Proto Sociology Vol. 17, 2002 - Semantic Theory and Reported Speech

ProtoSociology

ISSN 1434-4319

An International Journal of Interdisciplinary Research Volume17 – 2002 Semantic Theory and Reported Speech

Editor: Gerhard Preyer Editorial of the Vol. 17: Georg Peter Layout and Technical Conception: Georg Peter Editorial Office: ProtoSociology, Stephan-Heise-Str. 56, 60488 Frankfurt am Main, Germany; phone: (049)069-769461; e-mail: [email protected] P Homepage: http://www.protosociology.de P P P P

2

Autor

Die Deutsche Bibliothek – CIP-Einheitsaufnahme Ein Titeldatensatz für diese Publikation ist bei der Deutschen Bibliothek erhältlich. © 2002 ProtoSociology Frankfurt am Main http://www.protosociology.de [email protected] Erste Auflage / first published 2002 ISSN – 1434-4319 Alle Rechte vorbehalten. Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung der Zeitschirft und seines Herausgebers unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeisung und Verarbeitung in elektronischen Systemen. All rights reserved. No part of this publication may be reproduced, stored in a retrievalsystem, or transmitted, in any form or by any means, without the prior permission of ProtoSociology.

Titel

3

Proto Sociology An International Journal of Interdisciplinary Research

Vol. 17, 2002 Semantic Theory and Reported Speech Contents The Meaning of What is Said and the Ascription of Attitudes The Semantic Significance of What is Said ......................................... Emma Borg

7

Representing What Others Say ........................................................... Cara Spencer

26

The Things People Say ....................................................................... Jonathan Sutton

46

Reported Speech and the Epistemology of Testimony ........................ Sanford C. Goldberg

59

Reports and Imagination .................................................................... Eros Corazza

78

Propositional Content and Cognitive Structure On Representing Content .................................................................. 101 David Hunter Conceptual Realism and Interpretation .............................................. 119 Max A. Freund

4

Autor

Content Partialism and Davidson’s Dilemma ..................................... 138 Corey Washington Vagueness, Indirect Speech Reports, and the World ........................... Steven Gross

153

On Contemporary Philosophy and Social Sciences Epistemological Remarks Concerning the Concepts “Theory” and “Theoretical Concepts” ................................................ 171 Hans Lenk Social Science and (Null) Hypothesis Testing: Some Ontological Issues ..................................................................... 188 Steven Miller and Marcel Fredericks Contributors ....................................................................................... 202 Imprint ............................................................................................... 203 On ProtoSociology ............................................................................. 204 Published Volumes ............................................................................. 205 Bookpublications ................................................................................ 213 Subscription / Digital Version ............................................................. 232 ProtoSociology: Digital volumes vailable ............................................ 233

The Semantic Significance of What is Said

The Meaning of What is Said and the Ascription of Attitudes

5

6

Emma Borg

The Semantic Significance of What is Said

7

The Semantic Significance of What is Said Emma Borg

Abstract It is often held that a correct semantic theory should assign a semantic content, p, to a given sentence, s, just in case a speaker who utters s says that p – thus ‘what is said’ is taken to be a semantically significant notion. This paper explores what exactly such a claim amounts to and offers five versions of the relationship between a semantic theory and judgements of what is said. The first three of these versions embody the central claim of semantic significance; however, I argue that none of these versions are feasible. Thus, contrary to the initial proposal, I claim that ‘what is said’ is not a semantically significant notion. Assignments of semantic content do not turn on evaluations of what a speaker uttering a sentence says.

The question of how to correctly analyse sentences of oratio obliqua (indirect reported speech) has received a good deal of attention within the confines of philosophy of language and one reason for this, we might think, is that such sentences may prove interesting not only in their own right, but also because of the role they play in informing us about the meaning of the sentences they report. If a speaker, U, who utters a sentence, s, can be correctly reported by an act of oratio obliqua of the form ‘U said that p’ (where ‘p’ supplies the content sentence of the indirect speech act), then it seems that we have learnt something about the meaning of s. Now, exactly how we spell out this thought is open to question (as we will see in §1), but the intuitive idea which it embodies is that there is a move to be made from reports of what is said to the contents of a correct semantic theory. Such a stance, which I label the ‘standard view’, has a distinguished history, but it is a position which I wish to reject in this paper. I will argue that there is no semantically privileged notion of ‘what is said’, and thus no considerations concerning oratio obliqua should constrain or otherwise affect our semantic theorising. In no case of indirect speech reporting is it possible to move from the acceptability of ‘U said that p’ to a claim that p provides the correct semantic analysis of the original sentence s. The structure of the paper is as follows: in §1 I will set out five distinct ways in which the relationship between semantic theories and indirect speech reports might be construed. The initial three positions, (a-c), embody the key

8

Emma Borg

idea of semantically informative indirect speech acts, while the latter two, (de), do not. The simplest version of the standard view, given by (a), will quickly be seen to be implausible, but (b) and (c) initially remain live options. However, in §2, I will argue that this initial impression is mistaken; on closer inspection neither position is able to deliver the semantically informative notion of ‘what is said’ as promised.1 Yet, I will suggest in §3, this finding would be problematic only if the sole alternative to the standard view were given by position (e), for it too is problematic. This is not the case, however, for the theorist who rejects the semantic relevance of ‘what is said’ may instead opt for the more moderate position, to be given by (d), whereby there is a one-way dependence from semantic analyses to indirect speech reports. This position both avoids the earlier objections and, I will argue, is independently attractive.

1. The Relationship Between a Semantic Theory and Indirect Speech Reports Initially, at least, the idea that a semantic theory should tell us what information an utterance of a given sentence conveys seems quite unobjectionable, indeed it seems a mainstay of much formal semantic theorising. It seems entirely plausible to maintain that, if U can be reported as having said that p by her utterance of s, then p in some way gives the meaning of s. Although I don’t want to engage in an exegetical history of this position, I think it is not particularly controversial to label this stance the ‘standard view’.2 However, recently such a view has come under attack, perhaps most notably in Cappelen and Lepore 1997, where it is objected that facts about reported speech in ordinary language do not support the claim that the notion has any significant role in influencing the construction of a semantic theory. They write: ‘That a semantic theory should specify what is said by utterances of sentences seems innocent enough, but, when this assumption is embodied 1 In what follows I will use the notions of reported speech and judgements of what is said interchangeably. This might be thought to be problematic since judgements of what is said range wider than acts of reported speech. For instance, an interlocutor may make a judgement of what is said by making a judgement of equivalence between two or more sentences, or by answering questions about cancelability, etc. (I’m grateful to Kent Bach for stressing this point). However, even in these latter cases, it seems that the agent puts herself in a position to make or endorse a report of the form ‘U said that p’, even if such an indirect speech report is not actually produced. Thus it seems to me that some equivocation between the two notions is not here pernicious. 2 The ancestral home of this view is probably the work of Paul Grice (cf. Grice 1989).

The Semantic Significance of What is Said

9

by MA [the principle that an adequate semantic theory for a language L should assign p as the semantic content of a sentence S in L iff in uttering S a speaker says that p], semanticists both misconstrue the aim of semantics and unreasonably constrain the semantics for indirect speech’ (279). Ultimately, I want to side with Cappelen and Lepore as against what I have called the ‘standard view’, but to begin to get clear on what is at stake in the debate here, it seems that we need to start by establishing exactly what is being embraced by the former view and rejected by the latter – what is the relationship between a semantic theory and indirect speech reports that is under dispute? I think there are (at least) five ways in which we might construe the relationship here (these are illustrated in Fig.1): a) One-way dependency - semantic analyses of a sentence, s, are dependent on the content of indirect speech reports concerning s. On this approach the content of a correct semantic clause for s is constituted (at least in part) by the content of correct indirect speech reports for an utterance of s. b) Two-way dependency: semantic analyses of a sentence, s, are dependent on the content of appropriate indirect speech reports (‘what is said*’) concerning s, but which indirect speech reports are appropriate is itself a matter which is influenced by the semantic theory in play. c) One-way dependency: semantic analyses of a sentence, s, are dependent on the content of appropriate indirect speech reports (‘what is said*’), and which indirect speech reports are appropriate is determined by non-semantic features. d) One-way dependency - semantic analyses to indirect speech reports. Judgements concerning correct indirect speech reports are affected by the semantic analysis of the original sentence, but not vice-a-versa. e) No dependency: judgements of correct indirect speech reports and the output of a correct semantic theory are independent of one another – one concerns communication and the other linguistic meaning, and these are distinct domains.

10

Emma Borg Figure 1:

a)

what is said
semantic theory

d)

what is said  semantic theory

b)

what is said*
 semantic theory

e)

what is said
 semantic theory

c)

what is said*
semantic theory

What unites (a-c), as against (d-e), is that there is held to be some kind of connection from (certain) indirect speech reports to a correct semantic theory. This supports the idea that indirect speech reports can be genuinely semantically informative or significant, for we can use the former to determine in some way the content of the latter. This is the claim made by the standard view; so is any such stance plausible? Well, it seems fairly clear that the simplest construal of the standard view, given by (a), whereby fixing the facts about indirect speech reports fixes the facts for our semantic theory, is not plausible. For it is clear from even a momentary survey of the facts concerning reported speech in ordinary language that they diverge wildly from the kinds of features which can plausibly be taken to be the concern of semantics. To see this, imagine that Jim utters: 1)

Blair lives at No.10.

(1) is a simple subject-predicate sentence, containing no overt indexicals, and thus would seem one of the most elementary cases for the advocate of the semantic significance of indirect speech reports to deal with. Yet even here the problem is immediately obvious, for the utterance of (1) does not stand in a single, one:one relationship with some indirect speech report, but rather explodes into a plethora of possible indirect speech acts, all of which may be licensed by Jim’s production of (1). For instance, given the right sort of context of utterance and report, any of the following (amongst an indefinite number of others) may be acceptable: 2) Jim said that Tony Blair lived at No.10 Downing St, London, UK.

The Semantic Significance of What is Said

11

3) 4) 5) 6)

Jim said that the current Prime Minister lives at No.10. Jim said that that man lives there. Jim said that Baby Leo’s father lives in No.10. Jim said that the most right-wing Labour leader in the party’s history secured power where other, more left-wing, predecessors had failed. 7) Jim said that he knows about British politics.

Clearly, even in a basic case like (1), then, there are a vast range of potentially correct acts of reported speech, but intuitively we don’t want to countenance all of them as semantically informative as to the meaning of the original sentence. For one thing, allowing all indirect speech reports to be semantically relevant would mean rejecting the idea that a single proposition provides the content of a simple (non-indexical and non-ambiguous) sentence. On the current view we would have to allow that any such sentence type has a vast number of meanings, each one relating to something the sentence could be used to say. Such multiplying of meanings is an anathema to any kind of generalised semantic theory and would make entirely opaque how we ever learn or use a language. Secondly, however, it seems pretty clear that at least some of the above reports pick up not on what we might think of as the meaning of the antecedent sentence produced, but instead depend on features quite external to that sentence, like the conversational context in which it is produced and the (shared) background assumptions of speakers and hearers. Some of these indirect speech reports, though (potentially) perfectly acceptable, seem to be reporting things only tangentially connected with the meaning of the sentence produced. For instance, it is easy to imagine a context in which (6) is acceptable: say Jim produced (1) in an argument about the relative merits of right-wing versus left-wing Labour leaders, where it was mutual knowledge that Blair was the most right-wing leader to date. Then (6) might well be an accurate report of what Jim conveyed to his audience on this occasion by his utterance. Yet, although it is relatively easy to find a context in which (6) is acceptable, this does not seem to lead us towards taking the content sentence of (6) to tell us anything much about the meaning of the sentence Jim originally produced. (6) may well give us what Jim meant by his utterance of (1), but this seems a very different thing to what the sentence itself means. The problem we are recognising here is that some perfectly legitimate indirect speech reports pick up not on sentence meaning but on what Grice 3 Grice 1989: 117-37.

12

Emma Borg

termed ‘utterer’s meaning’.3 If this is correct, then there can be no general move from facts about indirect speech reports to facts about the correct content of our semantic theory. If we group together all the ordinarily correct instances of indirect speech reports (i.e. acceptable occurrences of the ordinary language locution ‘U said that p’) then we get a rag-bag class of cases, polluted by an array of non-semantic features, which cannot play a semantically informative role. What we need, it seems, is some way to select from this nebulous set some more refined class – to rule out, at the very least, acceptable indirect speech reports, like (6) or (7), which report not on the meaning of the original sentence, but merely on what the speaker who uttered it succeeded in conveying to their audience. It is no job of a successful semantic theory to tell us everything which a sentence may be used to say on a given occasion of utterance, and to adopt (a) is to overlook this basic fact. So (a) is untenable. Yet we have no reason to saddle an advocate of the standard view with anything so unappealing, for both (b) and (c) above capture the claim that there is a way to move from indirect speech reports to semantic analyses, without falling prey to the objection above. For both these more moderate views hold that the construction of a correct semantic theory for some language L will be affected by indirect speech reports, but also that which indirect speech reports are relevant is itself something which is constrained in some way. So, let us turn now to these more refined versions of the standard view.

2. A Privileged Notion of ‘What is Said’? The problem faced by advocates of a refined, semantically relevant, notion of ‘what is said’ is how they will select from amongst the entire set of ordinarily acceptable instances of indirect speech reports just those which are genuinely semantically informative. That is to say, faced with a set like (2-7) what principle can we use for selecting the semantically relevant sub-set, the ‘what is said*’ cases? A natural first move in answer to this question is to introduce the Gricean notion of ‘conversational implicature’.4 A proposition may be conversationally implied if it can be inferred from the meaning of the sentence produced together with the context of utterance, plus knowledge of some general principles of good communication. For instance, the speaker who utters ‘Someone hasn’t done their homework again’, in a context where it is clear to all parties that the recalcitrant student is Smith, might be taken to 4 Grice 1989: 31.

The Semantic Significance of What is Said

13

flout the maxim of quantity (by failing to convey the amount of relevant information which she possesses); thus her audience may be licensed in inferring some non-literal proposition as the one actually communicated, namely that Smith hasn’t done his homework again. For Grice, what is required to arrive at a conversational implicature is first to grasp the literal meaning of the sentence produced, then to see that this literal meaning flouts some quite general principle of good communication, and finally to infer that the literal proposition can’t be the one the speaker meant to convey and therefore it should be replaced with a more suitable proposition which no longer flouts the communicative principle in question. Relying on this inferential aspect to conversational implicatures, then, we might claim that an indirect speech report can be semantically informative just in case it reports a proposition which can provide the basis for an inference to a further proposition, but which is not itself inferentially arrived at.5 Although this principle is roughly sketched, let us assume for the sake of argument that it (or something similar) does indeed hold good, providing us with a clear way to individuate speech reports picking up on conversational implicatures from those which do not. The question then is: is this enough? Is the class of acceptable indirect speech reports sans those which pick up on conversational implicatures a class which we can take to be semantically relevant? Consider again Jim’s utterance of (1), and the small sample of acceptable indirect speech reports (2-7). Certainly some of these fit the model of conversational implicature reports, as just sketched. For instance, (6) and (7) might well be thought to arise as acceptable reports just in case Jim’s original utterance, if understood literally, does not meet the requirements of some Gricean principle of good communication (say, relevance) in the context in which (1) was uttered. However, it is not at all clear that any of the other reports fit this model of conversational implicature reporting, yet nor is it clear that we want to count all of the remaining reports as semantically informative. Take (5) for instance (‘Jim said that Baby Leo’s father lives in No.10’). Here the proper name ‘Blair’ has been replaced by the possessive ‘Baby Leo’s father’, while the relationship ‘living at No.10’ has been changed to ‘living in No.10’. Now we can easily imagine (5) being an acceptable report of what Jim said – say if the speaker of (5) wants to report Jim’s utterance to a fellow worker in Leo’s nursery, who is unaware that Leo’s father is Blair. Here the case seems very different to one of 5 Though this need not be taken as a claim about the actual cognitive procedures employed by a given interlocutor; cf. the debate over the ‘Availability Principle’ in Recanati 1989, Nicolle and Clarke 1999, Bach 2001, and Taylor 2001.

14

Emma Borg

straightforward conversational implicature; it is not as if the speaker of (5) had to work out the literal meaning of Jim’s utterance, see it as somehow conversationally improper (after all, we are imagining that Jim’s utterance was perfectly proper in the context in which it was uttered, it is only that, in the context of reporting, the expressions used would not be so apt), and infer her way to the content sentence of (5). Yet, on the other hand, it seems simply wrong to think that any part of the meaning of the sentence in (1) is given by the expression ‘Baby Leo’s father’ (after all, one could perfectly well understand the sentence in (1) without any grasp of the predicate ‘__ is the father of __’). If this is right then the refined notion of ‘what is said*’ in play at the moment (i.e. one which rules out only conversational implicature) is not strict enough: there are some perfectly acceptable indirect speech reports, which are not instances of conversational implicature reporting, which do not tell us about the correct semantic analysis of the original sentence. What we need is a more restricted version of ‘what is said*’, and again the question is ‘how is this to be arrived at?’ At this point, it seems to me that there are two broad approaches the advocate of semantic significance for certain indirect speech acts might pursue: either our theorist may take the appropriate cases of indirect speech to be delivered through their relationship to the correct semantic theory (position (b)), or she may take the appropriate set of cases to be delivered via some non-semantic (e.g. perhaps syntactic) route (position (c)). On the former approach, the idea is roughly that lurking behind the everyday notion of ‘what is said’ is a semantically informed notion – something like ‘literally said’ – and it is this semantically informed notion which is to play the key role in construction of our semantic theory. On the latter approach, the everyday notion is thought to be redundant for semantic theorising, and instead what is introduced is an independent, technical definition of ‘what is said*’, which can play precisely the semantically informative notion envisaged for it. In what follows I want to explore both these types of approach, but the claim will be, to the extent that either of them can deliver something semantically informative, the additional trip through ‘what is said’ is theoretically redundant. So, let us begin with approaches of type (b), where we come to judgements of what is said already armed with some independent conception of semantic content. We do not expect judgements of indirect speech reports to dictate semantic analyses (to be made in a semantic vacuum, as it were), but instead see them as already informed by semantic evaluations. Thus advocates of (b) are unscathed by the fundamental objection to (a) above, for the crucial notion of ‘what is said’ is not co-extensive with the ordinary language locu-

The Semantic Significance of What is Said

15

tion (which may pick up on speaker meaning rather than literal sentencemeaning) but is instead itself a semantically-informed notion. The important idea now is something like ‘literally said’ or ‘strictly said’, where this isolates the peculiarly semantic information conveyed by the utterance of a sentence.6 Now, initially at least, it seems that there is something of an air of circularity here: ‘what is said*’ is supposed to be a semantically informative notion, yet it itself is to be semantically informed, and this may seem worrying. However, advocates of this position can dissipate the perceived problem by distinguishing between epistemic and metaphysical claims. For it may be that the special relationship between privileged indirect speech reports and the semantic theory is a matter of metaphysics (e.g. facts about ‘what is said’ in this privileged sense supervene directly on semantic facts), while the sense in which facts about ‘what is said*’ are semantically informative is an epistemic one (e.g. they don’t, as (a) would have it, supply the raw data for our semantic theory, but can act as a guide or indicator of the contents of a correct semantic theory). In this way, judgements about what is said* are thought to be more accessible or intuitively obvious than direct reflection on the contents of a semantic theory, thus they are available to provide an initial epistemic guide. This idea of epistemic immediacy is, for instance, endorsed by Reimer, who defines ‘what is said*’ as ‘the meaning of the sentence uttered, relativized to a context of utterance’ (a definition she attributes originally to Grice), and suggests: This seems to be a ‘pre-theoretic phenomenon’. For one could easily explain to a non-philosopher (of language) what is meant by Grice’s notion of ‘what is said’ and then ask her to evaluate (for truth or falsity) statements of the form: In uttering S, a speaker ‘says that’ p. The intuitions to which such a person would appeal would be ‘pretheoretical’ in the relevant sense. For (by hypothesis) our speaker is not committed to any particular theory as to the meaning (the semantic content) of the contextually relativized S…For such reasons, her intuitions are properly regarded as potential evidence for semantic theories.7 6 Cf. Richard 1998: 606, “A literal utterance of S literally says p just in case (according to [a correct semantic theory for English]) the semantic content of S (in the utterance) is p”. 7 Reimer 1998: 602. Recanati 1989: 309, also endorses the claim of epistemic immediacy for judgements of what is said: “I [have] made two related claims: first, that sentence meaning is something more abstract and theoretical than what is said; second, that we have

16

Emma Borg

Now, although it seems correct to assign a degree of semantic immediacy or intuitive access to our ordinary judgements of ‘what is said’ (since competent speakers of a language have, in general, little difficulty in deciding whether to accept or reject putative reports of a given speech act), I think we must query Reimer’s suggestion that the more refined notion of ‘what is said*’ is one to which interlocutors in general have any such intuitive or unmediated access. For consider the range of putative counterexamples to the semantic relevance of what is said, as raised by Cappelen and Lepore 1997. They point out that, to take just a selection of cases, on occasion we may licence the exchange of synonyms from the original sentence to the indirect speech report (e.g. allowing ‘bachelor’ to be reported via ‘unmarried male’), or we may allow the exchange of co-referring terms (e.g. ‘Blair’ and ‘that man’), or exchange of descriptive phrases or relational properties (e.g. ‘the Prime Minister’ for ‘the father of Baby Leo’). We may also allow reports to pick out elements of the original utterance (e.g. an utterance of ‘p and q’ reported by ‘q’), logical entailments (‘if p then q, and p’ reported with ‘A said that q’), and paraphrases of the original utterance (e.g. reports of a long fiscal statement with ‘A said that inflation will fall’). I would suggest that, faced with a range of cases like this, we have very little reason to expect interlocutors to have any strong or consistent intuitions about which ones count as literal reports and which ones do not.8 Furthermore, to the extent that interlocutors are able to come to judgements in this area, it seems that they must surely be secondary to semantic judgements – we don’t, it seems, take s and p to mean the same because we judge a report of ‘A said that p’ as a literal report of some sentence s, rather we judge a report as a literal report just in case we think s and p mean the same thing.9 Consider a report of ‘that man lives at No.10’ by ‘A said that the man over there lives at ‘intuitions’ concerning what is said that serve as a starting point in the process of determining what the linguistic meaning of the sentence is” (though we should note that he also denies that there is a purely semantic notion of ‘what is said’). 8 See Nicolle and Clarke 1999 for some experimental findings which apparently support the variability of our reporting practices, even when asked to report ‘what is said*’. 9 The same sort of problem, I think, undermines Richard’s 1998 response to Cappelen and Lepore. Richard suggests that we will ultimately be able to construct a set of individually necessary and jointly sufficient conditions which yield all and only the semantically informative instances of the commonsense locution ‘what is said’. Yet, even if we agreed on the feasibility of such a project, it seems clear to me that to so much as get started on this kind of rule construction we already need some pretty robust semantic intuitions. That is to say, to decide that we will accept or reject a case as an instance of literal saying, we already need to have decided what we think the original sentence means. What drives the introduction of any of the necessary conditions is not, I would suggest, an intuition about what is literally said by a literal utterance of s, but simply a prior view about what s means.

The Semantic Significance of What is Said

17

No.10’. We will count this as a literal report only if we think that ‘that man’ and ‘the man over there’ mean the same (have the same semantic value); to assume that we can make an assessment of literal reporting prior to this kind of decision seems to ‘put the cart before the horse’, as it were.10 Or again, take the report of (1) (‘Blair lives at No.10’) with (2) (‘Jim said that Tony Blair lived at No.10 Downing St, London, UK’): does this count as an instance of literal reporting? It seems to me that this is something we simply lack any generally agreed intuitions about – in lieu of any claims about the meaning of s who knows whether p in this case counts as a literal report? The attempt to avoid circularity here fails for either we simply lack any intuitions over indirect speech reports like these, or those intuitions we can get hold of seem much better accommodated as direct semantic intuitions, concerning the meaning of the original sentence, as opposed to intuitions about some nebulous notion like ‘literal reporting’. Perhaps instead then we should look to non-semantic features to deliver the relevant set of indirect speech reports; this is the stance taken by (c) above. On this view there is still a move from appropriate indirect speech reports to the content of a semantic theory, but our earlier charge of circularity is avoided, since we are not appealing to semantic features in spelling out the notion of ‘appropriateness’. Now, obviously, for this to be a substantial position, the theorist owes us an account of which non-semantic factors come into play. Yet there seems to be a good candidate here, in the form of an appeal to syntactic features. For instance, such a theorist might think to adapt something like Bach’s ‘syntactic correlation constraint’ (SCC): According to which what is said must correspond to ‘the elements of [the sentence], their order, and their syntactic character’ [Studies in the Way of Words, p.87]. So if any element of the content of an utterance, i.e. of what the speaker intends to convey, does not correspond to any element of the sentence being uttered, it is not part of what is said.11 10 Reimer 1998: 602, fn.9, notes this and suggests that the key notion of ‘what is said*’ might remain silent on difficult cases like replacing names with definite descriptions; but, if this is the case, then the key claim of the standard view – that there is a genuinely semantically informative notion of ‘what is said’ – seems to me to have been significantly weakened. 11 Bach 2001: 15. A somewhat similar approach can, I think, be found in Reimer, 1989: 599: “The technical aspect of Grice’s notion lies in the fact that he uses ‘says’ in such a way that ‘what is said’ is very tightly constrained by the conventional meaning of the words uttered – considerably more so than it is in our ordinary, non-technical, sense of that locution.” The problems to be raised above carry over to this approach.

18

Emma Borg

Now, we should be clear that Bach himself rejects what I have called the standard view, for he denies that (certain) indirect speech reports can be used to dictate semantic content.12 But the pressing issue from our current perspective is whether something like SCC could be of help to the advocate of the standard view. Could we use such a principle to deliver a semantically informative notion of ‘what is said*’? Well, I think the answer to this question really depends on how exactly the technical definition itself is to be interpreted. For the syntactic correlation constraint (as deployed by the advocate of the standard view) tells us that to fix on the privileged class of speech reports, we need to limit our attention to just those cases where the content sentence of the indirect speech report matches the original sentence in the above respects (i.e. there is a one:one correspondence between the elements of each, their order and their syntactic character). Yet spelling out what exactly SCC itself amounts to remains, it seems, less than straightforward: it requires, first, an understanding of exactly what elements count as constituents of the original sentence, and, second, it requires that the notion of ‘correlation’ be spelt out. Bach himself concentrates on the first question – determining when an element counts as a constituent of a sentence and when it is better handled as a pragmatic introduction capturing speaker meaning instead of sentence meaning. For instance, he stresses that not all genuine syntactic constituents are voiced constituents. Thus in the exchange: A: Has Jim left? B: Yes, he has. A proper report of B’s utterance could include the verb phrase ‘left’, since this is taken to be a syntactically present but unspoken element of the sentence B produced (on the grounds that it is recoverable from the immediate linguistic context of B’s utterance). On the other hand, something like the implicit restriction to alcohol in an utterance of ‘I only drink Scotch’ is thought to be the result of pragmatic reasoning on the part of the hearer, rather than being supplied by a genuine syntactic constituent of the sentence produced.13 Clearly, this issue – of when an element can be traced to a syntactic but unvoiced constituent, and when it is a novel element introduced by pragmatic processes – is one which requires a great deal of attention.14 However, 12

Thus, since he wants to hold on to a semantically significant notion of ‘what is said’, he would resist the equivocation between ‘what is said’ and reported speech in this paper. 13 Bach 1981: 238. 14 See Bach 1994, Borg (forthcoming), Recanati (2002)

The Semantic Significance of What is Said

19

the issue I really want to concentrate on here is the second one noted above, namely how should we understand the key notion of ‘correlation’? For even if we have succeeded in determining the precise syntactic constituents of the original sentence, it remains an open question what constraints the correlation relationship puts on the elements of an indirect speech act. For instance, in some sense at least, the phrase ‘Baby Leo’s father’ in (5) correlates to the proper name ‘Blair’ in (1): both expressions play the role of determining an object – a person – as the subject of an ensuing predicative act, and both expressions turn out (given the way the world actually is) to deliver the very same object, but as we have already conceded (5) is not easily treated as semantically informative as to (1). Now of course an advocate of SCC may object that these two expressions – ‘Blair’ and ‘Baby Leo’s father’ – do not share a syntactic character, since the latter is (on most accounts) a quantified noun phrase and the former is (on most accounts) a referring term, but this move hardly seems to help. For what now is to stop an utterance of: 8)

The British Prime Minister lives at No.10

being reported with: 9) A said that the father of Leo lives at No.10. Here we have two descriptive phrases in subject position (indeed two descriptive phrases containing the same number of words), so intuitively two expressions of the same syntactic character, which are co-extensive (in the actual world), and which stand in a correlation relationship in the sense of ‘playing the same role in the sentence’, yet still it seems we don’t want to take part of the meaning of (8) to be given by (9).15 The problem emerging is that, as yet, we can’t have an informative notion of ‘what is said*’ because the technical terms involved in the definition of ‘what is said*’ themselves have to be spelt out. Yet of course it is still open to the advocate of a syntactic constraint on ‘what is said*’ to tell us more about the notion she has in mind here. For instance, it might be held that whether or not an item in the original sentence genuinely correlates to an item in the content sentence of the indirect speech report is itself a matter which can be 15

Furthermore, appeal to the extension of an expression across possible worlds can’t help either. Consider an utterance of ‘Blair is a cordate’ reported with ‘Jim said that Blair is a renate’.

20

Emma Borg

settled by intuitions (say about whether the two items count as ‘meaning the same’); however, on this approach we run the risk of collapsing back into the kind of direct appeal to semantic intuitions which we rejected above. Though we might end up with an indirect speech report which coincides with the semantic analysis assigned to the original sentence, this would be entirely parasitic on prior semantic judgements, and thus the key idea that judgements about what is said can help specify the contents of a correct semantic theory would be lost. So, then, a more appealing move would be to adopt a very tight correlation relationship, one which, say, demanded type identity – it is not enough that there be a one:one isomorphism between elements of the content sentence of the reported speech act and elements of the original sentence, what has to get mapped in each case is tokens of the very same type.16 Now, one consequence of this view, I think, is that we no longer have anything which approximates to the notion we first started with, i.e. the notion of indirect speech. We have in effect traded in oratio obliqua for oratio recta: the only kind of speech acts which will count as semantically informative are those which actually repeat the terms used in the original utterance. More worryingly for the advocate of the standard view, however, is that it is simply no longer clear what role the additional move through ‘what is said*’ is supposed to be playing. For the work of dictating the contents of a semantic theory is now being done by an independent principle relating syntax to semantics. What we have now is the claim that the contents of a semantic analysis for some sentence s is given by the syntactic contents of that sentence 16 This may seem too strong a constraint, for, as Grice was well aware, certain accommodations will have to be made for indexicals and demonstratives. An utterance of ‘I am cold’ should not (always) be reported with ‘U said that I am cold’. In these cases, the syntactic character of the expressions will (help to) determine a semantic value (e.g. the speaker of ‘I’) which should then be referentially indicated in the content sentence of the indirect speech report; see Recanati 1989: 297. Bach, however, wants to argue that even this move (at least in the demonstrative case) takes us too far from the semantically relevant notion of ‘what is said’, requiring as it does, appeal to speaker intentions to determine the referent of ‘that’; see Bach 2001: 31-3. In this case, the syntactic correlation constraint would indeed require something like type identity (so that an utterance of ‘That is G’ must always be reported with ‘U said that that is G’). In personal correspondence, Bach has indicated that the notion of an interpreted logical form might be used to spell out his correlation relationship: thus, since ‘Baby Leo’s father’ contains constituents and structure that ‘Blair’ does not, ‘they do not really correlate’, i.e. they don’t share a logical form. This move may help Bach (concerned as he is with a semantically informative notion of ‘what is said’ which is divorced from indirect speech reports), but, as we will see below, such a move cannot be of assistance to any advocate of the standard view.

The Semantic Significance of What is Said

21

– semantic meaning is extremely tightly tied to syntactic features. But this is a principle we could simply apply directly to those sentences themselves. The extra trip through ‘what is said*’ seems theoretically otiose. The point here is not so much that such appeals to syntax fail to deliver a notion which will pick out all and only the semantically relevant instances of ‘what is said’, but rather that delivering this set now seems entirely secondary to claims about how the syntactic features of our language map to semantic features. Yet, if this is the case, it seems that we have failed to deliver a genuinely semantically informative notion. It’s the determination of the relationship between syntax and semantics that is informative, and it is merely a by-product of this that we can then isolate certain instances of the locution ‘A said that p’ where p matches the semantic content of the original sentence s. To recap then, the claim is that neither of the accounts of the privileged notion of ‘what is said*’ provides us with the semantically informative notion we were after. On the one hand, there are accounts which take the set of privileged indirect speech reports to be constrained by semantic features (position (b)). Yet, to avoid circularity here, we need it to be the case that we have immediate, intuitive access to the set of semantically constrained indirect speech reports, i.e. intuitively judging whether ‘A said that the man there was happy’ counts as a literal report of ‘that man is happy’. It seems, however, that ordinary speakers are not in general able to make this kind of fine-grained judgement, and, to the extent that they are willing to make assessments of this kind, such judgements are parasitic on prior judgements about the meaning of the original sentence. Yet if this is the case, then we might as well look to these intuitions about the meaning of the original sentence as the basis for our construction of a semantic theory. On the other hand, we have accounts which seek to constrain the semantically relevant set of indirect speech reports by appeal to non-semantic features, such as syntax (position (c)). Here, however, even if the privileged, syntactic relation can be adequately spelt out, the appeal to what is said once again becomes otiose; for the work in semantic theorising is now being done by the relationship posited between syntax and semantics. Though the introduction of such a principle allows us to arrive at a set of (direct or indirect) speech reports which coincide with the content ascribed to a sentence by our semantic theory, the whole process is once again entirely secondary to the fundamental processes involved in construction of that semantic theory. The path to delivering a semantically informative notion of ‘what is said’ seems, then, fraught with difficulties; so why not simply reject the claim that indirect speech reports are in any way semantically relevant, i.e. why not reject the semantic significance of ‘what is said’? Well, one reason to resist this

22

Emma Borg

kind of move might be the thought that we need to preserve some kind of intimate connection between the two domains –indirect speech reports must have something to do with the content of a semantic theory. That is to say, it might seem initially that the only alternative to the standard view is a position like (e) above, whereby there is thought to be no relationship between indirect speech reports and a semantic theory. In the final section I want to argue that, were this the case, it might well throw us back to the so far fruitless search for a way to delimit a semantically significant notion of ‘what is said*’, but this is not the case, for opponents of the semantic significance of indirect speech reports have another option open to them.

3. The Proper Connection Between a Semantic Theory and Indirect Speech Reports To reject the standard view we must hold that there is no route from indirect speech reports to the content of a semantic theory, i.e. we must reject positions (a-c). Now, of course, one way to deny this claim is simply to deny that there is any relationship between the two realms at all; this is the claim made by (e). On (e) indirect speech reports and semantic theories belong to different domains of inquiry and each can proceed without any regard for the other. Yet it seems clear that (e) is not a very attractive position – it might not be logically impossible that there be a correct theory of meaning for some language L which plays no role at all in specifying the content of communicative acts involving the use of L (perhaps if L was a language used by a race who had direct, unmediated access to one another’s communicative intentions, so that language played for them some other, non-communicative role), but this certainly doesn’t seem plausible for our own cognitive condition. We simply cannot allow that there is no connection at all between what a correct semantic theory states for a given sentence and (correct) judgements of what that sentence may be used to say. It is perhaps the desire to avoid such an obviously unattractive position as (e) that motivates the search for a semantically relevant notion of ‘what is said*’. However, as has been evident since §1, (e) is not the only position available to those who want to deny the standard view. Instead they might pursue position (d), whereby there is a one-way connection between the two realms, but it runs only from semantic analyses to indirect speech reports. It is this position which I wish to advocate. The claim I want to make is that the output of a semantic theory is relevant to determining acceptable or correct indirect speech reports, but that this is

The Semantic Significance of What is Said

23

just because the output of such a theory provides the starting point for working out what given speakers communicate by an utterance of a given sentence. Since the semantic analysis of a sentence provides the initial input to constructions of indirect speech reports, these latter reports are clearly dependent on the output of a semantic theory, but what is important to note is that this does not entail that there is any clear or accessible route back from judgements about what is said to the output of a semantic theory. What precludes this is that, on the model I would like to promote, by the time we reach any judgement of what is said by a given utterance in a given context the peculiarly semantic contribution will have been swamped beneath the influence of a vast range of other (non-semantic) factors. Finding out what is said is always, I would argue, finding out what someone said, sometime, someplace; it is, at heart, a notion which subsumes much more than the merely semantic, and this is why we should not expect a semantically informative notion to emerge from such essentially pragmatic considerations. To sketch the kind of picture I have in mind, then, let me borrow from the framework made familiar by Chomsky, Fodor and others. I assume a broadly modular picture of the mind, containing discrete bodies of information, and encapsulated processes acting on that information, dealing with such subjects as such subjects as vision, hearing, touch, and, of course, language, commonsense physics, and, of course, language. The language faculty, as I conceive it, is comprised of at least three sub-domains: orthography and vocalised speech recognition, syntax and semantics. The semantic information the faculty contains is of quite a limited form; say just that required for generating the truth-conditions of complex linguistic items on the basis of the meaning of their parts and their mode of composition. It is this quite constrained item (the ‘minimal proposition’ in Recanati’s terminology) which feeds out of the language faculty and at this point becomes subject to a wealth of other information which the agent possesses. Judgements of what is said, I want to suggest, are essentially the output of this further process – what we get only once the purely semantic contribution has been subjected to a range of other (non-semantic) information the speaker takes to be relevant in the situation. Thus we cannot, contrary to the standard view start with indirect speech reports and move from here to the semantic content of sentences. It is, I suggest, the task of a successful semantic theory to account for certain semantic facts (such as the productivity and systemtaticity of our language) and not to account for the vast range of factors which are involved in making assessments of what is said. The proper domain of semantics is the meaning of simple and complex linguistic items, and shifting our focus from

24

Emma Borg

this to the tangential relation of ‘what is said’ serves only to confuse matters – indirect speech reports supervene on the output of a correct semantic theory, but it is a mistake to think such speech reports can reveal anything about the proper content of a semantic theory. Theorists like Recanati are, then, I believe, right when they suggest that pragmatic processes are necessary prior to any analysis of what is said.17 Thus, as Recanati writes: There is a single notion of what is said, and that is a pragmatic notion: saying, as Grice claimed, is a variety of non-natural meaning, characterised by the role which the conventional meaning of the sentence plays in the hearer’s intended recognition of the speaker’s communicative intentions … There is … no purely semantic notion of what is said.18 This recognition is at odds with the claims of a formal semantic theory (i.e. that we can deliver accounts of sentence meaning in a context-independent manner) only if ‘what is said’ is thought to be semantically relevant. Yet, as I have tried to argue, we have no reason to think this is the case. The facts about indirect speech reporting do not, from the start, support the idea that ‘what is said’ is a semantically significant notion (as Cappelen and Lepore have shown). Furthermore, if we try, in the light of this, to replace the ordinary locution with some technical notion, ‘what is said*’ which delimits all and only the semantically relevant cases, things do not seem to improve at all. For it seems the only option for yielding this refined notion of ‘what is said*’ is to appeal either to semantic properties or to syntactic properties. However, as I have argued, the first move is unable to avoid the charge of circularity, while the second approach makes the appeal to ‘what is said’ theoretically otiose. Thus we simply cannot deliver the technical notion necessary to support claims of semantic significance for indirect speech reports, but nor, I have suggested in this section, should this finding worry us. For so long as we can preserve the connection from semantic analyses to indirect speech reports we avoid collapse into the unpalatable position (e), yet without embracing the idea that indirect speech reports are semantically informative, i.e. without holding that there is a semantically significant notion of ‘what is said’.19 17 Where I think Recanati and other advocates of what he labels ‘truth conditional pragmatics’ (e.g. Sperber and Wilson’s 1986 Relevance Theory), go wrong is in thinking that this shows that the correct form for a theory of linguistic meaning must be one which is pragmatic ‘all the way down’ as it were. The argument against truth-conditional pragmatics is explored in more length in Borg (forthcoming). 18 Recanati 2001: 87-8. 19 Thanks are due to Kent Bach and Ernie Lepore. Research for this paper was made possible by the award of a Mind Fellowship.

The Semantic Significance of What is Said

25

Bibliography Bach, K. “Referential/Attributive.” Synthese 49 (1981): 219-44. —. “Semantic Slack: what is said and more.” In Foundations of Speech Act Theory, ed. S. Tsohatzidis. London: Routledge, 1994: 267-91. —. “You Don’t Say?” Synthese 128 (2001): 15-44. Borg, E. “Saying What You Mean: unarticulated constituents and communication.” In Ellipsis and Non-Sentential Speech, eds R. Elugardo and R. Stainton. Dordrecht: Kluwer Academic, (forthcoming). Cappelen, H. and E. Lepore. “On an Alleged Connection Between Indirect Speech and the Theory of Meaning.” Mind and Language 12 (1997): 278-96. —. “Reply to Richard and Reimer.” Mind and Language 13 (1998): 617-21. Grice, P. Studies in the Way of Words. Cambridge, Mass.: Harvard University Press, 1989. Nicolle, S. and B. Clark. “Experimental Pragmatics and What is Said: a response to Gibbs and Moise.” Cognition 69 (1999): 337-354. Recanati, F. “The Pragmatics of What is Said.” Mind and Language 4 (1989): 294328. —. “Unarticulated Constituents.” Linguistics and Philosophy 25 (2002): 299-345. —. “What is Said”, Synthese 128 (2001): 75-91. Reimer, M. “What is Meant by ‘What is Said’? A Reply to Cappelen and Lepore.” Mind and Language 13 (1998): 598-604. Richard, M. “Semantic Theory and Indirect Speech.” Mind and Language 13 (1998): 605-16. Sperber, D. and D. Wilson. Relevance: communication and cognition, Oxford: Blackwell, 1986. Taylor, K. “Sex, Breakfast and Descriptus Interruptus.” Synthese 128 (2001): 45-61.

26

Cara Spencer

Representing What Others Say Cara Spencer

Abstract The semantics of belief reports has recently received a great deal of attention.1 Speech reports have largely been left behind in this discussion. Here I extend a familiar recent account of attitude reports, the Russellian theory, to the special case of speech reports. I then consider how it compares to Davidson’s paratactic theory with respect to a few examples that raise special problems about speech reports. Neither theory accounts for everything we want to say about these cases. I suggest that the problem lies in an assumption common to both theories, that in reporting what others say, we aim to represent what was said exactly as the original speaker represented it, in so far as this is possible.

I. Davidson’s paratactic theory is offered as a response to a familiar problem about the logical form of speech and attitude ascription sentences. Since ‘Mark Twain’ and ‘Samuel L. Clemens’ are co-referring singular terms, we would expect that substitution of one for the other in a sentence would not change what the sentence says. Yet the speech reports in (1) and (2) seem to say different things, and as such seem like they can differ in truth value. (1) Mary said that Mark Twain wrote The Innocents Abroad. (2) Mary said that Samuel L. Clemens wrote The Innocents Abroad. Davidson’s response to the problem is the syntactically and semantically anomalous suggestion that the complementizer ‘that’ functions as a demonstrative that refers to an utterance. For Davidson, an indirect speech report (i.e., one of the form ‘A says that s’) says that the utterance to which the demonstrative ‘that’ refers stands in a certain relation to some utterance of the subject of the speech report. Although Davidson does not put it in precisely this way, his discussion clearly indicates the following reflexive semantics for speech reports: 1 See David Braun (1998); Mark Crimmins (1992); Francois Recanati (1993); Mark Richard (1990); Nathan Salmon (1986); Jennifer Saul (1997); and Scott Soames (1987), (1995). This brief list is far from exhaustive.

Representing What Others Say

27

(3) ‘A said that’ uttered in c is true IFF (›u)(A uttered u & Samesay (u, u1)) Here, u1 is supplied by the context c as the referent of the demonstrative ‘that.’ Typically, this will be the speech reporter’s utterance of the embedded sentence that immediately follows it, (making the account generally reflexive) but one could construct a case in which another utterance or inscription was selected as the referent of ‘that.’ For instance, one might write a sentence on a blackboard and indicate it while saying ‘He said that.’2 Other features of the context may help to determine what the demonstrated utterance means, so we may need to consider them in evaluating the truth of a speech report, but they are irrelevant to what the speech report says. How does this theory respond to the initial problem? Since an utterance of the embedded sentence in (1) is not identical to any utterance of the embedded sentence in (2), (1) and (2) say different things, and thus there is no logical or semantic requirement that they have the same truth value. Indeed, the paratactic account does not even require that different utterances of (1) must have the same truth value, since the utterances of the embedded sentences in each case will be different particular objects. To adequately address the problem, of course, the paratactic theory must also include an account of utterance synonymy on which utterances of ‘Mark Twain wrote The Innocents Abroad’ and ‘Samuel L. Clemens wrote The Innocents Abroad’ may fail to be synonymous. For if they are always synonymous, utterances of (1) and (2) will always have the same truth value even though they make different claims. To accommodate the typical intuition about (1) and (2), it is not enough for the theory to interpret them as saying different things. It must allow that the different things they say can differ in truth value.3 Thus, the paratactic account alone offers less of a solution to the problem than a promissory note 2 James Higginbotham has suggested that the assimilation of these cases is problematic. When ‘that’ uncontroversially functions as a demonstrative in a speech report, as it does in this case, it appears that the speech report has the same truth condition as a direct speech report instead of an indirect speech report. That is, the speech reporter must demonstrate a sentence that the original speaker actually uttered, rather than one that means the same thing as the original speaker’s utterance. See James Higginbotham (1986: 29-48). 3 David Braun has recently offered reason to doubt this. He suggests that to accommodate the intuition that (1) and (2) differ in truth value, it is enough to show that an interpreter can coherently believe that they differ in truth value, even if this belief is false. He offers this suggestion in defense of the Russellian theory, but the observation has a clear application here as well. See Braun 1998.

28

Cara Spencer

for a solution, which may or may not be provided by the account of utterance synonymy. Davidson’s view that the complementizer ‘that’ is a demonstrative in speech and other attitude reports has already come in for criticism.4 I think that many of these criticisms offer conclusive evidence against the demonstrative treatment of ‘that.’ I will not add to these criticisms here. Instead, I will focus on Davidson’s account of utterance synonymy, and consider whether he is correct to say that if a speech report of the form ‘A said that s’ is true, then the speaker’s utterance of ‘s’ and one of A’s utterances will in fact be synonymous in his sense of ‘synonymous.’ The two aspects of the paratactic approach, its demonstrative treatment of the complementizer and the account of utterance synonymy, are quite independent. Criticisms of one do not impugn the other. Even if semantic evidence does not support the demonstrative treatment of the complementizer, the account of utterance synonymy may still prove useful in a modified semantic account of speech reports. What does Davidson tell us about the relation of synonymy between utterances? Since Davidson denies that contents are entities, he does not construe the claim that utterance u and utterance u1 have the same content to require that utterance u has content c, utterance u1 has content c1, and c=c1. Rather, the relation expressed by ‘has the same content as’ is a relation between two utterances. In “Thought and Talk,” he offers the following suggestion about this relation: Suppose I say, “Herodotus asserted that the Nile rises in the Mountains of the Moon.” My second utterance--my just past utterance of “the Nile rises in the Mountains of the Moon”--must, if my attribution to Herodotus is correct, bear a certain relationship to an utterance of Herodotus’: it must, in some appropriate sense, be a translation of it.’5 The “appropriate sense” of translation is spelled out in Davidson’s account of radical interpretation. He tells us specifically that “radical interpretation, if it succeeds, yields an adequate concept of synonymy as between utterances.”6 The goal of radical interpretation is the construction of a theory of interpretation for a speaker. This theory entails what Davidson calls a ‘T-sentence’ for every sentence formable in the speaker’s idiolect. T-sentences in English are instances of the schema in (4). 4 See in particular Higginbotham 1986 and Tyler Burge (1986). 5 Donald Davidson (1984, 155-170, 167). 6 Davidson (1984, 93-108. 104 fn. 14).

Representing What Others Say

29

(4) ‘s’ is true-in-L if and only if p In (4), ‘s’ is a sentence of the speaker’s idiolect L and ‘p’ is replaced by a sentence of English that translates ‘s’ into the metalanguage, in this case English. An appropriate instance of (4) that accommodates explicitly context-sensitive expressions such as indexicals or demonstratives would have to specify both a sentence and a context, as in (5). (5) ‘s’ uttered in context c is true-in-L if and only if p Davidson does not specify more exactly how the T-sentences that compose a theory of interpretation generate a notion of synonymy for utterances. But presumably if the theory of interpretation generates a pair of sentences like (6) and (7), in which the right hand side of both biconditionals is the same, then we can count the utterances of s in c and s* in c* as synonymous. (6) “s” uttered in context c is true-in-L if and only if p (7) “s*” uttered in context c* is true-in-L if and only if p. If (6) and (7) are true, then Samesay (u, u*) is also true, where u is the utterance of s in c and u* is the utterance of s* in c*. The most general account of synonymy will also allow that utterances of sentences of different languages can be synonymous. A single theory of interpretation, which is specific to a language, will not entail both sentences, but different theories of interpretation can each yield a single sentence of the appropriate pair. One of these utterances (s or s*) is the speech reporter’s own utterance that he or she claims to be synonymous with the utterance under report, so the relevant theory of interpretation would simply be the speech reporter’s theory of interpretation for his or her own idiolect. With this, we have in hand a more complete account of what speech reports say according to Davidson. A speech report is true just in case the utterance the speech reporter demonstrates is synonymous with an utterance of the original speaker who is the subject of the speech report. Two utterances are synonymous just in case the speech reporter’s theory (or theories) of interpretation for the relevant idiolects entail two T-sentences of the general form of (6) and (7). How does a speech reporter know if his or her theories of interpretation are correct? Davidson thinks we are generally entitled to assume that other people speak largely as we do, in that they mean the same things as we would mean by their words. In principle anything the speaker says, not just the utterance under report, can undermine our entitlement to

30

Cara Spencer

this assumption. Since we interpret speech and attribute belief and desire on the basis of the same holistic evidence, we can do one of two things in the face of a highly anomalous assertion. We can suppose that the speaker really means by those words just what we would mean by them, and just has a bizarre belief. Or we can suppose that the speaker doesn’t mean by his words what we would mean by them. Just as it is impossible to interpret a person’s beliefs and desires piecemeal, it is also impossible to interpret their speech in this way. As Davidson notes, “Finding right words of my own to communicate another’s saying is a problem in translation. The words I use in the particular case may be viewed as products of my total theory (however vague and subject to correction) of what the originating speaker means by anything he says.”7 He illustrates the point with an example: Let someone say (and now discourse is direct), ‘There’s a hippopotamus in the refrigerator’; am I necessarily right in reporting him as having said that there is a hippopotamus in the refrigerator? Perhaps; but under questioning he goes on, ‘It’s roundish, has a wrinkled skin, does not mind being touched. It has a pleasant taste, at least the juice, and it costs a dime. I squeeze two or three for breakfast.’ After some finite amount of such talk we slip over the line where it is plausible or even possible to say correctly that he said there was a hippopotamus in the refrigerator, for it becomes clear that he means something else by at least some of his words than I do. The simplest hypothesis so far is that my word ‘hippopotamus’ no longer translates his word ‘hippopotamus’; my word ‘orange’ might do better. But in any case, long before we reach the point where homophonic translation must be abandoned, charity invites departures.8 For Davidson, the confused speaker’s initial utterance ‘there’s a hippopotamus in the refrigerator’ does not provide conclusive evidence for the speech report (8), at least not if the person who utters it uses ‘hippopotamus’ in the conventional way. (8) She said that there is a hippopotamus in the refrigerator. The confused speaker’s utterance offers some evidence for this assertion, but other things the confused speaker may say can defeat it. After we have gathered enough evidence that the speaker’s beliefs about what he calls a ‘hippo7 Davidson (1984, 100). 8 Davidson (1984, 100-101).

Representing What Others Say

31

potamus’ are more appropriate to oranges, he thinks it would be incorrect to assert (8). Any evidence that someone means something different from what I mean by a particular word is evidence that I should not use that word in reporting what he uses it to say. Davidson’s final remark of the quoted passage suggests that he takes the example to represent an extreme case--that even if we had less compelling evidence for this sort of divergence of meaning, the principle of charity would justify us in departing from the speaker’s actual words. A charitable interpreter in Davidson’s sense of that phrase will in general assume that those she interprets share a great many background beliefs with her, and that in general they believe largely what she thinks that she would believe if she were in their situations. Thus when we interpret the person who appears to claim to have a hippopotamus in the refrigerator, we find that we should reject the hypothesis that this is really what this person meant to say. It is of course possible that this person simply has bizarre beliefs about hippopotami, but this hypothesis stands in tension with the view that this person shares many of our background beliefs.

