here, defended by Crimmins (1992), Richard (1990) and others, which attempts to satisfy all three principles.
[8] See, e.g. Salmon (1986) and Soames (1989). Strictly speaking, Salmon and Soames focus on belief reports rather than indirect quotation.
[9] According to Salmon (1986), propositional attitudes sentences have a more complicated logical form than this, but these differences are not relevant for present purposes.
[10] I am assuming, of course, that the samesays relation is not simply a matter of expressing the same proposition.
[11] Of course, compositionality holds for each of the sentences that Davidsonians take indirect quotations to consist of. It only fails for surface-grammatically unit sentences that are the subject matter of analysis.
[12] Of course, we all alike to see our own theory as hard-done by and in need of defense. Moreover, while the Davidsonian view has been much discussed, not all of the discussion has been positive, to say the least.
[13] Kripke, 1980.
[14] Quine (2001) p. 357.
[15] Ludwig and Ray (1998) p. 142.
[16] There are, of course, exceptions. The question might arise as to who performed some particular action –say, the vandalization of a bus terminal – and an agent report might be used to identify the agent.
[17] If one were to draw an ontological distinction between sparse and abundant types, a decision would have to be made regarding into which category intensional types fall.
[18] The distinction between attributive and justificatory uses of quotation is even more striking in the case of direct discourse. In contrast to Paul’s actual reply in Dialogue II, consider:
Paul: [As] Bob Dylan says, “Peter is the most talented member of the trio.”
[19] Note: I do not mean to suggest that a Davidsonian could not give a unified treatment of attributive and justificatory uses of quotation sentences, only that that are (pre-theoretical) reasons for thinking they do not require a unified treatment.
[20] But it might yield problems for an innocent unified account of quotation, along the lines of Cappelen and Lepore (1997). More on this below.
[21] We might also stipulate that Mary and her conversants believe that the poisoner is distinct from the shooter.
[22] The bracketed ‘whoever he is’ is meant to indicate the definite description is used attributively.
[23] They also suggest that a general theory of quotation should give uniform treatment to pure quotation – sentences such as “The word “the” has three letters.” Although I think the unified Fregean account can easily accommodate pure quotation, I will not address this issue here.
[24] Cappelen and Lepore (1997), pp. 430-31.
[25] Strictly speaking, “says” ought to be treated as a three-place predicate satisfied by speakers, utterance types, and times. I am going to ignore this complication for present purposes.
[26] If one objects to treating quoted complements and “that”-clauses as names, one could instead treat expressions of the form ‘said that p” and ‘said “S”’ as complex one-place predicates satisfied by speakers.
[27] Note: the view presented here is similar in certain ways to Pietroski (1999).
[28] See, Kaplan (1971) for a similar view of belief sentences.
[29] Of course, it could turn that in some cases the relevant contextual parameters do not yield a unique interpretation of the utterance. This would result in genuine ambiguity.
[30] Strictly speaking, the parameters determine not how the complement sentence is to be interpreted but the meaning that an utterance has to have in order to be a token of the type named by the complement.
[31] I am relying here on the account of propositional structure developed by Crimmins (1992), pp. 99-124. Crimmins gives a more detailed analysis in terms of the roles entities fill within propositions, but I am going to ignore this complication for present purposes.
[32] If one identifies propositions with types of utterances, this will amount to the same thing.
[33] For those who eschew talk of propositional constituents and properties, the same point can be made in terms of epistemological relations to the extensions of subjects and predicates.
[34] See, e.g., Clapp (1995) and Alward (2003).
[35] I do not mean to suggest that conceptual relations are (entirely) non-causal or that perceptual relations are (entirely) non-conceptual. I suspect that (nearly) all cognitive relations to things include both conceptual and causal elements. They vary, however, in terms of conceptual richness and causal salience.
[36] I am not presupposing that the cognitive relations in question guarantee unique referents.
[37] Church (1951).
[38] Davidson (1968) p. 136.
[39] Heal (2001) p. 438.
[40] Heal (2001) pp. 443-4.
[41] Kaplan (1980).
[42] Cappelen and Lepore (1997) pp. 441-2.
[43] See, e.g., Burge (1986) and Stainton (1999).
[44] See, e.g., Salmon (1986) and Soames (1989). Although I have not been discussing it here, the direct reference theory also faces a cluster of difficulties stemming from its failure to satisfy opacity. See, e.g., Alward (2000), Richard (1990), and Crimmins (1992).
[45] Hornsby (1977). I have recast Hornsby’s analysis slightly to make it commensurable with the formulation of Davidson’s view we have been considering.
[46] Assuming, of course, that de re quotation sentences are sometimes true.
[47] Hornsby (1977) p. 180.
[48] Cappelen and Lepore would likely balk here. In response to a different objection, they insist “[it’s] the actual practice of making indirect reports of others that fixes that extension [i.e., of the samesays relation].” (1997, p. 445) This response is inadequate for two reasons: it presupposes the truth of the unified Davidsonian account when this is exactly what is at issue; and by tailoring the samesays relation to fit the linguistic data, it makes the theory untestable by this data.
[49] Cappelen and Lepore (1997), p. 436.
[50] One
might object, however, as follows: since ‘
[51] For more on the individuation of words, see Kaplan (1990) and Alward (manuscript).
[52] Burge (1968) p. 191.