II. Contemporary discussion of the semantics of propositional attitudes has focused on belief reports rather than speech reports, but the various theories on offer are typically extensible in a straightforward way to the case of speech reports. My concern here is with the Russellian theory but I contrast it with the Fregean theory to clarify some of its commitments. The Fregean and Russellian theories are primarily distinguished in their treatment of the typical speaker intuition that pairs of sentences like (9) and (10) can differ in truth value. (9) Mary believes that Mark Twain wrote The Innocents Abroad. (10) Mary believes that Samuel L. Clemens wrote The Innocents Abroad. The Fregean approach treats these intuitions as trustworthy semantic intuitions, and accordingly represents belief reports with different, albeit co-referring, expressions in the complement clause as expressing different propositions. The Russellian approach takes these intuitions to concern pragmatic rather than semantic phenomena. According to the Russellian theorist, (9) and (10) express the same proposition. Where the Fregean theorist wants to construct a theory that predicts these speaker intuitions, the Russellian seeks

32

Cara Spencer

to explain them away. Thus, the Fregean approach respects the common intuition that if Mary utters (11), whether or not she also utters (12), (10) is not merely misleading but false. (11) Samuel L. Clemens did not write The Innocents Abroad. (12) Mark Twain wrote The Innocents Abroad. If, as the Russellian approach has it, (9) and (10) express the same proposition, for which Mary’s utterance of (12) provides authoritative evidence, then the mere possibility of Mary uttering (11) presents trouble. It appears to substantially undermine our evidential support for (10), but it does not appear to do the same thing for (9). And presumably if a further piece of evidence undermines our support for (10) but not for (9), (9) and (10) express different propositions. The Russellian theorists reject this line of reasoning. They claim, first, that proper names and indexical and demonstrative pronouns are directly referential, that is, they contribute only their referents to the propositions expressed by sentences containing them.9 Second, they claim that propositions are structured entities whose structure mirrors that of the canonical sentences that express them. Verbs of propositional attitude such as ‘believes’ or ‘says’ express two place relations holding between a person and the proposition expressed by the sentence following the attitude verb. Russellian theorists disagree about how a proposition is structured, and what contributions the non-directly referential expressions make to the proposition. Disagreement about the contributions of these other components of a sentence will not affect the discussion here. For our purposes, it is enough to say that on the Russellian view, ‘Mark Twain’ and ‘Samuel L. Clemens’ make the same semantic contribution to (9) and (10) and that (9) and (10) express the same proposition. Since philosophers disagree about what propositions are, the claim that (9) and (10) express the same proposition tells us very little before we know how propositions are being conceived of. For the Russellian theorist, what matters is that (9) and (10) are truth conditionally equivalent. The difficult task for Russellians is to explain why we persistently have the intuition that (9) and (10) can differ in truth value. A widely adopted strategy towards this end is to claim that these intuitions concern pragmatic rather than semantic features of attitude ascription sentences. The suggestion is that attitude ascriptions typically carry some further information beyond the truth condition, and that sentences like (9) and (10), although truth 9 See Saul Kripke (1980) and David Kaplan (1989) for canonical arguments for this view.

Representing What Others Say

33

conditionally equivalent, nonetheless convey different information. If the information (9) conveys, but does not explicitly state, is true, and the information (10) conveys is false, then according to the Russellian we have an explanation for the typical speaker’s intuitions that (9) and (10) can differ in truth value. The Russellian theorists differ widely about what additional information attitude reports convey, and how it is conveyed. Nathan Salmon, for instance, has suggested that attitude reports may “involve the conversational implicature, suggestion, or presumption”10 that the subject of the attitude report believes that the sentence within the scope of the attitude verb expresses a truth, and that the subject would verbally assent to that sentence, or a translation of it, if asked about it under normal circumstances.11 So presumably Salmon would say that (9) conversationally implicates, suggests, or presumes (13) and (14). (13) Mary believes that a translation of ‘Mark Twain wrote The Innocents Abroad’ into her language expresses a truth. (14) Mary would verbally assent to a translation of ‘Mark Twain wrote The Innocents Abroad’ into her language if asked about it under normal circumstances. These same Russellian theorists have also suggested that the implicatures may concern representations associated with the sentence inside the attitude verb rather than the sentence itself, so that, for instance, (9) conversationally implicates something like (13*).12 (13*) Mary believes that Mark Twain wrote The Innocents Abroad under the representation associated with a translation of ‘Mark Twain wrote The Innocents Abroad’ into her language. Since utterances of (9) will have some or all of the implicatures in (13), (14) and (13*), and utterances of (10) will have different implicatures, the proposal is that speakers actually intuit that the implicated sentences can differ in truth value (or do differ in truth value in some hypothetical situation), and 10 Salmon (1986, 115). 11 See Salmon (1986, 115). Scott Soames suggests that the anti-Russellian intuition that (9) and (10) differ in truth value results from a confusion of the semantic content of a sentence relative to a context and the information conveyed by an utterance of that sentence. See Soames (1987, 219). 12 See Salmon (1986, 116) and Soames (1995).

34

Cara Spencer

they simply confuse their intuitions about the implicated propositions with their intuitions about the propositions literally expressed. Whether these propositions are suggested or in some sense conveyed to the audience is hard to assess in the absence of a specific account of how the sentence generates the suggested or conveyed proposition if it is not a part of what the sentence literally says. The proposal that (13), (14) and (13*) may be conversational implicatures as Grice conceived of them is more concrete in this respect and thus easier to assess. I call this the ‘Gricean pragmatic proposal’ in what follows. Future references herein to the Russellian theory should be understood to include the Gricean pragmatic proposal. Grice says that an implicature differs from the literal meaning of an utterance in several ways.13 First, a speaker can cancel an implicature of one of her utterances by immediately qualifying what she says. For instance, saying ‘they fell in love and got married’ typically implicates, but does not strictly say, that they fell in love before they got married. I can cancel this implicature by adding ‘but not in that order’ to my initial utterance. In so doing, I do not contradict myself, as I would if I tried to cancel the literal meaning of my utterance. Conversational implicatures are also nondetachable, that is, any utterance that expresses the same proposition as u will also have u’s conversational implicatures. The implicated content is thus tied to the proposition expressed rather than the means of expressing it. There is some intuitive plausibility to the view that we can cancel the supposed implicatures of (9) by following up with, for instance, ‘but Mary wouldn’t put it that way herself.’ It is less plausible that these implicatures are nondetachable in Grice’s sense. First, at least according to the account Salmon suggests, the implicated proposition typically is a function of the sentence uttered rather than the Russellian proposition it expresses. Since there is no route back from the proposition expressed to the sentence uttered, it is not clear how the proposition alone would determine any of the implicated propositions in (13), (14), or (13*). Furthermore, the Russellian supposes not only that sentences like (9) and (10) are truth-conditionally equivalent, but also that they are equivalent to any sentence ascribing to Mary a belief that Twain wrote The Innocents Abroad, where Twain (and the novel as well) are picked out by any directly referential term. According to the Gricean pragmatic proposal, many of these sentences will differ in their standard conversational implicatures. For instance, utterances of (10) will typically implicate that Mary recognizes ‘Samuel L. Clemens’ as a name for the author of The Innocents Abroad. There is thus no clear reason to expect 13

See H. Paul Grice (1975).

Representing What Others Say

35

that if an utterance u of a belief report sentence s implicates p, that there will be some other utterance u1 of another sentence s1 that will have the same truth conditional content and the same conversational implicature as u. Clearly, the Russellian will acknowledge that there are many utterances that have the same truth conditions as u but differ in their conversational implicatures. How would the Russellian approach apply to the case of speech reports? A straightforward extension of this approach would hold that (1) and (2) express the same proposition. (1) Mary said that Mark Twain wrote The Innocents Abroad. (2) Mary said that Samuel L. Clemens wrote The Innocents Abroad. Under what circumstances is an utterance of (1) or (2) true? Both sentences specify the content of something Mary said. So presumably (1) and (2) are true just in case Mary in fact said something with that content. The Russellian approach just instructs us to look to the semantic theory for the rest of the language, which allows us to determine the content of Mary’s utterances. Since defenders of the Russellian approach accept that proper names are directly referential, they will say that “Mark Twain wrote The Innocents Abroad” and “Samuel L. Clemens wrote The Innocents Abroad” have the same content, so an utterance of either one will make both (1) and (2) true. There will of course be utterances of other sentences that have the same semantic content as the complement clause of (1) and (2), such as those containing demonstratives or indexicals that refer to Twain. The extension of the Gricean pragmatic proposal is somewhat less straightforward. Since saying does not imply believing, the implicated sentence in (14) is clearly not implicated in a speech report like (1). Further, there is no obvious and simple revision to (14) that specifies what (1) plausibly implicates. The accounts of the implicated proposition in (13) and (13*) seem amenable to an extension to speech reports, but here again the extension is not straightforward. The result of simply substituting the ‘believes’ of (13) and (13*) with ‘says’ is palpably mistaken in (15) and barely coherent in (15*). (15) (15*)

Mary says that a translation of ‘Mark Twain wrote The Innocents Abroad’ into her language expresses a truth. Mary says that Mark Twain wrote The Innocents Abroad under the representation associated with a translation of ‘Mark Twain wrote The Innocents Abroad’ into her language.

36

Cara Spencer

Clearly, Mary can simply say that Mark Twain wrote The Innocents Abroad without saying anything about the truth value of a translation of a sentence into her language. So (15) cannot be right. (15*) is hard to interpret, but the guiding idea behind (13*) is that the implicated sentence conveys something about how the subject believes the designated proposition. We say something by uttering a sentence, so in this sense saying something is always saying it in some way just as believing something is always believing it in some way. Plausibly, we can understand the proposition implicated by a speech report to specify the way in which the content was said, just as the proposition that a belief report implicates simply specifies the way in which the subject believes the content. It specifies the way in which the speaker said something by specifying the sentence the speaker used to express it. Of course a speech reporter might use a different language from that used by the original speaker. Thus we might construe the proposition that (1) implicates along the lines of (16). (16) Mary uttered a translation of ‘Mark Twain wrote The Innocents Abroad’ into her language. The account suggested in (16) of the implicature of a speech report bears a close resemblance to Davidson’s view of the truth condition of the speech report in (1). There are of course some important differences between them. Centrally, they differ in syntactic structure. Davidson takes a speech report to be composed of two independent utterances, one of which contains a demonstrative referring to the other. On Davidson’s account, (1) comes out true in exactly the same circumstances as (16), provided that the notion of translation in (16) is understood as Davidson understands it. The Davidsonian understanding of translation is of course not implicit in (16), and this is a second difference between the (16) and Davidson’s account of (1).

III. My view, as I suggested at the outset, is that the idea that (16) is a part of the literally expressed or implicated content of (1) is what causes problems in treating two problems that arise particularly for speech reports. In this section, I present these two kinds of problematic cases and consider how each theory handles them. Davidson claims that in reporting speech indirectly, we should aim to capture what the speaker meant by his words in so far as this differs from what

Representing What Others Say

37

the typical speaker would mean by those words. Direct speech reports, in which the speaker’s words are enclosed in quotation marks, are used to report what words the speaker uttered. I agree with Davidson that there is a difference between reporting the words someone used and reporting what that person meant by them. But Davidson is wrong to suggest that indirect speech reports are always used to report what the speaker meant rather than what he or she said. Many examples fit awkwardly into Davidson’s model, and these examples suggest that indirect discourse reports can be used in both ways, to report the speaker’s words (or what those words conventionally mean) and what the speaker meant by them. For instance, George W. Bush, then Republican nominee for the Presidency of the United States, said (17) during a campaign stop in Bentonville, Arkansas.14 (17) They misunderestimated me. It is forced, at the least, to say that we can report what he said using (18) but not using (19). (18) He said ‘They misunderestimated me’ (19) He said that they misunderestimated him. Davidson has devoted considerable space to the discussion of malapropisms like this one. What he has to say about them requires a considerable revision to the account of radical interpretation on which the paratactic theory is based.15 On the earlier theory, our linguistic competence is represented in a static recursive theory of interpretation for the language(s) we understand. Our tacit understanding of these theories permits us to interpret one another. Examples like (17) raise a problem for the idea that a language is a fixed entity of which speakers have a prior understanding. The problem is that in context, we understand (17), but this understanding cannot result from any application of a theory of interpretation for English, since ‘misunderestimate’ is not a word of standard English, nor is its meaning in (17) compositionally derivable from the meanings of its component verb and prefix. It seems to be synonymous with ‘underestimate,’ but a static theory of interpretation would give interpreters no reason to expect that the prefix is semantically inert in this instance. Davidson’s response to the problem is that interpretation involves a kind of interplay between two theories of interpretation 14 See 15 See Davidson (1986).

38

Cara Spencer

for a language. The first, which he calls the ‘prior theory,’ more closely resembles the static theory of interpretation for the language from the earlier account of radical interpretation, and the second, called the ‘passing theory,’ incorporates context specific cues about how certain expressions are to be understood, which may be explicitly offered by the speaker or merely implicit in the speaker’s use, as in (17). These cues give interpreters a reason to revise the prior theory, but the revision is specific to the current conversational context. In another conversation, the prior theory will be revised in different ways, to generate a different passing theory, which is the theory actually used to interpret speakers. Davidson does not specify how this new account might be accommodated in the paratactic theory of speech reports, which relies on the old notion of radical translation that is called into question here. That theory, as explained above, requires that the utterance demonstrated in a speech report be a translation of one of the subject’s utterances into the speech reporter’s idiolect. The issue before us concerns the paratactic theory’s position about the correctness of (19) and (20) as reports of Bush’s utterance in (17). (20) He said that they underestimated him. Even if the speech reporter can recover the correct passing theory to interpret (17), and understands ‘misunderestimate’ to be synonymous with ‘underestimate,’ the speech reporter may, but need not, alter her own idiolect so that she means by ‘misunderestimate’ exactly what George W. Bush meant by it when he uttered (17). Problematically for Davidson’s account, how the speech reporter would express what Bush said in her own words does not appear to be relevant to a decision about the correctness of (19) and (20). What matters here is that the speech reporter and her audience recognize that there is an idiolect in which ‘underestimate’ and ‘misunderestimate’ are synonymous. They need not also speak this idiolect themselves. Such speakers and audiences will take (19) and (20) to be perfectly correct speech reports. The case also raises a problem for the traditional version of the paratactic theory that antedates Davidson’s distinction between the prior theory and the passing theory of interpretation. On a static theory of interpretation for English, ‘misunderestimate’ is meaningless. Thus, it offers no explanation of the fact that an audience can understand, to say nothing of report, the utterance of (17). One upshot of this example is that we are not always constrained to charitably report what others said, by glossing over their errors and reporting a touched-up version of their utterances. It is sometimes per-

Representing What Others Say

39

fectly correct to report what they said, errors and all, when we know that the speaker is conscious of the mistake after the fact. This point is also illustrated in a report of part of John F. Kennedy’s famous speech in Berlin, during which he said (21).16 (21) Ich bin ein Berliner. What Kennedy meant to say, of course, is that he considered himself in some sense with the people of Berlin. What he did not know beforehand is that the article ‘ein’ would not be used in German as the article ‘a’ would in the English construction ‘I am a New Yorker.’ To identify yourself as a resident of Berlin, you would say ‘Ich bin Berliner.’ In German, the construction ‘ein Berliner’ in fact refers to a particular sort of jelly doughnut widely available in Berlin. Thus, (22) expresses the literal meaning of the sentence Kennedy uttered. (22) I am a jelly doughnut. Since Kennedy uttered (21), and (22) expresses the literal meaning of (21) in English, it appears that we could correctly report what Kennedy said in English by uttering (23). (23) Kennedy said that he is a jelly doughnut. Here, we have not reported what Kennedy meant to say, but instead we report what he in fact said. If Davidson thinks we must report what the speaker meant to express, then he must think that (23) is incorrect. This simply runs counter to my intuition on the matter, according to which (23) is perfectly correct as a report of Kennedy’s utterance of (21), even if the speech 16 See the full text of his speech at the United States National Archives and Records Administration website <www.nara.gov/exhall/originals/kennedy.html>. There is some reason to believe that the reports of Kennedy’s verbal blunder in Berlin are at best exaggerated. Even though the phrase “ein Berliner” is used throughout Germany (though apparently not in Berlin) to refer to a particular sort of jelly doughnut, some suggest that the sentence “Ich bin ein Berliner” is ambiguous to some native speakers between “I am a jelly doughnut” and “I am, figuratively speaking, one with the people of Berlin.” If the utterance is ambiguous in this way, then it does not raise the problem I have indicated. See <www.wikipedia.com/wiki/Talk:Ich_bin_ein_Berliner>. At any event, the mere possibility that a speaker could make a mistake like the one Kennedy is commonly thought to have made is enough to raise a problem for these accounts of speech reports.

40

Cara Spencer

reporter understands that (23) is not faithful to Kennedy’s communicative intentions. These cases raise a difficulty for Davidson’s account only on a certain understanding of Davidson’s view that a translation of what someone says is simply a specification of what that person meant by uttering those words. I am assuming in my discussion of the Kennedy example that when someone makes the sort of mistake in expression illustrated in this example, what he means is not what he says. Thus if we are constrained in indirect speech reports to say what was meant, it will be a mistake to report the literal meaning of what was said. Davidson may reply that these cases do not present problems for the paratactic theory, simply because of an important difference between verbal error and systematic differences in idiolect. There may be an important difference for Davidson between a systematic idiosyncrasy in one person’s speech and an isolated mistake. If there is such a difference, then Davidson may want to say that systematic idiosyncrasies are to be translated into the speech reporter’s idiolect in indirect speech reports, but reports of mistaken speech are a special case that the paratactic theory simply does not address. Even if such a distinction could be adequately drawn, however, the problem still arises for the paratactic theory. Consider the teenager born in Boston, Massachusetts, who like many of her peers systematically uses the word ‘wicked’ as a synonym for ‘very.’ If this teenager utters (24), we can report this speech equally correctly with (25) or (26). (24) It’s wicked cold outside. (25) She said that it’s very cold outside. (26) She said that it’s wicked cold outside. As we saw already in the ‘misunderestimate’ case, the correctness of the report in (26) has nothing to do with the meaning of ‘wicked’ in the speech reporter’s own idiolect. A use of (26) may require that the speaker and audience understand this local idiosyncratic use of ‘wicked,’ but it does not require that either would actually express themselves in this way. On Davidson’s view, such speakers would have to use (25) rather than (26), and here again the view appears forced. These cases raise several problems for the paratactic theory as Davidson defends it as well as the revision that accommodates his distinction between the prior and passing theory. According to both theories, the speech reporter demonstrates an utterance that means, in his mouth, what the speech he reports meant in the mouth of its original utterer. This appears to require that the speech reporter take on the idiolect of the original speaker for the

Representing What Others Say

41

purposes of the speech report. It strikes me that this is mistaken. All that is required is that the speech reporter (and his or her audience) recognize that there is an idiolect, spoken at one time by the original speaker, on which the demonstrated utterance has a particular meaning, even if they would express that same meaning differently. The Boston teenager example shows that the point holds for systematic idiosyncrasies as well as isolated mistakes. Second, as is demonstrated in the Kennedy case, we can correctly report someone as having said p when we recognize he did not mean to say p, if what he in fact said conventionally means p. The Russellian theory does not treat these three cases uniformly. As Russellian theorists have had little or nothing to say about malapropisms and how they acquire meaning, it is not easy to discern how the view applies to the ‘misunderestimate’ case. If we suppose that the theory will respect the interpretive hypothesis that ‘misunderestimate’ means something like ‘underestimate,’ then the Russellian view will affirm that both (19) and (20) are correct and indeed equivalent reports of (17). As I construe the Gricean pragmatic proposal, (in (16)) (19) and (20) differ in their conversational implicatures. (19) correctly implicates that the speaker used the word ‘misunderestimate’ where (20) incorrectly implicates that ‘underestimate’ was used. Approximately the same account goes for the third example in (24)-(26). Since, in the Kennedy example, the Russellian proposition he literally expressed differs from the one he meant to express; the report in (23) is true. We would literally speak falsely if we reported him as saying what he meant to say. When reporting mistaken or idiosyncratic speech, we can correctly report that a speaker said what his words literally meant in the public language he or she was speaking. We can equally correctly report someone as having said what he meant to say, provided that the context of original utterance provided adequate cues to discern this intended meaning. It is difficult to make theoretical sense of this, but these appear to be the uncontroversial facts about everyday discourse.17 The common practice of reporting speech by summarizing what it says presents a second problem case for both the paratactic account and the 17 If, as in the Kennedy case, what the speaker literally said conflicts with what he or she intended to say, another difficult problem arises. We can report such speakers as having said that p, and we can report them as having said that q, where p and q conflict. Yet I do not think we would say that such speakers have contradicted themselves. Saying that each report is only appropriate in certain conversational contexts does not get around the problem, since the problem concerns the truth value of the speech reports rather than their contextual appropriateness. One might say in this case that one or both of the speech reports are literally false, but still convey relevant information to the audience. I suggest such a response to the second problematic case below.

42

Cara Spencer

Russellian theory. Suppose, for instance, that A is giving a paper at a conference. B attends the talk but C does not, and afterwards, C asks B what A said. Typically when someone asks this sort of question after a talk, one is not asking for a report of any particular sentence the speaker uttered. And though it would not be strictly speaking false for B to reply “A said that she would start by drawing a few important distinctions,” this is probably not the sort of answer C had in mind. Typically, one expects some sort of summary of the talk in response, something like “A said that recent discussions of the explanatory gap have all missed an important distinction.” This can be the right answer (or at least one of the right answers) to C’s question, even if A never uttered a sentence that exactly, or even roughly, matches the content of the complement clause of B’s assertion. If B’s answer is a correct answer to C’s question, then presumably it is true. This possibility presents a problem for the paratactic theory and the Russellian theory. What does the Russellian theory say about this example? If a proposition is a structured entity whose structure corresponds to the structure of the sentence that expresses it, then a summary will not express the same proposition as what it summarizes. The Russellian is compelled to say in this case that we do not correctly report what someone said by summarizing it, unless the summarizing sentence happens to be one very similar to a sentence the original speaker actually uttered. The Gricean pragmatic proposal does not help explain what is going on in this case. It says that a summarizing speech report just implicates, falsely in this case, that the original speaker uttered a particular sentence. So neither the semantic nor the pragmatic proposal explains the intuition that a summarizing speech report can be true. Would a Davidsonian radical interpreter find that a summary has the same content as what it summarizes? For Davidson, the content sentence of a true speech report will have the same content as one of the original speaker’s utterances. We typically think of utterances as utterances of sentences, but there is no reason that an hour-long lecture should not be construed as a single utterance. Does a single sentence summary have the same content as the hour-long lecture it summarizes? According to Davidson, two utterances have the same content if the relevant theory (or theories) of interpretation says that they are true under the same circumstances. These theories of interpretation are compositional, and since the parts and structure of the summary will invariably differ from the parts and structure of the summarized speech, the theory of interpretation will not calculate their truth conditions in the same way. Since the paratactic theory allows that utterances of structurally dissimilar sentences can be synonymous, it individuates meanings less finely than the Russellian theory. Still, the structural dissimilarity between a

Representing What Others Say

43

lecture and a summary of that lecture is extreme, and a lecture may well be about some topics or issues that the summary of it does not mention. Hence, we should expect that Davidson’s account of utterance synonymy would tell us that a lecture is not synonymous with any brief summary of it. Thus, the paratactic theory offers the wrong result about this case as well. An alternative for both the paratactic and the Russellian theorist is to claim that any speech report that summarizes what was said is false, and still grant that it conveys relevant information to the audience. One can imagine many cases in which someone could mislead her audience by speaking the literal truth, and can help the audience towards a better understanding of the matter at issue by saying something literally false. For instance, a high school science teacher might explain why a projectile moved in a particular trajectory by citing Newton’s laws of dynamics. These classical laws are strictly speaking false, yet they still explain the motion of middle-sized, slow moving bodies. The high school teacher may note that the laws she cites in her explanation are not strictly speaking true, since they make false predictions for very fast-moving objects, but there is no reason that she should have to do this to completely and correctly explain the motion of the projectile to her students. We might say something analogous about the summarizing speech report, that it is false but still offers a better explanation than any true report. I do think this is a response of last resort, and a better theory would allow that a speech report that summarizes rather than translates what someone said could be true.

IV. The paratactic theorist and the Russellian theorist agree that when we report what others say, we have an obligation to correctly identify not only what was said but also the way in which it was said. It is an obligation in that what we say is either false or pragmatically misleading unless we specify both the content of the speech act and the way the speaker represented that content. In the cases I have considered here, another consideration about how contents should be represented in speech reports appears to be at work. In these cases, speech reports seem incorrect or misleading when they represent the content of a belief so that the audience is unable to recognize it. In summarizing or explaining what someone said, we aim to make what that person said accessible to our own audience, even if in doing so we sacrifice some degree of faithfulness to the original speaker’s manner of representing that content. Similarly, a report of idiosyncratic speech that preserves the idiosyncrasy is

44

Cara Spencer

appropriate only when the speaker has reason to believe that his or her audience will interpret the idiosyncrasy correctly, as the original speaker intended. Neither the paratactic theory nor the Russellian theory adequately accommodates this consideration. According to Davidson, when reporting what someone said, we always aim to utter a content sentence that means, in our mouths, what the utterance under report meant in its speaker’s mouth. What I have suggested is that this is not our only aim in every case. In some cases, we should aim to utter a content sentence that our audience will best understand, even we could utter other content sentences whose meaning is closer to the original speaker’s intended meaning. Since the Gricean pragmatic proposal, as I have understood it, says that speech reporters pragmatically implicate something similar to what the paratactic theorist thinks they literally say, it is unsurprising that the same problem arises.18

Bibliography Braun, David. 1998. “Understanding Belief Reports.” Philosophical Review 107, no. 4 (October 1998): 555-595. Burge, Tyler. 1986. “On Davidson’s ‘Saying That’.” In Lepore, 1986, 190-210. Cole, P. and J. Morgan (ed.). 1975. Syntax and Semantics 3: Speech Acts. New York: Academic Press. Crimmins, Mark. 1992. Talk About Beliefs. Cambridge, Mass.: MIT Press. Davidson, Donald. 1986. “A Nice Derangement of Epitaphs.” In Lepore 1986, 43346. —. 1984. Inquiries into Truth and Interpretation. New York: Oxford UP. —. 1975. “Thought and Talk.” In Mind and Language. Edited by Samuel Guttenplan. New York: Oxford UP, 1975. Reprinted in Davidson 1984. page references from this reprint. —. 1968. “On Saying That.” Synthese 19 (1968-9): 130-46. Reprinted in Davidson 1984, 93-108. page references from this reprint. Grice, H. P. 1975. “Logic and Conversation.” In Cole and Morgan 1975, 41-58. Higginbotham, James. 1986. “Linguistic Theory and Davidson’s Program in Semantics.” In Lepore 1986, 29-48. “Ich bin ein Berliner” <www.wikipedia.com/wiki/Talk:Ich_bin_ein_Berliner> (2 May 2002). Kaplan, David. 1989. “Demonstratives.” In Themes From Kaplan, edited by Joseph Almog et al., 481-564. New York: Oxford UP. Kripke, Saul. 1980. Naming and Necessity. Cambridge, Mass.: Harvard UP. 18

I am grateful to Michael Glanzberg for helpful discussion.

Representing What Others Say

45

Lepore, E. (ed.). 1986. Truth and Interpretation: Essays on the Philosophy of Donald Davidson. Oxford, England and Cambridge, Mass.: Blackwell. National Archives and Records Administration, <www.nara.gov/exhall/originals/ kennedy.html> (2 May 2002). Recanati, Francois. 1993. Direct Reference. Oxford, England and Cambridge, Mass.: Blackwell. Richard, Mark. 1990. Propositional Attitudes: An Essay on Thoughts and How We Ascribe Them. Cambridge, England: Cambridge UP. Salmon, Nathan. 1986. Frege’s Puzzle. Cambridge, Mass.: MIT Press. Saul, Jennifer. 1997. “Substitution and Simple Sentences” Analysis 57: 102-8. Soames, Scott. 1995. “Beyond Singular Propositions?” Canadian Journal of Philosophy 25: 515-50. —. 1987. “Direct Reference, Propositional Attitudes, and Semantic Content” Philosophical Topics 15 (1987): 47-87. Reprinted in Scott Soames and Nathan Salmon, 1988: 197-239. Page references from this reprint. Soames, Scott and Nathan Salmon, editors. 1988. Propositions and Attitudes. New York: Oxford UP. Weisberg, Jacob and Bryan Curtis (ed.). 2002. “The Complete Bushisms” Slate. (2 May 2002).

46

Jonathan Sutton

The Things People Say Jonathan Sutton

Abstract It appears that the objects of belief and the objects of assertion are, often, one and the same. The objects of assertion must be communicable – if an assertion leads to successful communication, the audience grasps what the speaker said. There are good reasons for thinking that beliefs are relations to very fine-grained contents, however, which appear to be unsuitable for reliable transmission from speaker to audience. I consider two accounts of the apparent intersection of the objects of belief and the objects of assertion, and find them unable to embrace both of these claims. I defend the view that beliefs have multiple, truth-conditionally equivalent contents on the grounds that it is able to reconcile the apparently conflicting claims.

Introduction You can believe what I say, and you can say – or, more idiomatically, state – what I believe. I will say some things relating to the objects of belief and assertion in this article, and you can believe them; I do. For example, I will say that beliefs have multiple truth-conditionally equivalent contents, and I hope that you will believe what I say. I further hope that you will tell others what I believe about these matters. You might even believe everything that I say about belief and assertion, although I will not have the space to state everything that I believe about these issues. The statements that I made in the preceding paragraph suggest that the objects of belief (and other propositional attitudes) and the objects of assertion (and similar speech acts) intersect. At least some of the things that a thinker believes can be asserted by himself and others. At least some of the things that a speaker asserts (or questions, or denies) can be believed by himself and others. I will explore three ways in which we might try to explain appearances here. Each way will be grounded, naturally, in views about the nature of the objects of belief and assertion. I will start by examining a simple, natural account of the objects of belief and assertion and their intersection. Most philosophers nowadays will reject the simple account, and I am among their number. Having rejected the simple account, I will explore an account that coheres much better with views about the objects of belief

The Things People Say

47

held by many contemporary philosophers of mind and language. I will argue that the associated view of the objects of assertion is inadequate. I will then offer my own view of how the objects of belief and assertion intersect based on the idea that the contents of beliefs are multiple. Each of a thinker’s token beliefs is a belief in many things: truth-conditionally equivalent contents, some coarser- and some finer-grained. I cannot hope to survey even a fraction of the possible or actual views on the objects of assertion and belief. Nevertheless, I consider the defects of the two alternative views that I do consider to be representative of the similar defects that the many undiscussed views have, and by showing how my own view avoids these defects, I aim to provide a reason for adopting my own view.

The Simple Account The simple account starts by postulating that belief ascriptions of the form ‘S believes that p’ such as ‘Henry believes that cows are docile’ are made true by two-place relations between a believer, in this case Henry, and a proposition, in this case the proposition that cows are docile denoted by ‘that cows are docile’. The proposition denoted by ‘that cows are docile’ occurring in the belief ascription is identical to that expressed by an unembedded utterance of the sentence ‘cows are docile’. That proposition is the meaning of the sentence ‘cows are docile’ and is grasped by all who understand that sentence. Reports of speakers’ assertions of the form ‘S says that p’ such as ‘Henry says that cows are docile’ are made true by two-place relations between a speaker, in this case Henry, and a proposition, in this case the proposition that cows are docile denoted by ‘that cows are docile’. One important way of saying that cows are docile is saying ‘cows are docile’. By uttering a sentence whose meaning is the proposition that cows are docile one puts onself in ‘‘the saying relation’’ to that proposition; one who believes what you say puts themselves in ‘‘the belief relation’’ to that same entity. One who says what you believe puts himself in the saying relation to the proposition that is the content of your belief by uttering a sentence whose meaning is that proposition. Perhaps there are propositions that can be believed but are for some reason inexpressible; perhaps they can only be grasped by a single individual, and are the meaning of no actual or possible sentence. Perhaps there are propositions that are assertible and are the meanings of some sentences but cannot

48

Jonathan Sutton

be believed; perhaps they are too unspecific or vague to be objects of belief. Perhaps the proposition that cows are docile is identical to the proposition that female domesticated bovines are docile, and perhaps it is not (that is, propositions might be more or less Millian or more or less Fregean). Perhaps the term ‘cow’ has the same meaning as ‘female domesticated bovine’ and perhaps it does not. Perhaps propositions are purely abstract entities existing in Frege’s third realm and perhaps they are classes of sentences or types of mental representation. Resolving these issues and deciding which of these possibilities is genuine are not important here; the essence of the simple account is clear, and it is clear in outline how the simple account explains the overlap between the objects of belief and the objects of assertion. Many philosophers – in particular, most contemporary philosophers who lack Millian sympathies – will find the simple account unacceptable. In particular, many will reject the claim that ‘Henry believes that cows are docile’ expresses a two-place relation between Henry and the proposition that is the meaning of ‘cows are docile’ and which is expressed by unembedded occurrences of that sentence. There are two main reasons for rejecting the claim. The first is that belief contents, whatever they are, must be much more finely individuated than the meanings of sentences. The second, related reason is that the contribution made by ‘that cows are hostile’ to the truth-conditions of ‘Henry believes that cows are hostile’ is context-sensitive in a way that the meaning of ‘cows are hostile’ is not. The that-clause does not, then, contribute solely the meaning of the embedded sentence to the truth-conditions of the belief ascription. Both reasons can be illustrated by considering Kripke’s Paderewski puzzle (Kripke, 1979). Individuating propositions finely enough to meet the challenges posed by this development of Frege’s puzzle leads to the rejection of the simple account. A thinker, Peter, believes that there are two people named ‘Paderewski’, one of whom is a musician, the other a statesman; in fact, there is only one Paderewski who is both a musician and a statesman. If the thinker believes that Paderewski is going to be in town, then his belief will have distinct effects depending upon whether he believes that Paderewski the musician is coming or whether he believes that Paderewski the statesman is coming – in the former case, the thinker might try to purchase concert tickets, and in the latter, he might attempt to avoid the traffic problems that inevitably accompany politicians’ visits. It seems that we must handle the Paderewski puzzle in the way that we handle more orthodox variants of Frege’s puzzle such as the Hesperus/Phosphorus case. We will distinguish two contents, each of which would be true

The Things People Say

49

were Paderewski to arrive. One content is partially individuated by its association with the belief that its subject is a musician, the other by its association with the belief that its subject is a statesman. Two belief contents correspond to a single sentence, ‘Paderewski is going to be in town’, unambiguous in the relevant sense since the name ‘Paderewski’ is unambiguous in the relevant sense. Those contents cannot be the meaning of that sentence despite the fact that we might report beliefs with those contents by using that sentence; the contents are finer-grained than the meaning. Let us now consider two thinkers who believe there to be but one Paderewski, although one believes him to be a musician and is ignorant of his being a statesman, and the other believes him to be a statesman and is ignorant of his being a musician. It seems that we must deny that these two thinkers believe the same content, just as our isolated thinker seemed capable of believing two contents concerning Paderewski. One suspects that any difference in what two thinkers believe about an individual N gives rise to the possibility of a third thinker who can entertain two contents that he does not know are truth-conditionally equivalent, and that this entails that our two thinkers entertain distinct contents about N. The number of meanings of any name of N, and hence of meanings for sentences containing any name of N, is much fewer (and is often one); contents about N are much finergrained than the meanings of sentences about N. Kripke’s case also provides us with argument for the context-sensitivity of the truth-conditions of attitude ascriptions, and, in particular, for the context-sensitivity of the contributions that the that-clause makes to those truth-conditions. I will now briefly explain how Kripke’s puzzle about belief motivates the claim that attitude ascriptions have context-sensitive truthconditions. Much better explanations are found in, for example, Richard (1990) and Crimmins and Perry (1989). Recall that Peter believes that the man named ‘Paderewski’ who is a pianist is no politician, and that the man named ‘Paderewski’ who is a politician is no pianist. We seem to want to endorse both the claim: (1) Peter believes that Paderewski is a pianist. And the claim: (2) Peter believes that Paderewski is not a pianist. We do not, however, wish to say that Peter has contradictory beliefs – his problems derive from ignorance, not from irrationality. That is, we are reluctant to say that:

50

Jonathan Sutton

(3) Peter believes that Paderewski is and is not a pianist. Furthermore, we seem to have some inclination to assert that: (4) Peter does not believe that Paderewski is a pianist. And now we appear to be contradicting our own endorsement of the claim that Peter does believe that Paderewski is a pianist. If attitude ascriptions have context-sensitive truth-conditions, all of the claims ((1), (2), and (4)) that we want to endorse can be true without it being the case that Peter has contradictory beliefs (that is, that (3) is true). In different contexts, the term ‘that Paderewski is a pianist’ makes different contributions to the truth-conditions of the containing ascription (one popular way of fleshing out this claim will be outlined in the nest section). In some contexts, (1) is true and (2)-(4) are false. In other contexts, (1) is false, and (2) and (4) are true. All of the claims that we want to endorse are true in some context, although there is no context in which they are all true; hence, we do not contradict ourselves. Moreover, in no context is (3) true; Peter’s beliefs are not contradictory.

Hidden Indexical Accounts In light of both the apparent context-sensitivity of the truth-conditions of belief ascriptions and of the fact that fundamentally belief appears to be a relation to very fine-grained contents, the hidden indexical theory of belief ascription semantics has gained popularity (there are a number of variants of the theory offered by, among others, Fodor (1990); Schiffer (1979); Crimmins and Perry (1989); Crimmins (1992); Richard (1990)). In this section, I will explore whether an extension of the hidden indexical theory to reported speech might help explain how the objects of belief and the objects of assertion can coincide, and offer reasons for scepticism. Hidden-indexical theorists have claimed that attitude ascriptions contain a silent (that is, not phonetically realized) predicate satisfied by modes of presentation which is responsible for the context-sensitivity of the ascription. The logical form of an attitude ascription is something like ‘( ∃m)(P(m) & B(S, p, m))’ where ‘P’ is a predicate that specifies a type of mode of presentation of the Millian proposition p expressed by the sentence forming the ascription’s that-clause, ‘m’ is a variable ranging over modes of presentation,

The Things People Say

51

and B is the belief relation. Generally, the contribution that a word or phrase within a that-clause makes to determining the type of mode of presentation P of p is determined by pragmatic rather than semantic factors. That sentences (1) and (4) above are both true despite apparently contradicting each other is explained by the fact that the pragmatically determined hidden indexical predicate differs for each sentence. In case (1), Peter is said to think of the Millian proposition that Paderewski is a pianist under a Paderewski-thepianist mode of presentation type; in case (4), he is said (by use of a different hidden indexical predicate) to think of that Millian proposition under a Paderewski-the-statesman mode of presentation type.1 It is not necessary to appeal to hidden indexical predicates to explain the apparent context-sensitivity of attitude ascription truth-conditions. One might instead embrace a theory according to which the belief relation is the two-place relation between a thinker and a proposition that it appears to be rather than the three-place relation that hidden indexical theorists claim. However, the proposition denoted by the that-clause is what varies between one context of utterance and another. Propositions are identified with something more like Fregean thoughts than the Millian entities of the hidden indexical theorist – with what the hidden indexical theorist calls modes of presentation of propositions. (Recanati (1993) offers a theory of this kind.) Why, then, adopt the more complex hidden indexical account? Precisely because of the fine-grained character of belief contents (modes of presentation of Millian propositions in the hidden indexical theorist’s terminology). We are not generally in a position to refer to or otherwise specify the finegrained contents of others’ beliefs.2 It is hard to see, then, how such finegrained contents might contribute directly to the truth-conditions of our belief ascriptions; it is far more plausible to suppose that belief content types contribute to those truth-conditions, as the hidden indexical theory states. Is there a hidden indexical theory of reported speech? In principle, there is, and it runs as follows. The logical form of a ‘says that’ sentence is something like ‘(∃m)(P(m) & A(S, p, m))’ where ‘P’ is a predicate that specifies a type of mode of presentation of the Millian proposition p expressed by the sentence 1 More sophisticated versions of the theory (for example, Crimmins (1992)) take account of the fact that propositions and their modes of presentation are structured, and that individual words and phrases in an ascription’s that-clause correspond to modes of presentation that are constituents of the mode of presentation of the whole proposition expressed by the that-clause. 2 Crimmins once claimed otherwise (Crimmins, 1992); he has subsequently modified his view.

52

Jonathan Sutton

forming the that-clause, ‘m’ is a variable ranging over modes of presentation, and A is the assertion or saying relation. Once again, the contribution that a word or phrase within a that-clause makes to determining the type of mode of presentation P of p is determined by pragmatic rather than semantic factors. How might the objects of belief and the objects of assertion intersect for a hidden indexicalist? There are two natural strategies to explore. The first view has it that when I believe what you say, or you say what I believe, there is a specific mode of presentation of a type specified by a single hidden indexical predicate such that one of us bears the B relation to it (and a Millian proposition) and the other bears the A relation to it. The fine grain of modes of presentation should be a concern here. It is unlikely that different thinkers often share specific modes of presentation (indeed, if modes of presentation are token mental representations as Crimmins (Crimmins, 1992) suggests, it is impossible). Consequently, it is rare that one thinker bears the B relation to the same mode of presentation to which a different thinker bears the A relation. Believing what others say and stating what others believe is not so rare, however. The second strategy accepts that such a coincedence of modes of presentation is rare, but denies that it is required for me to believe what you say or for you to state what I believe. If we concentrate on the kinds of concrete case that motivated our enquiry in the first place, this becomes apparent. Suppose I say that cows are docile and you believe what I say. Well, this simply means that you believe that cows are docile. All that is (relevantly) required for the sentences ‘I said that cows are docile’ and ‘You believed that cows are docile’ to be true on the current view is that I bear the A relation to some contextually specified type of mode of presentation of the Millian proposition that cows are docile, and that you bear the B relation to some contextually specified type of mode of presentation of that same proposition. It need not be the same mode of presentation for each of us, nor even, depending upon how the view is fleshed out, the same type of mode of presentation. We can say that there is a regular intersection of the objects of assertion and belief if we take these to be Millian propositions. Or we can regard such intersection as merely apparent, an illusion generated by an overly simple understanding of the logical form of belief ascriptions and indirect speech reports (namely, an understanding of them as two-place relational statements). Further discussion of the two strategies is, I suggest, fruitless, since the proposal to extend the hidden indexical view of belief ascriptions to reported speech is not comprehensible as it stands, and there is reason to doubt that it

The Things People Say

53

can be made so. The problem lies in the A relation introduced as the counterpart of the B relation that figures in the truth-conditions of belief ascriptions. We have motivated the idea that belief is at root a relation to very finegrained modes of presentation of Millian propositions (and the propositions themselves – a qualification that will remain tacit hereafter), and seen how certain aspects of belief ascription truth-conditions can be explained thereby. But how could assertion be at root a relation to very fine-grained modes of presentation? One of the motivations for the quantificational structure of belief ascriptions that the hidden indexicalist posits is that modes of presentation are so exceptionally fine-grained that they are rarely, if ever, shared between distinct thinkers. What one says, on the other hand, had better be the kind of thing that an audience can grasp for themselves for successful communication seems to require an agreement on what is being said, and we would like to think that it occurs reasonably often. It might be suggested that the entities quantified over in reported speech are not identical to the entities quantified over in belief reports, and all that is required for one person to believe what another says (and suchlike) is that the believer’s cognitive mode of presentation present the same Millian proposition as the speaker’s assertive mode of presentation. However, until we are told what assertive modes of presentation are, this is no more than a gesture at a view, and cannot be discussed further. Alternatively, one might suggest that assertions are always caused by psychological states that are relations to fine-grained modes of presentation, and the A relation holds between the mode of presentation that is involved in the psychological state that causes the speaker’s assertion, the speaker, and the Millian proposition that the speaker ‘‘expresses.’’ In asserting that a speaker said that p, one is implicitly specifying the type of mode of presentation involved in the causation of an ‘‘expression’’ of a given Millian proposition. This is no improvement on the idea that, fundamentally, one asserts very fine-grained modes of presentation, however. The relation between finegrained mode of presentation and speech act cannot be causation simpliciter since indefinitely many of a speaker’s psychological states cause any given assertion, and the speaker clearly does not ‘‘express’’ the Millian propositions associated with all those states. We need a relation that holds in the typical case (in which a speaker is not exploiting ambiguity to ‘‘express’’ several Millian propositions with a single speech act) between a single mode of presentation, the speaker, and a Millian proposition. We should still be baffled at how a relation to such fine-grained modes of presentation could have anything to do with assertion.

54

Jonathan Sutton

I suggest that we must abandon the idea that either belief or assertion fundamentally involves a three-place relation between exclusively very finegrained entities, Millian propositions, and speakers or believers.

The Multiple Contents View In a nutshell, I propose that we abandon the view that belief is fundamentally a relation to a single fine-grained mode of presentation – the view that beliefs have unique, fine-grained contents. In its place, I suggest that token beliefs have multiple truth-conditionally equivalent contents, each located on a continuum between very fine-grained modes of presentation such as concerned the hidden indexicalist and very coarse-grained contents such as, at the limit, Millian propositions. Belief reports are made true by a traditional two-place relation between a believer and a content; which content that is, and, in particular, how fine-grained it is is a context-sensitive matter. ‘That p’ in ‘S believes that p’ picks out differently ‘‘grained’’ contents in the different contexts in which that sentence can be used – which of those contents is referred to on a given occasion is determined by the same pragmatic factors that, according to the hidden indexicalist, determine the hidden indexical predicate. What the hidden indexicalist regards as a type of content specified by the hidden indexical predicate I regard as a content simpliciter. I have argued for this view elsewhere on the basis of its ability to solve problems concerning psychological explanation (Sutton, 2001); here, I will argue for the view only implicitly, by showing how it can account for the intersection of the objects of assertion and the objects of belief. Before exploring how this picture helps explain the apparent intersection between the objects of belief and the objects of assertion, I should say what these truth-conditionally equivalent multiple contents of belief are. Firstly, I take the objects of belief to be mental representations rather than merely abstract entities; I will not argue for this claim here, although the reader will no doubt be familiar with a number of such arguments (I contribute one myself in Sutton (2002)). A token belief (or other propositional attitude) is a relation to a token mental representation or symbol; contents themselves are mental representation types. Which types? In the first instance, semantically individuated mental representation types. For example, in believing that Paderewski is coming to town, Peter is in a particular belief state that is a relation to a particular token of a formally individuated mental representation F. F is a token of a certain

The Things People Say

55

formal type, but it is also a token of many semantic types. Many of those semantic types are contents on my view. For example, F contains a subsymbol referring to Paderewski, and a subsymbol that expresses the property of coming to town, and those subsymbols are structurally related in a distinctive way. I have just specified a semantic type that F instantiates, and I claim that that type is one of the many contents of Peter’s belief state – indeed, it is equivalent to the coarse-grained Millian proposition that Paderewski is coming to town. The subsymbol that refers to Paderewski is also tokened in another belief state of Peter’s as part of a ‘‘mental sentence’’ that also contains a subsymbol that expresses the property of being a pianist, let us suppose. I have just specified another semantic type that F falls under – it contains a subsymbol with certain referential properties that also occurs in a distinct mental sentence with certain truth-conditional properties. Indeed, I have specified two semantic types that the Paderewski subsymbol instantiates – one of those is the Millian ‘‘concept’’ Paderewski, and the other is the concept that you have if you think of Paderewski as a pianist, a concept that is finer-grained than its Millian counterpart. In believing that Paderewski is a pianist, Peter is in a single belief state that has at least two contents. One of those contents contains the coarse-grained Millian concept Paderewski, and the other contains the finer-grained concept that requires its possessor to think of Paderewski as a pianist – that has a certain conceptual role. Much finer-grained conceptual roles will individuate much finer-grained concepts that other contents of Peter’s belief also have as constituents. For example, suppose that Peter believes that Paderewski studied at the Warsaw Conservatory. That belief enters into a conceptual role that defines an even finer-grained concept that enters into a third content that Peter’s belief has. The multiple contents that I am claiming propositional attitudes have vary along the dimension of fineness of grain; they do not vary in their truthconditions. Similarly, the more or less finely individuated concepts that are constituents of those more or less finely individuated contents share their extensions with corresponding constituents in other contents that a given attitude possesses. (I have said that many of the semantically individuated mental representation types under which token mental representations fall are concepts or complete contents. Just which ones? I suggest that the semantic symbol types that are contents (and concepts) are those that have some robust explanatory role in theorizing about the mind, and that we should understand that explanatory role in quite broad terms – if a semantic symbol type is explanato-

56

Jonathan Sutton

rily useful in philosophy or in commonsense folk-psychological explanation, for example, then it picks out a content or concept.) Our previous problems in accounting for the intersection of the objects of belief and assertion lay in the fact that the objects of belief appeared to be exceptionally fine-grained, which made them unsuitable as the kind of things that could be communicated, and therefore grasped by many minds. On the multiple contents view, token beliefs are indeed relations to very finegrained contents. They are also relations to coarse-grained contents, however, which are therefore suitable for serving in addition as the objects of assertion. Typical cases of reported speech – utterances of the form ‘S says that p’ – are made true by two-place relations between a speaker and a typically coarse-grained object of assertion. ‘That p’ on occasion might refer to a maximally coarse-grained entity – a Millian proposition. However, it need not do so; on other occasions, the very same complement clause will denote a finer-grained (although still communicable!) entity. The denotations of that-clauses in reported speech are as context-sensitive as are the denotations of belief report that-clauses. Suppose that you and I both believe that there are two people named ‘Paderewski’, one a pianist and one a statesman. There is but a single Millian proposition corresponding to ‘Paderewski was Polish’. However, there is surely a difference between my trying to tell you that the pianist is Polish and my trying to tell you that the statesman is Polish, and the difference remains even if I use the words ‘Paderewski was Polish’ in my attempt to communicate. The truth-conditions of an utterance of ‘I said that Paderewski was Polish’ depend on the context – am I reported to have made an assertion about the person I consider a pianist and not a statesman or the person I consider a statesman and not a pianist? My assertion of ‘Paderewski was Polish’ possesses one of the fine-grained contents indicated in the preceding paragraph; let us suppose that I intended to speak of the person that I think of as a pianist. Does my utterance of ‘Paderewski was Polish’ have the coarse-grained Millian content as well as the fine-grained content? That is, do token assertions have multiple contents just as token beliefs do? Indeed, we can imagine a whole range of contents my assertion of ‘Paderewski was Polish’ might have whose grain is intermediate between the Millian content and the fine-grained content. Is there a Paderewski-the-musician was Polish content that my assertion could have in addition to its Paderewski-the-pianist was Polish content? And, if so, does it actually have that content in addition to the finer-grained content? These questions are very important, clearly, but our explanation of how the objects

The Things People Say

57

of belief and assertion intersect is neutral between them. It rests only on the view that beliefs have multiple contents, not the view that assertions do so. I hence put these questions aside for a future occasion. In light of the fact that my assertion of ‘Paderewski was Polish’ can have one of two fine-grained contents, we should take the objects of assertion to be mental representations rather than the representations of a public language. There is only one public language name ‘Paderewski’, but two concepts that we are supposing that you and I have. We can account for the differing truth-conditions that ‘I said that Paderewski was Polish’ appears to have when uttered in different contexts by taking ‘that Paderewski is Polish’ to denote distinct mental reprentation types in different contexts. Just as on the simple account with which we began our enquiry, to say that p is to express a content. That content is not an abstract proposition (although there is no harm in calling it a proposition, provided we realize that it is composed of concepts rather than objects and properties, and is often finergrained than its Millian counterpart). It is, as the hidden indexicalist view of reported speech would concur, a psychological entity. However, there is no particular problem with understanding how such an entity could be expressed (and thereby communicated) since it is now clear how a content can be sufficiently coarse-grained for such purposes. The expression relation needs further explication, of course, but any problems that remain are not directly tied to the intersection of the objects of belief and assertion, I suggest. Moreover, we have preserved the parallel treatment of belief ascriptions and reported speech that was a feature of both the simple account and the extended hidden indexical view. Not only that, we have retained the simple account’s simple account of the logical form of belief and assertion reports – they are two-place relational sentences. We have at the same time allowed that reports of both kinds have context-sensitive truth-conditions, which allows us solutions to Kripke’s puzzles about belief (and, I indicated, reported speech) that parallel those of the hidden indexicalist. I cannot deny that many problems remain. The multiple contents view of belief reports is broadly Fregean, and inherits the problems that Fregean theories face – problems with compositionality, iteration, anaphoric reference, semantic innocence, and so on. I am confident that those problems can be resolved, but that resolution I leave to another occasion since the problems have no direct connection to the intersection problem which I have set out to resolve.

58

Jonathan Sutton

Bibliography Crimmins, M. 1992. Talk About Beliefs, MIT Press: Cambridge, MA. — and Perry, J. 1989.‘‘The Prince and the Phone Booth: Reporting Puzzling Beliefs.’’ Journal of Philosophy, 86: 685-711. Fodor, J. A. 1990. ‘‘Substitution Arguments and the Individuation of Belief.’’ In A Theory of Content, 161-176. MIT Press: Cambridge, MA. Kripke, S. 1979. ‘‘A Puzzle about Belief.’’ In Meaning and Use, A. Margalit A., ed., 239-83. Reidel: Dordrecht. Recanati, F. 1993. Direct Reference: From Language to Thought. Blackwell: Oxford. Richard, M. 1990. Propositional Attitudes: An Essay on Thoughts and How We Ascribe Them. Cambridge University Press: Cambridge. Schiffer, S. 1979. ‘‘Naming and Knowing.’’ In Contemporary Perspectives in the Philosophy of Language, edited by P. French, T. Uehling, and H. Wettstein, 61-74. University of Minnesota Press: Minneapolis. Sutton, J. 2001. ‘‘Russellian Thoughts, Fregean Thoughts, and the Unique Content Principle,’’ under review. —. 2002. ‘‘Are Concepts Mental Representations or Abstracta?’’ Forthcoming in Philosophy and Phenomenological Research.

Reported Speech and the Epistemology of Testimony

59

Reported Speech and the Epistemology of Testimony Sanford C. Goldberg

Abstract Speech reports of the form ‘A said that p’ are sometimes used by a speaker S as a reason in support of S’s own claim to know that p – in particular, when S’s claim to know is made on the basis of A’s testimony. In this paper I appeal to intuitions regarding the epistemology of testimony to argue that such ‘testimonial’ uses of speech reports ought to be ascribed their strict de dicto truth conditions. This result is then used as the basis for the claim that, no matter how they are used, all speech reports of this form ought to be ascribed their strict de dicto truth conditions. I conclude by offering a characterization of the content of the notions of saying and what is said, and by making some programmatic remarks regarding the role of these notions in semantic theory.

1. Introduction There are various uses for speech-reporting sentences of the form ‘A said that p’. Suppose that someone wants to know how to evaluate a speech act A made on a particular utterance-occasion; in this context, a sentence of the form ‘A said that p’ could be used to provide the semantic content of A’s speech act. But by no means is this the only use of speech reports of this form. Suppose A, a sententious and long-winded speaker, lectures for some time about the evils of lying. In this context, reporter S might then use ‘A said that lying is wrong’, not to characterize the semantic content of any particular speech act A made in the course of her lecture, but rather to summarize that lecture. Or (to take another example) suppose that A, a police chief in charge of a murder investigation, is asked whether Jones has been charged with the murder, and replies by uttering ‘No one has been charged.’ Under such circumstances, reporter S might use ‘A said that Jones has not been charged with the murder,’ not to characterize the semantic content of A’s utterance, but rather to identify the relevant implicature of A’s speech act. In fact, cursory reflection on the uses of speech reports would support two claims. First, there are many distinct uses for speech reports of the form ‘A said that p’, corresponding to the many distinct conversational aims one might have in making such a report (see e.g. Cappelen and Lepore 1997).

60

Sanford C. Goldberg

Second, for a good many of these reports, the report can seem true even if the sentence occurring obliquely expresses a proposition that was never presented-as-true in any speech act performed by source speaker A. These two claims suggest the possibility that no simple, unified account of the truth conditions of speech reports of the form ‘A said that p’ is to be had.1 In this paper I argue to the contrary that, notwithstanding the variety of uses for such reports, speech reports of the form ‘A said that p’ are to be assigned their strict de dicto truth conditions. This thesis, which I will call the broad de dicto thesis (‘broad’ because it covers all speech reports of the relevant form), asserts that a speech report made by way of the utterance of a sentence of the form ‘A said that p’ is true if and only if (1) there is an utterance U produced by speaker A such that ‘A said that p’ is offered as a report of U; and (2) ‘that p’ (as used on the occasion of the speech report) designates what is in fact the semantic (propositional) content of U, where the semantic (propositional) content expressed by a sentence is individuated in terms of the concepts expressed by the words making up the sentence.2 Though my overarching aim is to establish the broad de dicto thesis, my main focus will be on a narrower thesis. I will focus on those reports in which speaker S uses a report of the form ‘A said that p’ to manifest the testimonial support of S’s own claim to know that p. I will call this use of speech reports the testimonial use. The main burden of my paper is to argue that, in their testimonial use, speech reports of the form ‘A said that p’ should be assigned their strict de dicto truth conditions (the narrow de dicto thesis). After arguing for the narrow de dicto thesis, I will use this thesis as a basis from which to argue for the broad thesis.3

1 Something like this conclusion seems to be endorsed by Cappelen and Lepore 1997. 2 Suppose Jones utters ‘Water is wet’ and thereby expresses the proposition that water is wet – a proposition composed by the concepts WATER and WETNESS. Suppose that Smith, aware that water is H2O, reports Jones’ speech act by uttering ‘Jones said that H2O is wet’. On the strict de dicto reading, Smith’s report is false despite the identity of water and H2O. 3 In order to make the discussion as simple as possible, throughout it I will restrict the sentences occurring in oblique position to those in which there occur no referring expressions such as names, indexicals, or referentially-used pronouns. Broadening the scope of the present discussion to take account of such sentences would require complications not germane to the main topic.

Reported Speech and the Epistemology of Testimony

61

2. Testimonial Speech Reports and Epistemic Buck-Passing It should be uncontroversial that on some occasions sentences of the form ‘A said that p’ are used by S in reaction to the question “How do you know that p?” – namely, in those cases in which S takes herself to know that p on the strength of A’s word. This is simply a fact about one use of the verb ‘to say’: on such occasions it is used to represent another speaker (or another speaker’s speech act) as the source from which we acquired some belief (or on the basis of which we made some assertion). When used in this testimonial way, a speech report of the form ‘A said that p’ serves the function of passing the epistemic buck. To illustrate, suppose S asserts that p and is subsequently challenged to justify this claim. Since S takes herself to know that p on the strength of A’s word, she replies by uttering ‘A said so.’ (I take this to be shorthand for ‘A said that p’.) The point of S’s making such a speech report is to pass the justificatory burden on to A. Assuming S is entitled to pass the epistemic buck in this way, the justificatory burden – that is, the burden of providing epistemic support for the claim that p4 – is passed from S to A. We might say that (if true) S’s testimonial speech report discharges S’s epistemic obligation regarding her claim that p, passing that burden onto A. Of course, if A’s source testimony is somehow epistemically wanting – say, A lacks the relevant authority regarding the truth of her claim – or, alternatively, if for some reason S was not warranted in taking A at her word, then S’s claim that p fails to be justified in the testimonial way.5 I submit the foregoing as a schematic account of those cases in which one speaker S represents another speaker A as having said that p, as part of S’s defense of her own assertion or belief that p. Given that the passing of an epistemic buck involves the passing of a justificatory burden, it is worth enquiring into the conditions under which S would be entitled to pass the buck in this way. Our question is this: when is one speaker, S, entitled to pass to another, A, the burden of justifying S’s own assertion (or belief ) that p? The short answer is that S must be authorized to do so. In typical testimony cases, this authorization comes in the form of A’s having presented-as-true the proposition that p – for example, by having asserted that p. For, in asserting 4 An adequate answer typically takes the form of providing a warrant or justification (reasons, evidence, etc.), or else identifying yet another speaker whom one takes to be in a position to provide such a warrant or justification. (Throughout this paper I aim to remain neutral on issues of warrant and justification.) 5 For an account of the conditions for the transmission of epistemic status from testifier to hearer, see Goldberg 2001.

62

Sanford C. Goldberg

that p, A presents herself both as committed to the truth of the proposition that p and as being in a position to entitle others (including S) to do the same.6 In this way we can discern a connection between the testimonial use of speech reports, on the one hand, and the semantic content of the source utterance, on the other. The semantic content of the source utterance is a content that the source speaker A, in virtue of having performed a speech act in which that content is presented as true, makes available as a content to be believed and/or asserted by anyone who observed A’s speech act; and any belief acquired or assertion made on that basis is potentially testimonially warranted (where actual testimonial warrant is a matter of the satisfaction of the various conditions on testimonial transmission7). The flip side is that, having observed and understood A’s source speech act, an audience S who comes to believe and/or assert that content on the basis of A’s say-so can pass on to A the burden of justifying the testimonially-based belief or assertion. These considerations suggest a way to approach the semantics of the testimonial speech reports themselves. We begin by examining cases in which an audience S, having asserted that p, aims to pass the epistemic buck for that claim to a source speaker A; these will be cases in which S replies to the question “How do you know that p?” by saying that A said so (i.e., by saying that A said that p). Next we consult our intuitions regarding whether A’s source utterance authorized S to pass the epistemic buck to A in this way; and we treat S’s speech report as true if and only if the answer is affirmative. In short, the proposal is that we can discern the truth conditions of the testimonial use of speech reports by appeal to intuitions regarding the conditions for epistemic buck-passing. Such will be the methodology I follow in this paper.8 I will argue that in at least some cases in which speech reports of the form ‘A said that p’ are used in this testimonial way, the truth conditions of such reports are the strict ones, i.e., those embodied in the de dicto reading of such reports. Then I will argue that this result can be used to establish that testimonial speech reports as a class should be ascribed their strict de dicto truth conditions (= the narrow de dicto thesis). And finally I will suggest that the very same considerations can 6 The claim that the speech act of assertion is to be understood in this way is something I will simply assume here; others have argued for it at length (see e.g. Brandom 1994, especially pp. 171-5) or else for claims that entail it (see e.g. Williamson 1998). 7 See Goldberg 2001 for a discussion. 8 In so doing, I recognize that this will seem to be an unusual way to do semantics. But I ask for the reader’s patience as I develop the idea here, since I think that, despite its unusualness, this way to do semantics has some real virtues. I discuss some of these in the programmatic conclusion of this paper.

Reported Speech and the Epistemology of Testimony

63

be developed to yield the conclusion that , however they are used, all speech reports of the form ‘A said that p’ ought to be ascribed their strict de dicto truth conditions (= the broad de dicto thesis). Initial appearances to the contrary will then be explained away.

3. Application To fix ideas, I will begin with an illustration of a case in which the source speaker’s denial of having said what an audience reports her as having said provides grounds for the source speaker’s repudiation of the passed buck.9 A utters, ‘Ralph had the golden scepter in his possession.’ S observes and understands this utterance and takes A to be relevantly authoritative. What is more, S is aware of the custom that the golden scepter in question is always and only in the possession of the person of honor. Consequently, S comes to believe (and takes herself to know) that Ralph was the person of honor. Later, S asserts that Ralph was the person of honor. But Q, aware of the epistemic support for S’s assertion but alive to the possibility that Ralph came to possess the scepter in a devious way, challenges S: “How do you know that Ralph was the person of honor?” Clearly, S cannot pass the buck to A by saying that A said so. For, if S did so, A would be entitled to deny that she said that Ralph was the person of honor, and so to repudiate S’s attempt to pass the epistemic buck to her. I take this as patent. This example points to the need for a principled way to determine the conditions under which a source speaker A would be entitled to deny having said what she was reported to have said, in those cases in which another speaker S reports A as having said that p as part of S’s own attempt to pass the epistemic buck.10 To a first approximation, my claim is that the de dicto stan9 This example is modeled on an example from Goldberg 2001. 10 At this point an objection might occur in some quarters: why do we need a principled way to settle such issues? Why couldn’t we make such judgements on a case-by-case basis (i.e., without a general principle to cover all cases)? Anyone who is attracted to what below I will call the context-sensitive approach to the truth conditions of speech reports would be attracted to such a view. However, below I will argue that this view overlooks (or dismisses) the context-invariant nature of the intuitions that we bring to bear in evaluating buck-passing uses of speech reports. In particular, there are strong intuitions about what a given speaker would be entitled to repudiate in the way of a report of her speech, whether or not in the given context she chooses to do so. Below I argue that the truth conditions of speech reports (and the correlative notion of ‘what is said’) should be made to square with these intuitions. (I thank Ted Schatzki for indicating the need to make this point.)

64

Sanford C. Goldberg

dard is the appropriate standard to use: a source speaker A is entitled to deny having said what she was reported to have said, and in so doing to repudiate the passed buck, when and only when the semantic (propositional) content of A’s source utterance is distinct from the semantic (propositional) content expressed by the content-specifying sentence in the relevant speech report.11 11

I say ‘to a first approximation,’ since there is an issue that I will not have time to treat here, but which must be treated in a complete discussion of the testimonial use of speech reports. It concerns what Cappelen and Lepore 1997 call ‘partial overlap’ reports, i.e., those in which what is reported is part but not all of what is said: as when A utters ‘Margaret wore a spiffy pink suit’ and S reports this with ‘A said that Margaret wore a suit’. (The example comes from Richard 1998.) These reports raise a difficulty for my approach to the semantics of testimonial speech reports. Presumably, S could pick up a belief whose content is that Margaret wore a suit, from A’s testimony whose content is that Margaret wore a spiffy pink suit – in which case, if challenged regarding how she knows that Margaret wore a suit, S could respond with the speech report, ‘A said that Margaret wore a suit’. The problem, of course, is that this speech report does not satisfy the strict de dicto truth conditions. This suggests, contrary to my (‘first approximation’) thesis, that testimonial speech reports can be true even when the sentence occurring obliquely does not designate the very proposition expressed by the source utterance. There are two distinct ways in which this objection can be met, consistent with the spirit of the present approach to the semantics of speech reports. The first is to weaken the claim regarding the semantics of speech reports so that it reads as follows: a speech report is true so long as the reported portion of what source speaker A said answers to the strict de dicto standard. (Partial overlap reports would then be a special case in which the reported portion is part but only part of what the source speaker said.) The second option would be to surrender the claim that the partial-overlap speech report is true, while modifying the conditions for epistemic buck-passing in testimony cases so that they no longer require the truth of a speech report. We might propose such a modification in terms of the concept of conveying, as follows. Let us say that a proposition p is conveyed by an utterance U if and only if (i) p is entailed by the proposition expressed by U and (ii) p can be obtained from the proposition expressed by U by semantic subtraction. And let us say that the proposition that p is obtainable by semantic subtraction from the proposition that q just in case (one) there is a sentence S1, used by the speaker on occasion O, which on O expresses the proposition that q; and (two) the proposition that p can be obtained either (a) as the result of deleting (or ‘subtracting’) the semantic contribution of zero or more semantically significant expressions occurring in S1 (as these expressions are used on O), or else (b) by placing the result of this semantic subtraction in the scope of suitable quantifier-expressions. (A quantifier is suitable when and only when the result of placing an open sentence in its scope is a proposition that is entailed by the source proposition that q.) With these definitions in place we can then offer the following as our modified account of the conditions of epistemic buck-passing: S is warranted in passing the epistemic buck to A, in the form of a testimonial speech report of the form ‘A said that p’, iff A’s source utterance conveyed that p. I do not have time to develop these suggestions further here; further discussion e.g. of the notion of conveying can be found in Richard 1998 and in Goldberg 2001. The point is simply that, though partial overlap cases present a prima facie problem for he present proposal regarding the semantics of speech reports, they need not be taken to present an in-principle difficulty.

Reported Speech and the Epistemology of Testimony

65

I take it as uncontroversial that, if a buck-passing speech report is to count as true, then the content-specifying ‘that p’ must designate a proposition that agrees in truth conditions with the proposition expressed by A’s source utterance12 (although see footnote 11). I want to focus instead on the contention that the content-specifying ‘that p’ must designate a proposition that is (not merely equivalent in truth conditions, but also) equivalent in conceptual content to that of A’s source utterance. The brunt of the argument here turns on intuitions that are bound up with the testimonial transmission of knowledge and warranted belief. The following example (though not the point I want to make with it!) is based on one found in Cappelen and Lepore 1997. Suppose Francois utters (1) Chartreuse is Maria’s favorite color. Now, having heard Francois utter (1), Charlotte comes to believe (and takes herself to know) that chartreuse is Maria’s favorite color, and so (demonstrating a dress which she knows to be chartreuse) utters (2) The color of that [demonstrated] dress is Maria’s favorite color. Then Vanessa challenges Charlotte: “How do you know that the color of that [demonstrated] dress is Maria’s favorite color?” Charlotte responds by uttering (3) Francois said that the color of that [demonstrated] dress is Maria’s favorite color. (Charlotte’s response is informed by her belief that her audience includes many who are unfamiliar with the word ‘chartreuse’; this is why she employs another means for representing what Francois said.) But Francois (who is among those who do not know the color of chartreuse) protests this report: “I never said that. What I said is that chartreuse is Maria’s favorite color. I said nothing about that [demonstrated] dress; truth be told, I don’t know whether it is chartreuse or not!” I now want to provide two independent considerations in defense of the claim that in at least some cases such as this, where the source speaker fails to have some relevant auxiliary information (here, the information that the 12

This may be denied by some. In any case, the sorts of consideration I present below support my contention here.

66

Sanford C. Goldberg

demonstrated dress is chartreuse), a report such as is effected by (3) will be false. The first consideration in support of this contention involves a focus on how matters stand from the source speaker’s (here, Francois’) perspective; the second involves a focus on how matters stand from the generic third-person perspective. From Francois’ perspective, treating (3) as a true report will place an unfair epistemic burden on him. This is backed by the intuition that it would be wrong to use (3) to report Francois’ utterance of (1), given that the point of Charlotte’s report in (3) is to pass the epistemic buck to Francois, and that Francois does not know whether the demonstrated dress is chartreuse. I submit that the intuitive wrongness of such a report, in which the content-specifying sentence is one that the source speaker himself does not accept, reflects the unfairness of passing on a justificatory burden that, to that source speaker at least, appears to be a greater burden that what he is presently prepared to take up. In particular, to Francois, the task of justifying the claim that Maria’s favorite color is the color of the demonstrated dress involves the burden of establishing that the demonstrated dress is chartreuse, and this is a burden that (given his belief corpus) he is not presently prepared to take up. Given that Francois never commented about the color of that dress and that he doesn’t know whether that dress is chartreuse, it would appear that forcing him to take on this justificatory burden is patently unfair. It will be said in reply that Charlotte’s testimonial use of (3) is intended to pass the buck only with respect to the proposition that chartreuse is Maria’s favorite color, with Charlotte herself taking epistemic responsibility for the auxiliary claim that that [demonstrated] dress is chartreuse. But this reply only reinforces the hypothesis that we should regard the report made by way of (3) as false. Regarding it as true flies in the face of the considerations above: that Charlotte’s use of (3) aims to pass the epistemic buck, and that (3) (so used) has the effect of passing to Francois a justificatory burden that he is at present unprepared to shoulder. Short of an answer to why we should regard this use of (3) as true under these circumstances, the idea that Charlotte is only passing the buck with respect to the source proposition itself serves merely to reinforce the impression that (3), used testimonially under these circumstances, is false. This last point can be reinforced by shifting perspectives and reflecting on how matters stand from the generic third-person point of view. A good many of our beliefs are picked up in situations involving coordinated activity with multiple speakers, each of whom can attest to and pick up information by way of making or observing speech acts, and each of whom has his or her independent corpus of background beliefs in terms of which (s)he evaluates the first-blush plausibility of others’ claims. In such a state of affairs, it makes

Reported Speech and the Epistemology of Testimony

67

perfect sense for everyone to adopt a generic policy of employing – and, when one is an observer, taking as true – only those speech reports that are true on their de dicto reading. I will return to this point below in connection with my argument for the broad de dicto thesis. The point for now has to do with putting oneself in the best position to respond, in the eventuality that one discovers that one of one’s beliefs, in which another’s (or several others’) testimony might be implicated, is false. To return to the case above, even though Charlotte knows that the color of the demonstrated dress is chartreuse, Frank (or any other arbitrary third party to this matter) may not know this. And given the possibility that Maria’s favorite color is not the color of that dress – which, relative to Frank, remains an epistemic possibility – Frank may want to know how, were he to discover that such a possibility is actual, he should react to it. Assuming that he would want to react to it in the most epistemically conservative way (so as to preserve whatever true beliefs he has), Frank would be well-served to represent the distinct contributions of Charlotte and Francois to the envisaged epistemic predicament. In particular, Frank would do well to represent Francois as responsible for the propositional content that chartreuse is Maria’s favorite color, and to represent Charlotte as responsible for the propositional content that the dress in question is chartreuse. It is worth noting what he would surrender if he did not represent things in this manner. Two costs can be identified, both of which can be brought out by imagining circumstances in which Frank, who has not represented the distinct contributions of Francois and Charlotte, were to acquire evidence that the color of that dress is not Maria’s favorite color. Under such conditions (which again remain an epistemic possibility to Frank), Frank would foreclose on an epistemic option that he should want to keep open: he would no longer be in a position to treat himself as (potentially) epistemically entitled to retain the belief that chartreuse is Maria’s favorite color;13 and he would no longer be in a position to appraise the relative trustworthiness of Francois and Charlotte’s respective testimony: he would not be in a position to treat as untrustworthy only Charlotte’s testimony. Perhaps it is worth pointing out that, in arguing for the de dicto reading of (at least some) speech reports on the basis of considerations of the epistemology of testimony, I am putting a third-person spin on a phenomenon that has traditionally been presented in the first-person way. Traditionally, the de 13

Such information would be useful to him, for example, if he wants to know what, if anything, to tell the salesperson at the store at which he wants to buy Maria a blouse for her birthday.

68

Sanford C. Goldberg

dicto reading of speech reports (and of attitude reports in general) is motivated by appeal to the perspective of the person to whom the speech (or attitude) is being attributed: the idea is that if we aim to capture that person’s epistemic perspective then we need reports that are true on their de dicto reading. But in effect I am trying to motivate the role of the de dicto reading of speech reports by appeal, not to the perspective of the source speaker whose speech is being reported, but rather to the perspective of the reporter: the idea here is that, since the report is part of a highly-sophisticated and coordinated effort to pick up knowledge and warranted belief from others’ testimony, if the reporter aims to participate in this effort in a way that will maximize her chances at acquiring useful information in a reliable way (and preserving such information in the face of the need to revise her beliefs), she needs to employ and accept only testimonial reports that are true on their de dicto reading. My point has been that her doing so will enable her to react to the need for belief revision in the most epistemically conservative way, when confronted with evidence suggesting that at least one of her beliefs (in which others’ testimony may be implicated) is false.

4. Generalizing the results 4.1 So far I have merely argued that there are at least some testimonial speech reports that we will treat as false when the content-specifying sentence designates a semantic (propositional) content distinct from the semantic content of the source utterance, even when these contents agree in truth conditions. But this falls short of establishing even the narrow de dicto thesis, since that thesis seeks to ascribe de dicto truth conditions to all testimonial speech reports of the relevant form. In this section I suggest how to use the considerations above to establish the narrow de dicto thesis, and from there to support the broad de dicto thesis as well. Above I argued that, on the assumption that Francois does not know whether the demonstrated dress is chartreuse, Charlotte’s testimonial speech report using (3) is false. Here I want to argue that Charlotte’s testimonial speech report using (3) is false whether or not Francois knows whether the demonstrated dress is chartreuse. More generally, I want to argue that hypotheses regarding what auxiliary knowledge or beliefs the source speaker has are irrelevant to the truth conditions of testimonial speech reports. The crux of the matter is that, since the testimonial use of ‘A said that p’ serves the function of passing onto A the justificatory burden for the claim that p, we are not entitled to treat A as inheriting this burden unless it is the case that,

Reported Speech and the Epistemology of Testimony

69

in performing the speech act she did, she ipso facto committed herself (whatever her auxiliary beliefs may be) to inheriting that burden. The point can be made vivid as follows: even if Francois takes himself to know that the color of the relevant dress is chartreuse, he might be wrong about this, or alternatively he might come to think that he is wrong about this; in either case, he would be entitled to repudiate the testimonial speech report made in (3), for precisely the reasons given above. Let us describe a speech report as non-strict iff it is false on its de dicto reading. Then we can formulate the general point as follows. Suppose that A performs a speech act whose semantic (propositional) content is that q, that at time t S reports this speech act with a testimonial use of ‘A said that p’, that the propositions that p and that q differ in conceptual content but agree in truth conditions, and that at t A herself accepts (and takes herself to know) that these two propositions agree in truth conditions. Under such conditions, there would be no practical point in challenging the non-strict speech report ‘A said that p’. But I still would maintain that the non-strict speech report is false for all that. This result can be obtained by reductio. Suppose that S’s testimonial use of ‘A said that p’ were true. Above I noted that speech reports of this form, when used testimonially, serve to pass the epistemic buck for the claim that p. So if S’s testimonial report ‘A said that p’ is true, then S is entitled to pass to A the justificatory burden for the claim that p. But in that case A is not entitled to repudiate this burden. And this flies in the face of the intuition that, if for some (possibly idiosyncratic) reason A were to come to deny the agreement in truth conditions between the proposition that p and the proposition that q, then ipso facto A would be entitled to repudiate the justificatory burden S aimed to pass to her in having used ‘A said that p’. Now it might be thought that the proper reaction to the foregoing is to conclude, not that testimonial speech reports as a class should be ascribed their strict de dicto truth conditions, but rather that whether a non-strict testimonial speech report is true depends in part on the source speaker’s dispositions to accept or reject the report (e.g., on the basis of her auxiliary beliefs). For example, if Francois has the auxiliary belief that the color of the demonstrated dress is chartreuse, and so is willing to accept Charlotte’s nonstrict report of form (3), then Charlotte’s report is true; if not, not. Correspondingly, if Francois changes his mind at some point regarding the color of the demonstrated dress, it can be maintained that Charlotte’s report of form (3) is true for those intervals during which Francois accepts the relevant color claim, and false for those intervals during which Francois denies (or merely questions) that color claim.

70

Sanford C. Goldberg

But I think that such a reply is unacceptable, for two reasons. First, the proposed context-sensitive approach to the truth conditions of speech reports has consequences that should be counterintuitive to anyone not already committed to that approach. Imagine two speakers, A and A*, who are as alike as any two distinct speakers can be, save that A* has an idiosyncratic tendency to be very fussy with how his speech is reported. Now imagine that A and A* both assert the proposition that p; that S and S*, respectively, report those respective assertions with non-strict speech reports of the same form (where the same expression-form ‘that q’ occurs obliquely in both); and that A but not A* accepts the relevant report. In such circumstances, those who accept the context-sensitive approach to the truth conditions of speech reports will accept that S’s non-strict speech report is true while S*’s is false, merely in virtue of an idiosyncracy in A*’s personality. Moreover, if the speech reports are testimonial, then to accept this result is to accept that A inherits an epistemic burden regarding the claim that q, whereas A* does not – and this, merely because of A*’s fussy attitude towards speech reports! (I should add that a version of this kind of argument can be tailored to cover any proposal that treats the truth-value of non-strict reports as sensitive to contextual features – no matter what context-sensitive feature(s) the theorist takes to be relevant.) To those who are already committed to the context-sensitive approach to speech reports, these considerations will not convince; but to those not already so committed, these considerations should be seen as pointing out counterintuitive consequences of that view. A second objection to the proposed context-sensitive reply is that such a reply cannot provide a plausible explanation for an important asymmetry in the domain of speech reporting. The asymmetry, which holds for all speech reports (whether buck-passing or otherwise), is this: for any source utterance U and non-strict speech report R purporting to be a report of U, we can always imagine a context in which a report involving the very same sentencetype as was used in R is false as a report of U; whereas we cannot obtain such a result if we restrict ourselves to strict speech reports (i.e., those that satisfy the strict, de dicto truth conditions). The natural and simplest explanation of this asymmetry is that when push comes to shove, our use of ‘A said that p’ is governed by the rule that the expression substituting for ‘that p’ must capture the semantic (propositional) content of the source speech act itself. The asymmetry is then explained by saying that so long as this rule is satisfied there can be no basis for falsity for any strict speech report, whereas if this rule is not satisfied (as it is not by non-strict speech reports) then circumstances are imaginable in which we would reject the report. The difficulty

Reported Speech and the Epistemology of Testimony

71

facing the proponent of the context-sensitive approach to the truth of a speech report is now clear. If this proponent aims to explain the asymmetry, she will have to posit some feature – presumably a semantic feature – of speech reports, such that this feature accounts for why there are no imaginable circumstances in which a strict speech report is false (whereas the same is not true for non-strict reports). But, in order to explain the asymmetry, the semantic feature in question will have to be co-extensive with the property of being a true de dicto report of the source utterance. And in that case we can ask whether the proponent is accepting (albeit in different terminology) the de dicto conception of speech reports. The challenge faced by the proponent of the context-sensitive approach to speech reports, then, is the challenge of explaining the asymmetry without in effect endorsing the de dicto construal of speech reports. The point of this last paragraph can be made in an alternative way, in connection with what I take to be the context-invariant nature of the intuitions we bring to bear in thinking of buck-passing uses of speech reports. For any given non-strict speech report of the form ‘A said that p’ made in any context C, whether or not speaker A presently has a motive for rejecting the report, the intuition remains that if S were to have had such a motive then she would have been entitled to repudiate the speech report. And if we accept this intuition, then we have accepted something that holds of any speaker on any occasion of her having made a sincere assertoric utterance. That is, to accept the intuition is to accept the existence of a range of facts regarding what speakers would be entitled to repudiate (whether or not they have any actual motive for doing so). I submit that the relevant range of facts is something that any acceptable theory of the semantics of speech reports must accommodate. The strict de dicto account, on which buck-passing speech reports are true iff they satisfy the strict de dicto truth conditions, accommodates this range of facts quite easily, and does so without the need for any additional ad hoc assumptions. The burden of proof would then be on the proponent of the context-sensitive approach to speech reports: it must be shown how such a view can accommodate the facts in question. For the reasons given in the previous paragraph, I am skeptical that this can be done at all; but even if it can, we can ask whether the cost of doing so is the addition of further ad hoc assumptions. I tentatively conclude that, as a class, testimonial speech reports should be ascribed their strict, de dicto truth conditions. The general idea is that anything less than a strict de dicto reading will allow for the possibility that a speaker S could reply to a “How do you know?” challenge by passing the buck to another speaker A, claiming that A said that p, when A herself would

72

Sanford C. Goldberg

be entitled to repudiate the epistemic burden regarding the claim that p. My contention is that, in order to accommodate the context-invariant intuitions we bring to bear in the truth-evaluation of speech reports, we need to advert to the strict (de dicto) truth conditions of speech reports. So concludes my case for the narrow de dicto thesis. 4.2 We can use these results as part of an argument for the broad de dicto thesis as follows. No matter how it is initially used, a true speech report, made by way of the use of a sentence of the form ‘A said that p’, is available as a premise from which an audience who observes that report can draw inferences. Thus if T observes and understands S’s report ‘A said that p’, then T has at his disposal a piece of information that he can use as he goes about his business trying to acquire true beliefs. In particular, the piece of information he would thereby have at his disposal would be available for use in the following inference: S said that A said that p. S (or S’s testimony) satisfies all conditions on testimonial authority and trustworthiness. Therefore, A said that p. A (or A’s testimony) satisfies all conditions on testimonial authority and trustworthiness. Therefore, p. Of course, if T does draw such an inference and in so doing comes to believe that p, then T’s belief that p would be such that he would be entitled to pass the justificatory burden onto S, and (assuming nothing was amiss with S’s report) ultimately onto A. So long as S’s report of A’s speech (by way of S’s use of ‘A said that p’) is true, A inherits the justificatory burden regarding the claim that p. As a consequence, if a non-strict report of this form can be true, then A can be put in a position of inheriting the justificatory burden for a claim whose propositional content she never presented as true. And so we see that, however it is initially used, a speech report of the form ‘A said that p’ represents A as available to assume the justificatory burden for the claim that p – with the result that, if we are not to place an undue justificatory burden on A, then, no matter what aim the reporter had in making a report of this form, we must give speech reports of the form ‘A said that p’ their strict, de dicto reading. Doing so is a matter of holding A responsible for nothing more and nothing less than her word. From an epistemological point of view this seems only fair.

Reported Speech and the Epistemology of Testimony

73

5. Saying and What is Said I have just concluded an argument for the thesis that, no matter how such reports are used, speech reports of the form ‘A said that p’ should be ascribed their strict, de dicto truth conditions. But I began this essay by acknowledging that there are a variety of uses for speech reports of this form, and that at least some of these uses appear to be such that the resulting report can be true even when it fails to meet the strict de dicto standard. So it can now seem that, if the forgoing argument is successful, it leaves us with a paradox rather than an acceptable conclusion. In this final section I attempt to dissolve this apparent paradox. Return to the case of Officer A who, upon being asked whether Jones has been charged with the murder, responds by uttering ‘No one has been charged with the murder.’ Suppose that S reports A’s speech act with a report of the form, ‘A said that Jones has not been charged with the murder.’ Above I suggested that in cases of this sort there is some intuition behind treating the report as true despite its failure to satisfy the strict de dicto truth conditions. But this intuition (and the appearances that elicit it) should be handled carefully. To begin, it is easy to imagine circumstances in which A would be entitled to deny such a report. Suppose for example that A was under strict orders from the Police Commissioner regarding how to deal with the public dissemination of information regarding the present status of the murder investigation: A is told to provide as much information as possible to the public, but to do so in a way that avoids mentioning any potential suspect by name (lest he give off a false impression of the status of the named suspect in the present state of the investigation). A knows that, had he asserted the proposition that Jones has not been charged with the murder, the Commissioner would have seen this as in direct violation of the order not to mention anyone by name. Under such circumstances, A, who chose his words carefully, would be entitled to repudiate the speech report S offered.14 Indeed, there can be little doubt that, given any source utterance and any non-strict speech report, it would be easy to come up with circumstances in 14 This example makes clear that, beyond the norms involved in epistemic buck-passing, there are other norms that can be appealed to in defense of the de dicto reading of speech reports. This suggests that the norms involved in testimony cases are not fundamental to the case I want to make for the de dicto reading. Even so, I would continue to maintain the importance (if only rhetorical) of my focus on the testimony cases, by pointing out the vividness of the norms at play there: there is nothing quite like being asked to shoulder the justificatory burden for a proposition that you yourself deny having presented-as-true.

74

Sanford C. Goldberg

which the source speaker would be entitled to deny the speech report (and where intuition would support treating the report as false). This suggests that whatever intuitions support treating a non-strict speech report as true dissipate rather systematically whenever the speaker herself has a motive to deny the report. As above, one might choose to react to this by holding that whether a non-strict speech report is true depends on such contextual matters as whether (at the time of the report) the source speaker would repudiate the report. But again this reaction misses an important asymmetry between strict and non-strict reports, and it misses as well the crucial point that a source speaker A’s entitlement to repudiate a non-strict speech report is independent of whether or not she presently has reason to do so.15 Ascribing to speech reports their strict, de dicto truth conditions not only explains this fact about what A is entitled to repudiate, it also enables us to avoid the awkward conclusion that the truth-value of a non-strict speech report can turn on such things as the caste of mind of the source speaker (and in particular on how fussy such a speaker is regarding how she is reported). In arguing that non-strict speech reports are uniformly false, I am not denying that there is often much that recommends the making of such a report. One only has to think of the kinds of case described at the outset of this paper to see that this is so. Regarding such uses, I would say that the reports succeed in communicating something true, and that they do so in an efficient way. (For example, they succeed in identifying a relevant implicature of the source speaker’s speech act, without requiring the reporter to have grasped the technical notion of an implicature.) But we should guard against the conflation of the acceptable claim that non-strict speech reports sometimes succeed in communicating truths efficiently, with the unacceptable claim that nonstrict speech reports are sometimes true. Cases in which non-strict speech reports succeed in communicating truths without themselves being true are no more mysterious than other cases in which a speaker asserts something that is (strictly speaking) false in order to be brief in her attempt to communicate something. This practice makes perfect sense under circumstances in which (i) it is impractical or otherwise difficult to frame a proposition that is true and (ii) there is little chance given the context that the hearer will hold the speaker to the literal claim she made (even though the hearer is entitled to do so). 15

More generally, I submit that, since we can never tell in advance when circumstances will arise in which a source speaker will have a reason to repudiate a non-strict speech report, the considerations here suggest that we should always regard speech reports as true only when they satisfy the strict de dicto truth conditions.

Reported Speech and the Epistemology of Testimony

75

In this light it is interesting to frame the present point as a comment on the content of the verb ‘to say’ and that of the cognate notion ‘what is said.’ On the view being presently advocated, to say that p is (at least in part) to perform a speech act in which the proposition that p is presented as true, in such a way as to make oneself available for inheriting the related justificatory burden. In a parallel fashion, the notion of ‘what is said (by A’s utterance U)’ singles out that aspect of A’s utterance which is bound up in the practices whereby we evaluate the epistemic status of others’ beliefs and assertions which are grounded on U. More specifically (and recalling that we are dealing with utterances of sentences not involving indexicals and the like), ‘what is said by an utterance U’ is the maximum that any speaker A, having made an utterance involving the same sentence-type as that figuring in U, can be held epistemically responsible for justifying, were another speaker S to make an assertion on the basis of having observed U.16 That this aspect of an utterance does not exhaust the role that utterances play in communication more broadly construed is precisely why we should want to distinguish ‘what is said’ from ‘what is communicated’. An argument to the effect that the former ought to be conceived strictly and narrowly leaves this point, as it should be, intact.

6. Conclusion and Related Future Work The substantive conclusion of this paper is that speech reports of the form ‘A said that p’ should be ascribed their strict, de dicto truth conditions. This conclusion was argued for by appeal to the epistemic buck-passing role that speech reports of this form play, when such reports are used to manifest the testimonial basis of the reporter’s own belief or assertion that p. Admittedly, this conclusion depends on the prior plausibility of the programmatic remarks I made in 2, and in particular on the connection I sought to make between the semantic content of a source speech act (on the one hand) and 16 This same point can be made at the level of the sentence, as opposed to the utterance: the notion of ‘what is said by a sentence F’ singles out that aspect of the sentence which is bound up (when the sentence is uttered by a speaker on an occasion) with our practices of evaluating the epistemic status of beliefs and assertions grounded on utterances of that type. More specifically, ‘what is said by a sentence F’ is the maximum that any speaker A who used F would be epistemically responsible for justifying, were another speaker S to make an assertion on the basis of having both observed and accepted A’s use of F. (Again, we are restricting attention to sentences not employing indexicals and the like.)

76

Sanford C. Goldberg

the conditions for epistemic buck-passing by way of a testimonial speech report (on the other). In this paper I have more or less assumed that the semantic content of an utterance can be functionally specified in testimonial terms, as whatever it is that the speaker A could be held epistemically responsible for, in the event that an arbitrary audience S observes the speech act and acquires a belief (or makes an assertion) regarding which S would pass the epistemic buck onto A. Though I take this assumption to be very plausible, it might not strike everyone as obvious. Since space prohibits me from defending it here, I will briefly point out how this assumption might be put to the test, and what further work remains be done. The central assumption itself can be formulated as follows: (*) For all expressions E of a language L, the semantic content of E is determined by what a speaker A’s use of E in an assertoric utterance U adds to A’s justificatory burden, were another speaker S of L to acquire a belief or make an assertion on the basis of having observed U (and regarding which S would pass the justificatory burden onto A). (*) itself can be viewed from two distinct perspectives. On the one hand, there are those (like me) who will think that (*) is supported by reflecting on the strong intuitions we bring to bear in thinking about testimony cases. Such people can assume (*) in order to identify a potential source of evidence with which to confirm (or disconfirm) various hypotheses regarding the semantic content of expressions. In this respect the interest of (*) is in suggesting how intuitions regarding the epistemology of testimony might be exploited in the study of semantics (more on which below). Alternatively, there are those who will not yet be convinced of the truth of (*). Such people can put (*) to the test by reference to expressions whose semantic content they take to be well-understood. If (*) entails something incompatible with a claim that (all sides agree) expresses a truth regarding the semantic content of a given expression, then we can reasonably conclude that (*) itself is to blame. One final comment regarding the utility of (*) is in order. With (*) in hand, I can imagine going over some well-worn territory in the philosophy of language. Topics that might usefully be re-examined include the semantic significance of the referential use of definite descriptions, the semantics of incomplete descriptions, the semantic contribution (if any) of the sortal F in expressions of the form ‘that F’, and the semantics of proper names. Regarding the first of these, a question might be: given a speaker A who uses a definite description of the form ‘the F’ referentially, but where the object

Reported Speech and the Epistemology of Testimony

77

referred to is not in fact F, what is A epistemically responsible for justifying in virtue of this use of ‘the F’, in the event that another speaker were to acquire a belief or make an assertion on the basis of having observed A’s utterance?17 I believe that (*) will provide a fruitful way to inquire into these topics and perhaps others as well. At present I will be satisfied if I have succeeded at making such a project appear worthy of further pursuit.18

Bibliography Brandom. 1994. Making It Explicit. Cambridge: Harvard University Press. Cappelen, and Lepore. 1997. “On an Alleged Connection between Indirect Speech and the Theory of Meaning.” Mind and Language 12: 278-96. Goldberg. 2001. “Testimonially based knowledge from false testimony.” Philosophical Quarterly 51, 205: 512-26. Richard. 1998. “Semantic Theory and Indirect Speech.” Mind and Language 13, 4. Salmon.1990. “The Pragmatic Fallacy.” Philosophical Studies 63, 1990. Williamson. 1996. “Knowing and Asserting.” The Philosophical Review 105, 4: 489523.

17 This kind of question was first presented to me by Ernie Lepore, in comments he made on an earlier version of this paper. 18 I would like to thank Ernie Lepore, Brad Monton, Ziggy Rivkin-Fish, and Ted Schatzki for their comments on earlier versions of this paper.

78

Eros Corazza

Reports and Imagination Eros Corazza

Abstract The following thesis will be discussed and defended: An attitude ascription is an empathetic exercise resting on our, more general, imaginative faculty. Sentences of natural language are the best medium we have to classify someone’s mental life.The sentence used to classify one’s mental state is the one the reporter would use to express the attributee’s mental state if the reporter were in the attributee’s situation. A report of the form “A believes/desires/wishes/… that p” captures the attributee’s (A) mental life inasmuch as it conveys the sentence the reporter would use to express her mental state if the latter were in A’s situation.

My paper can be viewed as a long footnote to the following passage from Quine’s Word and Object: [I]n indirect quotation we project ourselves into what, from his remarks and other indications, we imagine the speaker’s state of mind to have been, and then we say what, in our language, is natural and relevant for us in the state thus feigned. An indirect quotation can usually expect to rate only as better or worse, more or less faithful, and we cannot even hope for a strict standard of more and less; what is involved is evaluation, relative to special purposes, of an essentially dramatic act. Correspondingly for other propositional attitudes, for all of one’s own imagined verbal response to an imagined situation. Casting our real selves thus in unreal roles, we do not generally know how much reality to hold constant. Quandaries arise. But despite them we find ourselves attributing beliefs, wishes and strivings even to creatures lacking the power of speech, such is our dramatic virtuosity. We project ourselves even into what from his behavior we imagine a mouse’s state of mind to have been, dramatize it as a belief, wish, or striving, verbalized as seems relevant and natural to us in the state thus feigned. (Quine 1960: 219, italics mine) The main idea behind this view can be tracked back to Aristotle: The people we pity are: those whom we know, if only they are not very closely related to us – in that case we feel about them as if we

Reports and Imagination

79

were in danger ourselves. (Aristotle; Rhetoric: 1378a31-b2, italics mine) Like Quine, and maybe Aristotle, I defend the view that an attitude ascription is an empathetic act. As such it rests on our more general imaginative faculty. In particular, it rests on our capacity to imagine ourselves in someone else’s shoes. Here is the way in which I shall proceed. In section 1 I discuss what is going on when we attribute states like pain to other people and argue that this activity rests on our empathetic capacity. I discuss a case of autism that seems to empirically sustain this view. In section 2 I defend the thesis that sentences are the best means we have to classify someone’s mental life and activity. I defend the view that, from the third person viewpoint, the attributee’s mental state can be captured by the sentence the reporter would use to express the relevant mental state if she were in the attributee’s shoes. Finally, in section 3 I discuss how attitude reports can allow us to classify the attributee’s mental life and I explain away some puzzles and problems related to attitude reports.

1. The Background: Imagination In this section I propose some general ideas concerning the way we capture other people’s mental life and activities. As a paradigmatic case I consider the attribution of pain. Descartes presented us with the following controversial thesis: animals are mere complex robots or machines.1 This thesis goes against our commonsensical views, for when we see someone beating her dog we almost “feel the dog’s pain” and we feel sorry for the dog. While if we see someone beating her car or computer we do not “feel the computer’s (or car’s) pain” and we do not feel sorry for the computer or car. We may feel sorry if an expensive car or computer gets damaged, but we do not feel sorry because the car or computer is suffering. Our feelings vis-à-vis the dog and the computer are of a different nature. How, though, can we explain this difference? Sure, the dog that is beaten behaves in a certain way while the computer does not react. Well, imagine that we face a dog and a robot-dog which behave the same way when beaten. In seeing the dog beaten we “feel her pain” while in seeing the robot-dog beaten we do not “feel its pain” so long as we know it is only a 1 Sometimes Descartes attributes sensations such as feelings, hope and joy to animals (see Cottingham; 1986: chapter 5). This contrasts with Descartes’ official viewpoint that animals are mere automata.

80

Eros Corazza

robot.2 If so, the behavioral reactions do not explain our feelings or, at least, are not sufficient conditions for our feelings. The behavioral manifestations need not be necessary conditions either, for two people undergoing the same torture and pain may behave very differently. While one starts shouting, crying and contorting, the other may well stay perfectly quiet. Yet we feel sorry for both people’s pain. Besides, if we know that someone is cheating and behaving as if she were in pain we do not “feel her pain”: we do not “feel the pain” simulated by two wrestling fighters.3 Behavior does not seem to be a good guide in explaining our reaction; at least, it is not always reliable. Hence the behavioral manifestations are not sufficient conditions which justify and trigger our feeling. This does not amount to saying that pain can be known only by its owner. Sure, only I can feel my headache. From the fact that I cannot feel someone else’s pain, though, it does not follow that I cannot know whether someone is in pain or not. Moreover: [O]ne would not have the concept of, say, being in pain, without having encountered organisms displaying pain-behavior. If I were the only advanced organism in the world, my concepts, if any, of pain, of experience, of subjects, and of myself, would be quite different from what it is now. This much, at least, is clear from Wittgenstein’s argument against private language. (Vendler; 1984: 18) How, though, do we feel entitled to attribute to someone the feeling of pain? A commonsensical answer is that we put ourselves in her position or situation and conclude that she is in pain from our being aware that we would be in pain in such a situation: if I were beaten like the dog or a tortured person I would feel pain. Imagination seems to be one of the key notions that allows us to attribute pain and the like. These kinds of attributions rest on our faculty to put ourselves into someone’s shoes and think what we would do/ feel/experience/… in her situation. This is a perspectival activity insofar as it rests on the attributer’s capacity to project herself into the attributee’s context, i.e., it rests on the capacity one has to adopt someone else’s viewpoint. 2 A clarification is in order, for it seems that simulation mechanisms already operate at a pre-conceptual level. It has been proved (Hutchinson et al.; 1999) that neurons in the human cingulate cortex do “simulate” other people’s neurons when perceiving what ordinarily causes pain. If one sees someone getting hurt, one’s neurons simulate the neurons of the wounded agent. Hence if our robot-dog is sufficiently similar to a real dog, then our neurons react in the same way as they would do in presence of a real dog. We should thus say that in seeing a robot-dog beaten we do not feel its pain conceptually. 3 Indeed, the situation is not so simple. Watching a movie we may “feel the actor’s pain” and may put ourselves in the actor’s shoes. In this kind of situation, though, we always have the conceptual capacity (or meta-capacity) to retreat and to tell ourselves that it is only a movie and that the actors are merely mimicking pain-behavior.

Reports and Imagination

81

It is because of this faculty that we can succeed in predicting someone’s movements, feelings, plans, etc. When playing chess, for instance, I can anticipate my partner’s move because I figure out what I would do in her position.4 The way we attribute pain to others, though, does not rest on analogy; i.e. it does not rest on an inference the reporter ought to make before being entitled to attribute pain. One need not project one’s own pain to someone else before making a pain attribution: If one has to imagine someone else’s pain on the model of one’s own, this is none too easy a thing to do: for I have to imagine pain which I do not feel on the model of pain which I do feel. That is, what I have to do is not simply to make a transition in imagination from one place of pain to another. As for pain in the hand to pain in the arm. For I am not to imagine that I feel pain in some region of his body. (Wittgenstein; PI: § 302) You say you attend a man who groans because experience has taught you that you yourself groan when you feel such-and-such. But as you don’t in fact make any such inference, we can abandon the justification by analogy. (Wittgenstein; Z: § 537) One reacts and attributes pain automatically, without reflection. There is, we can say, a pre-theoretical and pre-inferential foundation for our knowledge of other people having their own feelings and sensations.5 That is to say, that the capacity one has to make inferences about other people’s pains and sensations ultimately rests on one’s previous grasping that people do have feelings, pains, sensations, etc. This, though, does not contradict the idea that imagination and empathy play a crucial role when one comes to learn and master the language game of pain and sensation attributions. If one lacks these basic faculties one faces difficulties in understanding other people’s pain and the like, let alone attributing these kinds of sensation. It is worth mentioning that autistic people usually face many problems in reporting other people’s pain and the like. This is mainly due to the fact that they have difficulties in imagining themselves in other people’s shoes, as far as feelings are concerned. They cannot, so to speak, imagine other people’s 4 The general picture I have in mind bears some resemblance to the simulation theory in the philosophy of mind, and in particular to the position viewing our simulation activity as belonging to a family of mental faculties such as imitation, pretending, motor imagery, etc. For a defense of this view see Goldman (2000). 5 “[T]he basis for infants’ growing understanding of other people is their experience of reciprocal personal relations with others, what Hamlyn (1974) called ‘natural reaction of person to person’. … Another person’s state of mind is directly grasped in expressive phenomena.” (Hobson; 1990: 116)

82

Eros Corazza

experiences and sensations.6 They have, it seems, empathetic problems or impairment. They feel, to borrow Temple Grandin’s words, like an anthropologist on Mars.7 To handle these situations some (the less impaired) autistic people may recur to visual simulation: Temple constantly runs “simulations,” as she call them, in her head: “I visualize the animal entering the chute, from different angles, different distances, zooming in or wide-angle, even from a helicopter view – or I turn myself into an animal, and feel what it would feel entering the chute.” But if one thinks only in pictures, I could not help reflecting, one might not understand what nonvisual thinking was like, and one would miss the richness and ambiguity, the cultural presupposition, the depth, of language. All autistics, Temple has said earlier, were intensely visual thinkers, like her. … She feels she can have sympathy for what is physical or physiological – for an animal’s pain or terror – but lacks empathy for people’s states of mind and perspectives. When she was younger she was hardly able to interpret even the simplest expression of emotion; she learned to “decode” them later, without necessarily feeling them. (Sacks; 1993-4: 115-6) If my understanding of what is going on in Grandin’s case is right, we have some empirical evidence in favor of the picture I am putting forward.8 For, if autistic people face obstacles in reporting other people’s pain and the like, it is because they lack the empathetic faculty that I claim lies at the bottom of our pain and sensation reports. To borrow Wittgenstein’s terminology, we can further add that in order to play the language game of pain attributions and reports we need to share a form of life: [O]ne human being can be a complete enigma to another. We learn this when we come into a strange country with entirely strange traditions; and, what is more, even given a mastery of the country’s language. We do not understand the people. ... We can6 “A large number of studies have repeatedly demonstrated that children with autism have difficulties in shifting their perspective to judge what someone else might think, instead simply reporting what they themselves know.” (Baron-Cohen; 1993: 5) 7 See Oliver Sacks’ (1993-4) report on Temple Grandin, who was affected by the Asperger’s syndrome (a form of autism that does not fully affect the subject capacity of self-consciousness). People affected by this syndrome have some power to introspect and report. I thank Howard Wettstein for drawing my attention to this article. 8 Hobson’s (1993) discussion of autistic people brings even more evidence in favor of this picture. For Grandin, unlike the majority of autistic people, can simulate animals’ sensations.

Reports and Imagination

83

not find our feet with them. ... If a lion could talk, we could not understand him. (Wittgenstein; PI II: 223) It may be worth noticing that, if we follow Wittgenstein on this issue, we ought ultimately to accept that what is essential to the understanding of all our linguistic activity, and a fortiori the language game of pain attribution, is our form of life, for “to imagine a language is to imagine a form of life” (PI: § 7); “what has to be accepted, the given, is – so one could say – forms of life.” (PI II: 226) The form of life is the background against which our linguistic practice rests. As is often the case with Wittgenstein, these remarks can be understood in variegated ways. Among the possible understandings, the one I favor assumes that our form of life, qua foundation of our linguistic practice, can be given a socio-naturalistic interpretation. As such it is both part of our biological and anthropological nature insofar as they both concur in determining how we linguistically act and react: “commanding, questioning, recounting, chatting, are as much part of our natural history as walking, eating, drinking, playing.” (PI: § 25)9 To put it in a nutshell, it is because autistic people do not fully share our form of life that they feel like anthropologists on Mars and, therefore, do not easily master some of our linguistic practices: What is it, then, I pressed her further, that goes on between normal people, from which she feels herself excluded? It has to do, she has inferred, with an implicit knowledge of social conventions and codes, of cultural presuppositions of every sort. This implicit knowledge, which every normal person accumulates and generates throughout life on the basis of experience and encounters with others, Temple seems to be largely devoid of. Lacking it, she has insisted to “compute” others’ feelings and intentions and states of 9 Hence the understanding of someone’s language does not merely rest on shared beliefs; it ultimately rests on shared behavioral patterns: “The common behavior of mankind is the system of reference by means of which we interpret an unknown language.” (PI : § 206) For these reasons, we cannot feel at ease with a community that never expresses any feelings of pain: “Imagine that the people of a tribe were brought up from early youth to give no expression of feeling of any kind. … ‘These men would have nothing human about them.’ Why? – We could not possibly make ourselves understood to them. Not even as we can to a dog. We could not find our feet with them. And yet there surely could be such beings, who in other respects were human.” (Z: § 385, § 390). For: “knowledge and understanding of persons, or to put this differently, a conceptual grasp of the nature of minds, is acquired through an individual’s experience of effectively patterns, intersubjectively co-ordinated relations with other people. A young child comes to know about people’s psychological states through having subjective experiences that are shared with, opposed to, or otherwise articulated with the experience (and not merely the “behaviour”) of others.” (Hobson; 1993: 4-5)

84

Eros Corazza

mind, to try to make algorithmic, explicit, what for the rest of us is second nature. She herself, she infers, may never have had the normal social experiences from which a normal social knowledge is constructed. And it may be from this, too, that her difficulties with gesture and language stem – difficulties that were devastating when she was a near speech-less child, and also in the early days of speech, when she mixed all her pronouns up, not able to grasp the different meaning of “you” and “I,” depending on context. … She remains clearly autistic, but her new power of language and communication now gave her an anchor, some ability to master what had been total chaos before. … With the access of language, the terrible triad of impairments – social, communicative, and imaginative – began to yield somewhat. … At eight, Temple was starting to achieve the “pretend play” that normal children achieve as toddlers but the lower-functioning autistic child never achieve at all. … She is now aware of the existence of these social signals. She can infer them, she says, but she herself cannot perceive them, cannot participate in this magical communication directly, or conceive the many-leveled, kaleidoscopic states of mind behind it. (Sacks; 19934: 116) The moral I would like to draw is that we do not perceive someone’s pain as merely a bodily movement. Instead, we perceive it as we perceive actions and activities with specific aims and goals.10 The concept of pain is thus not confined to the behavioral manifestations when one is injured or sick. It is also linked with our capacity to attribute pain and to feel compassion, mercy, pity, commiseration, etc. It is also linked to anxiety, fear, apprehension, cruelty, atrocity, etc.: The concept of pain is characterized by its particular function in our life. (Wittgenstein; Z: § 532) Pain has this position in our life; it has these connexions; (That is to say: we only call “pain” what has this position, these connexions). (Wittgenstein; Z: § 533) Only surrounded by certain normal manifestations of life, is there such a thing as an expression of pain. Only surrounded by an 10 For more on this particularity and for some empirical results drawn both from humans and primates confirming this thesis, see Proust (2000).

Reports and Imagination

85

even more far-reaching manifestation of life, such a thing as the expression of sorrow or affection. And so on. (Wittgenstein; Z: § 534) Following these suggestions, it seems that a fruitful way to spell out the different kinds of abilities required by the use of a word like ‘pain’ (and in particular the mastery of the language game of pain attribution) can be given by considering what is going on in the perception of someone else’s pain. We perceive pain in a rather different way than that in which we perceive the computer’s screen in front of us, or the woman we are looking at. To stress this difference, consider the following: (1) (2) (3) (4)

Jon saw that Mary was in pain Jon saw Mary in pain Jon saw Mary’s partner *Jon saw Mary’s pain

While (1), (2) and (3) make sense, (4) does not. We cannot perceive someone’s pain, but we can perceive that someone is in pain. To spell out the difference between these kinds of attributions I borrow Dretske’s distinction between epistemic (seeing that) and non-epistemic seeing.11 We can see a leaf without seeing that the leaf is green. The former, unlike the latter, is non-epistemic: seeing a leaf in the non-epistemic way is like stepping on the leaf without realizing it. Another way to state the difference is to argue that non-epistemic seeing does not entail knowledge. One can see a lion without believing and, thus, without knowing that what one saw was a lion (and even without believing that it was an animal). Moreover, non-epistemic seeing is transparent: from “Sue saw Superman,” given that Superman is Clark Kent, we can infer that “Sue saw Clark Kent,” just as from “Lois kissed Superman” we can infer that “Lois kissed Clark Kent.” Epistemic seeing, on the other hand, does not seem to be transparent. From “Sue saw that Superman kissed Lois” we cannot infer that “Sue saw that Clark Kent kissed Lois,” as from “Lois wishes to kiss Superman” we cannot infer “Lois wishes to kiss Clark Kent.” The moral I would like to draw is that pain-reports can be rendered only using locutions of the form ‘seeing that.’12 These kinds of reports entail the knowledge of someone’s pain and, ultimately, the mastery of the word ‘pain.’ For these reasons I think that the exercise at play when we attribute pain to 11

See Dretske (1969: chapter 1).

86

Eros Corazza

someone is similar to the exercise at play when we attribute to someone a belief, desire and the like. In particular, the same imaginative and empathetic faculties are at work. To the discussion of this very idea I now turn.

2. Sentences qua Thought Classifiers If one is asked to report what one said, believes, thinks, wishes, etc. one would propose a self-attribution of the form: SA. I said/believe/think/wish/… that S where S is the sentence one uses in expressing what one said/believes/thinks/ wishes/etc. This sentence can be viewed as a reliable classifier of one’s though, i.e., the thought one would express in using that very sentence. I introduce the technical notion of acceptance to characterize the relation between the subject of an attitude report and the sentence S that classifies the attributee’s thought. Accepted sentences are thus taken to be thought classifiers: One has a belief by accepting a sentence. Which belief one thereby has also depends on who the believer is and when the believing takes place .... When we believe we do so by being in belief states. These states have typical effects which we use to classify them. In particular we classify them by the sentences a competent speaker of the language in question would be apt to think or utter in certain circumstances when in that state. To accept a sentence S is to be in a belief state that would lead such a speaker to utter or think S. (Perry 1980: 533) Here is the picture I have in mind: in accepting a sentence, an agent has a mental representation or thought of a certain type. An agent has the same type of mental representation whether she is in a belief, desire, saying, etc. relation to a proposition. Before going further, though, a world of clarification is needed. I am working within the framework of direct reference. I thus assume, without argument, that singular terms refer directly without the mediation of Fregean senses or modes of presentation. I further assume that the utterance of a 12

It is worth noticing that even if a report like (2), unlike (1), is transparent (as far as Mary is concerned) – If Mary is Jeff ’s partner we can infer “Jon saw Jeff ’s partner in pain” – it is not transparent as far as the perception of Mary’s pain is concerned. For from (2) we can only infer “Jon saw that someone was in pain,” we cannot infer “Jon saw someone’s pain.”

Reports and Imagination

87

sentence containing a singular term expresses a singular or Russellian proposition; that is, a proposition having the referents themselves as constituents. A singular proposition is an abstract, structured entity made up by objects, properties and relations. A mental state, on the other hand, is a mental representation plus the attitude (belief, desire, etc.) relation the agent bears to the proposition. I would like to think of mental representations as being embedded within attitude operators. A belief state, for instance, is a mental representation embedded within the BELIEF operator. So subjects who accept the same sentence share the same mental representation and may be in the same mental state. If they are in the same mental state they tend to behave in a similar way. Besides, two agents who are in the same mental state (e.g. belief state) may be related to different propositions. If Sue and Mary both accept the sentence “I am thirsty,” they may be in the same mental state but they are related to different propositions, one having Sue as constituent and the other having Mary. It is often the case that in order to be related to the same proposition two agents (and even the same agent) have to be in distinct mental states. Let us consider a version of an example made famous by Frege. On Monday, Sue accepts “Today I have to call Dr. Lauben.” After midnight, to be related to the same proposition Sue needs to accept “Yesterday I had to call Dr. Lauben.” Hence Sue’s mental states differ. If Sue were to accept on Tuesday “Today I have to call Dr. Lauben” she would be in the same mental state but related to a different proposition.13 The notion of acceptance captures the intuitive idea that an agent, if asked to express her attitude, uses the sentence the reporter attributes on her. I take this to be the paradigmatic case – this simple test, though, can be run only in “normal” situations. That is, an agent is disposed to utter the very same sen13

Perry argues (1980a) that the notion of acceptance is inadequate to handle the problem of cognitive dynamics, i.e. to explain what is going on in one’s mind when one continues to believe the same thing. To do so, he introduces the notion of internal identity, i.e., identity between mental files. So, one can be said to retain the same belief, say about Sue, when one stores the relevant information acquired on the same file. Although I am sympathetic with Perry’s treatment of cognitive dynamics, I do not think that in attitude ascriptions we need to make reference to files to deal with the problem of explaining cognitive significance. Besides, to attribute to someone the same belief across time we often have to attribute her the acceptance of different sentences. This is the case, for instance, with the attribution of temporal beliefs like in Frege’s ‘today’/‘yesterday’ example. An attribution involving proper names, on the other hand, usually does not need to attribute the acceptance of different sentences across time to stress that the attributee believes the same thing. But see Kripke’s (1979) Peter-‘Paderewski’s’ puzzle, where Peter wrongly takes this proper name to be about distinct individuals: Peter will have two files where she stores her information about the referent of ‘Paderewski’. More

88

Eros Corazza

tence whose acceptance the ascriber attributes her if the former speaks the language of the latter, is rational, and so on. It goes without saying that these conditions are not always satisfied: we attribute attitudes to pre-linguistic infants, to foreigner speakers, to animals, etc. I think that attitude ascriptions to non-linguistic organisms mirror the attribution of pain (and the like). As I argued in the previous section, we learn how to attribute pain on the basis of observable conditions in the presence of external stimuli and behavior. Without doubt I know when I am in pain: we self-attribute pain without hesitation. When we come to attribute pain to others we lack this kind of certainty or self-evidence. However, we often do not hesitate to say that an infant or an animal is suffering. When one is beating a man with a baseball bat and the latter is contorting, crying, etc. we do not hesitate to attribute him pain: we do not seem to lack evidence allowing us to attribute pain. In other words, even if the attribution of pain to others is defeasible (whereas self-attribution is not), we often do not hesitate in attributing pain and the like to others. Our assurance and certainty comes from the fact that we automatically imagine ourselves in their situation. This does not mean that this exercise is merely unconscious and/or confined to pre-conceptual activities. The notion of acceptance I have in mind is best spelled out on the basis of the following Transpositional Principle: TP. If the reporter were to express what the attributee believes, desires, says, etc., the reporter would utter the relevant sentence, i.e. the sentence whose acceptance the reporter is attributing to her.14 When we attribute an attitude to someone we often imagine ourselves in her situation: In saying what someone else believes, we describe his belief by relating it to one we ourselves might have. And we indicate this potential belief of our own by uttering the sentence we would use to express it. (Stich 1983: 79) on Kripke’s puzzle later. 14 Notice that this principle could be spelled out in a more straightforward way, as follows: if the reporter were the attributee, then in order to express what the latter believes, desires… the former would utter the sentence she accepts as expressing her belief, desire... That is, the reporter would make a self-attribution of the form “I believe/think/ wish/… that S” and in so doing she would express the sentence she accepts. Since, as a matter of fact, we cannot be someone else, all we can do is to imagine ourselves in the attributee’s context, i.e. to take one’s place. So this straightforward principle is just a shortcut to the transpositional one.

Reports and Imagination

89

It is not uncommon that we make attributions observing someone’s behavior or on the basis of inferences from other beliefs, desires, etc. we know the attributee holds. We do so by imagining what we would do, think, desire, etc. in these cases. Imagine, for example, that you and your partner are looking at Sue, who is behaving in a strange way. Your partner asks: “What do you think she is thinking?” The answer could be: “I guess she is thinking that her life is in danger. So she acts in that strange way.” On what basis are you able to answer yours partner’s query? It could be that you had spoken to Sue and she told you that she was frightened. But it could also be that you are making an inference of the form: (i) I know that Sue is frightened of a strange guy that just arrived, (ii) so she should think that her life is in danger. How then, are you entitled to infer that Sue is thinking that she is in danger? The natural answer that comes to my mind is that we do these kinds of inferences by imagining ourselves in the attributee’s situation. If I were Sue, given (i) I would be scared. It goes without saying that when we make this kind of attribution we are presupposing that the attributee is rational. This assumption, though, can always be cancelled by a paraphrase. A word of clarification may be useful. According to the picture I am putting forth, accepting a sentence does not always mean accepting it as true. Sure, if a subject is in a belief state, she takes the relevant sentence to be true. But one accepts a sentence when one is in a saying (desiring, thinking, etc.) relation, to a proposition as well. From “A says S” we cannot infer that A accepts S as true. The terminology, I agree, is not very felicitous. The notion of acceptance has been given several technical meanings: it is often assumed that an agent who accepts a sentence takes it to be true; otherwise she does not accept it. This is not the notion I have in mind. To simplify we can spell out the notion of acceptance this way: imagine that in front of an agent there is a set of indefinitely many sentences and a set of psychological verbs like ‘believe,’ ‘desire,’ ‘say,’ ‘doubt,’ etc.. Each time our agent entertains an attitude she is asked to do two things: (i) pick out a psychological verb and (ii) choose from among the sentences the one she would use to express her attitude. The sentence she picks out or points to is the sentence she accepts. So if our agent points to “Bob Dylan was born in Minnesota” she accepts this sentence whether or not she believes, desires, wishes, etc. that Bob Dylan was born in Minnesota. The sentence one accepts classifies the mental representation one has. So when you think that Bob Dylan is American, you have a mental representation classified by the sentence “Bob Dylan is American” that you accept.

90

Eros Corazza

Your mental representation cannot be classified by the sentence “Robert Zimmerman is American” even if Bob Dylan is Robert Zimmerman – in the aforementioned situation you do not point to this sentence. It may also be that if questioned you refuse to point to this sentence and, thus, you do not accept it. In short, I take the notion of having a mental representation and being in a given state to be a primitive one; I am not trying to cash out these notions in terms of belief. This way we can deal with attitudes other than belief in a straightforward way. We can also deal in an easy way with fiction and conscious illusions. An agent may be in a given state when she is seeing a movie without taking the story to be true. In the Müller-Lyer paradox, one line looks longer than the other even after we have been told that they are of equal length: we keep having the same representation even if we no longer believe it to be true. To avoid confusion it may be worth introducing an artificial term to distinguish the case when an agent accepts a sentence to be true from the case where she does not. Henceforth I shall use the notion of neutral acceptance (n-acceptance) to stress the fact that a subject does not need to take the sentence used to classify her mental representation to be true. Moreover, when we attribute someone the n-acceptance of a sentence (and thus classify her mental state), we are not necessarily presupposing that the attributee is aware of her mental state. For one can be in a given mental state – e.g. one may have an unconscious desire – one is not aware of. In short, sentences n-accepted are devices we use to classify mental representations. Two agents who n-accept the same sentence share the same mental representation type. The state they are in depends on the attribution, i.e. the psychological prefix. It is the latter which classifies the state the agent is in and not the sentence n-accepted. If the attribution is “A wishes that S,” then A is in a “wishing state” whose mental representation is classified by the sentence S that A n-accepts. If the report were “A says that S,” the sentence S A naccepts would still classify the same mental representation, but it would not attribute A the same mental state: A would be in a “saying state.” The state the attributee is in is expressed by the psychological prefix. Think of mental representations as being governed by sorts of operators like BELIEVE, THINK, SAY, WISH, etc. The operator that governs the agent’s representation gives us the agent’s mental state. When in a given mental state, the subject activates a token mental state, which I take to be a cognitive particular which may (but need not) be some neurological activity going on in her brain. In other words, when we attribute the n-acceptance of a sentence to someone, we classify the type mental representation instantiated by the corresponding token or cognitive particular she entertains. When someone is in a given mental state she activates a

Reports and Imagination

91

token mental state (a cognitive particular). The latter is what triggers the agent’s behavior. Two agents behave (ceteris paribus) the same way insofar as their cognitive particulars are of the same type. I now try to spell out the relation between a type mental representation as it is classified by a sentence and the corresponding cognitive particular. The n-acceptance relation can be cashed out as follows: n-ACC[A, S, t] iff (›mr)(›mr*){mr and mr* are token mental representations of the same type and mr would incline the attributer, in the attributee’s context at time t, to utter S & In-the-desire/belief...-box-Has[A, mr*, t]} This characterization fits with the transpositional principle. In particular, it should allow us to attribute the n-acceptance of sentences even to pre-linguistic infants and animals because the attributer projects herself into the attributee’s situation. This does not presuppose, though, that we are entitled to attribute mental states to pre-linguistic infants and/or animals. All I am committed to is the conditional thesis that if we feel entitled to attribute mental states to them, then we are presupposing a certain similarity between their mental life and ours.15 The notion of n-acceptance is spelled out in counterfactual terms: the attributee’s token mental representation (or cognitive particular) is similar to the one that would cause the attributer, in the attributee’s context, to utter S. As a matter of fact, the attributer and the attributee cannot share the same token mental representation. Nonetheless, token mental representations can be similar and be of the same type: two agents’ token representations are of the same type insofar as they n-accept the same sentences.

3. Reports and Mental Representations In (SA) the attributee and the attributer are one and the same person. Most of the attributions, though, are not self-attributions but attributions from the third person perspective where the attributee and the attributer diverge. These attributions would be of the form: TA. A said/believes/thinks/wishes/… that S 15

More on this in the next section.

92

Eros Corazza

In this case as well the sentence S helps classifying the attributee’s (A’s) thought. For if the reporter’s aim is to characterize the attributee’s mental life and state it is likely that she chooses, as a that-clause, a sentence which comes as close as possible to the sentence the attributee would utter in expressing her belief/desire/wish/…16 As I have argued, the notion of n-acceptance entails that only token representations of the same type can trigger agents to utter a given sentence or a translation of it. This sounds circular, for I claimed that A’s and B’s token representations are of the same type if A and B n-accept the same sentence and that if A and B n-accept the same sentence their token representations are of the same type. To break the circle we can focus on the way I spell out the transpositional principle and, in particular, on the fact that the sentence naccepted is the one the attributer would utter in expressing her attitude if in the attributee situation. In her attribution the attributer is thus assuming that the attributee holds a token representation similar to the one she would have in the attributee’s context. If the attribution is appropriate then the attributer is able to classify the attributee’s state. It goes without saying that in some cases, to get at what the attributee has in mind, the attributer needs to spell out in some detail the sentences which the former is disposed to naccept. This is, for instance, the case when we deal with Kripke’s (1979) Peter‘Paderewski’ puzzle, where the relevant sentences (whose n-acceptance is attributed to Peter) will be “Paderewski-the-musician is F” and “Paderewskithe-politician is F.” In so doing, the attributer is able to classify the attributee’s cognitive world sufficiently enough: in particular, she is able to point out that Peter’s token representations of Paderewski are of different types. It seems, then, that sentences n-accepted can cut sufficiently finely to classify one’s mental world and, thus, to distinguish causally different token mental representations. On the other hand, one could argue that the account I am proposing does not allow us to ascribe beliefs to animals, for the sentences whose n-acceptance we attribute them may classify mental states too finely. Since my cat, Felix, does not speak and, thus, does not belong to a linguistic community, he may not master the concept Doberman. So, in saying “Felix thinks that a 16 It goes without saying that in reporting indexical thoughts, some changes and adjustments need to be done. In reporting an ‘I’-thought, for instance, the reporter is likely to use what Castañeda (1966, 1967) called quasi-indicators. So an indirect report of “I am happy” is likely to be “A said that he (himself )/ she (herself ) is happy.” I discuss attributions involving quasi-indicators and how they can fit into a direct-reference framework in Corazza (1998). For a more exhaustive discussion of quasi-indicators see Corazza (1995: ch. 4 and 5).

Reports and Imagination

93

Doberman is around the corner” we cannot attribute him the n-acceptance of the sentence “A Doberman is around the corner.” Felix’s token mental representation is too different from the one that, in Felix’s context, would trigger me to utter “A Doberman is around the corner.” So, it cannot be of the same type. I do not deny this fact. But, if we do not think that a cat has the concept Doberman, then our attribution is inappropriate and we should look for a more accurate one which fits Felix’s cognitive world. In so doing, we attribute to him the n-acceptance of a sentence that classifies Felix’s state qua type of his token representation. I think that when dealing with a non-linguistic organism this is the best we can do. In other words, if the attributer assumes that Felix has a token representation which is similar to the one she would have in Felix’s context, then she is entitled to attribute him the n-acceptance of the sentence she would use, in Felix’s context, to state her attitude. If the attributer is not entitled to assume that Felix’s token representation is similar to the one which would trigger her to utter the relevant sentence, then she ought to look for another, more appropriate, token representation. This should be closer to Felix’s token representation and would trigger her to utter, in Felix’s context, the relevant sentence.17 Since the notion of n-acceptance is spelled out in counterfactual terms, it also allows us to accommodate so-called tacit beliefs such as the belief I held yesterday that there were no pink elephants flying on campus. In attributing this attitude to myself, I would say “Yesterday I did not believe that there were pink elephants flying on campus” and, in so doing, I attribute myself the n-acceptance of the sentence “There are no pink elephants flying on campus.” If I were to express this belief yesterday, I would have used this very sentence. The same story can be told about the attribution of tacit beliefs to others. It goes without saying that we have many tacit beliefs which will never become manifest. So, when I attribute a tacit belief, all I am doing is claiming that if our agent were to express her belief, she would use the sentence whose n-acceptance I am attributing to her. In such a case, she would be in the mental state classified by this sentence. I am not attributing the actual presence of a mental representation corresponding to the tacit belief. Moreover, as I said before, one need not be aware of the mental state one is in. Before going further I would like to address the following question: In attributing to Sue the n-acceptance of the English sentence, “Man is a rational 17 An alternative and more drastic way to deal with the attribution of attitudes to nonlinguistic organisms would be to argue that, since these organisms are conceptually too different from us, all we can do is to use de re locutions and, in so doing, we do not specify the mental state they are in.

94

Eros Corazza

animal,” do we ascribe to her the same mental representation as we would in attributing to her the n-acceptance of the French sentence “L’homme est un animal rationnel”? Do these sentences classify the same mental representation? If these sentences are translations of one another, my answer is affirmative. The very same attribution can be made in several languages. The question, as I understand it, amounts to knowing whether (and in which circumstances), two sentences translate each other. We feel comfortable in saying that “Sue believes that Tully is Roman” and “Sue croit que Tullius est romain” translate each other and attribute the same belief to Sue. We feel less comfortable in saying that “Sue believes that Tully is Roman” and “Sue croit que Cicéron est romain” translate each other and attribute to her the same belief. That is, we feel entitled to translate ‘Tullius’ by ‘Tully’ and not by ‘Cicero’, although these names are co-referential. The problem is most acute and arises when we do not face this kind of symmetry among languages as, for instance, when one language has more names for an individual than another. In Hebrew, there are two names for Germany: ‘Germaniah’ and ‘Ashkenaz.’ So, if Peter, a competent speaker of German, believes that Germany is his country, which sentence of Hebrew does he accept? This is a general problem having to do with the practice of translation and should not undermine the basic idea of the picture I am defending. You may recall that in a report, we attribute to the agent the disposition to n-accept the reporter’s sentence. We do so by means of the transpositional principle. This principle allows us to subsume under the same attribution people belonging to very different cultures. It could be that Sue belongs to a completely different community. It could also be that some of her beliefs are incommensurable with ours. The transpositional principle, though, makes her beliefs accessible to us. This principle stipulates, so to speak, a kind of similarity across cultures. But it is on the basis of this stipulation that attributions to people belonging to very different communities are possible. It may be worth mentioning that the analogy I was proposing between the attribution of pain to others and the attribution of beliefs may be misleading insofar as we distinguish between biological similarities and cultural ones. We do not hesitate to subsume under the same biological category a Mongolian and a Palestinian and attribute to them the feeling of pain. However, we may not subsume them under the same cultural category. So, when we attribute to them attitudes involving terms intrinsically tied to their culture we guess what is going on and we look for the sentence in our language which comes close to what we think represents their attitude. By virtue of the transpositional principle we attribute to them the n-acceptance of our sentence. This is the best we can do in our everyday practice:

Reports and Imagination

95

When native belief systems and “form of life” differ radically from our own, there is no shorter way to characterize their cognitive world. The descriptive apparatus of folk psychology is not designed to deal with the beliefs of exotic folks. ... in those contexts where ideological differences are both important and too great to warrant comfortable use of content sentences drawn from our own language, we are not forced into silence about the native doxastic world. We can opt for an anthropological description, using their labels for their beliefs and detailing how their beliefs fit into their own form of life. (Stich 1983: 102-3) I argued that someone is disposed to n-accept whatever sentence translates the sentence whose n-acceptance we actually attribute her. It seems then that my strategy crumbles under Kripke’s (1979) Pierre-‘London’/‘Londres’ puzzle. You may recall that Pierre assents to “Londres est jolie” and dissents from “London is pretty.” But, if Pierre n-accepts “Londres est jolie” he is disposed to n-accept “London is pretty” and, conversely, if he n-accepts “London is ugly” he is disposed to n-accept “Londres est moche.” In virtue of the transpositional principle, if our attribution is in French we attribute him the n-acceptance of a French sentence while if the attribution is in English the nacceptance of the English sentence. So it turns out that Pierre is disposed to naccept both “London is pretty” and “London is ugly.” This is the problem. The puzzle, as I understand it, includes the fact that Pierre does not know that ‘London’ translates ‘Londres.’ Notice that in stating the puzzle, we make reference to ‘London’ and ‘Londres’ and to the fact that Pierre does not know that they are the French and English words for the same thing. In so doing, we stress that Pierre is not disposed to n-accept the relevant sentence when translated, i.e. the sentences “London is ugly” and “London is pretty” cannot be successfully used to classify Pierre’s mental state. This is precisely the point of Kripke’s puzzle. If Pierre knew that ‘London’ translates ‘Londres’ he would certainly give up one of his beliefs or, at least, suspend his judgement. In saying that a subject who is disposed to n-accept a sentence is in a dispositional relation to each translation of the sentence whose n-acceptance we actually attribute her, I am saying that our subject is disposed to n-accept these sentences ceteris paribus. That is, modulo the fact that if she knew that they translate each other she would n-accept them. Pierre does not know that ‘London’ translates ‘Londres,’ so the ceteris paribus clause is not satisfied. In other words, a subject n-accepts a sentence of a foreign language either if: (i) she does not know the language in which the report is formulated, or (ii) she knows the language in which the report is formulated and, unlike Kripke’s Pierre, knows that the relevant sentences translate each other. The ceteris

96

Eros Corazza

paribus clause allows us to rule out cases where one of these conditions is not satisfied. As I understand Kripke’s puzzle, the reports “Pierre believes that London is pretty” and “Pierre croit que Londres est jolie” mean the same thing, ceteris paribus. If (i) or (ii) does not hold the reports do not mean the same thing, but in that case further information is needed: in Kripke’s story we have to point out that Pierre does not know that ‘London’ translates ‘Londres.’

4. Conclusion The picture I have presented should capture the rather intuitive and commonsensical view that when we attribute someone pain and the like, we automatically and without reflection project ourselves into their skin. The same empathetic exercise is at play when we attribute propositional knowledge as well. The attribution of propositional attitudes rests on the very same capacity, on the fact that we are essentially dramatic personas. The case of autism I discussed seems to support this picture, while the transpositional principle allows us to understand attitude reports and explain away some famous problems and puzzles. There is, though, another important issue I did not address. Attitude reports are audience-directed. For this reason, the attributer often needs to accommodate the report in view of the audience’s knowledge of the reported situation. Hence a report may not and cannot always be faithful in capturing the attributee’s mental state. In indirect discourse, for instance, we may need to change the words the attributee used because the audience would be unable to understand the report if the original words were preserved. If we are aware of the fact that our audience does not know that vixens are female foxes, we cannot report Jon’s utterance “A vixen made a mess in my garden” as “Jon said that a vixen made a mess in his garden.” We would rather say: “Jon said that a female fox made a mess in his garden.” For this reason our report may not be fully de dicto and thus capture the attributee’s mental state in its integrality. Because a report is audience-directed it often needs to be (at least partly) de re. Besides, it is often difficult, if not impossible, to detect whether a given report is de dicto or de re or a mixture of both. Thanks to the transpositional principle, the picture I put forward explains the mechanics at play when we try to capture someone’s mental life. It does not, however, presuppose that in each report we shall (and are even able to) capture them. Without doubt more should be said about the relationship and the tension between a report qua faithful attribution of someone’s mental state and a

Reports and Imagination

97

report qua audience-directed communicative act. But this is a tale for another occasion. Acknowledgment: This paper was written whilst visiting the Institute Jean-Nicod (June-July 2001), supported by a fellowship from the French Ministry of Research (No.303553L). To both institutions I express my gratitude. For comments I would also like to thank Jérôme Dokic, Elisabeth Pacherie and Joëlle Proust.

Bibliography Baron-Cohen, S. 1993. “Theory of Mind and Autism: A Fifteen Years Review.” In Understanding Other Minds: Perspectives From Developmental Cognitive Neuroscience. Edited by S. Baron-Cohen, H. Tager-Flusberg, and D. Cohen, 3-20. Oxford: Oxford University Press. Castañeda, Hector-Neri. 1966. “‘He’: A Study in the Logic of Self-Consciousness.” Ratio, Vol. 8 , No. 2: 130-157. —. 1967. “Indicators and Quasi-Indicators.” American Philosophical Quarterly, Vol. 4, No. 2: 85-100. Corazza, Eros. 1995. Référence, Contexte, et Attitudes. Paris/Montréal: Vrin/Bellarmin. —. 1998. “She*: Pragmatically Imparted or Semantically Encoded?” In Thought, Language and Ontology – Essays in Memory of Hector-Neri Castañeda. edited by Francesco Orilia, and William-James Rapaport, 217-234. Dordrecht: Kluver. Cottingham, John. 1986. Descartes. Oxford: B. Blackwell. Dretske, Fred. 1969. Seeing and Knowing. Chicago: University of Chicago Press. Goldman, Alvin. 2000. “Simulation Theory and Mental Concepts.” In Simulation and Knowledge of Action. Edited by Jérôme Dokic, and Joëlle Proust, 35-72. Paris: Biblioth que du CREA:. Hobson, R. P. 1990. “On Acquiring Knowledge About People and the Capacity to Pretend: Response to Leslie (1987).” Psychological Review, Vol. 97, No. 1: 114-21. —. 1993. Autism and the Development of Mind. Hove: Psychology Press, Erlbaum. Hutchinson, W. D., K.D. Davids, A. M. Lozano, R.R. Tasker, and J.O. Dostrowsky. 1999. “Pain Related Neurons in the Human Cingulate Cortex.” Nature Neuroscience, Vol. 2, No. 5: 403-5. Kripke, Saul. 1988. “A Puzzle about Belief.” In Meaning and Use. Edited by A. Margalit, 239-83. Boston: Dordrecht, 1979: 239-83. Reprinted in Propositions and Attitudes. Edited by Nathan Salmon, and Scott Soames, 102-148. Oxford: Oxford University Press. Perry, John. 1980. “Belief and Acceptance.” Midwest Studies in Philosophy, Vol. 5: 5342. Reprinted in John Perry. The Problem of the Essential Indexical and Other Essays: Expanded Edition, 45-56. Stanford: CSLI Publications.

98

Eros Corazza

—. 1980a/2000. “A Problem About Continued Belief.” Pacific Philosophical Quarterly, Vol. 61: 317-32. Reprinted in Perry John. The Problem of the Essential Indexical and Other Essays: Expanded Edition, 57-75. Stanford: CSLI Publications. Proust, Joëlle. 2000. “Can ‘Radical’ Simulation Theories Explain Psychological Concept Acquisition?” In Simulation and Knowledge of Action. Edited by Jérôme Dokic, and Joëlle Proust, 387-435. Paris: Bibliothéque du CREA. Quine, William van Orman. 1960. Word and Object. Cambridge Mass.: MIT Publications. Sacks, Oliver. 1993-94. “A Neurologist’s Notebook: An Anthropologist on Mars.” The New Yorker, 27 December–3 January: 106-25. Stich, Stephen. 1983. From Folk Psychology to Cognitive Sciences. Cambridge, Mass.: MIT Publications. Vendler, Zeno. 1984. The Matter of Minds. Oxford: Oxford University Press. Wittgenstein, L. 1958. Philosophical Investigations. Oxford: B. Blackwell, —. 1967. Zettel. Oxford: B. Blackwell.

On Representing Content

Propositional Content and Cognitive Structure

99

100

David Hunter

On Representing Content

101

On Representing Content David Hunter

Abstract I consider whether the content of a speech act is best represented by a set of possible worlds or by an ordered set containing the individual and properties the speech act is about. I argue that there is nothing in such contents that an ordered set can represent that a set of worlds cannot. In particular, both can be used to capture what is distinctive about singular propositions. But a set of worlds better represents content in cases where the content concerns individuals that no longer exist. It is also better at representing how content can be expressed in different ways, and how assertion relates to the pursuit of truth. Finally, representing content by a set of worlds allows for a clearer view of the puzzle about logical omniscience, even though it is often taken to founder on that puzzle.

Suppose Jones were to say something in assertively uttering (1). (1) Tony Blair is a proud father. Jones said things to be a certain way. A natural thought is that to report what Jones said one need only report how he said things to be: how things would be if things were as he said them to be. This natural thought supports a proposal about how we might represent what he said. Putting it picturesquely, we could represent what Jones said by a set of possible worlds: those where things are as he said them to be. I am inclined to think that this natural thought and the proposal it supports are right.1 But both are now almost universally rejected. All sides (or almost all) agree that whatever we use to represent what Jones said must determine that set of possible worlds. But most contend that this is not enough. There is, they claim, more to represent in what Jones said than is represented, or representable, by a set of possible worlds. According to one view, what Jones said can be better represented by an ordered set containing Tony Blair and the prop1 This proposal is familiar from the work of Robert Stalnaker. See the papers collected in (Stalnaker 1999). But it is also found, though less picturesquely, in Charles Travis’ work, especially (Travis 2000). What I say here, and especially in section 5, owes an important debt to their work.

102

David Hunter

erty of being a proud father.2 Some argue that even this is inadequate. But my aim here is to defend the natural proposal by showing that a set of possible worlds is in fact better at representing what Jones said than is that ordered set. This will not show that it is the best way to represent what he said. But it will be a substantial step towards this goal.

1. Truth What does an ordered set containing Tony Blair and the property of being a proud father represent? In a perfectly clear sense, it represents nothing at all. Sets have no representational properties of their own. We may use them to represent things. And we may use their having certain properties, including their having certain constituents, and even the properties of those constituents, to represent things or their properties. But neither sets nor their properties represent anything on their own. This is true for sets of possible worlds as well as for sets that contain people and properties. So if we are to represent what Jones said by a set we must say something about what it is that sets and their properties are to be used to represent. And to do this, we must first identify what we want or need to represent in what Jones said. We might start by noting some facts that we do not need to represent. (2) (a) (b) (c) (d) (e) (f ) (g) (h)

Jones referred to Tony Blair. Jones mentioned the property of being a proud father. Jones uttered the name “Tony Blair”. Jones uttered the predicate “is a proud father”. “Tony Blair” is a name for Tony Blair. “Is a proud father” expresses the property of being a proud father. Jones referred to Tony Blair in, or by, uttering “Tony Blair”. Jones ascribed to Tony Blair the property of being a proud father.

This is a list of facts concerning how Jones said what he did and some of the resources he employed. We could add more to this list. If we wanted to represent Jones’ saying what he did, the action he performed, then we might need to represent these facts. Perhaps it is essential to his saying what he did that he said it in the way he did, using the resources he did. But if we want to represent what he said, and not his saying it, then we need not represent 2 This view is traceable to Russell’s early work (Russell 1903). But it is developed most comprehensively in recent times by David Kaplan (Kaplan 1989), Scott Soames (Soames 1987), Nathan Salmon (Salmon 1986).

On Representing Content

103

these facts. For in saying what he did, Jones did not say any of these facts to obtain. He said nothing about his words, or about his actions, or about the resources he employed in saying what he did. And what he said does not entail any of these facts. So if there is something more in what Jones said than is representable by a set of worlds, it is not on this list. One fact that we do need to represent is that what Jones said depends for its truth on the ways things are. It is clear how to stipulate how a set of worlds is to represent this. The set contains those worlds where things are as Jones said them to be. That set will either contain the actual world or it won’t. If it does, that can represent that what Jones said is true. If it does not, then that can represent that what Jones said is false. In this way, facts about which worlds the set contains can represent that what Jones said depends for truth on how things actually are. The required stipulation is also straightforward in the case of the ordered set. That it contains Tony Blair and the property of being a proud father, and their order in the set, can together represent the fact that what Jones said is true just in case he has that property. In this way, the ordered set can represent that what Jones said depends for its truth on the facts. There is another thing in what Jones said that we need to represent. Suppose that Lucy were to say something in assertively uttering (3). (3) The Prime Minister of Great Britain is a proud father. And suppose that what she said would have been true even if John Major had won the last election, so long as he had been a proud father. In fact, she and Jones both said something true. And what each said was in fact made true by how things are with Tony Blair. But there is a sense in which what Jones said is more intimately about Tony Blair. What Lucy said could have been true even if Tony Blair had never been born. By contrast, what Jones said entails that Tony Blair exists. This is a fact about what Jones said that we need to represent. That the ordered set contains Tony Blair can represent this. And the fact that it contains the property of being a proud father can represent the fact that whether what Jones said is true depends on what has that property. To mark the difference between what Jones and Lucy said we could represent what she said by a set that does not contain Tony Blair, but that does contain that property. The fact that both sets contain that property would represent the fact that what each said depends for its truth on which things have it. These facts about what Jones and Lucy said are equally well representable by sets of possible worlds. The set in Jones’ case contains those worlds where

104

David Hunter

what Jones said is true. So each of these is a world where Tony Blair exists, and where he has the property of being a proud father. This fact about the worlds in that set can represent the fact that what Jones said is essentially about Tony Blair, and that it depends for its truth on whether he has that property. By contrast, the set of worlds where what Lucy said is true includes some where John Major is the Prime Minister of Great Britain and a proud father, and others where Tony Blair does not exist at all. But in each of the worlds someone is a proud father. These facts about the worlds in Lucy’s set can thus represent the fact that what she said is only contingently about Tony Blair, but essentially about the property of being a proud father. So far, then, there is nothing in what Jones said that can be represented by the ordered set that cannot be represented by the set of possible worlds. But there is a kind of case where what is said is better represented by a set of worlds than by an ordered set. Suppose that Amy were to say something in assertively uttering (4). (4) Aristotle was a proud father. And suppose that she said something whose truth depends on how things were with Aristotle, the student of Plato’s. Which ordered set could be used to represent what Amy said? Since Aristotle does not exist, there is no set that has him as a constituent. This is a serious problem for one who wants to represent what Amy said by such a set. There are two main options. Some have suggested that what she said can be represented by a gappy set, one that contains nothing in one of its positions.3 The difficulty is that this way of representing what Amy said fails to distinguish it from what would be said using (5). (5) Plato was a proud father. For since Plato no longer exists, there is also no set that contains him. Others have suggested that although there is now no set that contains Aristotle and the property of being a proud father, there once was, and though it no longer exists, it nonetheless now has properties.4 Perhaps among its properties is that of representing what Amy said. This is not the place to evaluate either option. 3 The idea seems to trace back to a footnote on page 496 of (David Kaplan 1989). For a recent development of it, see (Braun 1993). 4 See (Salmon 1998).

On Representing Content

105

For present purposes, it is enough to note that this problem is completely avoided if we represent what Amy said using a set of worlds. For a possible world is a complete history. To say that there is a possible world where Aristotle is a proud father is to say that the world is such that it might have been that way. We can represent what Amy said using a set of worlds in which Aristotle was a proud father. And we can represent what would be said using (5) by a set of worlds in which Plato was a proud father. This problem, of representing what is said when what is said is about things that no longer exist, is easily solved on a possible worlds account, but not at all straightforward on an ordered set account.

2. Reference One might think that facts about reference show that what Jones said can be better represented by an ordered set than a set of worlds. In saying what he did, Jones referred to Tony Blair. The ordered set can represent this fact. One could stipulate that the fact that that set contains Tony Blair is to represent the fact that Jones referred to him in saying what he did. And that the ordered set representing what Lucy said does not contain Tony Blair can represent the fact that she did not refer to him. Plausibly, the set of possible worlds cannot represent these facts. For Tony Blair is not the only things that exists in each of the worlds in that set. His parents do, and so does the number two. Plausibly, one could not tell, merely by examining the possible worlds in that set, which things Jones referred to in saying what he did. So, plausibly, that set cannot represent the fact that Jones referred to Tony Blair. Notice, though, that this fact is on list (2). That Jones referred to Tony Blair is a fact about how he said what he did. It is not one of the facts he said to obtain. Nor is it entailed by anything he said. To represent what he said we need not represent this fact about how he said it, any more than we need to represent that he said it in English, or that he used the name “Tony Blair”. So, even if the set of worlds cannot represent that fact but the ordered set can, this would not show that the ordered set better represents what Jones said. For that fact concerns how he said what he did, not what he said. But what if it were essential to what he said that it be said that way? What if one could not say what he said without referring to Tony Blair? This would be a fact about how to say what he said. If this were essential to what he said, and if the set of worlds cannot represent this but an ordered set can, this would show that the ordered set is better than the set of worlds at representing this aspect of what Jones said. But by ordinary standards, this is

106

David Hunter

not essential to what he said. What he said could, on some occasions, and by ordinary standards, equally well be said in assertively uttering the following. (6) Cherie Blair had a boy. What is more, this fact about what Jones said cannot be represented by an ordered set, but can be represented by a set of worlds. The set of worlds is thus better than an ordered set at representing this aspect of what Jones said. But I will postpone defending these claims until section 5.

3. Belief I One might think that facts about what it is possible to believe show that what Jones said can be better represented by an ordered set than by a set of worlds. Suppose that Kate said something in assertively uttering (7). (7) The actual Prime Minister of Great Britain is a proud father. The word “actual” in (7) has to be understood in a special way. It is contingent that what Lucy said using (3) is made true by how things are with Tony Blair. (3) The Prime Minister of Great Britain is a proud father. This is because the description she used expresses a property that is not essential to Tony Blair and that someone else might have had. Tony Blair need not have been Prime Minister, when Lucy said what she did. John Major might have been instead. Adding the word “actual” to the description is to have the effect of making it express a property that is essential to Tony Blair, and that only he could have had. No one but Tony Blair could have had that property and he has it essentially. This, in turn, is to have the effect of making what Kate said equivalent to what Jones said. What Kate said and what Jones said determine the same set of possible worlds. But one might think that they said different things. It is clear that they said what they did in very different ways. For one thing, Kate did not use the name “Tony Blair”. But this does not by itself show that they said different things. For all sides agree that there can be different ways to say the same thing. In particular, all sides agree that what Jones said could, in some contexts at least, also be said using (8), even though it is a different sentence than (1).

On Representing Content

107

(8) Anthony is a proud father. So both sides must agree that the fact that Kate and Jones used different resources to say what they did does not show that they said different things. But facts about what it is possible to believe might show this. If it is possible to believe what one person said while not believing what another said, then different things were said. So if it is possible to believe what Jones said and not to believe what Kate said, then they said different things. But since what they said determines the same set of possible worlds, this would show that there is something in what Jones (or Kate) said that cannot be represented by that set. And perhaps this difference could be represented by an ordered set. Perhaps what Kate said can be represented by a set containing the property of being Prime Minister. If so, this would show that the ordered set better represents what Jones said. So, is it possible to believe what he said and not to believe what she said? The following story might be thought to show that it is possible. Suppose that Max understands sentences (1) and (7). Suppose also that he believes that in uttering (1) Jones said something true but that in uttering (7) Kate said something false. If this were possible, wouldn’t it show that Max believes what Jones said and does not believe what Kate said? And wouldn’t this show that Jones and Kate said different things? It would, but only if the following principle were true. (P) If a rational person understands two sentences, and believes that an utterance of one on some occasion said something true and an utterance of the other on that occasion said something false, then different things were said. If (P) is true and if what was supposed about Max is possible, then Jones and Kate said different things. If so, then what Jones (or she) said cannot be represented by a set of possible worlds. So one who thinks that it can must either reject (P) or deny that what was supposed about Max is possible. Notice that one who believes that what Jones said can be best represented by an ordered set faces a parallel dilemma. For suppose that Hank says something in assertively uttering (8), in a context where it is clear he is referring to Tony Blair. (8) Anthony is a proud father.

108

David Hunter

Suppose that Max also understands (8). Suppose further that Max believes that in uttering (1) Jones said something true but that in uttering (8) Hank said something false. If all of this is possible, and if (P) is true, then Jones and Hank said different things. But according to one who thinks that what Jones said can be represented best by an ordered set, they said the same thing. So, one who thinks this must either deny that what was supposed about Max is possible or else deny (P). Most deny (P). To justify this denial, one must explain how it is possible to understand two sentences, believe that an utterance of one said something true and an utterance of the other something false, when in fact they said the same thing. Two strategies are familiar. First, some hold that in cases like Max’s the subject will have confused what was literally said with what was otherwise communicated.5 Suppose that in saying what they did, in the ways they did, Jones communicated something about the name “Tony Blair” and Hank communicated something about the name “Anthony”. And suppose that it is possible for someone to believe what Jones communicated and not believe what Hank communicated. If Max confused what Hank said with what Hank communicated, and if Max believed that what Hank communicated was false, this would lead Max to conclude that Hank said something false. But perhaps, for all that, Max might count as understanding sentence (8), and might believe that Jones said something true. He simply misunderstood what Hank said. This might well explain why Max believed that Jones said something true and Hank said something false, even if in fact they said the same thing. If all of this is possible, then (P) is false. A similar story could explain why Max believes that Jones said something true in uttering (1) but that Kate said something false in uttering (7), even if both said the same thing. (1) Tony Blair is a proud father. (7) The actual Prime Minister of Great Britain is a proud father. Perhaps Max confused what Kate said in uttering (7) with something she otherwise communicated about the words she used. And perhaps Max believes that what Kate communicated was false. If so, then this would explain why Max believes that Jones said something true and that Kate said something false, even if, in fact, they said the same thing. Others have tried to explain how (P) is false by appealing to ways of believing.6 All sides agree that there are different ways to say the same thing. Sup5 For a development of this, see (Soames 1987). 6 For developments of this, see (Salmon 1986) and (Braun 1998).

On Representing Content

109

pose that, by rough analogy, there are different ways to believe that same thing, corresponding to the ways there are to say it. Suppose one could believe what Jones said in a “Tony Blair” way but also in an “Anthony” way. And suppose one could believe it in one way but fail to believe it in the other. If Max believes what Jones said in the “Tony Blair” way, then he might be disposed to believe that Jones said something true. For Jones expressed something Max believes in words corresponding to the way Max believes it. But if Max does not believe it in the “Anthony” way, then he might not be disposed to believe that Hank said something true. For even if he believes what Hank said, Max might not recognize this if he does not believe it in the way corresponding to the words Hank used to say it. If all of this is possible, then (P) is false. A similar story could explain why Max believes that Jones said something true in uttering (1) but that Kate said something false in uttering (7), even if, in fact, both said the same thing. Perhaps Max believes what Jones said in the “Tony Blair” way, but not in the “The actual Prime Minister of Great Britain” way. If so, this would explain why Max believes that Jones said something true and that Kate said something false. For Jones said what Max believes in words corresponding to the way Max believes it. Kate did not. This is not the place to evaluate either of these strategies. What I want to make clear is, first, that all sides in the current debate face the same dilemma. Whether one thinks that what Jones said can be best represented by a set of possible worlds or by an ordered set containing Tony Blair, one is faced with denying (P). Second, all sides in the debate can agree on why (P) is false. There is nothing in the strategies I have sketched that would prevent a champion of sets of worlds from adopting them. These strategies aim to show that (P) is false. But even if (P) were true, I doubt that it could provide a useful test for deciding when two people have said the same thing. For using it as a test requires judgments about when people count as understanding sentences. But these judgments are not independent of ones about when two people have said the same thing. For, in general, to understand a sentence is to know what it could be used to say. So we cannot decide whether Max understands the sentences in question without deciding whether they can be used to say the same thing. This is reflected in the two accounts of how (P) is false. According to one, Max confuses what a sentence can be used to say with what it can be used to convey. In the other, he fails to see that a sentence can be used to say something he believes. These are both ways of failing to understand the sentence. So these strategies imply that Max does not really, or does not fully, understand the sentences in question after all.

110

David Hunter

What is more, there is independent reason to think this. For in both cases Max is suffering some semantic error or ignorance. He believes that in uttering (1) Jones said something true but that in uttering (8) Hank said something false. So he must believe that “Tony Blair” and “Anthony” were used to refer to different people (and that only one of them is a proud father). But the names were used to refer to the same person. So Max is mistaken about the semantic facts. Perhaps this mistake does not amount to a lack of understanding. But it would explain why he fails to see that the sentences were used to say the same thing, and why he believes Jones said something true and Hank said something false. Anyone who shared Max’s error would also conclude that Jones and Hank said different things. So there is an explanation of Max’s beliefs about what was said that appeals to facts that everyone must accept, and that does not entail that different things were said. All else equal, this is a better explanation. There is a similar, though more complex, story in the case of what Kate said. Max believes that in uttering (7) Kate said something false. So Max must believe that “The actual Prime Minister of Great Britain” does not denote the person “Tony Blair” is a name for (and that it denotes someone who is not a proud father). Assuming that he understands the name “Tony Blair”, it must be that he believes that that description does not denote Tony Blair. And this must be because he believes that Tony Blair is not the Prime Minister of Great Britain. Perhaps he believes that John Major is, or perhaps he does not know who is Prime Minister. Now, one can know what Lucy said using (3) without knowing who the Prime Minister is. But to know what Kate said using (7), one must know who the current Prime Minister of Great Britain is. (3) The Prime Minister of Great Britain is a proud father. (7) The actual Prime Minister of Great Britain is a proud father. This is clear if we consider a world where John Major won the last election. To decide whether what (3) says is true with respect to such a world one need not know that Tony Blair is in fact the Prime Minister. For the truth of what (3) says does not turn essentially on how things are with him. But to decide whether what (7) says is true with respect to such a world, one has to know that Tony Blair is in fact the Prime Minister. For the truth of what (7) says does depend essentially on how things are with him. What (7) says is true with respect to a world only if Tony Blair is, in that world, a proud father. So, one cannot evaluate whether what is said using (7) is true unless one knows who the Prime Minister is. Since Max does not know who it is, he is not able

On Representing Content

111

to evaluate its truth: this is as much as to say that he does not know what was said using it.7 And all of this explains why he believes that Jones said something true and Kate said something false. Anyone who shared Max’s semantic error would also believe that Jones said something true and Kate said something false. So there is an explanation of these beliefs that appeals to facts that everyone must accept, and that does not entail that different things were said. All else equal, this is a better explanation.

4. Belief II A second puzzle about belief is more troubling. Suppose that Zoë says something in assertively uttering (9). (8) Tony Blair is a proud father and 2 + 2 = 4. And suppose that what she said is logically equivalent to what Jones said. What Zoë said determines the same set of possible worlds as what Jones said. But it might seem that one can believe what Jones said without believing what Zoë said. And perhaps what Zoë said can be represented by a set containing, among other things, the number 2. So, one might conclude, this case shows that what Jones (or at least what Kate) said can better be represented by an ordered set than by a set of worlds.8 In part, the puzzle is the same as that from the last section. One can understand a sentence and yet fail to realize that it can be used to say something one believes. But there is more to the current puzzle than this fact about linguistic knowledge. It seems that one can believe something without believing everything that it entails. Ordinary believers are not logically omniscient. This is undeniable. But it is, or ought to be, somewhat puzzling how this could be. Intuitively, if one has a belief then one knows how things would be if one’s belief were true. For what one believes is precisely that things are as they would be if hey were as one believes them to be. So, if one 7 This effect of adding “actual” to the description makes it, in this respect, like an indexical. To know what is said in a use of “I” you have to know who is using it, and to know what is said in a use of “here” you have to know where the user is. Understanding the terms does not, of course, require knowing what they are used to refer to on some occasion. Likewise, one can understand the description ‘the actual prime Minister of Great Britain’ without knowing who the PM is. But to understand a use of it, to know what is said using it, one must know a non-linguistic fact about the world. 8 What I say here was influenced by the discussion of the problem of logical omniscience in (Stalnaker 1991), reprinted in (Stalnaker 1999).

112

David Hunter

were not sure how the world would be were it to be as one believed it to be, then, in a sense, one would not really understand one’s belief. Now, if what Jones said were true, then two plus two would be four. So anyone who believed what Jones said and yet did not believe that 2 + 2 = 4 would not really know how things would be if her belief were true. But if she doesn’t know this, then she does not really understand what she believes. The puzzle is that it seems possible to believe something one does not fully understand. Of course, if we represent what Jones said by a set of worlds the puzzle is very prominent. For on this way of representing what he said, one who believes what he said believes everything entailed by what he said.9 Thus, one who represents what Jones said by a set of worlds faces the puzzle in a stark way. But the puzzle is not avoided by representing what Jones said by an ordered set. For, as we have seen, that ordered set must nevertheless determine that set of worlds. The puzzle arises from the fact that what is believed entails that 2 + 2 = 4. This is so however we represent what is believed. Even if what is believed can be best represented by an ordered set, one is left with the question how one can have a belief and yet not know how things would be if it were true. Nothing in the idea of such a set says anything about this. A decision to represent the content of a belief in a certain way says nothing about the believer’s understanding of that content. But this is what a solution to the puzzle must say something about. So representing what Jones said by an ordered set does not avoid the puzzle. In fact, the puzzle is distorted if we represent what Jones and Zoë said by different sets. For this suggests that what they said differs in information. It makes it seem that what Zoë said contains more information than what Jones said. And this suggests that one who believed what Zoë said would believe more than one who believed what Jones said. It is right that the belief that Tony Blair is a proud father and is tall, contains more information than Jones’ belief. In that case, there is a difference in what is believed, a difference that it would be appropriate to represent using different sets. But the difference between Jones’ and Zoë’s beliefs is not a difference in information. One who believes what Jones said has the very same information as does one who believes what Zoë said. Representing what they said by different sets distorts the situation by assimilating their case to a difference in information, to a kind of disagreement. By avoiding this, representing what they said by a set of worlds allows for a clearer view of what is really puzzling. 9 Not quite: rather, he would believe the conjunction of what Jones said and everything it entails.

On Representing Content

113

The puzzle is that one can have a belief while not having access to all the information believed. This is a puzzle, not about what is believed, but about one’s relation to what is believed. It is thus a puzzle about the nature of belief. Ordinary believers are able to see some of what their beliefs entail, but not everything. The idea of different ways to believe the same thing might help to make sense of how this is possible. Some information might be clearer or more evident when what is believed is believed in one way rather than in another. But this idea is often modeled on different ways to say the same thing. This model is, I think, too restrictive to yield a general solution. For the puzzle is not essentially about how we represent or say what we believe. It is about how belief relates to action in general, and not just to linguistic action, or even to rational deliberation. Moreover, this model does not by itself explain how different ways of believing the same thing relate in different ways to what is believed. And it says nothing about how different ways of believing something play different roles in guiding one’s actions. I doubt that this kind of account can by itself provide a general or complete understanding of the puzzle. In fact, I think, the puzzle stems from a tension in our ordinary conception of belief. We want the concept of belief, and so our belief ascriptions, to play two roles. On one hand, in saying what someone believes we want to say what information they have about the world. On the other, we want what we say to explain and predict how they will act on that information. The problem, as Stalnaker points out, is that we typically do not say how that information is available to them to guide their actions. Our ordinary concept of belief does not itself distinguish different ways information can be available to guide action. The context of belief attribution sometimes helps, by helping to specify the information or the relation to that information that is being ascribed. But the phenomena is complex. Consider the difference between ascribing an arithmetical belief to a young child and to an adult. Failure to assert or assent to the belief ascribed need not falsify the ascription to the child, but might well undermine the ascription to the adult. How can this be? Were the two said to stand in different relations to the same information, such that the adult but not the child was said to have it accessible for purposes of assertion and assent? Or were they said to stand in one relation to different information, such that the adult was said to have information that was partly about how his language can be used to say what he believes? These questions might be answered in different ways in different contexts. In this way, facts about the context of belief ascriptions can help. But I doubt that the puzzle can be wholly resolved by contextual factors alone. For if information can be avail-

114

David Hunter

able to guide some actions and not others, and if our concept of belief says nothing about how this is possible, then what we say about what people believe will sometimes distort either the information they have at their disposal or else the ways they will act on that information. If the puzzle stems in this way from a tension in our ordinary concept of belief, then it may not be the kind of puzzle that can be solved. There is a methodological point to be made here. There is a temptation, when giving a philosophical account of belief and belief reports, to provide solutions to various puzzles about the phenomena. This leads some to provide complex accounts of the metaphysics of belief or the structure of language in hopes of showing that what seemed puzzling is not really puzzling after all. While such accounts may be valuable, they are not the only kind of clarification that philosophy can provide. Sometimes, philosophy makes progress by revealing how a puzzle stems from tensions in our ordinary conceptions. This may be the kind of insight philosophy provides in the case of the problem of free will, for instance, where insight is gained by seeing how the problem arises out of conflicting elements in our ordinary conception of agency. The methodological point is simply that we should remain open to the possibility of this kind of insight in the case of belief and belief reports, one that does not remove a puzzle, but which reveals the puzzle’s source.10

5. Assertion and Context Representing what Jones said by a set of worlds fits a picture of the nature of assertion that makes clear why what I called the natural thought is natural.11 We make assertions in the course of being engaged in various rational pursuits. Our linguistic acts, like our other acts, are typically intended to further those pursuits and are typically meant to be understood in the context of them. One of these pursuits is to identify the ways things are. To do this, we must distinguish among they ways things might be, ruling out possibilities in hopes of isolating the ways things are. What is distinctive of assertion, as compared to other linguistic acts, is the role it plays in that pursuit. The aim in asserting something is to propose a distinction among the ways things might be. To say that P is to say that ways things might be that are incompat10 I take this methodological point to be one of the themes in Kripke’s discussion of belief ascription in (Kripke 1979). 11 In sketching this picture of assertion I am drawing on (Stalnaker 1978), reprinted in (Stalnaker 1999), and on Travis (2000).

On Representing Content

115

ible with P are not ways things are. The force of an assertion is that one believes this. This picture abstracts from the means we might have, or that we might use, to say something. It draws a sharp distinction between what is said and how it is said. It allows that we might say something in different languages, and in different ways in the same language. It even allows that we can say something without words. For on it, to say something is simply to propose a distinction among ways things might be. One can do this with a picture or a map. By distinguishing what we say from how we say it, this picture focuses on the way assertion fits into the rational pursuit of truth. This picture of assertion can be accepted by all sides. But representing what is said by a set of worlds fits it better than does an ordered set. For it is not clear how to use the ordered set to represent the fact that in saying what he did Jones aimed to distinguish among the worlds. As we have seen, an additional stipulation would be needed to connect this way of representing what is said with the rational activity of saying it. For the set containing Tony Blair and the property of being a proud father does not, by itself, determine any set of possibilities. No such additional stipulation is needed if we represent what Jones said using a set of worlds, since that plainly represents the distinction he aimed to draw. This picture also helps us to see that one need not refer to Tony Blair to say what Jones said. What Jones said was also said by Kate, even though she did not refer to Tony Blair. But she did use a description that expresses an essential property unique to him. In this sense, she used a bit of language that denotes Tony Blair. But even this is not needed. What Jones said could, in some contexts, equally well be said using (6). (6) Cherie Blair had a boy. To show this, I want to develop this picture of assertion a bit more. The aim of an assertion, on this picture, is to propose a distinction among the ways things might be. But among which of those ways? A reasonable answer is that it aims to distinguish among the possibilities that are still open at the point in the discussion when the assertion is made. Max and his friends have made some progress narrowing down the possibilities. The point of a further assertion would be to propose eliminating some of the remaining ones. On this picture, what one says in uttering a sentence depends on which possibilities are open when one says what one does. Given the connection between assertion and the pursuit of truth, this dependence makes sense.

116

David Hunter

With this in mind, suppose that Max and his friends are discussing Tony Blair, and that the following is common knowledge among them. Tony Blair is married to Cherie Blair; she has just given birth, but the baby’s sex is unknown to them; Tony Blair really wants a son: nothing would make him prouder than to be a father of a little boy. What they wonder is whether the baby is a boy. Suppose Jones hangs up the phone and assertively utters (1). (1) Tony Blair is a proud father. If his friends believe what Jones said, they have ruled out the worlds where Tony Blair is not a proud father. But, given their common knowledge, they have also ruled out worlds where Cherie had a girl. The only worlds that remain are ones where both he is a proud father, and she had a boy. So they have their answer. And all of this is common knowledge. Jones knew that uttering (1) would have this effect on their beliefs. It is clear how referring to Tony Blair and ascribing to him the property of being a proud father would effect this. In fact, this is the very effect he intended his utterance to produce. Now suppose that Lucy joins them, and assertively utters (6). (6) Cherie Blair had a boy. It would be natural for Max to reply: “Yes, we just heard.” For what she said is not news. There is nothing in what she said, or in what she intended to convey, that was not in what Jones said. What she said had already been said by Jones. Max’s imagined reaction reflects, I think, how we ordinarily judge when two people have said the same thing. In his own way, Jones said what Lucy did. And in her own way, she said what he did. But she did not refer to Tony Blair, and made no mention of being a proud father. He did not refer to Cherie and made no mention of a baby boy. But they said the same thing. And given the context, it is clear how the means they used could effect the very same thing. This shows that saying what Jones said does not require referring to Tony Blair. There are different ways to say what he said, differing not only in the words used, but also in what is referred to or mentioned. The picture of assertion helps to make sense of this fact. In saying what he did, Jones was proposing to rule out some of the open possibilities. To do this, he had to find some fact in the worlds he wanted to retain to distinguish them from those he was proposing to rule out. But, in this case, the worlds he wanted to keep differ from the eliminated ones in two ways. They are the worlds where Tony Blair is a proud father. They are also the ones where

On Representing Content

117

Cherie Blair had a boy. Given this, there are two ways he could identify those worlds. And this is clear to all in the context. In general, whenever a distinction among possibilities marks more than one difference among the worlds, there will be more than one way to express it. That what Jones said can be said without referring to Tony Blair cannot be represented by an ordered set. There is no way to tell from its constituents or their order that what Jones said can be said by referring to Cherie Blair. In part, this reflects the fact that representing what he said by such a set disconnects what he said from what he is trying to achieve in saying it, from what made his utterance a rational action. In part, it also reflects the fact that representing what he said by such a set ties what he said to the means he used to say it too closely. But this context-dependence can be represented by a set of worlds. For, in this case, the worlds in the set are the ones where Tony Blair is a proud father and where Cherie Blair had a baby boy. Neither of these facts obtains in any of the eliminated worlds. So this fact about the worlds in that set can represent the fact that what Jones said could, in that context, be said using either (1) or (6). So the set of worlds can better represent what Jones said than can the ordered set. I have argued that it is not essential to what Jones said that to say it one must refer to Tony Blair. It does not follow that one will, in every context, be able to say what he said without referring to or denoting Tony Blair. In some contexts, drawing the distinction Jones proposed may require referring to or denoting Tony Blair. That would be so when, in that context, there is no other contingent feature that distinguishes those worlds from the rest. If there is none, then to say what Jones said would, in such a context, require referring to or denoting Tony Blair and ascribing to him the property of being a proud father. In general, what it would take to say what Jones said depends on context. And the picture of assertion I sketched makes sense of this dependence. Let me summarize my discussion. I have been considering whether what Jones said is best represented by a set of possible worlds or by an ordered set containing Tony Blair and the property of being a proud father. There is nothing in what Jones said that an ordered set can represent that a set of worlds cannot. In particular both can be used to capture the fact that Jones expressed a singular proposition about Tony Blair. But there are several respects in which a set of worlds is superior to an ordered set. First, a set of worlds is better able to represent what is said in cases where what is said is about individuals that no longer exist. Second, a set of worlds is better able to represent the fact that what is said can be said in different ways. Third, representing what he said by a set of worlds fits better into a natural picture

118

David Hunter

connecting assertion to the pursuit of truth. Finally, representing what Jones said by a set of worlds allows for a clearer view of the puzzle about logical omniscience, even though it is often taken to founder on that puzzle. Sometimes, what seems like a failing is actually a virtue.12

Bibliography Braun, David. 1998. “Understanding Belief Reports.” Philosophical Review . —. 1993. “Empty Names.” Nous 27, no. 4. Kaplan, David. 1989. “Demonstratives.” In Themes from Kaplan, edited by John Perry Joseph Almog, and Howard Wettstein, 481-614. Oxford: Oxford University Press. Kripke, Saul. 1979. “A Puzzle about Belief.” In Meaning and Use, edited by A. Margalit, 239-83. Dordrecht: Reidel. Russell, Bertrand.1903. The Principles of Mathematics. London: Allen & Unwin. Salmon, Nathan. 1998. “Nonexistence.” Nous 32, no. 3. —. 1986. Frege’s Puzzle. Cambridge, MA: MIT Press. Soames, Scott. 1987. “Direct Reference, Propositional Attitudes, and Semantic Content.” Philosophical Topics 15. Stalnaker, Robert. 1999. Context and Content. Oxford: Oxford University Press. —. 1991. “The Problem of Logical Omniscience, I.” Synthese 89. —. 1978. “Assertion.” In Syntax and Semantics. New York: Academic Press. Travis, Charles. 2000. Unshadowed Thought. Cambridge, MA: Harvard University Press.

12

I would like to thank Andrew Botterell, David Braun, Andrew Hunter, Cara Spencer and Robert Stainton for helpful comments and discussion on earlier versions of this paper.

Conceptual Realism and Interpretation

119

Conceptual Realism and Interpretation Max A. Freund

Abstract Conceptual realism (a logico-philosophical semantic theory) has introduced a logical distinction between the cognitive structure of an assertion and its truth-conditions. We shall argue that the cognitive structure is part of the meaning of an assertion and that, consequently, should be taken into account when interpreting a natural language. We shall also explore this topic in relation to the problem of radical interpretation. The distinction will be made evident by first formulating a logical system (having conceptual realism as its philosophical background) and then exhibiting it in the formal system. This will be preceded by a description of the main philosophical features of conceptual realism.

An interpretation for a natural language is a theory that should provide, in any given situation and for every possible expression of the language, a meaning to that expression in that situation. As remarked by several authors, the theory must fulfill further conditions. It might be required, for instance, that the theory be sufficiently enough for semantic mastery of the language or that it should satisfy compositionality 1. It might also be demanded that the adequacy of the theory be ascertained by standards set by a theory of radical interpretation, that is, by a theory determining how interpretation should be carried out under very restrictive epistemic conditions. They are conditions under which the (radical) interpreter would not have access to informants knowing his language or to the informants’ linguistic intuitions concerning their native tongue. A theory for this situation should specify, among other things, the quality or quantity of the evidence a radical interpreter must take into account to construct an interpretation of the totally alien tongue. Indeed, there are approaches that limit these data to the informants’ observable behaviour only. Since there are convincing arguments for not adopting this position2 and because of the philosophical standpoint of the present article, we shall not here endorse that restrictive view on the evidence. 1 See, for example, Forster (1976). 2 See Chomsky ( 2000, chapter 3); and J. Fodor and E. Lepore (1992, chapter 3).

120

Max A. Freund

Conceptual realism3, on the other hand, is a logico-philosophical semantic theory, which has introduced a logical distinction between the cognitive structure of an assertion and its truth-conditions. It is this distinction and its relevance to the issue of interpretation the main focus of this paper. We shall argue that the cognitive structure is part of the meaning of an assertion and that, consequently, should be taken into account when interpreting a natural language. We shall also explore this topic in relation to the problem of radical interpretation. The distinction will be made evident by first formulating a logical system (having conceptual realism as its philosophical background) and then expressing the distinction in the formal system. This will be preceded by a description of the main philosophical features of conceptual realism.

1. Conceptual Realism: Predicable, Sortal and Referential Concepts; Concept Formation Conceptual realism can be initially characterized as a philosophical theory of predication and nominalized predicates, which (unlike nominalism and logical realism) looks at concepts as the semantic grounds for the correct or incorrect application of predicate expressions.4 Concepts should be here understood as dispositional entities, more precisely, as cognitive capacities, or cognitive structures otherwise based upon such capacities. Among these sorts of capacities, we want to single out first those concepts predicate expressions stand for. They are called predicable concepts and they are those capacities the possession of which allows one to characterize and relate objects in various ways. They correspond to those features of thought that determine the truthconditions of predicate expressions in different contexts of use and, in this way, they constitute the cognitive capacities that underlie our rule-following abilities in the correct use of predicate expressions. Associated to the use of predicate expression, there is the human capacity to nominalized predicate expressions, that is, the capacitiy to transform predicate expressions into abstract singular terms, such as in the transformation of “human” into “humanity” and “red” into “redness”. As singular 3 Professor Nino B. Cocchiarella originally formulated conceptual realism. He and the author of the present paper have intensively inquired into the logical features of this philosophical theory. 4 Two sorts of realism compose conceptual realism: intensional and natural realism. The former kind of realism deals with the denotata of nominalized predicates and propositional forms. The latter kind is concerned with natural kinds and natural properties and relations.

Conceptual Realism and Interpretation

121

terms, nominalized predicates might denote and when they do it, they denote objects. However, there is no assumption concerning the nature of such an object, apart from the requirement that a link should exist between the object denoted by the nominalized predicate expression and the concept such a predicate expression stands for. This link should be based on the object being somehow related to the truth conditions determined by the possible applications of the concept. For this reason, possible denotations of nominalized predicate expressions are also called concept-correlates. The denoted object might be intensional, in which case we would be talking of a variant of conceptual realism called conceptual intensional realism (if the ontological nature of the object is not independent of the mental process) or Conceptual Platonism (if the object is viewed as an entity ontologically independent from the mind). It is important to note that the process of nominalization provides to conceptual realism an explanation of how we have cognitive access to abstract objects. Predicable concepts and their correlates are not the only entities posited by conceptual realism; it also postulates other kinds of concepts. These capacities constitute, according to this philosophical theory, the semantic basis for linguistic expressions other than predicates. So conceptual realism is also a general philosophical semantic theory and, since its method involves construction of a logical formal system (in which interpretations of natural language expressions are to be carried out), this theory aims at becoming a logical semantics for natural languages. Some minimal formal logical systems with this purpose in mind have been constructed, which in this paper we shall not describe5. Rather, we shall here formulate a new formal logical system that incorporates a complete and consistent set of axioms and rules for sortal quantifiers, sortal identity and (second order) universal quantifiers over sortal concepts, something that have not been contemplated before6. The formal system is the product of adding this set to another set of axioms and rules (concerning concepts in general) that have been previously characterized together with axioms connecting both sets. We shall display the entire formal system in the third section; in the present and next section, we shall continue depicting the most important philosophical features of conceptual realism. We have stated above that, apart from predicable concepts, conceptual realism posits other kinds of concepts that are supposed to underlie our use 5 See for example, the basic formal system in Cocchiarella (1986, chapter II). 6 For a completeness and consistency proof for this set of axioms and rules see the author’s “A Complete and Consistent Formal System for Sortals”.

122

Max A. Freund

of linguistic expression (other than predicates). Among them, we want to call attention first to sortal concepts. These are intersubjectively realizable cognitive capacities whose use in thought and communication are associated with certain identity criteria, i.e., criteria by which we are able to distinguish, count and classify objects. Expressions such as cow, horse and insect are paradigmatic examples of linguist terms standing for sortal concepts. We shall call these linguistic expressions sortal terms. So, as the reader might have guessed, conceptual realism separates itself from the standard view of grouping, in the same logical category, linguist expressions supplying criteria enabling us to distinguish and count objects that have not been previously individuated with linguistic expressions in which the criteria presuppose that the objects have been already individuated. For example, according to the present view, adjectives and common nouns belong, in general, to different logical categories. Incidentally, proper names and definite descriptions are interpreted, in conceptual realism, as special cases of sortal terms. That is, they are considered to stand for sortal concepts (providing criteria for identifying at most one object). Sortal terms might form part of more complex linguistic constructions as in some cow, every horse and any insect. These constructions are called sortal quantificational phrases or (for short) sortal quantifiers and they stand for referential concepts, that is, cognitive capacities that allow us to refer to one or more individual of a certain sort. In this connection, we should notice that proper names and definite description, when used in a context of a sentence, are viewed as carrying a referential content and so they should be translated to the formal system as sortal quantifiers. For example, within the context of the sentence Pegasus is a white horse, the proper name Pegasus is a proper name which might stand for a referential concept in which it is assumed that the unique object identifiable by the identity criteria provided by the sortal concept associated to Pegasus exists. In another interpretation, it can be interpreted as a referential concept in which no existential content is presumed. Referential concepts will not always involve sortal concepts tough, as in those represented by the quantifier phrases anything, some thing and every individual. These phrases stand for referential concepts, even though “thing” and “individual” do not represent sortal concepts; rather, they stand for concepts not involving any identity criteria. On the other hand, among referential concepts, a distinction can be drawn between simple and complex. Quantifier phrases involving complex sortal names, as in the sentence “Some snakes that are poisonous are being extinguished”, are clear cases of expressions standing for complex referential concepts. In that sentence, it is the quantifier phrase “Some snakes that are poisonous” that stands for a complex

Conceptual Realism and Interpretation

123

referential concept, which is constructed on the basis of the complex sortal concept represented by “snake that is poisonous”. We should also notice that quantifier phrases does not always stand for concepts and this is when they appear as part of complex predicate expressions. In these contexts, any of such phrases represents the intensional content of the referential concept it would otherwise stand for. For example, in the statement John looks for some entertainment when he goes to the movies, the component sentence John looks for some entertainment contains the quantifier phrase some entertainment as part of the predicate to look for some entertainment. This phrase, according to conceptual realism, does not stand for a referential concept but rather for an intensional object. In other words, depending on the context of use, the quantifier phrase some entertainment might stand either for a concept or for the intensional content of this concept.7 The twofold roles of quantifier phrases offers the evidence for a distinction between active and deactivated referential concepts. To understand this double role, we should consider speech acts of assertions. Referential and predicable concepts relate themselves in assertions. In this context, the former are complementary to the latter in the sense that when exercised in a speech act of assertion, a referential concept is the element informing that act with a referential nature (that is, referential concepts are what accounts for the aboutness or directness to objects in an assertion) and predicable concepts are responsible for the predicable nature of the act. It is in this role that we say that the referential concept is active or actively exercised. However, when a quantifier phrase occurs in the context of a complex predicate expression, it does not longer stand for an entity having that referential role, because the concept the complex predicate expression stands for is informing the speech act only with a predicable nature. In this situation, one says that the referential concept the quantifier phrase represents has been deactivated. This means that the predicable concept is formed on the basis of the intensional content of the referential concept represented by the complex predicate expression containing the quantifier phrase. So far we have listed some of the different concepts posited by conceptual realism, which form the semantic basis for our use of several linguistic expressions. We want to close the present section by giving a brief outline, from the perspective of conceptual realism, of how formation of these and any other concept is supposed to evolve. It is one of the main tenets, of the present philosophical theory, that concepts are constructed in a process of different stages in which conceptual 7 For details on this topic, see Cocchiarella (1998, 169-202).

124

Max A. Freund

structures at a later stage are in general not explicitly definable or reducible to those at the earlier they presuppose (as conceptually prior bases for their construction). Moreover, at any given stage of conceptual development, the earlier stages are not thrown out but rather retained as still important and useful parts of the overall conceptual framework. There are two major forms, not necessarily incompatible, of how the different phases of cognitive development can be conducted; they are associated to two variants of conceptual realism: holistic and constructive conceptualism. The former subscribes to the widest form of concept formation. It assumes that a phase can be reached at which every predicate expression of a given language would stand for a concept and so it allows for impredicative concept-construction, i.e., for the formation of concepts whose content presupposes a totality to which they belong. Constructive conceptualism constitutes the other major view of how concept formation evolves. Unlike holistic conceptualism, this philosophical variant maintains that the only concepts, which can be formed, are the predicative ones, i.e., those concepts fulfilling Poincare-Russell vicious circle principle. Notwithstanding that this phase of concept formation is very restrictive, holistic conceptualism somehow assumes it, since it presupposes construction of predicative concepts as the basis leading to the process of impredicative concept formation. There is indeed a cognitive mechanism that takes us from the predicative to the impredicative stages of concept development. In constrast, there is a strict version of constructive conceptualism that rejects impredicative concept formation as being cognitively impossible.

2. Conceptual Realism: Concept Formation and Logical Notions We have seen that referential concepts are formed in a gradual process of concept construction (Holistic or otherwise). Other logical notions are also formed in the same process and at different stages. We should first mention the concept of relative identity or identity relative to a sortal such as “a is the same horse as b” (in symbols, a =S b, where S stands for the sortal concept horse); this notion should be formed after sortal concepts have been constructed. Another important logical notion is (second order) quantification over sortal concepts. This notion is considered to follow the construction of referential concepts, since we can only learn to refer to sorts after we have learned to refer to individuals of different sorts.

Conceptual Realism and Interpretation

125

Second order quantification over sortals together with relative identity constitute the cognitive ground for the construction of another logical concept: absolute identity, i.e., the notion of an object a being simpliciter the same as another object b (in symbols, a = b). The view is that absolute identity is initially constructed as the notion of identity with respect to some sortal or other. This means, in formal terms, that absolute identity is implicity understood, at an intermediate stage of concept development, as (x = y) = def (∃sA )(x = A y) where the symbol ∃sA stands for (second order) existential quantification over sortal concepts that might, in principle, be constructed. Similar to absolute identity, absolute quantification ( i.e., the concept expressed in words such as “everything”, “every individual” and “every object”) requires antecedent formation of sortal quantification and second order sortal quantification. That concept is formed as the notion of reference to all objects of any sort (i.e. reference to all objects that might fall under sortals which could in principle be formed), that is, at an intermediate stage absolute quantification is implicitly understood as (∀x)ϕ = def (∀sA )(∀xA)ϕ where the symbols ∀xAϕ stands for sortal quantification with respect to the sortal A (and should be read as “for every S”) and the symbol ∀sA stands for universal (second order) quantification over sortal concepts which might, in principle, be constructed and so should be read as “for every sortal concept”. Reference in this latter expression should be understood implicitly counterfactual. In the conceptualist philosophical framework, it is left open whether there is a level of concept formation at which the notion of absolute quantification as reference to all objects in general (i.e., reference to all objects irrespective of whether or not the objects fall or might fall under a sortal concept) has been constructed. In other words, the possibility that there is a more mature stage of conceptual development at which we come to refer in an absolute way that is independent of sortals (i.e., to refer to things that might never be of any sort at all) is not precluded, even though the notion presupposes the philosophically problematic view that things might exist for which there will never be identity criteria. The same applies to the concept of absolute identity between objects in general. For this reason, we shall introduce absolute

126

Max A. Freund

quantification and identity, as primitive symbols, in the formal system characterized in the next section. We have not mentioned yet the operations corresponding to propositional connectives. They too are viewed as part of the gradual process of concept formation. Moreover, some of them are considered to be basic and the others derivative. Now, concerning these and other basic logical notions (such as relative identiy and sortal quantification), the question might naturally arises of whether or not they are constructed on the basis of another cognitive capacities. This is a question for which there is not a definite answer in conceptual realism, we should say. The present philosophical view can accommodate different perspectives regarding the nature of those basic concepts, ranging from total innatism to an approach in terms of innate capacities that will be awaken by the individual in its interaction with nature and that will yield in this process, sooner or later, the primitive notions. There are many other logical notions that have not been alluded to here (for example, quantifiers such as many and few), but our main objective was to show how logical concepts might depend on their formation on a prior development of another logical concepts and how they fit within the entire framework of conceptual realism. Also, we were interested in introducing, in an intuitive way, some of the logical concepts we shall employ in the formal system we shall characterize in the next section. However, we should say that conceptual realism offers a theoretical framework that accounts also for those logical notions.

3. Conceptual Realism: the Minimal Formal System In the two previous section, we have offered a brief description of important features of the philosophical framework of conceptual realism. We shall now represent these features in a second order logic having sortal quantifiers, sortal identity and, universal quantifiers over sortals, in its logical syntax, and in which predicate expressions can be nominalized and treated as abstract singular terms. The system is minimal in the sense that it has principles consistent with the different variants of conceptual realism. Indeed, the two variants mentioned above (Holistic and constructive conceptualism) can be represented by adding to the system certain comprehension schemata 8. We proceed now to characterize the system. 8 For example, holistic conceptualism might be represented in the formal system by adding the schema (∃F)([λ x1…xn ϕ] = F ), where no restrictions are attached to ϕ, and

Conceptual Realism and Interpretation

127

We presuppose a countable set of sortal term constants and predicate constants and assume the availability of denumerable many individual variables, sortal term variables and n-place predicate variables (for each natural number n). We shall use x, y, z and w, with or without numerical subscripts, to refer in the metalanguage to individual variables, Fn, Gn and Rn to refer to nplace predicate variables or constants, and A as well as B to represent sortal terms variables or constants. We shall usually drop the superscript when the context makes clear the degree of a predicate variable or when otherwise does not matter what degree it is. For convenience, we shall also use u in order to refer to individual or predicate variables in general. As primitive logical constants we take →, =, -, ∀, λ and /. We shall intuitively interpret the first four symbols, as the material implication, identity, classical negation and the universal quantifier, respectively. The lambda operator λ will be employed to construct complex predicate expressions and / to construct complex sortal names (with a role similar to that one used in English to form relative clauses). We recursively define simultaneously the set of individual terms, sortal terms, predicate expressions and formulas, as follows: (1) Every individual variable and n-place predicate variable or constant is an individual term. Every sortal variable and constant is a sortal term. (2) Every n-place predicate variable or constant is an n-place predicate expression. (3) If a, b are individual or sortal terms, then (a = b) is a formula. If A a sortal term, then (a = A b) is a formula. (4) If π is a predicate expression and a 1,..., a n are individual terms, then π( a 1,..., a n) is a formula. (5) If δ is a formula, -δ is a formula. (6) If δ, ϕ are formulas, then (δ → ϕ) is a formula. (7) If A is a sortal term and ϕ is a formula, then A/ϕ is a sortal term. (8) If δ is a formula, x is an individual variable and A is a sortal term, F is a predicate variable, then ∀xδ, ∀xAδ, ∀Fδ and ∀SAδ are formulas. (9) If δ is a formula and x1,..., x n are pairwise distinct individual variables, then [ λx1, …, x n δ] is a predicate expression. (9) Every predicate expression is an individual term. constructive conceptualism by adding a restricted form of this schema in which it is required that ϕ should not have as sub-expressions any second order quantifier (over concepts in general or sortal concepts). On the different possible schemata, see, for example, M. Freund (1996, 483-505) and Cocchiarella (1986).

128

Max A. Freund

We shall call the set of predicate expressions, individual terms, sortal terms and formulas the set of meaningful expressions. We shall use δ, σ, γ and α to refer to any member of this set. We shall use a, t and b, with or without numerical subscripts, to refer to individual terms in general. The concepts of a bound and free occurrence of a (predicate, sortal or individual) variable are understood as usual. An occurrence of a term b in a wff or term δ is said to be a bound occurrence of b in δ if some occurrence of a variable in b is a free occurrence of that variable in b but a bound occurrence of that variable in δ. If a and b are terms, then by ϕa/b we shall mean the wff or term that results by replacing each free occurrence of b in ϕ by a free occurrence of a (ϕ is wff or a term), if such a wff or term exists, and otherwise we take ϕa/b to be just ϕ itself. We shall say that a is free for b in ϕ, if ϕa/b is not ϕ unless a is b. The universal quantifier ∀ when applied to predicate variables will be intuitively interpreted as the universal quantifier whose range is the set of all concepts, i.e., ∀F should be read as “for every concept F” and when applied to individual variables should be read as “for every individual”. On the other hand, ∀xA should be read as “for every A” and ∀sA as “for every sortal concept” Existential quantifiers are defined in terms of universal quantifiers as usual. Whenever a formula of the form B/ϕ is found, it should be read as “B (who, which) that is (are) ϕ”, where B is either a simple or a complex sortal term. We proceed now to formulate the formal system. Axioms: A0. All tautologies A1. (a =S b) → (a = S a) A2. a = a A3. (∀u)(ϕ → γ) → ((∀u)ϕ → (∀u)γ) A4. (∀xA)(ϕ → γ) → ((∀xA)ϕ → (∀xA)γ)) A5. ϕ → (∀u)ϕ, provided u does not occur free in ϕ A6. ϕ → (∀xA)ϕ, provided x does not occur free in ϕ A7. a =S b → (ϕ → ϕ *), where ϕ* comes from ϕ by replacing any free occurrence of a by a free occurrence of b A8. a = b → (ϕ → ϕ*), where ϕ* comes from ϕ by replacing any free occurrence of a by a free occurrence of b A9. (∀xA)(∃yA)(x =S y) A10. (∀F)(∃G)(F = G) A11. (∀x)(∃y)(x = y) A12. (a =A a) → ∃yA(y = A a)

Conceptual Realism and Interpretation

129

A13. (∀xA/ϕ)γ ↔ (∀xA)(ϕ → γ) A14. λx1,..., xnRx1,..., x n] = R A15 . [λx1,..., xn δ]( a1,..., an) ↔ (∃x1)... (∃xn )( a1 = x1 &... & an= xn & δ), provided x1,..., xn do not occur free in a 1,..., a n A16. [λx1,..., xnϕ] = [λy1,..., y n ϕ y1 / x1..., y n / xn], provided y 1,..., y n are free for x 1,..., x n , respectively, in ϕ, and y 1,…,yn do not occur free in ϕ A17. (∀SA)(ϕ → γ) → ((∀SA)ϕ → (∀SA)γ)) A18. ϕ → (∀SA)ϕ, provided A does not occur free in ϕ A19. (∀x)ϕ → (∀xA)ϕ A20. (∀F)ϕ → (∀SA)ϕ Rules: Modus Ponens (MP): from ϕ → γ and ϕ infer γ Universal Generalization (UG): from γ infer ∀uγ UsGI: from γ infer (∀xA)γ UsGC: from γ infer (∀sB)γ As the reader might have noticed, except for the common noun object or thing (represented in the syntax by the absolute quantifier), the minimal formal system does not allow for a formal representation of non-sortal common nouns, that is, common nouns standing for concepts on which referential concepts could be based but which do not provide identity criteria. However, a system having conceptual realism as its philosophical background that allows for a formal representation of those nouns has been recently formulated9 and the minimal formal system can be enlarged to accommodate this other system. The other strategy would be to introduce into the logical syntax of the system, as needed, quantifiers representing referential concepts involving those concepts represented by non-sortal common nouns. We should note that the logical analysis of conceptual realism is in a process of development. Apart from other determiners (such as many and few), the formal system might need to be supplied with additional logical categories and a logic of events and state of affairs to obtain a logical theory complete enough for the interpretation of any natural language within the framework of conceptual realism10.

9 For a logic of common nouns, see Cocchiarella (2002). 10 For these points, see Cocchiarella (1997, 175-179).

130

Max A. Freund

4. Meaning and Cognitive Structure The view that the truth-conditions of a statement should be deemed part of its meaning is confirmed by the thought experiment of imaging a possible world in which we knew all of its facts and, at the same time, we were unable to decide the truth-value of the statement. In this case, anybody would say that we do not know or understand the meaning of such a statement11. The above view is welcome by the framework of conceptual realism. The fact of this acceptance can be elucidated by the circumstance that conceptual realism is an heir of a paradigmatic logical theory of meaning: Montague’s intensional semantics. This paradigm, as is well known, constitutes an alternative to Donald Davidson’s extensional approach towards a theory of meaning as a theory of truth-conditions. Montague’s intensional semantics involves elements (pragmatic ones, represented in the semantics as contextual indices) not involved in the Davidsonian perspective. Clearly, knowledge of meaning implies knowledge of truth-conditions. It has been claimed, however, that the converse of this implication does not necessarily holds. A theory of meaning must go beyond a theory of truthconditions. Among those following this line of criticism, we should specially mention philosophers coming from the Austin-Searle line of research. They have demanded construction of a theory of speech acts as supplement to a theory of truth-conditions in a semantics for a natural language. This is because a theory of truth cannot account for illocutionary acts 12. Conceptual realism agrees with speech-act theorists that the story concerning meaning does not reduce to a report of its truth-conditions and so that a semantics for natural language does not amount to a theory of truth for a language only. It concedes the importance of illocutionary acts in the theory of meaning but it adds to this discussion a further point, not contemplated by speech act theorists: the distinction between the conceptual structure of speech acts of assertions and their truth-conditions (or propositional content), that is, the form of the referential and predicable concepts involved in an assertion, on one hand, and the state of affairs that would turn the assertion true and those that would turn it false, on the other hand. This distinction can be illustrated more clearly by considering particular translations to the language of the minimal formal system. Consider any assertion of the sentence All Swans are white. It can be translated to the formal language as (∀xS)[λxWx](x), where W and S stand for the 11

This thought experiment was devised by W. G. Lycan (2000, 153), based on an interpretation of remarks made by D. Lewis (1970). 12 See Searle, J. and D. Vanderwerken (1985, chapt. 1).

Conceptual Realism and Interpretation

131

concepts of being white and of being a Swan, respectively. According to the minimal formal system, the bi-conditional (∀xS) [λxWx] (x) ↔ (∀x)([λx(∃yS)(x = y) → W(x))](x) is a logical truth13. This means then that “(∀x)[λx(∃yS)(x = y) → W(x))](x)” and “(∀xS)W(x)” have the same truth-conditions. On the other hand, the left formula makes reference to all S and predicates of them that they are W while the formula on the right makes reference to all objects in general and predicates of them that they are W, if they are an S. In other words, the formula on the left represents an assertion whose referential and predicable nature is informed, respectively by referential and predicable concepts of a form different from those in the formula on the right side. Consequently, the formula on the left side of the bi-conditional represents a cognitive structure different from the formula on its right side. Another example: consider possible assertions of the sentences Some reptiles are poisonous and small , Some reptiles that are poisonous are small, Some reptiles that are small are poisonous and There is a reptile that is poisonous and small. Those speech acts when translated to the formal language will reveal respectively the following cognitive structures: 1. 2. 3. 4.

(∃xR)[λy(P(y) & S(y))](x) (∃xR/P(x))S(y) (∃xR/S(x))P(y) (∃xR)/P(x) & S(x))[λy(∃z)(y = z)](x)

These formulas have the same truth conditions as shown by the following theorems14: 1*. 2*. 3*. 4*.

(∃xR)[λy(P(y) & S(y))](x) ↔ (∃xR)(Py & Sy)) (∃xR/P(x))S(y) ↔ (∃xR)(Py & Sy)) (∃xR/S(x))P(y) ↔ (∃xR)(Py & Sy)) (∃xR)/P(x) & S(x))[λy( z)(y = z)](x) ↔ (∃xR)(Py & Sy))

However, the form of the referential and predicable concepts involved, in each case, are different and this means that they have a different cognitive 13 Its proof includes use of propositional logic, A7, A15, U.G., A5, A19 and A9. 14 The proof for 1* involves propositional logic, USGI ,A2, A4, A8, A15 ; for 2* and 3*, propositional logic, definition of existential quantifiers and A13 ; and for 4*, propositional logic, A2, A8, A15, U.G. and USGI

132

Max A. Freund

structure. For instance, the referential concept in 2 and 1 are, respectively, the concepts that (∃xR/P(x)) and (∃xR) stand for and they have different forms. The first concept is formed on the basis of two concepts (being a reptile and being poisonous), but the second one is constructed on the basis of just one concept (being a reptile). The above examples reveal that it is possible for different assertions to have the same propositional content or truth-conditions but different cognitive structures. The question is whether a difference in the cognitive structure implies a difference in meaning; whether an assertion of Some reptiles are poisonous and small would have a meaning different from an assertion of Some reptiles that are poisonous are small, notwithstanding the fact that they have the same propositional content. We think that an affirmative answer should be given to these questions. In other words, even if we were clear of the propositional content of those assertions, we think that a determination of the form of the referential and predicable concepts involved will be needed to decide whether or not they have the same meaning. The cognitive structure of an assertion is a cognitive form in which several concepts concur to make possible construction of its propositional content. Consequently, different cognitive structures of assertions having the same propositional content establish different cognitive conditions under which that content can be construed. For instance, the cognitive structure of the sentence Some reptiles are poisonous and small, represented in the formula 1 above, shows that by the exercise and mutual interaction of a cognitive capacity to refer to objects identifiable by a sortal concept (reptile) and a cognitive capacity to classify individuals by means of two other concepts (poisonous and small) the propositional content of an assertion can be formed. The cognitive structure of Some reptiles that are poisonous are small, in which the form of the referential and predicable concept involved is different, shows another way of constructing a propositional concept identical to the first sentence. How the speaker, having certain cognitive capacities, can exercise and relate them in order to construct the propositional content of an assertion is information conveyed by an assertion. This information is what the cognitive structure of the assertion furnishes and it is objective. The objectivity consist in their inter-subjectivity: any speaker having the same capacities can construct the same proposition by exercising them in the way demanded by the cognitive structure. Indeed, one of the aims of conceptual realism is the construction of a formal theory encompassing all formal conditions for or logical forms which make possible construction of the propositional content of any assertion. So, the formal theory would display the different possible

Conceptual Realism and Interpretation

133

cognitive structures of assertions. Clearly, empirical factors will determine the referential and predicable concepts a speaker might possess and natural languages the linguistic devices s/he can appeal to in order to express the cognitive structures involving those concepts. Discovery of these syntactic devices in a particular natural language is the task of the empirical linguist. Conceptual realism, however, has not established whether or not there are cognitive structures, postulated by the formal theory of conceptual realism, that might find no expression in a particular natural language. In other words, conceptual realism has not ruled out the possibility that a natural language might exist that does not posses linguistic devices to express some cognitive structures. Whoever can specify the propositional content of an assertion (that is, under which state of affairs the assertion is true or false) but cannot determine how the individuals involved in the propositional content are being referred and how these individuals are being classified or related would not qualify for someone who has fully understand what is conveyed by the assertion. Each one of these referential and predicable concepts involved has a specific form (apart from their content), which should have been grasped when those concepts were intuited in the assertion. But those forms are what the cognitive structure of the assertion transmit us. So grasping the cognitive structure is part of grasping the meaning of the assertion.

5. Radical Interpretation and Cognitive Structures On the basis of above considerations, we want to conclude that an imperative for the interpreter of a natural language is to develop, apart from a theory of truth-conditions and a theory of speech acts for the native language, a theory that would lay open the cognitive structure of speech acts of assertions performed by their speakers. How the interpreter, in particular, the radical interpreter would have to proceed? An adequate extension of the minimal formal system of conceptual realism together with its formal language would function as a guide-book (so to speak) for a program of activities to be followed by the radical interpreter. After all the interpretation of the native tongue will be constructed in the formal system. So, following this lead, one of the obvious first steps would be to discern whether there are linguistic expressions in the native tongue which can be interpreted in the formal language as basic sortal terms, basic predicate expressions and relative identity. S/he will continue discerning whether

134

Max A. Freund

there are linguistic expressions that can be interpreted in the formal language as sortal quantifiers and second order quantifiers over sortal concepts, and so on. What would be the evidence to which the radical interpreter should appeal, following the guidelines of the logical theory? For a radical interpreter trying to build an interpretation along lines suggested by Donald Davidson, he has specified15 the type of linguistic evidence that must be taken into account and the order of how that evidence must considered. As a first step, he recommends taking as evidence classes of sentences always held to be true or always held to be false and patterns of inference. Next come sentences held sometimes true and sometimes false according to discoverable changes in the world. The remaining sentences, those on which there is no uniform agreement, or whose held truth value does not depend systematically on changes in the environment would be the last step. Whether the Davidsonian strategy would actually work in the field or not is something to be tested; yet for the moment let us suppose it might yield the required results for a radical interpreter trying to establish the truth-conditions of the possible utterances in the native tongue. Clearly, this strategy assumes as basic just one speech act (the act of assertion)16 and does not seem to appeal to the other ones. Obviously, if a theory of speech acts were being looked for, the strategy would be insufficient. Another linguistic data will clearly have to be considered. In this case the conceptualist radical interpreter would not have to restrict the data to assertions and s/he can find an important source for constructing the theory of the cognitive structure for the native tongue in the initial construction of a theory of speech acts. This is because speech acts, in general, contain a propositional content and a linguistic device to express it. It is in these devices that the interpreter would find evidence for constructing his/her theory of cognitive structures17. Although the conceptualist radical interpreter might make use 15 See Davidson (1984, 125-40). 16 We should qualify this point. Davidson actually speaks of the attitude of holding a sentence true, which is a mental state that does not necessarily have to be overtly expressed in a speech act. That attitude when overtly expressed is just a speech act of assertion, which is the only thing of the propositional attitude a radical interpreter can have direct access to . That speech acts of assertion are involved is more clearly seen in what Quine calls radical translation. Here he speaks of acts of assenting and dissenting as the evidence available to a radical translator. 17 In so called explicit performative sentences (i.e., linguistic expressions in which a performative verb is used in the first person present tense of the indicative mood with a statement as complement clause as in “I promise that I will come tomorrow”) the interpreter finds a mean to connect cognitive structures with speech acts in general.

Conceptual Realism and Interpretation

135

of any speech act, for the purpose of the present discussion we shall restrict ourselves to assertions only. How is the radical interpreter going to assign cognitive structures to assertions? Imagine, for example, an interpreter of Spanish in the epistemic situation of radical interpreter. Let us say that he has been able to construct a theory (following the Davidsonian approach or maybe one of the other two methods described by David Lewis for radical interpretation18) that can correctly assign the truth-conditions to every assertion of this language. The theory, then, would assign the same truth-conditions to any utterances of the following sentences: U1). U2) U3) U4)

Algunos reptiles son pequeños y venenosos Algunos reptiles que son pequeños son venenosos Existen reptiles pequeños y venenosos Algunos reptiles pequeños son venenosos.

The correct theory of truth-conditions devised by the radical interpreter would establish that all these utterances are true when and only when some reptiles are small and poisonous. In other words, their propositional content is the same as of the English sentence “some reptiles are small and poisonous”. The question now is, for the radical interpreter, whether or not they have the same cognitive structure. How s/he is to find out? We have shown, in the previous section, that for an assertion having that propositional content there might be, at least, four possible cognitive structures. Which of these possible cognitive structures will be the cognitive structures of the assertions U1-2? If none, which other cognitive structures there are that might be used to construct the propositional content and corresponding to those assertions? These are all questions that a conceptualist radical interpreter must answer. In the construction of a theory of truth-conditions a la Davidson, certain states of mind (beliefs) are assumed and then on their basis propositional contents of sentences are being search for (once it has been identified when a sentence is held true). This inquiry proceeds assuming a certain charity principle19, which can be spelled out as the presumption that, to a large extent, most of the informants beliefs form a consistent set composed mostly of beliefs which the radical interpreter think are true. In a wider approach (Lewis’) the assumption is more general: given the experience to which the informants have been subjected certain beliefs are to be expected (even though the radical interpreter might consider them to be false). This latter possibility is due to learning mechanisms and values, which are postulated to be in the informants cognitive apparatus.

136

Max A. Freund

The situation for the radical interpreter seems to be the inverse of the Davidsonian scenario: we are certainly in possession of the propositional content of assertions but we have to look, on this basis, for certain states of mind that have determined their propositional content: the cognitive structures. We should recall that the conceptualist radical interpreter approaches interpretation with a formal theory giving him prior knowledge of the possible cognitive forms a language user should adjust to, that is, with possible perspicuous representations of the possible cognitive structure of speech acts of assertion. For this no access, directly or indirectly, to the speakers’ linguistic intuitions is needed. But this clearly is not enough. What then needs to be supplied? The radical interpreter can certainly have access to the different syntactic devices accompanying assertions that have been determined, by the theory of truth-conditions, to have the same propositional content. Some of those syntactic devices, however, might not be related to the cognitive structure of assertions. How is the conceptualist radical interpreter must look for the devices that are, in fact, linked or associated with the use of particular cognitive structures? For this purpose, we think he should appeal to an analogue of a principle of charity: assume that the informant will assign to assertions (under certain circumstances) the same cognitive structures that the radical interpreter (or speakers of another languages) would assign to assertions having the same propositional content, under the same or similar circumstances20. In other words, part of the initial evidence for construction of a theory of cognitive structures would come from the interpreter’s intuitions and/or the data he (or other linguists) has previously gathered from users of other languages, regarding the circumstances under which a cognitive structure is being associated to an assertion. But this is just to say that the radical interpreter should attempt to develop a theory of context, that is, a theory capable of predicting the contextual situations in which any speaker would use a certain cognitive structure. The interpreter must go beyond the truthmakers in Nature to discover under which contexts speakers of his own language (or any other languages to which access to the linguistic intuitions of the speakers exist) employ different cognitive structures and project the results to speakers of any language. Once developed, s/he will have to treat the theory of contexts as any empirical theory and test it. Whether this theory is possible is an empirical question to which we do not have a ready answer now. Clearly, given a theory of truth-conditions for the language, the theory of contexts would provide the necessary hints to connect syntactic devices to cognitive structures.

Conceptual Realism and Interpretation

137

Bibliography Chomsky, N. 2002. “On the logic of classes as many. “ Studia Logica 70, no. 3. —. 2000. New Horizons in the Study of Language and Mind. Cambridge: Cambridge U.P. —. 1998. “Reference in Conceptual Realism.” Synthese 114. —. 1997. “Conceptual Realism as a Theory of Logical Form.” Revue Internationale de Philosophie 2. —. 1986. Logical Investigations of Predication Theory and the Problem of Universals. Naples: Bibliopolis. Davidson, D. 1984. “Radical Interpretation.” In Truth and Interpretation. Oxford: Clarendon Press. Fodor, Jerry and Ernest Lepore. 1992. Holism, A Shopper's Guide. Cambridge, Mass.: Blackwell. Forster, J.A. 1976. “Meaning and Truth Theory.” Truth and Meaning, edited by G. Evans and J. McDowell, Oxford: Oxford U.P. Freund, M. A. 2000. “A Complete and Consistent Formal System for Sortals.” Studia Logica 65, no.3: 367-81. —. 1996. “Semantics for Two Logical Systems: RRC and Cocchiarella’s RRC.” SNotre Dame Journal of Formal Logic, Vol. 37, No. 3. Lewis, D. 1983. “Radical Interpretation.” In Philosophical Papers, Vol. 1. Oxford: Oxford U.P. —. 1970. “General Semantics.” Synthese 22. Lycan, W. G. 2000. Philosophy of Language. London: Routledge. Searle, J. and D. Vanderwerken. 1985. Foundations of Illocutionary Logic. Cambridge: Cambridge U.P.

138

Corey Washington

Content Partialism and Davidson’s Dilemma Corey Washington

Abstract Hartry Field, Jerry Fodor and others differ with Donald Davidson over the question of how a theory of content should be structured. Field and Fodor maintain that a theory should begin by following the compositional structure of a sentence in reducing the semantic properties of complex expressions to the semantic properties of their simplest parts and complete the job by reducing the semantic properties of the parts to non-semantic ones. Davidson describes this approach as the ‘Building-Block method’ and maintains that it cannot possibly succeed. He holds that a theory of content should ‘give up reference’ by treating the semantic properties of basic expressions as a purely technical devices with no direct relation to non-semantic phenomenon. In this essay, I examine what I call “Davidson’s Dilemma”, the conflict between the apparent soundness the arguments for the view that a theory must treat reference as a point where linguistic and non-linguistic reality meet and the equally apparent soundness his argument that reference cannot possibly play this role. I propose a resolution to the dilemma that grants the validity of Davidson’s arguments against the building-block theories but is, I believe, more palatable to mainstream semanticists than Davidson’s solution. This solution, which I call ‘content partialism’ treats the reference of terms as regularities across propositional contents. I show how content partialism is consistent with a Kripkean theory of reference fixing, the touchstone of those who advocate building-block theories.

There is no reason why a semantical theory should assign information content independently to each expression in a symbol system. It will do if contents are assigned only to the atomic expressions, the semantics for molecular symbols being built up recursively by the sorts of techniques that are familiar from the construction of truth definitions. In what follows, I will in fact assume that the problem of naturalizing representation reduces to the problem of naturalizing it for atomic symbols. Jerry Fodor (“A Theory of Content I”) If we could give the desired analysis or reduction of the concept of reference, then all would, I suppose, be clear sailing. Having explained directly the semantic features of proper names and simple predicates, we could go on to explain the reference of complex sin-

Content Partialism and Davidson’s Dilemma

139

gular terms and complex predicates, we could characterize satisfaction (as a derivative concept), and finally truth. This picture of how to do semantics is (aside from details) an only one and a natural one. It is often called the Building-Block theory. It has often been tried. And it is hopeless. Donald Davidson, “Reality without Reference” [219-220]

Content Partialism and Davidson’s Dilemma Many philosophers subscribe to three assumptions about the nature of content: that linguistic expressions have content, e.g. that names refer to objects, predicates express properties, sentences express propositions, etc.; that the semantic relations connecting expressions to their contents supervene on more basic relations; and that the relationship between semantic relations and their founding relations is comprehensible. Together, these assumptions give rise to the problem of how to explain semantic relations in more basic terms. This problem, a linguistic version of “Brentano’s problem”, has been at the heart of the philosophy of language for much of its history. While believing that the linguistic version of Brentano’s problem is real, Donald Davidson raises important questions about the viability of conventional methods of solving the problem. In his article “Reality without Reference” (1984, 215-225), he brings these doubts together as a dilemma for what he calls ‘building-block’ approaches. The dilemma is this: on the one hand, it seems that the compositional nature of language forces us to take a particular route to explaining sentence content: the building-block method. One must reduce the semantic relation between a sentence and its content to the relations between the sentence’s constituent words and their contents and then explain the relations between the words, taken individually, and their contents in non-semantic terms. On the other hand, there are compelling reasons to think that the second half of this explanation the account of the contents of words taken in isolation – cannot be accomplished. This conflict is the essence of Davidson’s Dilemma. I believe there is much to learn from Davidson’s Dilemma and the effort to resolve it and, moreover, that that the dilemma has not received the kind of attention it deserves. There are, I believe at least, three reasons it has been overlooked in the literature. The first is Davidson’s failure to provide a serious argument for the horns of the dilemma. Second, Davidson formulates the dilemma exclusively in extensional terms and only with respect to natural language, whereas most contemporary discussions of content such as those

140

Corey Washington

carried out by Fred Dretske and Jerry Fodor concern mental representation and assume that representations have intensional entities as their contents. The third problem lies with Davidson’s resolution. To resolve the dilemma one must give up one of its horns. One must either show that there is another way to explain content aside from the building-block method or show that it is possible to reduce isolated instances of reference directly to non-semantic counterparts. Davidson takes the first approach, a route he describes as ‘giving up on reference’. Since many writers believe that reference is real and the point at which semantic and non-semantic reality meet, they find this resolution unpalatable and may be inclined to reject the reality of the dilemma that apparently motivates it. My goal in this paper is to try to give Davidson’s Dilemma some welldeserved attention by setting out an argument for it, showing how far it generalizes, and proposing a resolution that saves reference. I maintain that the dilemma extends to explanations of all kinds of content, extensional and intensional, possessed all language-like systems of representation, be they mental or spoken. Like Davidson, I reject the first horn of the dilemma. I argue that while an explanation of content must follow compositionality, the building-block method is not the only compositional option. I advance a compositional approach to explaining content that respects the motivation for the building-block method but eschews its implausible conception of word content. I argue that it is possible to explain reference in non-semantic terms but that it is not possible to explain the reference of a word in isolation from propositional units of which that word may be a part. I call the conception of content on which this resolution rests ‘content partialism’. The rejection of the building block approach has significant implications for work in the philosophy of language and mind. Davidson wrote “Reality without Reference” in reply to Hartry Field’s arguments in “Tarski’s Theory of Truth” (1972), which advances what may be the classic modern statement of the building-block theory. If the argument against the building-block approach is sound, Field’s proposal, and others, like those in Devitt (1981) and (1984), that follow his lead, must be mistaken. Field (1978) extends the building block approach to mental representation, and Fodor’s remarks at the top of this article are a faithful rendition of this extension. If the buildingblock approach does not work, the theories of content in Dretske (1981) and Fodor (1990) that are modeled on Field’s architecture are also doomed.

Content Partialism and Davidson’s Dilemma

141

Davidson’s Dilemma In this section, I give an exposition of Davidson’s dilemma and set out an argument for its reality. I then indicate the scope over which it extends and explain how far I think it can be generalized. Davidson’s presents the dilemma as arising for certain types of theories of meaning for natural language, theories that seek to explain semantic properties of/ relations involving linguistic entities in terms of more basic properties/ relations, e.g. semantic properties of more basic entities or more basic non-semantic properties of linguistic entities (Davidson 1984, 219). In what follows I focus on theories that seek to explain semantic relations between expressions and their contents, ‘theories of content’, in the classical sense. The theories Davidson thinks are problematic follow the building-block approach, according to which, one can explain semantic relations in the same way one might explain the physical properties of any complex object. To explain such properties, one would divide the object into its immediate parts and then explain the properties of the whole in terms of the physical properties of the parts and their physical relations to one another. One would then repeat the process for the immediate parts until one reaches the basic parts, the building blocks. Proceeding analogously in the case of language, one first explains the propositional content of complex sentences in terms of their immediate constituent structure and the semantic contents of their immediate parts. One then continues, explaining the properties of successively smaller parts, until one reaches semantically basic expressions, the semantic building blocks. While an example of a building-block theory might help in grasping Davidson’s point, he does not offer concrete cases, but contents himself to say only that such examples can be found in the British empiricists (220). Luckily, however, there is a clear example of a building-block theory lurking in the shadows of “Reality without Reference”. As noted, Davidson wrote the paper, in part, in reply to Hartry Field’s arguments in “Tarski’s Theory of Truth.” That article takes its start from Tarski’s claim to have brought the concept of truth into line with physicalism by reducing it to physicalistically respectable concepts, using the apparatus of his truth definitions (Tarski 1990: 408). Field takes Tarski to task, arguing that theories of truth based on truth definitions fail to reduce truth to nonsemantic concepts because they do not give reductions of the semantic properties of basic expressions. To succeed in the reductive task Tarski sets for himself, Field thinks that Tarskian theories must be supplemented with

142

Corey Washington

theories of reference, theories of property expression and theories for other basic expressions.1 Field believes that an adequate explanation of truth conditions might take the form of the conjunction of the theories in (1) and (2). The clauses in (1) are standard recursive and predicative clauses from Tarskian theories of truth. Those in (2) are Kripke-style accounts of reference for proper names and predicates: 1a. jP and Qk is true iff P is true and Q is true. 1b. jX is Pk is true if and only if the referent of ‘X’ and satisfies ‘P’. 2a. A’s utterance of jYk refers to Z if Z is an individual and there exists a causal chain stretching from a baptism of Z as jYk to X’s utterance of jYk. 2b. A’s utterance of jQk expresses P if P is a property and there exists a causal chain stretching from a characterization of an object (or group of objects) possessing P as jQk to A’s utterance of jQk. Let us use the term ‘theory of combination’ for a set clauses, like those in (1), that assigns properties to complexes on the basis of the properties of their parts and ‘basic theories’ for accounts, like those in (2), of the contents of simple substantive expressions. It is important to underscore that basic theories explain the semantic properties of words, considered in isolation from larger units of which they may be a part. Such theories are ‘atomistic’ in a common sense of the term.2 Field’s idea is that a theory of combination will reduce the concept of truth to the relations between names and objects and predicates and properties. Basic theories will then reduce these relations to non-semantic ones in an atomistic fashion. A building-block theory is a collective made from a set of basic theories and a theory of com1 Scott Soames argues that even with these supplementations, Tarskian theories will still not provide the desired explanations. He maintains that a thorough explanation of the semantic properties of sentences must also contain explanations of logical notions like universal quantification and predication (See Soames 1984, 411-429). I believe that Soames is entirely correct but because his point does not affect the argument here, I set it aside. 2 Kripke’s theory is sometimes described as a ‘causal’ theory and sometimes as a ‘chain of communication’ theory. Field sees the theory under the former guise, for it is in this light that it seems plausible that the theory could further the project of reducing reference to purely physical relations. If the concepts of baptism and ‘chain of communication’ can be made precise and spelled out in physical terms, Field would be right. I believe, however, that there is a significant difference between these two conceptions of the theory. As we will see below, viewing communication as essential to the theory’s goals is key to understanding how my account helps to resolve Davidson’s dilemma.

Content Partialism and Davidson’s Dilemma

143

bination. The basic theories are the building blocks, the atoms, while the theory of combination is the cement, which binds the building blocks together. Though Davidson and Field focus on the explanation of material truth conditions, Field’s own discussion shows that the building-block paradigm also encompasses theories of propositional content. For if names refer to objects and predicates express properties, as Field’s basic theories presuppose, sentences will possess structured propositions as contents. While a theory that purports to explain the relation between sentences and such contents will contain a theory of combination different than (1), it will still qualify as a building-block theory. To the extent that the dilemma is real, it covers theories of propositional content. The principle of compositionality is the main premise in the argument for the building block theory. The principle maintains that the content of a complex grammatical sentence is a function of the contents of its parts, with the identity of the function depending on the structure of the sentence. The second step remarks on an analogy between semantic and physical composition. It appears that sentences divide into their parts semantically in the same way that complex physical objects divide into their parts physically. That is, the properties of words appear to be independent of larger units in the same way that physical objects, like individual building blocks, are independent of the larger constructs they may be used to build. Just as a building block could possess its physical properties even if there were no buildings, the buildingblock approach assumes that a word may possess its semantic properties even if there were no things that express propositions. Moreover, if an expression can possess semantic properties independently of wholes of which it is a potential part, then it should be possible to explain those properties independently of those wholes. It should, in other words, be possible to carry out a building-block explanation whose coda is an account of the reference of words in isolation. Let us turn to the other horn of the dilemma: the claim that an atomistic account of reference is impossible. Davidson uses the term ‘reference’ as a catchall term for the semantic relationships between sub-sentential expressions and their contents. So, when he claims that explanations of reference inevitably fail, he means to condemn basic theories of name content and basic theories of predicate content. He believes building-block theories are doomed because their entire second tier is not viable. Though it is clear Davidson thinks the weak point of building-block theories is in the blocks, he is not very precise about why he thinks these explanations will fail. What hints he does give come in historical illustrations of what

144

Corey Washington

he takes to be the problem. Speaking of building-block theories, he writes (1984, 220, emphasis original): The ambitious attempts at behaviouristic analyses of meaning by Ogden and Richards and Charles Morris are not clear cases, for these authors tend to blur the distinction between words and sentences (‘Fire!’, ‘Slab!’, ‘Block!’) and much of what they said really applies intelligibly only to sentences as the basic atoms for analysis. Quine, in chapter II of Word and Object, attempts a behaviouristic analysis, but although his most famous example (‘Gavagai!’) is a single word, it is explicitly treated as a sentence. Grice, if I understand his project, wants to explain linguistic meaning ultimately by appeal to non-linguistic intentions – but again it is the meaning of sentences, not words, that are to be analysed in terms of something else. Theories of the kind developed by Ogden, Richards and Morris see the meaning of expressions like ‘fire’, ‘slab’ and ‘block’ as resting on their utterance conditions. A theory of this kind might claim that it is sufficient for an utterance of ‘fire’ to express the property fire (or refer to the natural kind fire) for it to be uttered when and only when something is on fire in the area. To understand what Davidson is saying about these theories it is important to note that, in this context, he is using ‘sentence’ to identify a semantic rather than a syntactic category. An expression token qualifies as a sentence if it is used to express a proposition or perform a speech act with propositional content. So, when he asserts that what the authors say “applies intelligibly only to sentences”, he means that the utterance conditions they postulate are possessed only by things that express propositions. The only utterances that we make when and only when there is a fire present are utterances that mean there is a fire (or something along those lines). Similar utterances with ‘slab’ and ‘block’ would also express propositions. If utterance conditions are a guide to content, then, quite unintentionally, Ogden and co. end up analyzing one-word sentences (‘fire!’, ‘slab!’, ‘block!’) rather than predicates (‘fire’, ‘slab’, ‘block’). More sophisticated writers, like Quine and Grice, Davidson suggests, recognize the futility of going after sub-sentential expressions in isolation. Instead, these writers make propositional units the focus of analysis from the beginning. Thus, Quine’s term ‘Gavagai’ is a one-word sentence meaning something like there is a rabbit. Grice’s goal is to explain what it is for a speaker to mean something with the use of a term, and the meanings he has in mind are propositional. So, the utterances he handles are Davidsonian sentences.

Content Partialism and Davidson’s Dilemma

145

Davidson thinks there is a moral in his historical sketch and this is that accounts of reference will always fail because the conditions they postulate for an expression to refer to an entity or express a property are, if they are sufficient for anything, always sufficient for that expression to have a proposition as its content. Making the notion of an utterance and proposition explicit, we can rewrite this as the first premise of Davidson’s argument for the second horn: Premise 1: If the conditions offered by an explanation of content are sufficient for a token to have content, they are sufficient for it to have propositional content. Why is this a problem? It is a problem because a theory of meaning must account for the contents of the full range of grammatical sentences (a syntactic category) that make up a language, and a building-block theory can do this only if it contains basic theories of reference that explain the contents of names and predicates. That is, only if the basic theories do this can a building block theory built from them exploit compositionality to give an account of the full range of sentences. As Davidson writes (1984, 216): For well-known reasons … a theory can’t begin by explaining truth for a finite number of simple sentences and then assign truth to the rest on the basis of the simples. It is necessary … to analyse sentences into constituent elements – predicates, names, connectives, quantifiers, functors – and to show how the truth value of each sentence derives from features of the elements and the composition of the elements in the sentence. The only kinds of sentences that can be built up from simple sentences are those with the structure of a propositional calculus. The ‘well-known reasons’ to which Davidson alludes have to do with the fact that, with only a few exceptions, such as time and weather reports, the structure of natural language sentences always exceeds that of a propositional calculus. Sentences like ‘John has three slabs’ are grammatical and meaningful, and their meaning can only be explained by a theory that contains components that tell us what it is for a predicate to express the property slab and a name to refer to John. This is Davidson’s second premise. Premise 11: To account for the contents of arbitrary sentences of a language, a theory must contain constituent theories that explain the contents of names and predicates.

146

Corey Washington

Given that building-block theories try to explain word content in isolation, Premise I implies that they will assign propositional content to simple expressions. This fact and premise II imply that Inference: Building-block theories cannot explain the contents of arbitrary grammatical sentences of a language. A theory of combination must work with the material the basic theories provide for it. If the latter only handle one-word sentences, then the former can only build up complexes whose simple components are themselves sentences. For example, if one-word sentences of the Ogden and Richards variety are at the base, a theory will only be able to explain the contents of sentences like ‘Slab!’, ‘Slab! and Block!’, ‘fire!’, ‘fire! or slab! or block!’, etc. Clearly such a theory will miss sentences with the subjectpredicate structures that make up the vast majority of natural language sentences. The next premise specifies a standard of adequacy for theories of meaning. As Davidson writes: “whatever else it embraces, a theory of meaning must include an account of truth – a statement of the conditions under which an arbitrary sentence in the language is true.” The important part of this quote is its reference to the range of sentences. An adequate complete building-block theory must be able to account for the contents of arbitrary sentences of a language. Now, Davidson would be the first to admit that this condition sets the bar rather high and that, in practice, actual theories will fall short of this goal. Being reasonable, he would not condemn a theory as inadequate that was on the way to accumulating the full range of sentences even if it had yet to reach the end. A nuanced adequacy condition that sees most theories as works in progress should recognize as adequate partially complete theories. To allow for these cases I suggest that we take the following premise as Davidson’s condition of adequacy: Adequacy Condition: A theory of meaning is adequate if and only if it can explain, or be extended to explain, the contents of arbitrary sentences of the language. The second horn follows directly. Conclusion: Building-block theories are inadequate as theories of meaning. The dilemma consists in the fact that building block theories appear, on the

Content Partialism and Davidson’s Dilemma

147

one hand, to be the only option as explannations of content and, on the other hand, to be totally unworkable as explanations.

Reference and Propositional Content I believe that the argument against the building-block theory is sound and that the argument for it is not. In this section, I defend this critique, but, for the sake of brevity, not each step. I limit myself instead to a defense of the first, and most controversial, premise: that any effort to explain reference in isolation from propositional content will fail by virtue of endowing the target expression with propositional content. Davidson’s discussion of his historical examples is brief and, perhaps, not fully convincing. So before giving an argument for the premise one, I would like to offer another example in support of its truth. The theory is a Skinnerian account of property expression discussed by Fodor (1990, 53-57). As Fodor presents the account, it claims that a word expresses a property if utterances of the word are caused by instantiations of the property. So, the theory holds that DOG expresses the property dog if utterances of DOG are, by law, caused by dogs. Fodor thinks that there are a number of problems with this theory, but he never questions the nature of the contents it assigns. He assumes these to be properties. However, since this theory purports to explain the semantic properties of an isolated word in non-semantic terms, Davidson’s first premise predicts that the contents it supports will be propositional. If a causal relation between dogs and utterances of DOG is sufficient for DOG to have content, it should be sufficient for the expression to have propositional content. If the premise is true, DOG will not be a predicate but a one-word sentence. Evidence that this prediction is born out can be found in some of Fodor’s own comments. Fodor maintains that the theory is inadequate, in part, because it is unable to account for misrepresentation. The contents the theory assigns are uniformly true and misrepresentation occurs only when an utterance expresses a false proposition (Fodor 1990: 59). Since only propositions can be true or false, these remarks are comprehensible only if tokens treated by the theory express propositions, e.g. only if tokens of DOG express contents like that is a dog. We get further support for the prediction if we eschew behavioristic assumptions, as Fodor does, and go inside the speaker’s head for a look at the chain of psychological states that mediate between events external to the speaker and her utterances. In other writings, Fodor notes that the psycho-

148

Corey Washington

logical states involved in law-like relations have propositional contents. For example, if a person has her eyes open and is properly oriented with nothing blocking her view, then the presence of a brown object two feet in front of her face will, ceteris parabis, cause her to perceive the object as brown. She will have a perception with the content that x is brown, where x is the object. No analogous laws connect appearances of objects to perceptual elements that refer to the objects or express properties without expressing a proposition. Similarly, laws that connect perceptions to beliefs relate states with propositional content. If a series of perceptions of an object as brown, four-legged and slobbering are stable and coherent, and one has background beliefs that suggest the perceptual situation is normal, then a belief may be caused with the content that the perceived individual is a dog. These perceptions, background beliefs and newly formed beliefs, all have propositional content. Finally, any utterance caused by a series of propositional states in such a way that it expresses one or more of those states will itself be propositional. As Fodor writes (55): “Verbal behavior is ‘cognitively penetrable’, as one says these days: whether one utters ‘DOG’ in the presence of a dog depends on things like whether one thinks one’s audience would be interested to hear that there’s a dog about, whether one is desirous of telling them what one thinks they would be interested to hear, and so forth.” With these remarks Fodor is granting that, if appearances of dogs cause utterances of DOG, then they do so by means of intervening psychological states with propositional content. The desire Fodor is describing is the desire to inform one’s audience of the truth of the proposition that there is a dog present. This desire is normally realized by making an utterance with a content like there is a dog present. Since precisely this type of chain between appearance event, perception, belief and utterance will occur in any normal case in which the presence of a dog causes a person to say DOG, these considerations make it reasonable to infer that utterances of DOG in this situation have propositional content in line with the first premise. If this example is convincing, then we can ask why things turn out this way. Given that Davidson does not actually state premise one, we should not be surprised to find that he does not directly address the issue of why it is true. Still, he makes comments that are relevant: “words have no function save as they play a role in sentences: their semantic features are abstracted from the semantic features of sentences, just as the semantic features of sentences are abstracted from their part in helping people achieve goals and realize intentions” (220). A key question in understanding this claim is knowing what Davidson means by ‘sentence’. Is he using the term in the syntactic or semantic sense? If he is using it in the semantic sense, then I believe that the claim

Content Partialism and Davidson’s Dilemma

149

is entirely correct. Content of all kinds is intrinsically connected to the notion of representation and representations are, by nature, propositional. It is possible to overlook this connection when theorizing abstractly, but it makes itself felt in our intuitions about the content in hypothesized situations. If the conditions are sufficient for an expression to have content, we reflexively wrap it in propositional content. If intuitions are an infallible guide to content, then any attempt to explain reference in isolation will fail because intuitions will always insert a proposition to fill out the representational situation.

Content Partialism Davidson’s solution to the dilemma starts with his belief that it is impossible to explain reference in non-semantic terms. If this is true, then one cannot relate the propositional content of sentences to non-semantic reality through a reductive compositional explanation by means of the parts of sentences. Davidson proposes that propositions are themselves the level of content at which semantics meets non-semantic facts and that the non-semantic facts it meets are simply evidence for the interpretation of an expression as having a particular proposition as its content. Because propositions meet evidence as wholes, Davidson describes his view as ‘holistic’ (1984, 221). Compositionality enters the explanation, not in servicing the semantic/non-semantic interface, but as a purely semantic-internal principle that connects the contents of simple expressions to the contents of grammatical sentences. No further downward explanation of the contents of simple expressions is possible or necessary, in Davidson’s eyes. I believe that this solution goes astray in its very first step. The argument against reference does not show that it is not possible to explain reference in non-semantic terms, but only that it is not possible to give a non-semantic explanation a referring expression in complete isolation from propositional units of which it is a part. Nothing in the argument rules out explanations of word content when words occur in the context of larger expressions with propositional content such as grammatical sentences or propositional utterances in which parts of the proposition are unarticulated. The type of explanation I have in mind for reference is not a conventional contextual explanation because it does not seek to account for a word’s reference by treating, as a single unit, whole grammatical sentences of which it is a part. Rather the account seeks to provide conditions for an expression to refer to an object when that expression occurs in a context with propositional content. The proposal is not original. In fact, I think it is the proposal that

150

Corey Washington

Kripke intended to present in Naming and Necessity, though it differs from what Field and others have taken away from that work. The key to this reading is Kripke’s characterization of reference as being fixed by a ‘chain of communication’. We communicate, in the sense in which Kripke uses the term, only by means of propositional utterances. Utterances that serve as links in the chain express propositions.3 Thus, it is an essential part of the account I envisage that an utterance of a word refers to an object only when it is part of a communicative act and only when it connects to the object through a series of propositional utterances containing tokens of the word. The following diagram illustrates such a network of connections:

John gave Bill a cookie

John went to McLain High

Meet my friend Bill

There are three cookies on the tray

Ed is John’s brother

Bill is 10 years old

Would you like a cookie?

. . Let’s call him ‘john’

. . I baptize you ‘Bill’

. . Why don’t we call it a ‘cookie’?

The original utterance and links in each of the three chains consist of propositional utterances. Links are tied together by referring uses of names and predicates, which are regularities across the utterances. One can state the theory of reference as follows: 3a. A’s utterance of jYk refers to Z if jYk is uttered as part of an utterance with propositional content, Z is an individual, and there exists a chain of communicative utterances containing referring uses of jYk stretching from a baptism of Z as jYk to X’s utterance of jYk. 3 I am allowing that the utterance of a single word or a sentence fragment in response to a question qualifies as propositional, e.g. “yes” in response to “Is that you john?” or “Bill” in response to “Who are the cookie?”

Content Partialism and Davidson’s Dilemma

151

3b. A’s utterance of jQk expresses P if jQk is uttered as part of an utterance with propositional content, P is a property, and there exists a chain of communicative utterances containing referring uses of jQk stretching from a characterization of an object (or group of objects) possessing P as jQk to A’s utterance of jQk. This take on Kripke differs from Field’s in two salient respects. First, it does not purport to explain reference in purely non-semantic terms. The intensional concept of a proposition occurs ineliminably in the explanans. Second, it does not treat the referring utterance in isolation nor does it see the links in the chain as involving isolated uses of a referring term. Reference occurs and plays a functional role only as part of a propositional act. Both of these differences stem from conceiving of the theory as a ‘chain of communication’ account rather than as a ‘causal’ theory. This account brings with it a concept of word content that I call partialism. Each causal chain linking a word token to an object gives a partial characterization of the content of the utterance of which the name is a part. Above, the chain running through tokens of ‘john’ determines the identity of the agent of the proposition, while the chain running through tokens of ‘bill’ determines the identity of the direct object and the chain running through tokens of ‘cookie’ determines the nature of the indirect object. Each chain contributes a part of the sentence’s content and together they determine the whole content. How does this account resolve the dilemma? It gives compositionality a place by showing how the content of a complex sentence can be a function of the contents of its parts. It does this by reducing the contents of complex sentences not to the contents of isolated words, but to the contents of words in contexts. They can be represented by elements I call ‘lexical schema’, which take the form of a term surrounded by an indeterminate context that fills out a proposition, i.e. [----X----]p. In the example above, the content of (4a) is reduced to the contents in (4b)-(4d): 3a. John gave Bill a cookie expresses the proposition >. 3b. [----john----]p refers to John. 3c. [----bill----]p refers to Bill. 3d. [----a cookie----]p expresses the property cookie. At the same time, the contextual nature of schema amounts to an acknowledgement of the intrinsic connection between reference and propositional

152

Corey Washington

content. A word is an abstraction from a propositional context in the way that some metaphysicians see properties as abstractions from objects. In this way, I believe that partialistic accounts wed the desirable features of holism, the taking of propositions as basic in a certain sense, with the desirable features of atomism, the taking of words as basic in a certain sense, while avoiding their undesirable features.

Bibliography Davidson, D. 1984. Inquiries into Truth and Interpretation. New York: Clarendon Press. Devitt, D. 1981. Designation. New York: Columbia University Press. —. 1984. Realism and Truth. Princeton, NJ: Princeton University Press. Dretske, F. 1981. Knowledge and the Flow of Information. Cambridge, Ma: MIT Press. Field, H. 1972. “Tarski’s Theory of Truth.”Journal of Philosophy 69, no. 13 (July): 347-375. —. 1978. “Mental Representation.” Erkenntnis, 13 (January): 9-61. Fodor, J. 1990. A Theory of Content and Other Essays. Cambridge, MA: MIT Press. Kripke, S. 1980. Naming and Necessity (2nd ed.). Cambridge, MA: Harvard University Press. Soames, S. 1984. “What is a Theory of Truth?”Journal of Philosophy 81, no. 8 (August): 411-429. Tarski, A. 1990. Logic, Semantics and Metamathematics, J. H Woodger (trans.), John Corcoran (2nd edition ed. and intro.). Indianapolis IN: Hackett Publishing.

Vagueness, Indirect Speech Reports, and the World

153

Vagueness, Indirect Speech Reports, and the World Steven Gross

Abstract Can all truths be stated in precise language? Not if true indirect speech reports of assertions entered using vague language must themselves use vague language. Sententialism – the view that an indirect speech report is true if and only if the report’s complement clause “same-says” the sentence the original speaker uttered – provides two ways of resisting this claim: first, by allowing that precise language can “same-say” vague language; second, by implying that expressions occurring in an indirect speech report’s complement clause are not used. I reject the first line of resistance, but argue that the second is successful if one accepts sententialism.

I. Introduction Vagueness is a pervasive feature of natural language. This is not surprising, given our sensory and cognitive endowments and limitations: vague expressions are easier to learn and apply than precise replacements would be, and bestow a beneficial flexibility. So, perhaps we shouldn’t expect or desire to eliminate vagueness from natural language generally. Still, it might be thought that at least in principle all truths can be stated in precise language. Why might one think this? One reason might be an antecedent commitment to the supervenience of the vague on the precise.1 But does it follow generally, from the supervenience of truths stated using X-expressions on truths stated using Y-language, that the former could be stated using only Ylanguage? It would be very controversial to claim on the basis merely of the relevant supervenience claim, for example, that all truths stated using nonphysical vocabulary can be stated using physical vocabulary. Indeed, supervenience relations are of interest precisely because they promise naturalistic acceptability without reduction. It has been argued that this promise is not fulfilled, that supervenience of the relevant variety entails reduction.2 But the 1 Compare Timothy Williamson (1994: 108): “… vague properties supervene on precise ones. A tempting conclusion to draw, but not a necessary one, is that the world can be completely described without resort to vague language.” 2 See Jaegwon Kim (1989: 31-47).

154

Steven Gross

argument remains controversial, in part owing to its reliance on infinite disjunction. Moreover, even if it succeeds, the reduction that supervenience is supposed to yield arguably involves entities individuated more coarsely than “truths”. And, even if it is “truths” that have been reduced, it doesn’t yet follow that the reduced truths can be stated in the base vocabulary: even if properties or propositions may be infinitely disjoined, sentences we can state must be finite. All this might be beside the point: even if the move from supervenience to statability in the base vocabulary doesn’t hold generally, someone might try shouldering the burden of showing that there are auxiliary premises, underwritten by features particular to the case, that do allow one to infer that all truths can be stated in precise language from the supervenience of the vague on the precise. Some variety of realist convictions might provide a second route to the claim that all truths can be stated in precise language. Very crudely, one might hold that the world just is all the truths, or what makes all the truths true; and that (up to inter-translatability) there exists a unique true and complete description of the world as it is independent of our conception of it in virtue of a “correspondence” between the description and the world. Suppose one also holds that it is incoherent, or at least false, that the world is or could be itself vague; and that vagueness is a kind of semantic indeterminacy, not a matter of in-principle ignorance as epistemicism would have it.3 Then the unique true and complete description, if it is to correspond with the world, must also be free of vagueness. Assuming the unique true and complete description is something we can state (whether or not we could know it to be true),4 it would follow that all truths can be stated in precise language.5 This might be given a scientific twist. Here’s one way (again, crudely): suppose one holds that the only truths there are – perhaps, the only objective truths – are scientific truths. But science aims to carve nature at the joints, to describe things as they are independently of our conceptions of them. So, there is an intention that the expressions used to describe these kinds of things have their content fixed by the nature of that which they aim to describe. But it is incoherent or at least false that those things or their natures are themselves vague. So, the content they fix isn’t either. So, insofar as we 3 See Williamson: 185-215. 4 It’s thus assumed that there’s no in-principle obstacle to our minimally grasping and thus expressing whatever concepts are necessary to formulate the description. 5 Vagueness is pressed as a problem for such realists by, for example, Hilary Putnam (1983: 297-314).

Vagueness, Indirect Speech Reports, and the World

155

succeed in stating truths at all, those truths can be stated using precise language. Whatever reasons one may have, whether some more developed version of some portion of the above or some other reasons, they must be wrong if vague language is required to describe the very real phenomenon of vagueness itself. That vague language is required to describe the phenomenon of vagueness is a theme of Timothy Williamson’s Vagueness. Williamson constructs a persuasive case, in particular, for the claim that vague language is required to state the properties of vague expressions. He also, however, suggests that vague language must be used to report what speakers assert using vague language. This note critically examines this more specific suggestion.6 Even if one agrees with Williamson’s conclusion that not all truths can be stated precisely, one wants to know, so far as possible, precisely what truths yield the ineliminability of vagueness.

II. The Argument Williamson writes: “… since people use vague language, vague language is needed to report what they have said. … a well-chosen ‘that’-clause reproduces what was said …” He doesn’t explicitly endorse the suggested argument, nor does he pretend to have developed it in detail. But he does say that it presents a “serious difficulty for the idea that all truths can be stated in precise language.” 7 The suggested argument would take something like the following form. Natural language users enter assertions using vague expressions. In addition to this general truth, there are all the specific truths reporting such assertions – that is, indirect speech reports of the canonical form ‘S said that P’.8 In order for such reports to be true, they must accurately report what was asserted. This places a constraint on the expressions appearing in a report’s complement clause. What is this constraint? Williamson – perhaps not speaking in his own voice, but rather in the voice of an imagined promulgator of the argument – requires that the complement clause in some sense reproduce what was said. This doesn’t mean that the actual words uttered (in6 Both the suggested argument and the critical discussion could be generalized to other attitude ascriptions. 7 Williamson (93). 8 Read quotation marks as corner-quotes where necessary. I follow others in using the pasttense, since examples can sometimes sound funny in the present-tense. But I’ll ignore tense when glossing truth-conditions.

156

Steven Gross

cluding the vague ones) must be reproduced, but presumably that their content – or relevant aspects thereof – must be reproduced, perhaps using other words. If among the relevant aspects to be reproduced is the vagueness of some vague expression actually uttered, and if this can be reproduced only by employing an expression that is also vague,9 then the complement clause itself, and thus the report as a whole, will need to employ vague language. It would thus follow that there are truths – specifically, true indirect speech reports of assertions entered using vague language – that can only be stated using vague language.10 Clearly, the crucial step in the argument is the assumption that there is a condition on the truth of indirect speech reports from which it follows that indirect speech reports of assertions entered using vague expressions must themselves use vague expressions. In what follows, I examine how this claim fares in the context of one account of indirect speech reports: sententialism.11

III. Indirect Speech Reports: Sententialism The basic idea of sententialism is to take the truth of an indirect speech report ‘S said that P’ to depend on the relation in which S stands to the sentence ‘P.’ In what relation must S stand to ‘P’ in order for the report to be 9 One might require more weakly that it is sometimes the case that the relevant aspect of the used vague expression can be reproduced only by using a vague expression in the complement clause. See below. 10 This argument does not require that there be any vague non-embedded sentence (or utterance thereof ) that itself expresses a truth. It’s perfectly consistent with the premises and conclusion of the suggested argument that all vague non-embedded sentences express falsehoods, or have some other truth-status other than truth, or even fail to express anything truth-evaluable at all. (Perhaps then we would speak of making as if to assert.) So, the argument would be available even to a “nihilist” about non-embedded vague sentences – at least if this nihilist’s reasons for maintaining her position didn’t force her to be nihilistic as well about indirect speech reports containing vague expressions. 11 Versions of sententialism have been endorsed by various philosophers, but contemporary sententialists trace their views in particular to Donald Davidson (1968-9: 130-46). My discussing the suggested argument in light of sententialism is not meant to constitute an endorsement of sententialism. Insofar as sententialism has any plausibility, however, and sententialism doesn’t support the suggested argument’s crucial premise, then that premise will have been shown to be controversial. Correlatively, if sententialism can be rejected, much of my discussion will be drained of interest. (For replies to some standard objections to sententialism, see Ernest Lepore and Barry Loewer (1989: 338-356.) Of course, the rejection of sententialism wouldn’t establish the crucial premise. For example, a version of the “interpreted logical form” approach to attitude ascriptions might also be used to challenge the crucial premise. I briefly comment on this below.

Vagueness, Indirect Speech Reports, and the World

157

true? It can’t just be that S in fact uttered that sentence, or even uttered that sentence assertorically, because we can accurately report in one language what someone said in another. Rather, it’s required that S uttered some sentence that is relevantly similar to P. As it’s often put: P must “same-say” the sentence S actually uttered. There is much room for refinement here: for instance, we might want to understand same-saying as a relation between sentence tokens rather than types.12 But nothing in my discussion will turn on such matters of detail. For our purposes, we can work with the following statement of the truth-conditions for indirect speech reports: ‘S said that P’ is true iff there’s a sentence s such that S assertorically uttered s and s samesays ‘P.’ But what constitutes same-saying? Surely that’s a detail we can’t brush aside? It’s important to note that for the statement of truth-conditions to be correct, it is not required that we be able to produce an analysis of samesaying. ‘Steven walks’ is true iff Steven walks: whether this approximates a correct assignment of truth-conditions by an empirically verified, compositional semantic theory for my idiolect seems quite independent of there being an analysis of walking. Still, in order to understand the statement of truth-conditions, one needs to understand sufficiently the words used. Whether or not an analysis of same-saying could be given, it is supposed that speakers’ grasp of same-saying is displayed in their successfully engaging in the practice of reporting speech and assessing such reports. That said, it would indeed be useful in the discussion to follow if there were a non-controversial criterion of same-saying ready-to-hand that one could wield in thinking through hard cases. Lacking one, I’ll explore what one can say in any event. Before proceeding, however, I should note one aspect of same-saying that reflection on the practice of reporting speech and assessing such reports can suggest: namely, that what counts as same-saying – what counts as being relevantly similar – can vary with conversational context.13 For example, on some occasions, the interests of the parties to the conversation may be such 12

13

This provides one way of accommodating the potential relevance of various non-semantic properties, such as phonological properties, pragmatic features relevant to the determination of the semantic values of context-sensitive expressions, etc. Davidson and Lepore and Loewer advert to utterances of sentences in stating the truth-conditions of indirect speech reports. Of course, not all aspects of Davidson’s own position must be retained for an approach to count as sententialist. In particular, one needn’t accept his assimilation of the complementizer ‘that’ to the demonstrative ‘that.’ See below. For expositional ease, I will sometimes allow myself to speak in this manner of what counts as same-saying contextually varying instead of speaking meta-linguistically – and perhaps more properly – of what satisfies ‘same-says’ as contextually varying.

158

Steven Gross

as to allow the reporter to utter a different proper name in the complement clause than that uttered by the original speaker, or even to replace the name with an appropriate definite descriptions – for instance, if what’s of interest to them is what the original speaker said of some person, as opposed to how she indicated that it was that person about whom she was speaking. On some other occasion, however, how the original speaker indicated the subject of predication might matter to those receiving the report. If this is so (in this case or in others), and if the statement of truth-conditions is not to be itself context-sensitive,14 then this dependence on context must be registered in the statement of truth-conditions. How this ought to be done – for example, whether a contextual parameter must be introduced into a syntactical representation of the report itself – is not pertinent to our concerns. But, in supplying the defender of the suggested argument a response to the first of two lines of resistance I’ll examine, I will assume that ‘same-says’ (and thus ‘says’) is context-sensitive.

IV. First Reply The main point of the first line of resistance is this. Suppose S assertorically utters a vague sentence, a sentence with at least one vague constituent. Samesaying requires only relevant similarity. The possibility is thus raised that similarity with respect to vagueness/precision is not among those aspects that must be preserved. So, could there be a precise sentence ‘P’ appropriately similar to the uttered vague sentence such that someone could accurately report what S said by uttering ‘S said that P’? For example, suppose Sally assertorically utters ‘Bob is heavy.’ And suppose I’m conversing with someone who knows Bob and knows that Bob is a bit under six feet tall and weighs about 240 pounds; and, what’s more, there is nothing about our conversational context that has shifted the relevant comparison class for men’s weights away from what one might expect the default to be. Could I accurately report Sally’s assertion by saying, say, that Sally said that Bob weighs over 220 pounds? If I could report what Sally said using a vague expression without myself employing a vague expression, and if it were the case generally that this was so, then the existence of truths concerning what was said 14 For the view that a statement of truth-conditions, even for a context-sensitive sentence, ought not itself be context-sensitive, see Davidson (1976: 37); and James Higginbotham (1988: 25). For the suggestion that one might allow truth-conditional semantic theories to be themselves context-sensitive, see Steven Gross (2001: 70-5) and Peter Ludlow (1999: 63).

Vagueness, Indirect Speech Reports, and the World

159

using vague expressions would not entail the existence of truths that cannot be expressed in precise language.15 I’m not sure that same-saying is sufficiently undemanding to allow such reports as ‘Sally said that Bob weighs over 220 pounds’ to pass as true. It’s indeed not to hard to imagine circumstances in which no one, having learned that the original speaker deployed vague language, would censure the reporter whose complement clause was precise. But the inappropriateness of censure doesn’t show that ‘S said that P’ is true: perhaps the report is false, but provides grounds for something that is true as well as conversationally salient, such as that S believes that P. Suppose, though, that we grant that at least in some cases one may truly report what was said using a vague expression via a report whose complement clause is precise – and thus that we reject the suggested argument’s crucial premise. The possibility would remain, however, that in some cases one could not. If the crucial premise is thus weakened – from requiring in all cases that accurate reports preserve vagueness to noting that in some cases they must – the suggested argument still stands. How might one support the weakened premise? To begin with, note that it’s unclear whether, for any assertion entered using a vague expression, there will always exist even a candidate same-saying complement devoid of vagueness. It was noted, after all, that vagueness is a pervasive phenomenon. The difficulty here is perhaps best seen by trying oneself to formulate precise replacements for a variety of cases. Consider, for example: John is reclining beside the red table. How might one report in precise terms an assertion entered via an utterance of this multiply vague sentence? What should replace the vague ‘table,’ for instance? In reply, one would like a general answer to such questions, or at least a general reason for thinking that an answer could be given in each particular case. Might the opponent of the (revised) suggested argument advert here to the supervenience of the vague on the precise? Perhaps a candidate precise complement clause can always be found in a base claim that would necessitate what was vaguely asserted.16 For all that’s been said, however, the determining base sentence might involve concepts rather dissimilar to those deployed by the original speaker. What if the supervenience base must be described in the language of physics?17 It might seem that the precise replace15

One might object that ‘weighs over 220 pounds’ (even adding: as used in that context) remains vague, at least to the extent that we could apply it. Let’s let this worry pass. I return in a moment to the difficulty of finding precise replacements. 16 If the vague assertion is false, one might take the modally “closest” base claim that would necessitate it. 17 Perhaps not the language of current physics, which is not obviously devoid of vagueness.

160

Steven Gross

ment for ‘John is reclining beside the red table’ would be too recondite for anyone reasonably to consider the report true. Mere supervenience doesn’t yet assure us that the candidates thereby produced will enable reports that same-say the vague sentence originally uttered, even if same-saying is relatively undemanding. A better reply might run as follows. One demonstrates that an expression is vague by describing a sorites series for it along some dimension. Once such a description is given, it is then always possible to introduce a sharp ersatz by specifying a complete sharpening of the expression along that dimension. So, for any expression whose vagueness we can in principle demonstrate, we can in principle introduce a sharp replacement for it. If such sharpened replacements are arguably sufficiently similar in content to the vague originals to satisfy the demands of same-saying, then we can always accurately report an assertion entered using vague language via a precise complement clause. There is a problem, however. What if there are vague expressions that are multi-dimensionally vague? Then it is not clear that we could stipulate a sharp cut-off for each dimension along which a sorites series could be generated. Obviously, no problem is posed by an expression that is vague along just a few dimensions. But consider ‘is a table.’ Its vagueness is often demonstrated by a sorites of “decomposition.”18 One imagines repeatedly removing in some set order a small bit of the matter that constitutes the table. But there are many such orders one could use, with no grounds to think that it would be reasonable to stipulate that one stop, no matter the order, after the same number of removals. Again, it has been suggested that some expressions – perhaps ‘nice’, ‘intelligent’ – are indefinitely multi-dimensionally vague.19 Suppose we were to waive this last worry and, further, grant that precise replacements generated in this fashion satisfy the demands of same-saying. Then it would seem that even the weakened crucial premise could be rejected. There is another revision to the suggested argument, however, that would nonetheless nullify the line of resistance we’ve been exploring. Resources remain to defenders of the suggested argument not willing to press these worries. It was noted that ‘same-says’ is often held to be context-sensitive, that what counts as a relevant similarity may vary from one conversational setting to another. If so, and if similarity as to vagueness/precision is among the features whose relevance can shift, then whether the precise replacements samesay the original vague sentences may likewise be context-sensitive. Suppose 18 See Peter Unger (1979: 236-242). 19 See, for example, Rosanna Keefe (2000: 12 and 95).

Vagueness, Indirect Speech Reports, and the World

161

then that the defender of the suggested argument grants – even if only for the sake of argument – that at least in some contexts the standards for same-saying20 are sufficiently undemanding that ‘Bob weighs over 220 pounds’ counts as relevantly similar to ‘Bob is heavy.’ Suppose she grants further that, for any assertion entered using vague language, there is at least one context such that by its standards for same-saying the assertion can be reported via a precise complement clause. Indeed, suppose she even grants that there exists a context in which any assertion can be reported without using vague language. Would it then follow that the truths about what people say can be reported in precise language? Well, that question deploys the context-sensitive expression ‘says’.21 So, our answer will depend on what ‘says’ happens to express as used there – that is, on what standards for same-saying are in play. In other words, insofar as ‘says’ is context-sensitive, so is the sentence ‘not all truths about what someone says using vague language can be stated using precise language.’ Now, suppose that, in addition to what was just granted, it’s also the case that there is a context sufficiently demanding with respect to same-saying such that in it no assertion entered using vague language could be accurately reported via a precise complement clause. This seems a reasonable supposition: it’s hard to see what could render illegitimate conversational contexts that, for one reason or another, rendered this a relevant similarity. But then we can introduce by stipulation a new term ‘same-says+’ that context-insensitively expresses the relation ‘same-says’ does in that context; and we can introduce ‘says+’ to express (context-insensitively) the relation in which someone stands to a sentence that same-says+ the sentence the person actually assertorically uttered.22 There would then be truths about what people 20 That is: the standards for the relation that ‘same-says’ expresses in those contexts. 21 I am assuming that the context-sensitivity of ‘relevantly similar’ and of ‘same-says’ in the statement and explanation of the truth-conditions of indirect speech reports arise from the context-sensitivity of ‘says.’ 22 It’s a consequence of an approach to vagueness that I have explored elsewhere that vague expressions are context-sensitive in such a manner that one cannot stipulatively introduce a context-insensitive expression that expresses what the vague expression expressed on some particular occasion of utterance. If ‘says’ is vague, and if this approach to vagueness is correct, then the move suggested above is not available. The approach to vagueness that has this consequence is hardly uncontroversial, however. In addition, I will mention below the possibility of replacing vague ‘says’ with a sharpened ersatz. One could then introduce ‘says+’ to express context-insensitively what this sharp ersatz expresses in particularly demanding contexts. (The controversial approach to vagueness is sketched in Steven Gross (2001: Chapter IV; and (2000). The consequence mentioned above is indicated in Gross (Forthcoming), and more fully discussed in unpublished work.

162

Steven Gross

say+ that cannot be stated using precise language – viz., what they say+ by using vague language. So, it seems that the suggested argument goes through after all, once it’s been properly modified: one must replace context-sensitive ‘says’ with a term that context-insensitively expresses what ‘says’ expresses in contexts in which “reproducing” vagueness matters for what counts in such contexts as same-saying. And so it seems that not all truths can be stated using precise language. Why the hedging “it seems”? Because there is a complication to note. If one holds that ‘says’ is in this way context-sensitive, one might find it difficult to deny that ‘states’ is likewise context-sensitive. But then so is the sentence ‘all true indirect speech reports can be stated using precise language’ (as well as, of course, ‘all truths can be stated using precise language’). If we are unclear what the contextually relevant standards are for having stated a particular truth, then we are unclear what claim the suggested argument is meant to deny. Once the suggested argument is formulated in terms of ‘says+,’ however, it seems charitable to assume that the standards for having stated a particular truth are likewise fairly demanding: perhaps we might even make this explicit by using ‘states+.’ With the argument and its conclusion so understood, this first line of resistance can thus itself be resisted.

V. Second Reply Sententialism also opens up a second line of resistance to the suggested argument. The suggested argument would have it that not all truths can be stated using precise language, since (at least some) truths concerning assertions entered using vague language themselves must be stated using vague language.23 But does the mere appearance of a vague expression in an indirect speech report’s embedded complement clause constitute use? It’s not clear that it does, according to sententialism. If it doesn’t, then the suggested argument fails to establish its conclusion. It is often suggested that sententialism in effect assimmilates indirect speech reports to direct speech reports: sententialism can be viewed as treating indirect speech reports as akin to a kind of quotation. As we saw, this assimilation isn’t complete, since it is required of indirect speech reports only that the embedded material be sufficiently similar to the sentence actually 23 The qualification “at least some” is inserted in light of the previous discussion. I’ll henceforth omit such qualifications, considering them understood.

Vagueness, Indirect Speech Reports, and the World

163

uttered, not that it be identical.24 But it might be suggested that the assimilation is sufficient to raise the question of whether in the matter of use indirect speech reports are on all fours with direct speech reports. Perhaps not all kinds of quotation exclude use, but arguably direct speech reports – S says “P” – paradigmatically do. What is needed to take this further is some criterion of use. Use is often contrasted with mention; so, it might be thought that an independent grip on mention might supply an account of use. Suppose an expression is mentioned if it is referred to.25 A used expression would then be one that occurs without being referred to. But an expression can be both used and mentioned, as in ‘Giorgione is so-called because of his size.’26 Use at best contrasts with mere mention, and there’s no reason to think an independent explication of that – one that doesn’t appeal to the expression’s not being used – would be any easier to obtain than one of use itself.27 If what’s required to exclude expressions from use is the explicit occurrence in surface structure of surrounding quotation marks, then vague expressions are indeed used in indirect reports. But, even putting aside cases of mixed quotation in which it appears that an expression inside quotes is both used and mentioned,28 this seems at best a superficial indication of use. Spoken, as opposed to written, language typically has no explicit “non-use” markers (unless one includes such accompanying gestures as scratching the air with two fingers). If one’s impressed by the assimilation sententialism suggests, then, absent some more fundamental understanding of what they are supposed to indicate, one will see whatever distinction explicit quotation marks mark as within the realm of the mentioned. The use/mention distinction is sometimes explicated by reference to the principle of semantic innocence. According to this principle, an expression retains its normal semantic properties no matter the linguistic context in 24 Actually, even direct speech reports arguably only require sufficient similarity, with context-sensitive standards for similarity: for example, in some contexts it might be required that phonological features be “reproduced,” in others not. So, it would perhaps be more accurate to say that the difference between direct and indirect speech reports consists in the kind of similarity required. 25 Perhaps an expression can be mentioned without being referred to – for example, by validating an existentially quantified claim. Perhaps an expression is mentioned if it’s what a sentence is “about.” 26 See W. v. O. Quine (1960: 153). 27 I say “at best” to allow for the possibility of occurrences that are neither used nor mentioned. 28 Consider Davidson’s well-known example: Quine says that quotation “has a certain anomalous feature.” See Davidson (1979: 27-40).

164

Steven Gross

which it occurs. This principle can seem to be violated, however, in some linguistic contexts, for example in quotational contexts. It can seem, for example, that in the sentence ‘‘cats’ has four letters,’ the term ‘cats’ lacks its normal semantic property of denoting cats. (If inter-substitutability salva veritate with co-denoting expressions is considered a normal semantic property, or something that follows from them, then one can point here to its violation relative to ‘domestic felines.’) One might then distinguish uses of an expression according to whether semantic innocence is maintained. A defender of the suggested argument who explicated use in this way would then need to show that embedded expressions in indirect speech reports retain their normal semantic properties. How would such a criterion be applied in the setting of sententialism? Suppose one sees the semantic properties of expressions as being given by a neo-Davidsonian truth-conditional semantic theory. Such theories consist of a finite number of axioms and production rules, from which can be derived theorems stating the truth-conditions for each sentence in the target language. A neo-Davidsonian about semantics needn’t be in particular a sententialist about the semantics of indirect speech reports; she needn’t think that the relevant theorems will assign truth-conditions like those specified above. But I am supposing that our sententialist is a neo-Davidsonian. It is not obvious how, in such a setting, one might argue that an occurrence of an expression violates semantic innocence by failing to retain its normal semantic properties. Let’s first put to one side the case of lexical ambiguity. Orthographically similar expressions can of course diverge in their contributions to semantic content – a phenomenon that could appear to yield cases of one term failing to retain its normal semantic properties in certain linguistic contexts. But if we can trace back the apparent divergence to distinct lexical axioms, then we have mere lexical ambiguity, not a violation of innocence. What if a lexical axiom assigns semantic properties to an expression as a function of the linguistic context in which it occurs? The expression is not lexically ambiguous, because there is but one axiom for it. But it seems that its semantic properties differ in different linguistic contexts. Do they? Suppose the expression is a noun whose axiom says that it refers to A, if the expression is immediately preceded by adjective X, and to B otherwise.29 Still, this semantic specification is true (supposing it is) no matter what pre29 Perhaps such an axiom distinguishing multiple cases could instead be replaced by multiple axioms, each utilizing a distinct valuation-function with appropriate restrictions for the invocation. See Richard Larson and Gabriel Segal (1995: 436).

Vagueness, Indirect Speech Reports, and the World

165

cedes the expression. In this sense, the expression retains its semantic properties wherever it occurs, and so innocence is upheld. It might be reasonably suggested, nonetheless, that it’s in the spirit of the principle of semantic innocence to consider this a violation – at least, in the supposed case, if either A or B could be designated the normal referent. Conceding this, however, has upshot for matters at hand only if the semantics of direct speech reports – our paradigm of non-use – violate innocence in this way. For, if they don’t, then either expressions embedded in the complement clauses of direct speech reports are used, or the semantics of direct speech reports must violate innocence in some other way. The former would provide sufficient reason to reject an explication of use in terms of innocence; the latter, given the sententialist’s proposed assimilation of indirect to direct speech, would direct our attention to this other way of violating innocence. So, does the semantics of direct speech reports violate innocence in this way? It’s by no means clear that it does. Theorems to the effect that ‘S said “P”’ is true iff S said ‘P’ can be generated without having to introduce lexical axioms in violation of innocence. Here’s one possibility. Suppose the quotation marks have no semantic function, but rather serve the pragmatic function of instituting particularly demanding standards for same-saying. Then nothing in the syntactical representation of the report that is fed into the semantics will correspond to the surface quotations marks: one will have a noun, a verb, and a sentence with the syntactic structure [S [N S] [VP [V said] [S P]]]. Suppose an ordered pair <x,y> satisfies ‘said’ iff x said y; and x satisfies a verb phrase consisting of a verb followed by a sentence iff, for some y, <x,y> satisfies the verb and y same-says ‘P.’30 This assignment of LF (the syntactic representation relevant to the assignment of truth-conditions), together with these axioms, generate the target theorems.31 Suppose, then, that we must either locate the semantics of direct speech reports’ violation of innocence elsewhere or abandon the attempt to explicate use via innocence. The only place left to look for a violation of innocence, however, is in the axioms for determining the semantic properties of 30 I adapt Larson and Segal (420). The adaptation consists in the remark about the pragmatic function of quotation marks. Otherwise, this is their rendering of a sententialist approach to indirect speech reports. They themselves reject such an approach, however, in favor of an “interpreted logical form” approach. See below. 31 Obviously, it would take much work to defend this proposal. Note, by the way, that Davidson’s own version of sententialism, according to which the quoted material doesn’t even appear in LF (similarly for the complement clause of indirect reports), also does not require the introduction of special lexical axioms. The defender of Davidson’s approach might be tempted to explicate use in terms of occurrence in LF.

166

Steven Gross

syntactically complex expressions from their constituents. But it is unclear how they could give rise to a violation of innocence, since these axioms don’t have the power to reassign semantic properties to the constituents. At best, they could assign differing semantic properties to the complex expressions themselves as a function of the larger linguistic contexts in which they could be embedded, just as we noted with lexical axioms. In any event, the sketch of a semantics for direct speech reports given above involves no such thing. We have thus failed to locate a way innocence would be violated that places direct speech reports properly on the non-use side. The discussion, however, suggests a better way of explicating use. Arguably, what distinguishes occurrences in which an expression is not used is not a difference in the expression’s semantic properties, but rather their irrelevance. When an expression is merely mentioned, the expression may retain its semantic properties, but those properties are inert: they don’t contribute to the determination of the truth-conditions of the sentence in which they occur. This is captured by the just proposed derivation of truth-conditions for direct speech reports: note that, in deriving the truth conditions for sentences of the form ‘S said “P,”’ one nowhere needs to invoke the axioms for the constituents of ‘P,’ since one nowhere needs to assign a semantic value to ‘P’ itself. Suppose, then, that an expression is used iff a derivation of the truth-conditions of the sentence in which the expression occurs must assign the expression a semantic value. This explication of use properly sorts the complements of direct speech reports into the category of non-use. More importantly for us, however, is that, given sententialism, it also sorts the occurrence of an expression in the complement clause of an indirect speech report – including an occurrence of a vague expression – into the category of non-use. For observe that the semantics proposed above for direct speech reports works equally well (sans the claim about the pragmatics of quotation marks) for indirect speech reports.32 So, even if it’s necessary that 32 Again, this goes for Davidson’s version of sententialism as well. Note, further, that one gets a similar result for a version of the “interpreted logical form” approach to indirect speech reports. A sentence’s logical form, on this usage, is the syntactic representation of a sentence for which one specifies its truth-conditions. An interpreted logical form is a logical form with the appropriate semantic value specified at each node. On the interpreted logical form approach, an indirect speech report is true iff the speaker said something expressed by a certain interpreted logical form – viz., that of the embedded sentence in the report. In Larson and Segal’s presentation, the final step of the derivation of truth-conditions for indirect speech reports involves applying the algorithm for generating interpreted logical forms in order to substitute the embedded clause’s interpreted logical form for what’s in effect the definite description “the interpreted logical form of ‘P’” (compare the truth-conditions just given). (See Larson and Segal (446).) Applying

Vagueness, Indirect Speech Reports, and the World

167

a vague expression occur in the complement clause of a true indirect speech report of an assertion entered using vague language, it hasn’t yet been shown that such a report must use vague language; and so it hasn’t yet been shown that not all truths can be stated using precise language.

VI. Concluding Remark Even if the (revised) suggested argument doesn’t establish that not all true indirect speech reports can be stated using only precise language, it may yet be the case. Perhaps ‘says’ (‘says+’) is vague—because ‘same-says’ (‘same-says+’) is vague, or because there are borderline cases of uttering assertorically, or even of uttering (suppose you barely perceptibly mumble to yourself something approximating ‘P’). On the other hand, if its vagueness isn’t indefinitely multi-dimensional, then maybe ‘says’ (‘says+’) could be replaced by a sharp ersatz in the manner discussed above. In any event, the suggested argument concerned only vague terms occurring in the complement clause. And even if all true indirect speech reports could be stated using precise language, it of course wouldn’t follow that all truths could be. I earlier indicated my sympathy with the thought that there are truths about vague expressions that can only be stated using vague language.

Bibliography Davidson, Donald. “On Saying That.” Synthese 19 (1968-9): 130-46. —. “Reply to Foster.” Pp. 33-41 in Truth and Meaning, ed. Gareth Evans and John McDowell. Oxford: Oxford University Press, 1976. —. “Quotation.” Theory and Decision 11 (1979): 27-40. Gross, Steven. Essays on Linguistic Context-Sensitivity and its Philosophical Significance. Ph.D. Dissertation, Harvard University, 1998. Published in Studies in Philosophy: Outstanding Dissertations. London: Routledge, 2001. —. “Vagueness in Context.” Pp. 208-13 in Proceedings of the Twenty-Second Annual Conference of the Cognitive Science Society, ed. Lila Gleitman and Avarind Joshi (Mahwah, NJ: Lawrence Erlbaum, 2000).

this algorithm involves invoking the axioms for constituents of ‘P’ in order to assign semantic values to the logical form’s nodes. If one were to omit this last step from one’s derivation of the truth-conditions for indirect speech reports (which is not to say one couldn’t carry out this inference extrasemantically), then the embedded clause’s constituent expressions would not count as used, according to the above proposal.

168

Steven Gross

—. “Is Context-Sensitivity Eliminable? Some Remarks.” Forthcoming in the proceedings of the Third Annual Italian/American Philosophy Conference, ed. Riccardo Dottori. Higginbotham, James. “Contexts, Models, and Meaning: A Note on the Data of Semantics.” Pp. 29-48 in Mental Representations, ed. Ruth Kempson. Cambridge: Cambridge University Press, 1988. Keefe, Rosanna. Theories of Vagueness. Cambridge: Cambridge University Press, 2000. Kim, Jaegwon. “The Myth of Nonreductive Materialism.” Proceedings and Addresses of the American Philosophical Association 63 (1989): 31-47. Larson, Richard, and Gabriel Segal. Knowledge of Meaning. Cambridge: MIT Press, 1995. Lepore, Ernest, and Barry Loewer. “You Can Say That Again.” Midwest Studies in Philosophy XIV (1989): 338-356. Ludlow, Peter. Semantics, Tense, and Time. Cambridge: MIT Press, 1999. Putnam, Hilary . “Vagueness and Alternative Logic.” Erkenntnis 19 (1983): 297-314. Quine, W. v. O. Word and Object. Cambridge: MIT Press, 1960. Unger, Peter. “I Do Not Exist.” 236-42 in Perception and Identity, ed. G. F. Macdonald. Ithaca, NY: Cornell University Press, 1979. Williamson, Timothy. Vagueness. London: Routledge, 1994.

Epistemological Remarks …

On Contemporary Philosophy and Social Sciences

169

170

Hans Lenk

Epistemological Remarks …

171

Epistemological Remarks Concerning the Concepts “Theory” and “Theoretical Concepts” Hans Lenk

Abstract Current interpretations of the concept of theory by philosophers of science of different orientations are sketched out, in particular the traditional statement view (theories as systems of statements) and the formal axiomatic interpretation (theories as calculi) as well as the historical or historicist interpretation of theories as series of successive research programs or paradigms etc. (theory dynamics). In addition, the model-theoretical nonstatement view (the so-called structuralist interpretation) of theories is roughly delineated and discussed as regards the status and interpretation of theoretical (T-theoretical) concepts. Finally, technological and action-theoretical approaches defining the modeltheoretic view are presented as particularly apt for a methodology of technology. A number of recommendations are developed from this, particularly the idea of satisficing in construction theory and the methodological idea of optimum fit instead of approximation to truth. The interplay between action, experiment and gaining knowledge is highlighted from a methodological and schema-interpretationist perspective. This approach seems to be conducive to a methodology and epistemology of engineering sciences, in particular construction theory and design theory.

Introduction Let me start with some rather ironic aphorisms: “A theory is nothing but the skin of truth – propped and stuffed” (Henry Ward Beecher). (We may infer from this “that theory has something to do with truth and should be related to practical applications” – as stated in the second bon-mot): “Theory should never forget that it is nothing but applied practice” – that is the late Gabriel Laub’s rather pointed ironic version of the proverbial insight that nothing would be more practical than a good theory. Or does A. J. Carter’s sarcastic remark apply:truth would just be a conjecture with academic education or qualification? In some sense these aphoristic insights make sense and all of them lead to different conceptions of theories.

172

Hans Lenk

Substantive and Operative (Instrumental) Theories It seems to me to be most important to distinguish between substantive and operative or instrumental technological theories (after Bunge, II, 122, I, 502ff ): the first ones are empirically contentful (notably to be found in real science) empirical and informative and testable, whereas the latter ones comprise instrumental or operative concepts of methods representing consistent methodologies. They represent structural concepts as, e.g., mathematical “theories” which may be applied in a kind of instrumental manner to specific realms, e.g. of empirical or ideal or even normative provenance. While the substantive theories imply empirical content and substantial information necessary for scientific or other substantial explanations and predictions, operative or instrumental theories do not comprise the respective explanatory nomological hypotheses, but are used for the precision and explanation or prediction in a secondary manner by being applied for practical reasons to given theoretical or structural presuppositions. They are but instrumental calculi or formal and procedural structures usually represented by mathematical operations or systems of operators. Operative theories are formal constructions of rules, calculi, structural interconnections which in their instrumental character are like language forms or mathematical formulae which do not represent substantial nomological hypotheses with empirical content although they are indispensable for formulating laws in precise scientific form. They have to be supplemented by nomological hypotheses with content: a calculus per se has no empirical content. Only by attributing to it observation statements, measurement procedures and testing techniques as well as application procedures will a formal operative theory together with these substantive ingredients have an empirical content and become so to speak a substantial or substantive total (of an operative and substantive theory). Substantive and operative theories are also models of quite different kinds and manners of acceptance. In the narrower sense one should only speak of genuine theories with respect to substantive theories. The so-called operative theories are but instrumental models or frameworks of models and procedures or procedural rules. (Many of the traditional mathematical instruments and “theories” are of this operative character. This also applies to most of the axiomatised theories as far as structural framework and the formalised basic axioms are concerned. Also formal system-theories, axiom systems of logics and mathematics are to be conceived of in this sense as instrumental or operative theories.)

Epistemological Remarks …

173

The distinction between operative and substantive theories is rather frequently not seen and acknowledged clear enough in axiomatically oriented natural science or in practice oriented variants of applied disciplines (like technology and organization or management sciences). The first practical maxim to be taken into account from a designtheorist would derived there from: it is necessary clearly to distinguish between substantive (substantial or contentful, e.g. empirically contentful) theories and operative or instrumental “theories” (which would rather figures instrumental concepts of methodologies and methods). This is advisable even though the latter ones somehow “penetrate” the first ones by providing their “logical framework”. Only the first ones (or in connection with the later ones) can claim for (empirical or contentful) truth. The latter ones are rather like instruments selected and taken from a cupboard of more or less useful instruments or appliances for structuring and formalizing or handling statements with substantial content.

From Truth towards Fitting/Feasibility or Function From what was said we can gain another insight which is known from traditional conceptions of philosophy of science: scientific theories have to be testable and truth-oriented, i. e. they have to be tested by confirmation or corroboration (total verification of universal hypothesis is not possible!) or by falsification (empirical rejection) taking into account extant criteria of truth. Theories not oriented at factual truth are not theories of empirical science. (Logical or mathematical ‘truth’ is another concept: such formal theories are really proved as valid (formally true or logically true) by proving their consistency, i. e. non-availability of contradictions. For empirically contentful theories non-contradiction is but a necessary, though not a sufficient condition for truth.) As a special note for design-theorists let me state: in so far as “theory” does not deal with truth or empirical or factual truth claims, but e.g. with “goodness”, it cannot be an empirical scientific or science-based theory, but rather methodical or even meta-methodical or methodological principles, i. e. general structural maxims of actions. These would structure theprocesses and steps of actions at times even functioning normatively, but they would not be related to just nomological truth in the empirical sense. Axiomatic design theories are actually rather normative instructions or maxims. More precisely, they are generalised and abstracted models of interpretive constructs

174

Hans Lenk

of procedural proveniance which are to be seen or applied under specific normative criteria like “goodness”, “optimisation”, “adequate problem solution”, “satisfaction of functional requirements” as well as “simplicity”, “feasibility”, “cost” etc.

What are Scientific Theories? In current approaches to the philosophy of science we can distinguish different conceptions of theories and their interpretations which are all derived from the traditional conception of analytic philosophers of science (like the positivists of the Vienna Circle, but also from Popper’s critical rationalism). Traditional Approaches 1. The traditional interpretation conceives of theories as logically interconnected systems of statements or sentences consisting of universal law like hypotheses which are taken as perfect or completed theories represented in an axiomatised form. Their reference to reality or empiricality is realised by attributing to the concepts of the theories the corresponding observation statements observing special corresponding rules or attribution rules, connecting the theoretical concepts with observation predicates of a lower level which can be satisfied or fulfilled by observation or measurement. This traditional two-level conception of empirical scientific theories is however today assessed very critically – or at least controversely – and has to be supplemented or substituted by several other conceptions. (It is common knowledge now that even observation statements or statements about measurement and their results are theoretically “infested”, “theory-laden”, “theory-impregnated” – be it by a theory of measurement or so.) 2. Intimate connected with a traditional conception of theories is the primacy or paragon function of mathematical theories in the form of axiomatisation (as it was devised by mathematical logical formalism (Hilbert) at the beginning of the 20th century. Accordingly, mathematical theories are such axiom systems which would define their uninterpreted theoretical concepts “implicitly” by their respective structural axioms themselves and which are able completely to cover a realm to be comprised by the theory (e.g. classical algebra). (Through the well known meta-mathematical results by Gödel und

Epistemological Remarks …

175

Church these expectations have been proven to be impossible: highly complex axiomatisations1 comprise a kind of incompleteness and the provability of consistency is restricted: in addition, there are no absolute guarantees and mechanisms and algorithms of proofs even for some logical or mathematical valid theorems within the formal system.) Attempts of total axiomatisation of theories have therefore proved definitely unsuccessful. The axiomatic method can generally only be conceived of as a didactical useful construction. As such it is certainly still important and applied everywhere especially with respect to purely formal, e.g. mathematical, theories. But axiomatisation and “scientificness” cannot be identified, not even in the purely formal sense, not to speak of the real empirical scientific sense. Axiomatic systems are but aids – maybe practically necessary ones –, but they are not the real empirical scientific theories themselves. This leads to the conclusion: axiomatisation cannot be the only criteria of being or going scientific, as necessary in the practice of teaching science axiomatic systems might be. In particular, formal axiomatisation by itself, that is the mathematical structure of the calculus, of a theory cannot be mistaken as “the theory” in its scientific as well as in its practical sense. (This should be taken into consideration by construction engineers and theorists of design oriented at formal axiomatisation.) Dynamics of Theories and their Succession 3. Since 40 years or so historisation of the conception of science is characteristic for the philosophy of science. After Feyerabend and Kuhn philosophers of science started to take into account factual history of science. The real development of the sciences is largely shaped by the groups of the respective relevant scientists and the values and changes of conceptions: so called “normal science” is distinguished from “ revolutionary” phases (Kuhn). The latter ones can lead towards a new “paradigm” within a special realm of research: the consequence is a certain (or even absolute) incommensurability of paradigms or the respective theoretical fundamental interpretations (e.g. classical mechanics vs. quantum-mechanics). From this a third conception of theories as group-supported basic paradigms developed according to which the “normal” internal development of a scientific paradigm is so to speak science-politically supported and authoritatively launched or put through, only 1 Being at least of the complexity of number theory and quantified logic (predicate logic) of first order including identity.

176

Hans Lenk

to be later superseded by a new paradigm. Comparisons of theories, competing results and data as well as scientific progress lead in such an extreme sociological or even sociologistical version of the approach like (e.g. according to the “strong” conception of the so called Edinburgh School of the Sociology and History of Science) to a theoretically insoluble dilemma between a rational assessment of theoretical progress and but rational scientific change. 4. Imre Lakatos (like Feyerabend a creatively deviant disciple of Popper) developed a concept of historical succession of theories consisting of “research programs” which supersede one another according to criteria. This conception replaces Popper’s original naive falsificationalism by a sophisticated one. Theories are still systems of universal hypotheses (“laws”) in their logical-deductive interconnection, but they are only indirectly falsified, in so far as the interplay between the explanatory theory and the theoretically impregnated observation statements will only lead to a statement of inconsistency (e.g. between measurement theory and the explanatory theory). Theories are always to be assessed within the historical succession of predecessors and successors: succeding theories are “better” than their predecessors, if and only if they have more empirical content, i. e. allow to predict some novel facts, or contribute to the avoidance of some anomalies of the old theory (although any theory comprises some anomalies still) and/or renders an integrating theory-fusing process of generalisation. Transition to a better theory is called “theoretically progressive” if it allows to predict novel facts which are not yet empirically confirmed. (A change is called “empirically progressive” if some of these predicted novel facts are really confirmed.) An example would be the transition from Newton’s theory of gravitation to Einstein’s general theory of relativity due to Eddington’s experiments regarding the deviation of light rays close to the sun (1919). Model-theoretic Approaches Non-statement View 5. The so called non-statement view of theories as set-theoretical predicates or model sets was mainly developed by Sneed and Stegmüller – after ideas by Suppes. This approach which is also called a structuralist view of theories understands a “theory” as a totality or net of theory elements which are partially ordered2 by specialising relations or constraints, i. e. by edding of spe2 In the sense of a logically mathematical partial ordering relation, i.e. reflexive, asymmetric, transitive and identitive relation.

Epistemological Remarks …

177

cial laws. (Occasionally and originally the ordered pair of the mathematical structural core K and the set of partial potential models being the intended possible applications – the ordered pair {K, I} – is called a theory – or nowadays “theory-element”. The partial potential models of a theory comprise all the applications conceived of which are not yet described by theoretical functions and observables, namely the models of intended applications.) By definition, the structural core (defined only by mathematical relations) comprises the potential and partial models as well as constraints (i. e. theoretically required interconnections between the partially overlapping partial potential models) and the models themselves (i. e. the factual that systems are really successfully described by the theory). Empirical statements and hypothesis of a theory are now met in the statement that the respective intended applications of the theory belong to the applications of the net (or the structural core) fulfilling the constraints. The set of partial potential models (i. e. the real systems which are intended for application of the theory without theoretical functions) are supplemented by the respective theoretical functions leading to this set of potential models. To be sure the adding of theoretical functions and specialisation (by added special laws) has to lead to a partial set of “fulfilled” models (M) such that the whole sequence of theoretical functions has to satisfy the constraints. According to the structuralist non-statement approach a theory consists simply of an ordered pair of a mathematical framework of formulae (structural core) and a set of possible intended applications and constraints. The possible applications are object systems or real systems which are up for application of the theory and which are usually given by certain paradig-matic initial models usually proposed by the founder of the theory. The theory then is an ordered set of theoretical predicates (after Sneed and Stegmüller). “Formally, a core K may be represented either as a quadrupel K = [Mp, Mpp, M, C] or as a quintupel K = [Mp, Mpp, r, M, C]. Here, Mp, Mpp, and M are the sets mentioned … C is a set of constraints, i. e., a subset of the power set of M; and the restriction function r: Mp – > Mpp transforms an element of Mp, i. e. the potential model, into an element of Mpp, i. e., into a partial potential model by “lopping off ” all theoretical functions” (Stegmüller 1979, 25). The total theory might be an exented core leading to a theory net by adding specialised laws and respective new constraints, restriction functions and new intended models to be inserted into the set of potential models and fulfilled models. In short, a theory is therefore a relation predicate defined over the set of potential models of applications. The predicate “… is a theory” states the existence of a relation between the mathematical structural

178

Hans Lenk

core and the set of at the time specified set of intended applications of a theory, the set of potential models being extendable: further potential, previously not intended applications can be integrated (e.g. extension of Newtonian classical dynamics to include gravitational systems by adding a law of gravitation or the extension by adding Hook’s law of a linear restoring force to include harmonic motion). A series of interesting results can be derived from this new conception: it is possible to speak of one and the same theory, even if the set of special laws of the theory and the set of intended models is expanded, as long as the structural core (the mathematical basic laws) of the theory are maintained. E.g., Newtonian mechanics of the Newton’s first three axioms is specialised by adding special laws like Hook’s law or the law of gravitation while remaining still the same theory, but differentiated and specified. T-theoretical Concepts In addition, the property of being a theoretical concept is now relativised to the respective theory and can be governed by a criterium independent of the traditional (principally thwarted) distinction between observable and non-observable magnitudes. Beside the opposition of “theoretical” vs. “observable” (amounting to the distinction of theory language and observation language according to logical positivists like Carnap) now a relativesed conception of “pre-theoretic” concepts and “theoretic” concepts of the respective theory T (more precisely, T-theoretic concepts) is introduced: for instance the concepts of “mass” and “force” in Newtonian mechanics and theory of gravitation are theoretical concepts, whereas “space” and “time” are not (they might however be theoretical concepts in a more basic physical geometry of Euclidean structure). A T-theoretical concept is distinguished by the fact, that it can only be instantiated or attributed by using the theory itself, i. e. by measuring the respective variable. Pragmatically, it is possible now to define the concepts of “using a theory” as an action concept which is not only logically characterised, but also impregnated by pragmatic, action theoretical and agent related aspects (cf Stegmüller, e.g. 1980). Extensions of Structural Cores Progress in handling theories by extending the structural core via adding further intended models can be realised in various manners: by extending the set of intended applications, i. e. partial potential models, by refinement of the net, of the structural core as well as by ramification with respect to mu-

Epistemological Remarks …

179

tually exclusive intended potential models characterised by non-compatible special laws. The structural approach renders a certain autonomy of the structural core as regards remifications and extensions as well as a certain immunity with regard to relativised falsifications. From T-theoriticity (the relativity with respect to the theory T) of the theoretical concepts and due to the theory impregnatedness of observations a theory so to speak defines its own facts to be governed and covered by it leading to a holism of falsifications and confirmations in such a manner that only a theory net in total is confronted with experience or experiment and confirmed, corroborated or falsified by it. Practical Modelling and Axiomatisation Again a piece of advice for design theorists: this structuralistic approach of the theoretical structural cores seems to be at first view rather abstract and far from practical considerations. However it is a considerable advantage of it that theories are conceived of sets of mathematical structural frameworks and models and that one and the same structural core (the same basic axioms) can be related to new intended models (or sets of these, respectively), without being urged to talk of a new theory. In addition – which is of importance for design theory – one may transfer the constitutive connection between mathematical structural cores and intended applications (partial potential models) also to contexts in which one cannot speak of the truth of a substantive theory: for instance this would apply to general principles of technological fields to be described, characterised and covered under the point of view of a criterium of “fitting” or “goodness” or functional requirements etc and the respective specialisations by adding further more specific “laws” or “rules of thumb”. In fact, axiomatic design theory after Suh and Chang may be presented as a set of precised fundamental principles (“axioms”) characterising a structural core supplemented by plurifunctional requirements (FR) instead of “truth”. The structural view of theories would favour this combination of an axiomatic approach regarding the mathematical structural core to be maintaining and the practical selection of models from an open and extendable set of partial potential models. (This approach can also be rendered compatible with historical developments of principles and their specialisations.) Generally speaking the structural view of theories would support the interpretation of design theorists that “theories” (more exactly: methodologies, general design principles and their structural interconnections) can be conceived of as precise (or to be precised) structural cores including plural functional requirements of satisfaction and different

180

Hans Lenk

models etc. Axiomatic design theory is rather a normative generalised methodology respecting specific functional requirements and design parameters (PR) which are characterised by multiple and changing optimal or satisfactory model solutions, than a substantive scientific theory with empirical content in the narrower sense. Such pragmatic approaches of the relation between structural cores and models seemed to be especially suited for technological “methodologies” in the sense of a set of systematic interconnected methods and the intimate relationship between technological rules on the one hand and scientific substantive basic theories on the other. (Unfortunately, engineering sciences have traditionally not developed a general methodology in the philosophy of science sense with criteria for theoretical justification, although the development of a “general technology” (“Allgemeine Technologie”) and its methodology has been asked for by some philosophers of technology (Ropohl 1991, Lenk/Moser 1973). Technology-oriented and Action-theoretic Approaches 6. The model theoretic approach can be extended in order to better cover technological models. Ronald Giere, in his book Explaining Science (1988), developed an interpretation which combined the structural view with technological approaches (similarly also Hacking 1983). Giere (1988, 85f ) understands “theory as comprising two elements: (1) a population of models, and (2) various hypotheses linking those models with systems in the real world. Thus, what one finds in the textbooks is not literally the theory itself, but statements defining the models that are part of the theory.” Important is the relation of similarity between the models and their real systems to which the models apply as cases of application: “the links between models and the real world … are nothing like correspondance rules linking terms with things or terms with other terms. Rather, they are again relations of similarity between the whole model and some real system. A real system is identified as being similar to one of the models.” (ibid. 86) Giere conceives of a theory “as a family of models”, or still better, was a family of families of models” (ibid. 82, 91) which can be related to the world by similarity and fitting of the models with the respective real system as mentioned above. Theories in that sense are no linguistic entities or just frameworks of formulae but heterogeneous sets consisting of abstract constructs, the theoretical models, and linguistic entities like hypotheses about the fitting character of these models and their similarity with reality susceptible to grading and perspectives: “a real system is identified as being similar to one of the models. The interpretation of terms used to define the models does not appear in the picture; neither do the

Epistemological Remarks …

181

defining linguistic entities, such as equations” (ibd. 86) “When approaching a theory, look first for the models and then (only, H. L.) for the hypothesis employing those models. Don’t look for general principles, axioms, or the like.” (ibid. 89) That will mean with respect to the above mentioned example that one has to look for the models and their similarity with real systems regarding pre-theoretically characterised representations of the model of the planetary system or the model of the earth – moon as original or primary models of Newtonian theory of gravitation. With respect to relating and combining theoretical models with real systems to be covered technology now plays a decisive role. Like Hacking (1983) Giere’s constructive realism sees a proof of reality in the successfully managed technologies in handling entities (e.g. electrons) which earlier had the status of a theoretical entity, if they are applied to cover and characterise new models or other theoretical entities. (If we routinely use nowadays electron rays in accelerators or in electronic microscopes successfully to resolve other scientific tasks, we understand in this technological sense the theoretically postulated electrons which were earlier mere theoretical entities now as scientific-technological real entities.) In so far as electrons and protons are manipulated and applied in big technology measurement instruments and appliances to probe and prove the structure of other elementary particles like gluons, quarks etc., these electrons and protons now are “real” indeed (Hacking 1983): thus “some of what we learned today becomes embodied in the research tools of tomorrow” (Giere 1988, 140). Fitting as Satisficing Giere states that scientists are constructive realists who relate models by technological applications and interventions to reality. Thereby they are blared to an experimentalistic realistic conception of improved models in the sense of a relativised (not necessarily optimally) fitting or suiting after H. A. Simon called “satisficing”: one does not maximize the model adjustment, but would optimize it in the sense of rendering a satisfying result, the relative optimisation of goal attainment so that a satisfying result for experimental and somehow functionally restricted but relatively best accordance of models with reality would occur. Scientists are according to Giere optimizers or “satisficers”, but no absolute maximizers with regard to a correspondence of their models to reality. Advice for design theorists: this view of technological theoretical models as relatively well-fitting, not necessarily the absolutely best models with regard to problem solutions from the point of view of fulling functional require-

182

Hans Lenk

ments (FR) would render the technological (“technologistic”) and constructive, realistic model as developed by Giere and Hacking specially suited for technological designs, for the methodology of construction etc. Giere states explicitly (1988, 137): “The main connector between our evolved cognitive capacities and the microworld of nuclear physics is technology.” Certainly this is valid more generally: “The development of science depends at least as much on new machines as it does on new ideas”: “It is technology that provides the connection between our evolved sensory capacities and the world of science” (eb. 138). Technology is conceived of as the example of ways of experimenting, development of instruments that are “man-made extensions of the senses” and all our capacities including the instruments and procedures as well as operations are metaphorically speaking, concretised “embodied knowledge”, i. e. “embodied in the technology used in performing experiments” (ibid. 140) We have to add technological developments of new instruments, procedures, technological systemsa and the interface between these latter ones and action systems, if not today social and socio-technological systems in socio-technical networks. Therefore we have to add the essential reference to structured actions and action systems in their worldly and social embeddings. Even experimenting and the devicing developing of theories as also the utilizing of theories is a sort of acting, a kind of action.

Action, Experimenting, Knowledge In the treaties of epistemology and the philosophy of scientific theories as well as knowledge in general the present author (1998 shows the insoluable interconnection between knowledge, experimenting and action. Insofar Giere’s combination of scientific models and their relation to real systems by technology and technological manipulation and intermediate operators like measurement instruments and machines has to be extended by an action theoretical interpretation. This would be of utmost interest for design theorists, since the design of hardware structures and real systems and the respective structuring manipulations as well as the design of software in software models would be covered by such a view. The pragmatical model theoretic approach with respect to technical instruments has to be supplemented or expanded by an action theoretical perspective which is notably suitable for design theories. Whereas traditional conceptions of theories in science emphasized too strongly the rather pure theoretical elements and hypothesis as linguistic en-

Epistemological Remarks …

183

tities, traditional axiomatic as well as the pure structuralistic conception suffered from too formalistic an orientation conceiving of theories and their structures exclusively as mathematical structures or even subtheoretical complex predicates. The philosophy of this sometimes so called new experimentalism, of a pragmatic technology-oriented provenance and the action-theoretic perspective may avoid these over-simplifications and at the same time refine the structural interconnection between idealized cognitive models or intented partial potential models of theories by stressing technological realisations and materialisations and action-theoretical as well as operational sequences (as to be found in the design of operations and experiments etc.). It is in such way that also the design theorist may relate his methodology or metamethodical conception of operative principles of design to the fulfilment of functional requirements and the optimisation or satisficing of plurifunctional conditions – somehow independent of absolute truth claims (in the sense of substantive empirical true or truth-like theories). That seems to be typical for design tasks. In this sense the normative component is taken into consideration and accounted for within this kind of general methodology or doctrine of principles determined by plurifunctional requirements (e.g. in Suh’s sense). This amounts basically to an extension of the old objective of traditional construction systematics of the fifties and the sixties (as of e.g. Hansen 1965, Müller 1967) now only in a model theoretically refined way. One could even go further and attach a theory of creatively ordered and structured activities in general as insinuated for instance by social psychology and philosophy of creativity (Lenk, 2000a).

Schematheoretic and Interpretationist Perspective Pragmatic philosophy of science has much to learn from technological and action-theoretical approaches. Similarly also the methodology of engineering sciences or the general technology still to be developed may gain much from insights of methodological character and differentiated considerations of refined developments and novel insights of philosophy of science and general methodology including theories of action. Epistemologically speaking these methodological approaches can be embedded in a general “theory” or methodology of schema-interpretation (Lenk 1993, 1995). This approach understands any grasping of real systems as methodologically and epistemologically dependent on specific perspectives, teleofunctional requirements,

184

Hans Lenk

theoretical constructions and approaches as well as practical action routines or social conventions and institutions, respectively. A new unity of the sciences and technologies as well as the understanding a knowledge of the world and of the manipulations as well as interpretations of reality by acting and utilising theoretical interpretational as well as experimental and action-practical models gains relief on a metatheoretical level characterised by general methodological requirements of any active processes of “grasping” external or mental entities as well as ideal structures. Acting, grasping and realizing as well as forming and shaping and rendering normative structures is in this sense to be understood and analyzed fromthe point of view of schematisations (the development and activation as well as utilisation of schemas of partially hereditary, mostly however learned provenance) and by interpretations, manners and patterns of “grasping” something in the rather active or passive sense under specific perspectives. The main question – generally and ironically speaking – seems not to be “Dasein oder Design” (“being or design”) – to put it in a pun not translatable to English! –, but that any “grasp” of structured real “beings” (real entities and their relational patterns) is always also dependent on a kind of “design” of in part primary biological or hereditary, in part conventional and higherlevel interpretations (cf. figure). To come back to Beecher’s statement on truth at the beginning of this paper: a theory is more than “the skin of truth – propped and stuffed”: beyond fitting and satisficing as well as fulfilling some correspondance conditions for reality or –based restrictions it is the complex interpretational construct (out of and consisting in many schemata and interpretations mutually subordinated or coordinated), a way of dealing by procedures and operations, actions and techniques in more or less practical concrete steps and measures with external world sections, world factors (entities and processes), potential models, real systems as well as meaningful semantic entities (mental entities, ideal constructs etc.). Theories and generally speaking, methodical and methodological concepts as well as normative structures of actions and procedures would guide us in the form of interpretations and schematisations. Methodological interpretative constructivism (schemainterpretationism) was developed by the present author as a higher-level methodological as well as epistemological concept comprising the special cases of scientific theories, technological designing, blue-printing and prototyping as well as all other kinds of procedures to structure action and thought patterns. Interpretaions are constructions in a narrower or a wider sense (cf Lenk 1995). Theories are interpreta-

Epistemological Remarks …

185

tional constructs which as substantive theories claim for truth, at least approximate truth or for empirical relidity – or as operative or instrumental theories for methodical or methodological validity. Norms and values are interpretative constructs, too. So, there are normative interpretative constructs in any process of designing and decision making. Designing and problem-solving in design are completely of an interpretatory character. In short: there is nothing more practical than clear-sighted interpretations.

References Bunge, M. 1966. Scientific Research, I, II. Berlin – Heidelberg – New York: Springer. Carnap, R.: Philosophical Foundations of Physics. New York . Chang, S. 1990. “A Scientific Approach Towards Developing An Engineering Design Theory.” In International Conference on Engineering Design ICED 1990 (Proceedings). Dubrovnik. Giere, R. N. 1994. “The Cognitive Structure of Scientific Theories.” Philosophy of Science 61 (1994): 276-296. — . 1988. Explaining Science: The Cognitive Approach. Chicago – London: Chicago University Press. — . 1985. “Constructive Realism.” Pp. 75-98 in Images of Science, ed. by D.M. Churchland, and C.A. Hooker. Chicago: Chicago University Press. Hacking, I. 1983. Representing and Intervening. Cambridge – New York : Cambridge University Press. Hansen, F. 1965. Konstruktionssystematik. Berlin: VEB Verlag Technik. Kesselring, F. 1954. Technische Kompositionslehre. Berlin – Heidelberg – Göttingen: Springer. Kuhn, T. 19702. The Structure of Scientific Revolutions (1962, mit Postskript 1969). Chicago: Chicago University Press. Lakatos, I. 1968. Criticism and the Methodology of Scientific Research Programs. Proceedings of the Aristotelian Society 69 (1968): 149ff. Lakatos, I., and A. Musgrave (ed.). 1970. Criticism and the Growth of Knowledge. Cambridge: Cambridge University Press. Lenk, H. 2000. Erfassung der Wirklichkeit. Würzburg: Königshausen und Neumann. — . 2000a. Kreative Aufstiege. Frankfurt am Main: Suhrkamp. — . 1998. Einführung in die Erkenntnistheorie: Interpretation – Interaktion – Intervention. München: Fink – UTB. — . 1995. Schemaspiele. Über Schemainterpretationen und Interpretationskonstrukte. Frankfurt a. M.: Suhrkamp. — . 1994. Macht und Machbarkeit der Technik. Stuttgart: Reclam.

186

Hans Lenk

— . 1993. Interpretationskonstrukte. Frankfurt a. M: Suhrkamp. — . 1993a. Interpretation und Realität. Frankfurt a. M: Suhrkamp. — . 1991: “Zu einem methodologischen Interpretationskonstruktionismus”. In: Zeitschrift für allgemeine Wissenschaftstheorie (Journal for General Philosophy of Science) 22: 283-302. — . 1986. Zwischen Wissenschaftstheorie und Sozialwissenschaft. Frankfurt a. M.: Suhrkamp. — . 1982. Zur Sozialphilosophie der Technik. Frankfurt a. M.: Suhrkamp. — . 1975. Pragmatische Philosophie. Hamburg: Hoffmann & Campe. — . 1971, 19722. Philosophie im technologischen Zeitalter. Stuttgart: Kohlhammer. Lenk, H. und M. Maring (Hg.). 1992. Wirtschaft und Ethik. Stuttgart: Reclam. Lenk, H. und S. Moser (Hg.). 1973. Techne – Technik – Technologie. Pullach bei München: Dokumentation Saur. Lenk, H. und G. Ropohl (Hg.). 1987, 19892. Technik und Ethik. Stuttgart: Reclam. Müller, J. 1967: “Probleme einer Konstruktionswissenschaft.” Maschinenbautechnik 16: 338ff, 394ff, 454ff. Nagel, E. 1961. The Structure of Science. New York: Hartcourt. Popper, K. R. 1966. Logik der Forschung (1935). Tübingen: Mohr & Siebeck. Rapp, F. 1978. Analytische Technikphilosophie. Freiburg – München: Alber. Ropohl, G. 1991. Technologische Aufklärung. Frankfurt a. M.: Suhrkamp. — . 1979. Eine Systemtheorie der Technik. München: Hauser. Sneed, J. D. 1971. The Logical Structure of Mathematical Physics. Dordrecht: Reidel. Stegmüller, W. 1980. Neue Wege der Wissenschaftsphilosophie. Berlin – Heidelberg – New York: Springer. — . 1979. The Structuralist View of Theories. Berlin – Heidelberg – New York: Springer. — . 1976. The Structure and Dynamics of Theories. Berlin-Heidelberg-New York: Springer. — . 1973. Probleme und Resultate der Wissenschaftstheorie und analytischen Philosophie. Band II: Theorie und Erfahrung, Studienausgabe Teil D: Logische Analyse der Struktur ausgereifter physikalischer Theorien. Ein ‘Non-Statement-View’ von Theorien. Berlin – Heidelberg – New York: Springer. Suh, N. P. 1990. Principles of Design. New York – Oxford : MII. Suppe, F. 1972. “What’s Wrong With the Received View on the Structure of Scientific Theories”. Philosophy of Science: 39, 1-19.

Epistemological Remarks …

187

Diagram of the Levels of Interpretation IS1: practically unchangeable productive primary interpretation (“Urinterpretation”) (primary constitution or schematization, respectively) IS2: habit-shaping, (equal) forms-constituting pattern interpretation (ontegenetically habitual(ized) form and schema categori(ali)zation and preverbal concept-formation) IS3: conventional concept formation transmitted by social, cultural and normregulated tradition IS3a: … by non-verbal cultural gestures, rules, norms, forms, conventions, implicit communicative symbols IS3b: … by verbal forms and explicitly representing communicative symbols, metasymbols, metaschemata, and so forth IS4: applied, consciously shaped and accepted as well as transmitted classificatory interpretation (classification, subsumption, description by “sortals,” generic formation of kinds, directed concept-formation) IS5: explanatory and in the narrow sense “comprehending” (“verstehende”), justifying, theoretically or argumentatively substantiating-interpretation, justificatory interpretation IS6: epistemological (methodological) meta-intcrpretation (plus metameta-interpretation, and so forth) of methods, results, instruments, conception of establishing and analyzing interpretative constructs themselves

188

Steven Miller and Marcel Fredericks

Social Science and (Null) Hypothesis Testing: Some Ontological Issues Steven Miller and Marcel Fredericks

There appears to be a continuing interest in the applicability of the so-called null hypothesis testing issue to a variety of concerns in the philosophy of social science. Thus, Douglass (2000) has recently argued that statistical testing (NHST) is a form of “inductive risk” whereby non-epistemic values do enter into interpretations of scientific models. For example, in the testing of dioxin levels on alleged cancer rates, the values guiding the choice of levels of statistical significance in how model generated data are eventually interpreted become crucial. As Douglas (2000: 567) indicates in this regard, the setting of standards of statistical significance must take into consideration the balancing of false negatives and false positives in terms of the policy consequences of such decisions.While the issue of what values (epistemic or non-epistemic) may mitigate “inductive risk” is specifically central to discussions of NHST, we wish to generalize the idea of inductive risk to an examination of how such risk may figure into the evaluation of theories, both in the natural and social sciences. What we have in mind is not to argue for the desirability of critical or “severe tests” (Mayo, 1997, 1996, 1991) based on error statistics – we take this as a given – but rather that the application of NHST as a way of providing for severe tests for theory assessment presents additional difficulties. These difficulties lie in the epistemological neglect of the implications of the fuller NHST model, one that must take into consideration the associated ideas of “power” and others in relation to the notion of so called Type II error. Our argument is two fold: (1) that the appropriateness of the NHST model for theory assessment must, especially, take the “power” dimension explicitly into consideration but, paradoxically, doing so undermines the possibility of meaningfully addressing hypotheses from both “strong” and “weak” theories. We need to point out that our concern is not centered around the epistemological adequacy of the NHST model itself, which has been reviewed elsewhere (e.g., Cohen, 1994), but how it plays out in what kinds of “ontological commitments” remain given the NHST assumptions. Our concern, then, is what can be “said” about a theory (whether “strong” or “weak”) once a more complete view of NHST is itself taken as a background theory. In doing this, the intent is not to enter into either the thicket of “semantical”

Social Science and (Null) Hypothesis Testing: Some Ontological Issues

189

theories or those of the realist-anti-realist persuasion (see, generally, Devitt, 1991 for a good review of these), but only to adhere to a “minimal” realist view (e.g., Lynch, 1998) concerning the “existence” of theories; namely, whatever theories are, they exist independently of us as knowers. Although not implied by even a minimal realism, there is the interesting (derivative) issue of whether the NHST model can in some sense be viewed as a “correspondence” theory and, thereby, indirectly give some support to this form of Realism. We do not believe this to be the case and will comment on it briefly later. Our principal motivation for trying to look at the NHST model from an ontological perspective lies in the recent important work of Trout (1998a, b). Our beginning focus will be on Trout’s definition of Measured Realism. We will then attempt to show that how, with a fuller analysis of NHST, the measured realism notion remains problematic and indeterminant for both well established and weaker theories generally.

Trout’s “Measured Realism” Trout is rightly concerned with the general utility of NHST in, especially, the testing of theories, or parts of them, in the social sciences - here, those in psychology. While Trout focuses on the distinction within psychology between the “hard” areas such as cognition and perception versus the “soft” theories of personality and counseling, his remarks are in principle generalizable to any theoretical concerns within the philosophy of social science. After laying out some of the general problems of NHST, such as the overdependence on the .05 level of significance (S 264), Trout’s focus, following Meehl (1996), is on the appropriateness of using NHST in the following situations: (1) In early stages of research examining possibly important trends, (2) in technologically defined intervention contexts (e.g., drug efficacy), (3) in theoretical contexts where a competing (false) theory can be refuted, and (4) in theoretical contexts where the failure to falsify can be grounds to corroborate (e.g., Popper, 1962). Trout’s major concern, however, is not when NHST is appropriate to use, but rather how its over-use has led to a false (epistemological) reliance on its being the best, and perhaps only, way to test theoretical claims. That is, NHST over-reliance has diverted our attention away from the real purpose of any such testing, which is the importance of theoretical (ontological) commitments. Moreover, the diversion from the ontological has clouded such important perceptions that it is the robustness (or lack of ) a theory, to begin with, that determines how, where

190

Steven Miller and Marcel Fredericks

and when a “severe” test is appropriate.Trout (S 266) brings this point out forcefully in saying, “with its exclusive focus on methodological conditions of theory testing, confirmation theory attempts to address, in a metaphysically neutral way, questions about the goodness of a test. In doing so, ontologically consequential questions are set aside, concerning the way in which the quality of a test depends upon the epistemic reliability of background theoretical commitments.” He goes on to illustrate this claim by arguing that unless there is, for example, already a well developed MRI (magnetic resonance imaging) technology in place and supported by well established underlying principles in physics, the probability of correctly picking out a brain tumor as a cause of epileptic seizures (i.e., the hypothesis) would be low. Measured Realism for Trout, then, is based on the assumption that only true theories or approximately true theories, or their relevant parts, make for sensitive (i.e., “severe”) tests. Again, this important point is summarized by Trout (S 266) as follows: The severity of a test routinely depends on the epistemic reliability of sophisticated bodies of theoretical information. Just because severity can be analyzed in statistical terms does not mean that probability assignments can be made independently of theoretical commitment. The methodological import of test severity, like that of experimental design, can only be fully explained extra-statistically, even if certain of its features, can be elucidated in statistical terms. These extra-statistical considerations concern background theories. The above strategies are abductive, because the inference that an outcome is genuine depends on the sufficient reliability of background theories, and provides the best explanation for performance. Measured Realism for Trout, then, lies squarely within Realism as an ontological doctrine, but as applied to the social sciences it is more “modest” and “measured” (S 269). To clarify more precisely what a Measured Realism consists of, he posits four theses (S 269-270) as follows: 1. At least some of the quantitative methods and practices of science, including NHST, reliably detect and identify some of our central posited entities, processes, states, properties, and events in the social and behavioral sciences. 2. Sometimes behavioral and social scientific theories, more often the generalizations that constitute them, are confirmed by the application of accepted methodological standards in the field. This process of confirmation is most evident in such activities as the refinement of measurement procedures and in the successful introduction of controls.

Social Science and (Null) Hypothesis Testing: Some Ontological Issues

191

3. In our best psychological and social theories, confirmation relations important to theoretical progress in those sciences are epistemically bitesized and, accordingly, adipose or extraneous. The methodological independence (or “diverse testability”) of theoretical objects, processes, events, states, and properties provides evidence that laws about those theoretical kinds are relatively detachable from theories, and so confirmable relatively independently of the particular theories in which those laws were expressed. 4. The reality described by our best psychological and social theories is independent of us as observers of, or as thinkers about, that reality. Of the theses for Measured Realism, the one that needs some additional attention is (4). While Trout acknowledges realist ontology as being central to understanding how NHST may apply to the testing of theories (or some relevant parts of them), his ontological commitment needs additional clarification. While “strong” theories do in some sense describe that reality more accurately than “weak” theories, the reality itself is not affected by the distinction. However in (4), Trout seems to suggest that only our “best” theories can serve as a basis for describing an independent reality. But whether such theories are “best” (or even weak), they must be describing an independently existing reality if even the most basic realist claim is to be supported.In other words, if Trout is claiming that the defect of the NHST model is that it does not adequately take into account the way ontological (realist) assumptions count in terms of theory development, then there is nothing wrong in this as long as the independent-reality assumption is not made contingent upon this fact. However, the role of NHST then becomes even more problematic: our weakest theories - even if existing independently - remain so because the hypotheses (null) put up for testing will never admit of severe testing by the very fact that they are weak to begin with; while well developed theories are not only better candidates on the independent-reality criterion but their hypotheses are capable as well of severe testing. But, then, severe testing is not needed since such theories are already quite capable of “passing” such tests. From such a perspective, the process of theory growth or theory refinement remains unclear - weak theories remain in an “ontological limbo,” perhaps independently real, but consigned to remain so in any case since the very purpose of severe testing is not meaningful, hence they cannot be refined. Conversely, strong theories do not need such testing now, but it begs the question of whether they previously had to undergo less severe testing to get where they are now. By the above, perhaps not. Trout is correct in claiming the traditional NHST model does not adequately take (realist) ontological considerations into account in terms of

192

Steven Miller and Marcel Fredericks

theory testing, but when it does the ontological status of any theory (social or natural science) still remains unclear in light of NHST.

Severe Tests and Ontological Commitment There are a number of interrelated reasons why the NHST model is not adequate for theory-testing; however, these reasons go beyond what Trout is suggesting. Some of these will be mentioned here and expanded in the following section. An often unforseen consequence of NHST is that it results in a type of “ontological ambiguity.” By this we mean there results an atypical bifurcation of what is assumed can be known, given how NHST is used. For example, under NHST a given hypothesis is “accepted” or “rejected” at some specified (e.g., p < .05) level of statistical significance. Now, under Trout’s construal, let’s assume the theory (or some part of it) is “weak” to begin with, but that the investigator believes, with acceptable reasons, that a more severe test (i.e., p < .01), should be attempted, and indeed, the (null) hypothesis is not rejected under the more stringent test. Ontologically, does the theory (or some part of it) now alter its status? Conversely, under a weaker test, is the theory (or some part of it) more independently present? Parallel considerations apply for a “strong” theory (or some part of it) but in somewhat of an inverse direction: a weak test will more likely confirm, but is not needed since it will not (presumably) affect the theory’s ontological status. A severe test might, and if the test is “successful”, the theory’s status is “strengthened”; although, as mentioned above, this may not be conclusive since the theory has already, on other grounds, been judged to be so.Again, we are suggesting that the reason why NHST may be appropriate or not appropriate to use is not only a function of a previously assumed ontological status, but because such status may be rendered indeterminate given how NHST is utilized. Thus the real issue is whether the correct and appropriate use of NHST supports, strengthens, or weakens the originally held ontological commitment. It seems this would be the case but more needs to be said. What needs to be further analyzed, of course, is how test severity plays out against the backdrop of Type I and II errors, and especially Type II. Before this issue is examined, we wish to mention a distinction that is often overlooked in discussions of NHST. This is whether “severe test” is used as a verb or a noun. While Trout is using it in its verb sense, its bearing as a noun should not be disregarded. That is, the conducting of a severe test is not completely independent of the form of the type of test itself. The underlying

Social Science and (Null) Hypothesis Testing: Some Ontological Issues

193

assumptions of the test (e.g., normal distribution, levels of variable measurement, size and characteristics of the sample, etc.) impact the characterization of the hypothesis and how it plays into the range and types of theory tests and thus ontological commitments. Here, the distinction between a parametric and non-parametric test is relevant. Briefly, parametric tests (e.g., t-tests) require a set of stricter assumptions concerning the parameters of the population from which samples are drawn, such as equal variances, normal distributions and so on. Additionally, it is assumed that relevant variables will be at the interval and ratio levels of measurement. Conversely, non-parametric tests (e.g., Pearson chi-square) do not make equally rigorous assumptions (Blalock 1972; McCall 1970). One important additional distinction between the two is that parametric tests are often favored over non-parametric because of their power efficiency; a concept which will be explored in detail below.However, the point to be made in terms of our concern with NHST and ontological commitments is that the idea of a “severe” test needs to be clarified as it relates to the level of significance chosen versus the type of test (parametric vs. non-parametric) chosen for a given hypothesis. Even with the severity of tests issue outlined above, the real issue of NHST and ontological commitment may lie in the related assumptions of NHST itself, namely power, Type II errors, sampling, and effect size (ES). That is, we are suggesting not that Trout’s emphasis on the ontological is incorrect vis-avis NHST, but that the epistemic nature of NHST must be viewed as a totality first. Doing so, may reveal that NHST is still misused (as he argues), or, perhaps, that it even further clarifies those instances where ontological commitment is most probable.

The Related Concepts The concepts related to the NHST model are basic and rather well known (Cohen 1988). However, we wish to look at them briefly because they are usually not discussed in the context of ontological commitments. Moreover, they are often not considered as a single epistemic system. For example, Trout (S 264) briefly defines a Type II error but does not analyze its significance as a possible constraint to even a legitimate (i.e., ontological/theoretical) application of NHST. The related concepts include Type I and II errors, “power”, sample and effect size (Cohen 1988: 1-17). Type I and II errors must, of course, be balanced against one another and, overall, constitute the most important consideration in the types of eventual ontological commitments made. Thus, a

194

Steven Miller and Marcel Fredericks

“severe” test (e.g., p < .01) will, as compared to a less severe one (e.g., p < .05), reduce the chances of a Type I but increase the probability of a Type II error.However, Type I and II errors are parts of a “family resemblance” in the way the other factors relate to one another. Cohen (1965: 97-101) states these relationships as follows: Four parameters of statistical inference have been described: power, significance criterion (a), sample size (n), and effective size (ES). They are so related that any one of them is a function of the other three, which means that when any three of them are fixed, the fourth is completely determined. By way of mention, effect size (ES) refers to the stipulation (or finding) of a specific nonzero value in the population under study: if a correlation coefficient is of a certain magnitude (e.g., r = .80) then that is the ES; or if a value is assumed for the population (e.g., IQ = 100) and the hypothesized IQ of some particular group is, say, 110, the ES = 10; or if in an experiment, the mean score difference between the experimental and control group is divided by the standard deviation of the control group, the resulting value is the ES. Of course, ES can simply be the statement (as in the null hypothesis) of “no difference”, hence ES = 0. The importance of noting the ES in the context under discussion is not only that it is one specific dimension that needs to be included in the full NHST model, but its value, given the other three, is then a reflection of the “strength” of a particular ontological commitment.Of these four interlocking factors, perhaps the most central to understand is that of “power”. Power is defined by Cohen (1988: 4) as follows: “The power of a statistical test of a null hypothesis is the probability that it will lead to the rejection of the null hypothesis, i.e., the probability that it will result in the conclusion that the phenomenon exists” (italics added). The last portion of the definition is highlighted not only as a means of focusing these issue to our own concern of ontological commitment, but also to note that it is phrased this way by a leading statistician non-philosopher. Indeed, Cohen consistently refers to the interaction of power, alpha, sample, and effect size as a means of establishing the existence (and degree) of a phenomenon. The concept of “power” plays a defining role in the type of ontological status assigned to the phenomenon in the sense that it judges the ability of the testing procedure to reject (or accept) the null hypothesis, and therefore the type of commitment made. A “severe” test, following Trout, is most likely the “power” concept. Thus, a test’s severity is not simply about assigning a given level of “statistical significance” but rather how any assigned level relates to the probability that a severe test is actually that. For example, if the investigator chooses a “high”

Social Science and (Null) Hypothesis Testing: Some Ontological Issues

195

level of power, i.e., .85, she is saying that there is a 85% chance that, given the experimental procedure being utilized, the null hypothesis will be successfully (perhaps truly) rejected 85% of the time. The power concept is especially significant to the NHST approach since the complement of power (1power) turns out to be “B”, the probability of making a Type II error. In general, then, a “high” power value will reduce the probability of a Type II error, e.g., (1-.85=.15), but the catch is that if a stringent (i.e., .001) level of rejection is selected, doing so may reduce the power to .10, hence increase (1.10=.90) rather dramatically the possibility of a Type II error (Cohen 1988: 5).It is important to note, however, that the interactions of the related concepts set threshold effects. That is, through the use of so-called Power Tables (Cohen 1988; Hays 1981; Hsu 1980a, b), one can determine the power of a test given selected parameter values of the other factors: sample size, significance level and effect size. There are a variety of these power tables worked out for a number of statistical tests and approaches. Thus, the NHST procedure, in its entirety, should incorporate power analyses and the other concepts. And it is exactly here that the whole notion of NHST related to the ontological status of a theory or a part of it (i.e., hypotheses) raises additional complexities similar to, but going beyond, those addressed by Trout. Specifically, the basic idea of a “severe test” amounts to more than simply conducting one, and conducting one, for that matter, on “stronger” theories. Under the power analysis construal, the ontological status of a theory consists of pinpointing, if you will, a metaphysically moving target whose very “existence” is a function of the type of power analysis one chooses to undertake.These potentially shifting choices indicate the presence (or degree) of inductive risk but do not tell us much how our theory or hypothesis really is. The issue of the “ontological sensitivity” of tests can be further complicated by such considerations as the “directionality” of our assumptions (i.e., socalled one and two tailed tests) and, for instance, whether several hypotheses are being tested. On the first issue, if a specific null hypothesis with an effect size and specific “direction” is given, its “acceptance” or “rejection” is a function of its being directional, as well as the other power table considerations. Likewise, if several hypotheses are being tested, they will usually be taken as constituting a “set”; however, the probability of making a Type I error may somewhere within the set exceed the risk of any one hypothesis. To handle this situation an adjustment strategy, such as the Bonferroni technique, which utilizes the desired Alpha-level for all hypotheses, is adjusted by dividing it by the number of hypothesis testing procedures used (generally, A/N) (Huck 2000: 220-225). The difficulty is that using such a technique usually raises the standard for rejection, making it more stringent. But, again, such

196

Steven Miller and Marcel Fredericks

“adjustments” may change our ontological commitments - ordinarily rejecting the null hypothesis but not now: the metaphysical moving target has once more shifted. What is at stake in NHST is not only a general problem of induction, but a special problem unique to NHST. The general problem is simply the one that pertains to any inductive inference: given the nature of induction, how are any putative claims of warrantability to be sustained? For NHST, a “grue-like” (Goodman 1995) problem, for instance, could be formulated in many ways, depending on the degree of specificity desired and the addition of what we have called the related concepts. One simple formulation, however, could be as follows: (1) Generalization: “All null hypotheses have a specified % (alpha) level.” (2) “Nill”: “A null hypothesis has a given % (alpha) level if examined at To, but different if it has not been examined.” The use of the term “nill” is adapted from Cohen (1994:1000) where he uses it to describe the tendency to routinely reject the null hypothesis in social science research. He labels this as the “nil hypothesis” (one “l”). We have added one more letter to make it “nill” and closer to “null.”Our grue-like “nill” simply describes the fact that NHST is inherently unstable in specifying a specific ontological commitment. Its instability is reflected by “nill”, where any alpha level can (in principle) be used or “adjusted.” The significance of this fact is not undermined by arguments for “conventional” use (e.g., “by convention, most members of the social scientific community have agreed to use P = .05 …”) or by the very notion of “inductive risk.” Both cases appeal to a particular way of characterizing “epistemic norms,” whether conceived of as either “internal” or “external” to the scientific process itself. That is, the act of choosing an alpha level has no independent privileged justificatory status. Put differently, the standard research textbook arguments for choosing and sticking to a predetermined alpha level (i.e., the so-called “decision rules” for “accepting” or “rejecting” the null), for example, are nothing more than an agreement to play by these rules, as some type of tacit agreement that doing so is the “rational” thing to do (Sainsbury: 91-93).But it could be otherwise, as “nill” illustrates. What if we decide to change the rules in the “middle of the game”; i.e., we declare alpha = .01 but then switch to .05 in analyzing the data? The argument that one should not do so because the rule of sticking to your original choice is (to use Goodman’s term) the better “entrenched” does not have much prima facie merit since “nill” could also be argued from a variety of “possible worlds” perspectives equally well (Lewis, 1986; Plantinga, 1993). Likewise, if an instrumentalist or consequentialist premise is added to the above – e.g., “If we switch alpha levels the conse-

Social Science and (Null) Hypothesis Testing: Some Ontological Issues

197

quences may be drastic in such areas as medical research or giving support to weak theories” – this may be a way of justifying certain actions in terms of their inductive risks, but such assessments in no way justify the underlying structure of NHST itself, especially under “nill.” Thus all of these concerns do not give us clear criteria for demarcating the ontological issue, because NHST (whether the theory and/or hypothesis is “strong” or “weak”) permits us to make conflicting commitments depending on how we apply the rules. Under certain conditions (theory-based) reality - whether “strong” or “weak” - is (presumably) accessible, while under others, where the null hypothesis is not rejected, that same reality is not accessible.

NHST as Event-Markers Trout’s claim that test severity is not simply an assignment of statistical significance but a function of certain theoretical commitments is well founded. However, as these further aspects of the NHST model have shown, what is (or can be) said about such theoretical commitments is at least partially constrained by how we choose to say whatever we decide ought to be said. The ambiguities in NHST produce uncertainty in how theories ought to be viewed under realist assumptions. Of course, they are in some “weak” or “strong” sense present, but NHST has shown us that this may differ depending on construal. One way to overcome such ontological shifting is to view NHST in a different way. The advantage being we still retain realist ideas about our theories, but override the undesirable effects of NHST, i.e., in terms of those possibilities where changing levels of test severity determine accessibilityThe NHST issue may be recast as one related to the ontology of events. There is a rich literature on events (e.g., Laurence and Macdonald, eds., 1998), which, while controversial on several issues (e.g., nature of “properties”), does have a solid conceptual grounding. Events (like theories) admit of no one easy definition because of varying accounts on what constitutes their actual “makeup.” Thus, Chisolm’s (1976) early view holds that events are states-of-affairs that can be characterized as abstract universals, meaning that more than one event can occur at the same time to one individual (abstract), and because they can recur, they are universal. Others hold events to be species of “particulars”. On Quine’s (1960) classic view, events are concrete particulars whereby they are similar to physical objects, non-repeatable entities. Events are also characterized as abstract particulars, a view broadly held by Kim (1999) and Davidson (1969) but who differ on other dimensions. Generally, under an abstract particular construal, any given event can only

198

Steven Miller and Marcel Fredericks

take place in one place at the same time, but more than one event is possible of occurring in one place, simultaneously. For our purposes, Kim’s general outline of what it means to be an event is most useful. Roughly, Kim holds that events are composed of objects which exemplify some type of property in a given time period. Events are not only about objects, in the sense of being constituted by them, but they (events) are the objects of explanation also, i.e., events as “units” or “wholes” necessitate their own explanation. This characterization of events as “objects of explanation” will figure into our NHST analysis. According to Kim, an event, then, is “some object’s O’s coming to have (or having) a property P at a time, t” (Lombard, 1999: 284). Moreover, events are identical only if they are the exemplification(s) of the same property(ies).Lombard’s (1999: 289-290) take on the central role of “exemplification” in defining events is related to Kim’s but with some important distinctions. For example, Lombard presents a straightforward definition – “… events [are] the changes that objects undergo when they change” (289). And, “an object changes just in case it goes from the having of one property to the having of another, contrary property” (289). The emphasis on how properties are exemplified is important in his analysis. For Lombard, property exemplification is a dynamic not a static process. Under Kim’s construal, “exemplification” is the having of a property that results in an event. This is contrasted with a dynamic characterization of a property as “exemplifying” an event. The object in exemplifying the property does so as the consequence of undergoing some type of change. This very brief analysis of event-ontology has not dealt with on-going issues such as whether “events” are different from “propositions” (Loux, 2002), or if two events which are claimed to be “identical” are so in virtue of having the same cause and effect structure (Davidson, 1980). The intent here has only been to suggest in a preliminary way that event-ontology may be a useful and different way to begin to think about NHST and ontological commitment. Some parallels will be briefly discussed below.As we have been discussing it, the drawback of the NHST model lies not exclusively in its inappropriate applications vis-a-vis strong and weak theories, but rather on the types of counterintuitive results it produces. The ontological consequences are that a given hypothesis can both assert and deny a “reality” (i.e., the accessibility to the phenomenon) as a function of how NHST thinking is utilized. For example, as Trout rightfully points out, for weaker theories in the social sciences “severe tests” are often not conducted because the null is more likely to be accepted. Conducting less severe tests may result in more rejections of the null but with the price being the theory remains weak.

Social Science and (Null) Hypothesis Testing: Some Ontological Issues

199

What does the event-ontology position have to offer? To begin with, our focus shifts from worrying about the ontological status of theories to a characterization of them as “event-states.” That is, whatever we are testing at a given time is a type of event related to a theory. This recognizes that a theory (or a part of one) is a set of event-states which at a given point in time represents the outcome of one test. More precisely, a given theory may be at one point in time capable of many possible tests; tests possibly producing dissimilar results, and hence, different event-states. Or, a given event state at t1 changes with a further test at t2. Thus, each test, severe or not, represents an event-status in some mode of “presentness”. Event-states thus depicted are also consistent with the definition of events as “abstract particulars.” Importantly, however, the NHST model, whether in its basic “null form” or more extensive “power” sense, is co-extensive with the exemplification (or exemplifying) idea of event-ontology. But it also goes beyond traditional notions of property exemplification because it specifies under what (exact) conditions the property is or is not exemplified. NHST, then, is an exemplification process operating on (or instantiating) object/property characteristics of event-states. Put differently, NHST procedures “permit” properties to exemplify themselves. Without this function, a theory would constitute a single event. NHST has, then, a constitutive function whereby the properties (and their relations) in a theory are constituted as unique events. When a null hypothesis is “rejected” at a certain statistical level, that is an event-state. When it is accepted, that is an event state. Moreover, the inclusion of the related concepts (i.e., power, sample, effect size, significance level) instantiates event-states in more complex ways. Indeed, a full Power Table gives the total number of possible event-states for whatever property(ies) and relations are hypothesized. Also, there are no covert Idealist assumptions here; NHST is simply the process that gives the probabilities of an hypothesis - the event-states. NHST is, furthermore, not a causal process, since it does not comment directly on how the properties to be exemplified by an event have come about. It simply “comments” on what sense and (perhaps) to what degree these properties are (or are not) constitutive of the event. Theories, then, are events, some of which are “simple” and some of which are complex. The NHST controversy on this construal is not so much about how our theories/hypotheses are (possible), but more about how they are. The idea of events is compatible with minimal or Trout’s Measured Realism, but it does not commit us to defending (using NHST) a realism which is “shifting” or “constrained” or with the possibility of non-accessibility. Again, the severity of a test is simply a way of characterizing an event in a particular way. Some “events” are “stronger”, “truer”, since they have been tested more

200

Steven Miller and Marcel Fredericks

“severely”; but these things do not imply that there is a one-sided commitment to methodological (rather than ontological) issues in the use (or misuse) of NHST. In fact, under an event construal NHST can never be deliberately misused; although accidental misuse is certainly a possibility.By this we mean, for a given hypothesis as an event-state, under a full power analysis, the full set of events can be specified. If one of those event-states is chosen because it is amendable to a severe test-based, let us say, on the knowledge that this type of event state has previously determined well established properties - then such a test ought to be conducted. If for whatever reason a less severe test is chosen, that event is likewise describable. It may turn out not to be a prudent choice, but the choice is simply that: it does not affect the ontological status. We should also not think about the ontological landscape as strewn with the multiplication of unwanted and unneeded entities. We still have one theory, under multiple possible characterizations. Or, alternatively, in thinking of events as abstract particulars, any one given event, for a time and place, can be objectively examined as that event, while not precluding the objective examination of other versions of the event “occupying” that same eventspace at the same time. Events, then, are “real” but simply under different characterizations. NHST gives us these possibilities without “causing” them by way of determining their “strength” or “weakness.” If nothing else, the NHST issue is important for the social sciences not because of its instrumental value alone, but in how it alerts social scientists to view their efforts not only epistemically but also ontologically.

Bibliography Blalock, H. M. 1972. Social Statistics. New York: McGraw-Hill. Chisolm, R. 1976. Person and Object. LaSalle, IL: Open Court Publishing. Cohen, J. 1988. Statistical Power Analysis for the Behavioral Sciences. Hillsdale, NJ: Lawrence Erlbaum Associates. —. 1994. “The Earth Is Round (p < .05).” Educational Psychologist 49, 12: 997-1003. Cohen, J., and J. S. Hyman. 1979. “How Come So Many Hypotheses in Educational Research are Supported?” (A modest proposal). Educational Researcher 8: 12-16. Davidson, D. 1980. “The Individuation of Events.” In Essays on Actions and Events, 216-34. Oxford: Oxford University Press. Devitt, M. 1997. Realism and Truth (2nd ed). Princeton, NJ: Princeton University Press.

Social Science and (Null) Hypothesis Testing: Some Ontological Issues

201

Douglas, H. 2000. “Inductive Risk and Values in Science.” Philosophy of Science 67: 559-579. Goodman, N. 1955. Fact, Fiction and Forecast. Cambridge, MA: Harvard University Press. Hays, W. C. 1981. Statistics (3rd ed). New York: Holt-Saunders. Hsu, L. M. 1980. “On the Power of Multiple Independent Tests When the Experiment-Wise Error Rate is Controlled.” Educational and Psychological Measurement 40: 31-40. Huck, S. W. 2000. Reading Statistics and Research (3rd ed). New York: Longman. Kim, J. 1976. “Events as Property Exemplifications.” In Action Theory. Edited by M. Brand and D. Walton, 159-77. Dordrecht, Holland: D. Reidel Publishing Co. Laurence, S., and C. Macdonald, eds. 1998. Contemporary Readings in the Foundations of Metaphysics. Oxford: Blackwell. Lombard, B. 1998. “’Ontologies of Events’: State of the Art Essay.” In Contemporary Readings in the Foundations of Metaphysics. Edited by S. Laurence and C. Macdonald, 277-94. Oxford: Blackwell. Loux, M. J. 2002. Metaphysics: A contemporary introduction (2nd ed.). New York: Routledge. Lynch, M. P. 1998. Truth in Context: An Essay on Pluralism and Objectivity. Cambridge, MA: The MIT Press. Mayo, D. G. 1997. “Duhem’s Problem, the Bayesian Way and Error Statistics, or ‘What’s Belief Got To Do With IT’.” Philosophy of Science 64: 222-44. —. 1991. “Novel Evidence and Severe Tests.” Philosophy of Science 58: 523-52. McCall, R. B. 1970. Fundamental Statistics for Psychology. New York: Harcourt, Brace and World. Plantinga, A. 1993. Warrant and Proper Function. Oxford: Oxford University Press. Popper, K. 1962. Conjecture and Refutations. New York: Basic Books. Quine, W. v. O. 1960. Word and Object. Cambridge, MA: MIT Press. Trout, J. D. 1998. “Measured Realism and Statistical Inference: An Explanation for the Fast Progress of ‘Hard’ Psychology.” Philosophy of Science (Proceedings of the 1998 Biennial Meeting of the Philosophy of Science Association) 66, 3: 260-72. —. 1998. Measuring the Intentional World: Realism, Naturalism, and Quantitative Methods in the Behavioral Sciences. New York: Oxford University Press.

202

Published Volumes

CONTRIBUTORS Emma Borg, Philosophy Department, University of Reading, Reading RG6 6AA, Great Britain. Eros Corazza, Department of Philosophy, University of Nottingham, University Park, Nottingham, NG7 2RD, Great Britain. Marcel Fredericks, Department of Sociology and Anthropology Loyola University Chicago, Damen Hall-934, Lake Shore Campus 6525, N. Sheridan Road, Chicago,Il. 60626, USA. Max Freund, Departmento de Filosofia, Universidad Nacional de Costa Rice, Apdo 86-3000, Heredia, Costa Rica. Sanford Goldberg, Department of Philosophy, College of Arts and Science, 1415 Patterson Office Tower, Lexington, KY 40506-0027, USA. Steven Gross, Department of Philosophy, University of Pennsylvania, 433 Logan Hall, Philadelphia, PA 19104, USA. David Hunter, Philosophy, Buffalo State College, Suny, 1300 Elmwood Avenue, Buffalo, NY 11422, USA. Hans Lenk, Universität Karlsruhe, Fakultät für Geistes und Sozialwissenschaften, Institut für Philosophie, 76128 Karlsruhe. Steven Miller, Loyola University. Department LFCD, School of Education, Mallinckroth Campus, 1041 N. Ridge Road, Wilmetta IL. 60091, USA. Cara Spencer, Philosophy Department, 226 Locke Hall, Howard University, Washington, DC 20059, USA. Jonathan Sutton, Philosophy Department, Southern Methodist University, PO Box 750142 Dallas, TX 75275_0142, USA. Corey Washington, Department of Philosophy, University of Maryland, 1123 Skinner Bldg. College Park, MD 20742-7615, USA.

Published Volumes

203

IMPRINT ProtoSociology: An International Journal of Interdisciplinary Research ISSN 14344319 Editor: Gerhard Preyer Johann Wolfgang Goethe-Universität Frankfurt am Main, Dep. of Social Sciences Editorial Staff: Georg Peter Editorial of the Vol. 17 2002: Georg Peter Layout and digital publication: Georg Peter Editorial Office: ProtoSociology, Stephan-Heise-Str. 56, 60488 Frankfurt am Main, Germany, Phone: (049)069-769461, Email: [email protected], [email protected] Bank: Dresdner Bank AG, Frankfurt am Main, account: 44 121 168 01, BLZ: 500 800 00 – SWIFT-BIC: DRES DE FF IBAN DE60 5008 0000 4412 1168 01 The Vol. 17 is sponsored by GASS INTERNATIONAL, Vorlandstr. 4, D-77756-Hausach, www.gass-logistics.com )))))))))))))))))))))))))))))))))))))))))) Die Zeitschrift soll 1/2jährlich erscheinen. Die Anzahl der jährlich erscheinenden Hefte und Sonderhefte bleibt jedoch vorbehalten. Copyright: Die in dieser Zeitschrift veröffentlichten Beiträge sind urheberrechtlich geschützt. Alle Rechte sind vorbehalten. Übersetzungen, Nachdruck, Vervielfältigung auf fotomechanischem oder ähnlichem Weg oder im Magnettonverfahren, Wiedergabe durch Vortrag, Funk- und Fernsehsendungen sowie Speicherung in Datenverarbeitungsanlagen, auch auszugsweise, sind nur mit Genehmigung des Herausgebers möglich. Für eingereichte Beiträge wird keine Haftung übernommen. Weitere Publikationsrechte von Artikeln bleiben vorbehalten. Zweitpublikationen seitens des Autors werden eingeräumt. Bei einer Zweitpublikation ist das Heft (Nummer, Titel, Erscheinungsjahr) der PROTOSOCIOLOGY zu zitieren. Für unaufgefordert eingesandte Manuskripte wird keine Haftung übernommen. Gerichtsstand ist Frankfurt am Main. Copyright: All rights reserved. This publication may not be reproduced, stored or transmitted in any form or by any means without the prior permission in writing of the publisher. Additional publications of the articles are reserved. The authors retain the personal right to re-use their own articles. Authorization to photocopy items for internal or personal use, or the internal or personal use of specific clients is garanted by PROTOSOCIOLOGY, provided that the base fee is paid directly to VG Wort, Goethestr. 49, 80336 München RFA.The publisher accepts no responsibility for submitted manuscripts. ProtoSociology im Internet: www.protosociology.de

204

Published Volumes

ON PROTOSOCIOLOGY Protosociology occupies an important position in the European intellectual scene, bridging philosophy, economics, sociology and related disciplines. Its volumes on rationality bring together concerns in all these topics, and present an important challenge to the cognitive sciences. Donald Davidson, Berkeley (USA) Protosociology publishes original papers of great interest that deal with fundamental issues in the human and social science. No academic library is complete without it. Nicholas Rescher, Pittsburgh (USA) Protosociology has been remarkably successful in publishing interesting work from different tradition and different disciplines and, as the title signals, in giving that work a new, eye-catching slant. Philipp Pettit, Canberra, Australia Protosociology is a truly premier interdisciplinary journal that publishes articles and reviews on timely topics written by and for a wide range of international scholars. The recent volumes on rationality are remarkable for their breadth and depth. Protosociology would be a great addition to any library. Roger Gibson, St. Louis (USA

Published Volumes

205

PROTOSOCIOLOGY An International Journal of Interdisciplinary Research

Vol. 16, 2002 Understanding the Social: New Perspectives from Epistemology

CONTENTS FROM MENTALIZING FOLK SOCIAL EPISTEMOLOGY

TO

GROUP RESPONSIBILITY AND THE CONSENSUS ACCOUNT OF JUSTIFIED BELIEF

Alvin I. Goldman The Mentalizing Folk

Pekka Mäkelä and Raimo Tuomela Group Action and Group Responsibility

Margaret Gilbert Believe and Acceptance as Features of Groups

Lamber Royakkers Collective Commitments: a Theoretical Understanding of Human Cooperation

Antonie Meijers Collective Agents and Cognitive Attitudes

Fred Schmitt Justification and Consensus: The Peircean Approach

Deborah Perron Tollefsen Challenging Epistemic Individualism

ON CONTEMPORARY PHILOSOPHY EXTERNALISM, EVENTS AND INSTITUTIONAL FACTS Gerhard Preyer From an Externalistic Point of View: Understanding the Social Steven Miller Are „Context“ and „Event“ Equivalent? Possibilities for an Ontological Symbiosis Frank Hindriks Institutional Facts and the Naturalistic Fallacy: Confronting Searle (1964) with Searle (1995)

Roger F. Gibson How I Came to Know Quine: A Reminiscence Hans Lenk and Matthias Maring Responsibility and Globalization

206

Published Volumes

PROTOSOCIOLOGY An International Journal of Interdisciplinary Research Vol. 15, 2001 On a Sociology of Borderlines: Social Process in Time of Globalization Gerhard Preyer, Mathias Bös (eds.) CONTENTS Introduction: Borderlines in Time of Globalization (Gerhard Preyer, Mathias Boes) I RECONCEPTIONALIZATIONS OF THE GLOBAL: BORDERLINES IN WORLD SOCIETY Shmuel Noah Eisenstadt The Continual Reconstruction of Multiple Modern Civilizations and Collective Identities Christopher Chase-Dunn: Globalization: A World-Systems Perspective Thomas D. Hall: World-Systems, Frontiers, and Ethnogenesis: Incorporation and Resistance to State Expansion Richard E. Lee: After History? The Last Frontier of Historical Capitalism II DEFINING BORDERLINES IN WORLD SOCIETY: THE EMERGENCE OF NEW MEMBERSHIPS Gerhard Preyer: Globalization and the Evolution of Membership Barrie Axford: Enacting Globalization: Transnational Networks and the Deterritorialization of Social Relationships in the Global System Mathias Bös Immigration and the Open Society: The Normative Patterns of Membership in the Nation State Uta Gerhardt, Birgitta Hohenester: A Transformation of National Identity? Refugees and German Society After World War II Marja Keränen: Citizenship, UniversalSpeak, and Local-Speak

III THE GLOBAL AND THE LOCAL: THE COLLAPSE AND RECONSTRUCTION OF BORDERLINES Christie Davies, Eugene Trivizas The Collapse of the National Morality and National Moral Boundaries of Small Peripheral Countries: not Globalisation but the Imposition of Liberty Walter L. Bühl: Former GDR between “Transformation” and “Social Evolution” F. Peter Wagner: Beyond “East” and “West”: On the European and Global Dimensions of the Fall of Communism Ramón Grosfoguel ‘Cultural Racism’ and ‘Borders of Exclusion’ in the Capitalist World-Economy: Colonial Caribbean Migrants in Core Zones Francisco Entrena: Socio-Economic Restructurings of the Local Settings in the Era of Globalization ON A SOCIOLOGY OF VIOLENCE Konrad Thomas Ein anderes Verständnis von Gewalt: Der gesellschaftsanalytische Beitrag des Literaturwissenschaftlers René Girard CRITICAL REVIEW: ProtoSciology, Vol. 12, 1998-Special Edition: After the Received View. Developments in the Theory of Science George N. Schlesinger

Published Volumes

207

PROTOSOCIOLOGY An International Journal of Interdisciplinary Research

Vol. 14, 2000 Folk Psychology, Mental Concepts and the Ascription of Attitudes: On Contemporary Philosophy of Mind

CONTENTS Alvin I. Goldman Folk Psychology and Mental Concepts Philip Pettit How the Folk Understand Folk Psychology Jane Heal Understanding Other Minds from the Inside Christopher S. Hill From Assertion to Belief: The Role of Linguistic Data in the Practice of Belief-Ascription David Rosenthal Content, Interpretation, and Consciousness Jay L. Garfield Thought as Language: A Metaphor Too Far Robert M. Gordon Sellars’s Ryleans Revisited Louise Röska-Hardy Self-Ascription and Simulation Theory Brian P. McLaughlin Why Intentional Systems Theory Cannot Reconcile Physicalism With Realism about Belief and Desire Gerhard Preyer Primary Reasons: From Radical Interpretation to a Pure Anomalism of the Mental

Rebecca Kukla How to Get an Interpretivist Committed David Pitt Nativism and the Theory of Content Raffaella De Rosa On Fodor’s Claim that Classical Empiricists and Rationalists Agree on the Innateness of Ideas Consuelo Preti Belief and Desire Under The Elms Erwin Rogler On David Lewis’ Philosophy of Mind Barbara Von Eckardt, Jeffrey S. Poland In Defense of the Standard View

ON CONTEMPORARY PHILOSOPHY OF LANGUAGE, ONTOLOGY AND MORAL THEORY Thomas Baldwin Nearly Logic James E. Tomberlin Logical Form, Actualism, and Ontology Arend Kulenkampff, Frank Siebelt What a Noncognitivist might tell a Moral Realist

www.protosociology.de

208

Published Volumes

PROTOSOCIOLOGY An International Journal of Interdisciplinary Research Vol. 13 1999: Reasoning and Argumentation D. Mans, G. Preyer (eds.) CONTENTS Ralph H. Johnson: Reasoning, Argumentation and The Network Problem Leo Groarke: The Fox and the Hedgehog: On Logic, Argument, and Argumentation Theory J. Anthony Blair: Presumptive Reasoning/Argument: An Overlooked Class Robert C. Pinto: Argument Schemes and the Evaluation of Presumptive Reasoning: some Reflections on Blair’s Account Douglas Walton: The New Dialectic: A Method of Evaluating an Argument Used for Some Purpose in a Given Case Manfred Kienpointner: Comments on Douglas Walton’s Paper Christopher W. Tindale: The Authority of Testimony John Woods: Peirce’s Abductive Enthusiasms Henry W. Johnstone, Jr.: “‘Any,’ ‘Every,’ and the Philosophical Argumentum ad Hominem” ON CONTEMPORARY PHILOSOPHY Hans Lenk Interdisziplinarität und Interpretation Ellery Eells Causal Decision Theory

www.protosociology.de

209

Published Volumes

PROTOSOCIOLOGY An International Journal of Interdisciplinary Research VOL. 12, 1998 – Special Edition AFTER THE RECEIVED VIEW – Developments in the Theory of Science Gerhard Preyer, Georg Peter, Alexander Ulfig (Eds.) in memoriam Wolfgang Stegmüller CONTENTS Introduction: Developments in the Theory of Science (Gerhard Preyer, Georg Peter, Alexander Ulfig) LOGICAL OPERATIONALISM – SIGNIFICANCE AND MEANING Wilhelm K. Essler: Truth and Knowledge. Some Considerations concerning the Task of Philosophy of Science Gerhard Preyer: The Received View, Incommensurability and Comparison of Theories – Beliefs as the Basis of Theorizing Robert Schwartz: Reflections on Projection Jeffrey E. Foss: The Logical and Sociological Structure of Science STRUCTURALISM – MEANINGFUL MEASUREMENT – THE CONCEPTION OF PHYSICAL LAW C. Ulises Moulines: Structuralism vs. Operationalism Nicholas Rescher: Meaningless Numbers R. I. G. Hughes: Laws of Nature, Laws of Physics, and the Representational Account of Theories James R. Brown: Einstein’s Principle Theory INDUCTIVE INFERENCES – INTERPRETATION OF PROBABILITY – GAME THEORY Kevin T. Kelly, Cory Juhl Transcendental Deductions and Universal Architectures for Inductive Inferences

Howard H. Harriott: R.A. Fisher and the Interpretation of Probability Brian Skyrms: Evolution of an Anomaly PROPERTIES – UNDERDETERMINATION – SCIENTIFIC REALISM George N. Schlesinger: Degrees of Characterizations Carl A. Matheson: Observational Adequacy as distinct from the Truth about Observables Thomas R. Grimes: Scientific Realism and the Problem of Underdetermination Paul C. L. Tang: On Paul Churchland’s Treatment of the Argument from Introspection and Scientific Realism RATIONALITY – METAPHORS – VALUES IN SCIENCE David Resnik: Scientific Rationality and Epistemic Goals Aldo Montesano: Rationality in Economics: A General Framework Joseph Agassi: Science Real and Ideal: Popper and the Dogmatic Scientist Michael Bradie: Models and Metaphors in Science David Gruender: Values and the Philosophy of Science

210

Published Volumes

PROTOSOCIOLOGY An International Journal of Interdisciplinary Research

Vol. 11, 1998 Cognitive Semantics II – Externalism in Debate CONTENTS RADICAL INTERPRETATION, ONTOLOGY AND THE STRUCTURE OF LANGUAGE Gerhard Preyer, Michael Roth: On Donald Davidson’s Philosophy: An Outline Richard Manning: All Facts Great and Small Barbara Fultner: Of Parts and Wholes: The Molecularist Critique of Semantic Holism Louis Goble: Re-Evaluating Supervaluations David Simpson: Interpretation and Skill: On Passing Theory Wulf Kellerwessel: Katz on Semantics and Pragmatics

EXTERNALISM

AND THE INDIVIDUATION OF

CONTENT Ron Wilburn: Knowledge, Content, and the Wellstrings of Objectivity Anthony Brueckner: Content Externalism and A Priori Knowledge Consuelo Preti: The Irrelevance of Supervenience Michael Liston: Externalist Determinants of Reference Arnold Silverberg: Semantic Externalism. A Response to Chomsky Gerhard Preyer: Interpretation and Rationality: Steps from Radical Interpretation to the Externalism of Triangulation

www.protosociology.de

Published Volumes

211

PROTOSOCIOLOGY An International Journal of Interdisciplinary Research

Vol. 10, 1997 Cognitive Semantics I – Conceptions of Meaning

CONTENTS MEANING, TRUTH AND THE USE OF LANGUAGE Herman Cappelen, Ernie Lepore: Semantic Theory and Indirect Speech Kirk Ludwig: The Truth about Moods Louise Röska-Hardy: Language, Use and Action Jeffrey King: The Source(s) of Necessity Filip Buekens: The Genesis of Meaning (a Myth) REFERENCE, INDEXICALS AND SPEAKER MEANING Robert Hanna: Extending Direct Reference Peter Ludlow: Semantics, Tense, and Time: a Note on Tenseless Truth-Conditions for Token-Reflexive Tensed Sentences Gerhard Preyer: Verstehen, Referenz, Wahrheit. Zur Philosophie Hilary Putnams

Reinaldo Elugardo: Descriptions, Indexicals and Speaker Meaning RATIONAL DIALETISM OF ANTINOMIES Mark Sainsbury: Can Rational Dialetheism Be Refuted By Considerations about Negation and Denial? CONTEMPORARY DEBATES IN THE PHILOSOPHY OF MIND Klaus Sachs-Hombach: Philosophy of Mind: Die Simulationstheorie IN RETROSPECT Joseph Agassi: Wittgenstein – The End of a Myth Kent A. Peacock, Richard Feist: The Einstein-DeSitter Controversy

www.protosociology.de

212

Published Volumes Forthcoming

PROTOSOCIOLOGY An International Journal of Interdisciplinary Research Vol. 18, 2003 Indirect and Reported Speech CONTRIBUTORS (WITH

TENTATIVE TITLES)

1. Paul M. Pietroski: Meaning before Truth, Character before Context 2. Jason Stanley: Relevant Alternatives Theories: A Comparison 3. Peter Pagin: Compositionality and Context 4. Zoltan Szabo: Coming and Knowing 5. Michael Glanzberg: Truth Values, Presuppositions, and Expressing Propositions 6. Jonathan Schaffer: The Locus of Epistemic Shiftiness: Context, Standard, and Alternatives 7. Peter Ludlow: Five Grades of Contextual Involvement 8. Ernie Lepore ... 9. John Hawthorne: Is the Verb 'know' context dependent? Some Syntactic and Semantic Considerations 10. Kent Bach: The Semantic Incoherence of Contextualism

Bookpublications

LOGICAL FORM

AND

213

LANGUAGE

Gerhard Preyer, Georg Peter (eds.) To Donald Davison

This volume brings together some of the world’s most distinguished linguistically-minded philosophers and philosophically-minded linguists in an outstanding collection of new papers on what logical form is supposed to be and why it matters for the study of language. - Zoltan Szabo, Cornell University

CONTENTS Gerhard Preyer, Georg Peter: Introduction PART I: THE NATURE

LOGICALFORM

Stephen Neale Logically Unformed

Norbert Hornstein A Grammatical Argument for A Davidsonian Semantics

Ernie Lepore, Kirk Ludwig What is Logical Form

Jason Stanley Nominal Restriction

OF

Paul M. Pietroski Function and Concatenation Jeffrey King Two Sorts of Claim about “Logical Form” Peter Ludlow LF and Natural Logic. The Syntax of Directional Entailing Environments Robert May/R.F. Fiengo Identity Statements PART II: INTENSIONALITY, EVENTS SEMANTIC CONTENT

AND

James Higginbotham Why is Sequence of Tens Obligatory? Richard Larson The Grammar of Intensionality Barry Schein Events and the Semantic Content of Thematic Relations

PART III: LOGICAL FORM, BELIEF ASCRIPTION AND PROPER NAMES Bernard Linsky Russell’s Logical Form, LF and Truth Conditions Lenny Clapp, Robert Stainton “Obviously Propositions are Nothing.” Russell and the Logical Form of Belief Reports Robert Matthews Logical Form and the Relational Conception of Belief Marga Reimer Ordinary Proper Names Reinaldo Elugardo The Predicate View of Proper Names

Oxford University Press, Oxford 2002 www.protosociology.de/Books.htm

214

Bookpublications

The Contextualization of Rationality Gerhard Preyer, Georg Peter (eds.)

Introduction: Problems, Concepts and Theories of Rationality

RADICAL INTERPRETATION, NATURALIZED EPISTEMOLOGY

AND

NORMATIVITY

John Heil The Propositional Attitudes

Roger F. Gibson Stich on Intentionality and Rationality

David K. Henderson Epistemic Rationality, Epistemic Motivation, and Interpretive Charity

Paul K. Moser and David Yandell Against Naturalizing Rationality Harvey Siegel Naturalism, Instrumental Rationality, and the Normativity of Epistemology

Gerhard Preyer Interpretation and Rationality: Steps from Radical Interpretation to the Externalism of Triangulation INTENTIONS

AND THE

SOCIAL ASPECT

OF

RATIONALITY

Alfred R. Mele Rational Intentions and the Toxin Puzzle

Raimo Tuomela Rational Cooperation and Collective Goals

Peter Gärdenfors The Social Stance

Peter French Rationality and Ethics

Timo Airaksinen and Katri Kaalikoski Instrumental Rationality

Julian Nida-Rümelin The Plurality of Good Reasons and the Theory of Practical Rationality

CONCEPTS

OF

EXPLANATION, JUSTIFICATION

AND

REALITY

Philip Pettit Three Aspects of Rational Explanation

Alexander Ulfig Validity, Justification and Rationality

Keith Lehrer Rationality and Trustworthiness

Nicholas Rescher Reason and Reality

250 Seiten mentis Verlag, Paderborn 2000

215

Bookpublications

Analytische Ästhetik Eine Untersuchung zu Nelson Goodman und zur literarischen Parodie Georg Peter I. ANALYTISCHE ÄSTHETIK:

DIE

ZEICHENTHEORIE NELSON GOODMANS

1. Eine ‚andere‘ Fragestellung / 2. Zu Nelson Goodman in allgemeiner Absicht / 3. Zeichenbedeutung: Sprachliche und nichtsprachliche Denotation / 4. Von der Probe zum passenden Etikett: Exemplifikation und Ausdruck / 5. Die Nebenbeschäftigung der Symbole: Ein bedeutungstheoretischer Exkurs zu Metapher und Wahrheit / 6. Ästhetische Bedeutung und kognitive Funktion: Das Kunstwerk als Probe / 7. ‚Wann‘ oder doch ‚Was ist Kunst?‘ II. SCHEMAINTERPRETATION

UND DIE IDENTIFIZIERBARKEIT ÄSTHETISCHER

ZEICHEN

1. Schemainterpretationismus und Symboltheorie: Vom Symbol zum Schema / 2. Der konstruktive Schemainterpretationismus von Hans Lenk / 3. Hermeneutik und Schemainterpretation / 4. Die soziale Verankerung der Interpretation: Ästhetische Kompetenz und Kultursoziologie / 5. Thesenhaftes zu einer schemaorientierten Symboltheorie: Wann ist (es) Parodie III. ROBERT NEUMANNS PARODIEN: DIE PROBE

AUF DIE

EXEMPLIFIKATION

1. Literaturwissenschaft und Parodie: eine vorbelastete Beziehung / 2. Die Parodie als eine Form des Komischen / 3. Das Problem der Wertung: Die kritische und die (bloß) komische Parodie / 4. Nachahmung, Bezugnahme und Robert Neumanns Theorie der Parodie / 5. Von der Funktion der Zeichen zur Funktion der Parodie: Goethe und Brecht parodiert / 6. Komplexität der Parodie und die komplexen Parodien Robert Neumanns / 7. Zeigen und Sagen: Funktion und Qualität der Parodie anhand einer Benn-Parodie / 8. Die Grenzen der Parodie und der Parodierbarkeit / 9. Parodien und andere Formen der Nachahmung IV. ZUR METHODIK

DES

ANSATZES

UND SEINER

WEITERFÜHRUNG

ca. 330 Seiten, Erschienen in der Reihe: PHILOSOPHISCHE ANALYSE / PHILOSOPHICAL ANALYSIS, Herausgegeben von / Edited by Herbert Hochberg · Rafael Hüntelmann et al., Deutsche Bibliothek der Wissenschaften, Dr. Hänsel-Hohenhausen AG, Frankfurt a.M., München, New York 2002. www.protosociology.de/Books.htm

216

Bookpublications

Borderlines in a Globalized World New Perspectives in a Sociology of the World System Gerhard Preyer, Mathias Bös (eds.) To Walter L. Bühl Introduction Gerhard Preyer, Mathias Bös Borderlines in Time of Globalization: New Theoretical Perspectives

I RECONCEPTIONALIZATIONS

OF THE

GLOBAL: BORDERLINES

IN

WORLD SOCIETY

Shmuel N. Eisenstadt: The Continual Reconstruction of Multiple Modern Civilizations and Collective Identities Christopher Chase Dunn: Globalization: A World-Systems Perspective Thomas D. Hall: World-Systems, Frontiers, and Ethnogenesis: Incorporation and Resistance to State Expansion. Richard Lee: After History? The Last Frontier of Historical Capitalism

II DEFINING BORDERLINES

IN

WORLD SOCIETY: THE EMERGENCE

OF

NEW MEMBERSHIPS

Gerhard Preyer: Globalization and the Evolution of Membership Barrie Axford: Enacting Globalization - Transnational Networks and the Deterritorialisation of Social Relationship Mathias Bös: Immigration and the Open Society: The Normative Patterns of Membership in the Nation State Uta Gerhardt, Birgitta Hohenester: A Transformation of National Identity? Refugees and German Society after World War II

III. THE GLOBAL

AND THE

LOCAL : THE COLLAPSE

AND

RECONSTRUCTION

OF

BORDERLINES

Christie Davies, Evgenios Trivizas: The Collapse of the National Morality and National Moral Boundaries of Small Peripherial Countries not Globalization but the Imposition of Liberty F. Peter Wagner: Beyond “East” and “West”: On the European and Global Dimensions of the Fall of Communism Francisco Entrena:Socioeconomic Restructurings of the Local Settings in the Era of Globalization Kluwer Academic Publishers Social Indicators Research Series Vol. 9 Dordrecht, Netherlands

Bookpublications

CONCEPTS

OF

MEANING

Framing an Integrated Theory of Linguistic Behavior ed. by Gerhard Preyer, Georg Peter, Maria Ulkan Introduction Gerhard Preyer, Georg Peter, Maria Ulkan I Speaker Meaning, Communication, and Intentions 1 Communicative and Illocutionary Acts Maria Ulkan 2 Language Acts and Action Louise Röska-Hardy 3 Reflections on the Intentionality of Linguistic Behavior Jan Nuyts 4 Descriptions, Indexicals, and Speaker Meaning Reinaldo Elugardo 5 Informatives and/or Directives? (A New Start in Speech Act Classification) Georg Meggle, Maria Ulkan 6 Constructive Speech-Act Theory Dirk Hartmann II Truth, Semantic Content, and Externalism 7 The Truth about Moods Kirk Ludwig 8 Semantic Theory and Indirect Speech Herman Cappelen and Ernie Lepore 9 All Facts Great and Small Richard N. Manning 10 Knowledge, Content, and the Wellsprings of Objectivity Ron Wilburn 11 Interpretation and Skill: On Passing Theory David Simpson Contributors Index Kluwer Academic Publisher, Philosophical Studies, Dordrecht, Netherland

217

218

Bookpublications

PROTOSOZIOLOGIE IM KONTEXT “Lebenswelt” und “System” in Philosophie und Soziologie Gerhard Preyer, Georg Peter, Alexander Ulfig (Hrsg.) in memoriam Paul Lorenzen Einleitung: “Lebenswelt” und “System” in Philosophie und Soziologie, Gerhard Preyer, Georg Peter, Alexander Ulfig

ZUM BEGRIFF

DER

LEBENSWELT

Ernst W. Orth: Lebenswelt als unvermeidliche Illusion? Husserls Lebensweltbegriff und seine kulturpolitischen Weiterungen, Walter Biemel: Gedanken zur Genesis der Lebenswelt, Alexander Ulfig: Lebenswelt und Reflexion. Anhang: Lebenswelt als Fundament der Wissenschaft, Gerhard Preyer: Hintergrundwissen: Kritik eines Begriffs, Hubert A. Knoblauch: Soziologie als strenge Wissenschaft? Phänomenologie, kommunikative Lebenswelt und soziologische Methodologie

LEBENSWELT – BEGRÜNDUNG – WISSENSCHAFT Jürgen Mittelstraß: Das lebensweltliche Apriori, Peter Janich: Die Rationalität der Naturwissenschaften, Jürgen Mittelstraß: Rationalität und Reproduzierbarkeit, Elisabeth Ströker:Lebenswelt durch Wissenschaft, Paul Janssen: Lebenswelt, Wissen und Wissenschaft, Richard T. Murphy: E. Husserl’s Phenomenology of Reason LEBENSWELT/LEBENSFORM – SPRACHE Pierre Kerszberg: Lifeworld and Language, John F.M. Hunter: The Motley Forms of Life in the Later Wittgenstein, Peter A. French: Why did Wittgenstein read Tagore to the Vienna Circle? Georg Peter: Die Nebenbeschäftigung der Symbole. Zu Wahrheit und Funktion der Metapher

SYSTEM – SOZIALSYSTEM – GESELLSCHAFT Niklas Luhmann: Die Lebenswelt nach Rücksprache mit Phänomenologen, Niklas Luhmann: Observing Re-entries, Gerhard Preyer: System-, Medien- und Evolutionstheorie. Zu Niklas Luhmanns Ansatz, Richard Münch: Autopoesis per Definition, Hans Zitko: Codierungen der Kunst: Zur Kunstsoziologie Niklas Luhmanns, James Bohman: The Completeness of Macro-Sociological Explanations: System and Lifeworld, Göran Ahrne: Outline of an Organisational Theory of Society, Anhang: Karl Otto Hondrich: Zu Göran Ahrnes Ansatz

392 Seiten Verlag Königshausen & Neumann, Würzburg 1996 Digitale Neuauflage bei: Humanities Online, Frankfurt am Main2001 www.humanities-online.de

219

Bookpublications

INTENTION – BEDEUTUNG – KOMMUNIKATION Kognitive und handlungstheoretische Grundlagen der Sprachtheorie Gerhard Preyer, Maria Ulkan, Alexander Ulfig (Hrsg.) Einleitung: Zu kognitiven und handlungstheoretischen Grundlagen der Sprachtheorie, Gerhard Preyer, Maria Ulkan, Alexander Ulfig I INTENTIONEN

UND KOMMUNIKATIVE

HANDLUNGEN

Maria Ulkan: Kommunikative und illokutionäre Akte; Georg Meggle/Maria Ulkan: Grices Doppelfehler. Ein Nachtrag zum Griceschen Grundmodell; Jan Nuyts: Intentionalität und Sprachfunktionen II INTERPRETATION UND BEDEUTUNG Gerhard Preyer: Kognitive Semantik, Anhang Sprechaktsemantik: J.L. Austin, J.R. Searle, H.P. Grice, P.F. Strawson; Louise Röska-Hardy: Sprechen, Sprache, Handeln; Frank Siebelt: Zweierlei Holismus. Überlegungen zur Interpretationstheorie Donald Davidsons; Peter Rothermel: Semantische Implikaturen; Volkmar Taube: Referenz und Interpretation. Zur Theorie nichtsprachlicher Symbolisierung; Georg Peter: Zu Richtigkeit und Interpretation der Metapher: Kognitive Funktion und rekonstruktive Schemainterpretation III KLASSIFIKATION

VON

SPRECHAKTEN

Maria Ulkan: Informations- und Aufforderungshandlungen; Dirk Hartmann: Konstruktive Sprechakttheorie; Volkmar Taube: Bildliche Sprechakte IV KOMMUNIKATIVES HANDELN

UND INTERSUBJEKTIVE

GÜLTIGKEIT

Jürgen Habermas: Sprechakttheoretische Erläuterungen zum Begriff der kommunikativen Rationalität; Karl-Otto Apel: Illokutionäre Bedeutung und normative Gültigkeit. Die transzendentalpragmatische Begründung der uneingeschränkten kommunikativen Verständigung; Peter-Paul König: Kommunikatives und strategisches Handeln. Kritische Bemerkungen zu zwei zentralen Begriffen der “Theorie des kommunikativen Handelns” von Jürgen Habermas; Alexander Ulfig: Präsuppositionen und Hintergrundwissen. Eine Kritik am formalpragmatischen Präsuppositionsbegriff V DIALOGSTRUKTUR

UND

ARGUMENTATION

Wilhelm Franke: Konzepte linguistischer Dialogforschung; Franz Hundsnurscher: Streitspezifische Sprechakte: Vorwerfen, Insistieren, Beschimpfen; Dieter Mans: Argumentation im Kontext, Exkurs: Zu Christoph Lumers “Praktische Argumentationstheorie” 408 Seiten Westdeutscher Verlag, Opladen 1997 Digitale Version: Humanities Online 2001

220

Bookpublications

STRUKTURELLE EVOLUTION

UND DAS WELTSYSTEM

Theorien, Sozialstruktur und evolutionäre Entwicklungen Gerhard Preyer (Hrsg.) Einleitung: Gerhard Preyer: Strukturelle Evolution und das Weltsystem: Theorien, Sozialstruktur und evolutionäre Entwicklungen ZENTRUM

UND

PERIPHERIE –

INSTITUTIONELLE

ENTWICKLUNG –

ASKRIPTIVE

SOLIDARITÄT

Shmuel Noah Eisenstadt: Social Division of Labor, Construction of Centers and Institutional Dynamics: A Reassessment of the Structural–Evolutionary Perspective, Bernhard Giesen, Kay Junge: Strukturelle Evolution, Gerhard Preyer: Mitgliedschaftsbedingungen. Zur soziologischen Kerntheorie einer Protosoziologie, Anhang: Die modernen Gesellschaften verstehen. Richard Münchs Entwicklungstheorie moderner Gesellschaften, Erwin Rogler, Gerhard Preyer: Relationslogische Darstellung der sozialen Gesetze, Gerhard Preyer: Die modernen Gesellschaften “verstehen”. Zu Richard Münchs Entwicklungstheorie moderner Gesellschaften, Dieter Claessens: Bemerkungen zur Entstehung der modernen Ökonomie: Das Organistionsproblem, Richard Pieper: Strukturelle Emotionen, elementare Strukturbildung und strukturelle Evolution DIE EVOLUTION

POLITISCHER

ORDNUNGEN

Rainer C. Baum: Parsons on Evolution of Democracy, Mathias Bös: Zur Evolution nationalstaatlich verfaßter Gesellschaften, Konrad Thomas: Das Ethnische und das Staatliche, Volker Bornschier: Die westeuropäische Integration als Gesellschaftsmodell im Zentrumswettbewerb ZUR SOZIOLOGIE

DES

WELTSYSTEMS

Immanuel Wallerstein: Evolution of the Modern World-System, Christopher ChaseDunn, Thomas D. Hall: The Historical Evolution of World-Systems: Iterations and Transformations, Albert Bergesen: Postmodernism: A World System Explanation, Richard Münch: Modernity and Irrationality: Paradoxes of Moral Modernization, Walter L. Bühl: Transformation oder strukturelle Evolution? Zum Problem der Steuerbarkeit von sozialen Systemen STATE

OF THE

ART: Michael Schmid: Soziologische Evolutionstheorien

Suhrkamp Verlag, Frankfurt am Main 1998, Suhrkamp Taschenbuch Wissenschaft

Bookpublications

221

LEBENSWELT – SYSTEM – GESELLSCHAFT Konstruktionsprobleme der “Theorie des kommunikativen Handelns” von Jürgen Habermas Gerhard Preyer I DIE ENTWICKLUNGSLOGIK VON WELTBILDERN 1. Die Rationalisierung von Weltbildern 2. Strategien der Analyse von Weltbildern 3. Folgeprobleme und Kritik II GESELLSCHAFT ALS LEBENSWELT UND SYSTEM 1. Die Strukturen der Lebenswelt 2. Allgemeine Bezugsprobleme der soziologischen Evolutionstheorie 3. Die Verständigungsformen 1. Weltbilder und soziale Integration 2. Zur Durkheim-Interpretation 4. Kommunikationsmedien und generalisierte Kommunikationsweisen 5. Folgeprobleme und Kritik III DIE FORMAL-PRAGMATISCHE BEDEUTUNGSTHEORIE 1. Die sprechakttheoretische Grundlegung 2. Interpersonal geltende Bedeutungskonventionen 3. Regelbewußtsein und Handlungskompetenz 4. Die Expansion des semantischen Gehalts 5. Verständigung und die Herstellung interpersonaler Beziehungen IV DER ERWERB DES MORALISCHEN BEWUßTSEINS 1. Die sozial-kognitive Grundausstattung 2. Diskurs und moralisches Bewußtsein 3. Folgeprobleme V KONSTRUKTIONSPROBLEME UND KRITIK 1. Rekonstruktionshypothesen 2. Zu den Konstruktionsproblemen ANHANG: Max Webers Religionssoziologie als eine Typologie des Rationalismus 290 Seiten digital bookpublication: Verlag Humanities Online, Frankfurt am Main 2000 www.humanities-online.de

222

Bookpublications

Die globale Herausforderung Wie Deutschland an die Weltspitze zurückkehren kann Gerhard Preyer Der Begriff Globalisierung hat bereits Schlagwortcharakter bekommen und ist in aller Munde. Globale Märkte haben die Grundsituation wirtschaftlichen und politischen Handelns grundsätzlich verändert. Das heute entstehende globale Weltsystem und der Medienverbund auf dem es beruht, fordert uns alle heraus. Was bedeutet Globalisierung? Worin bestehen die globalen Herausforderungen, die zu bewältigen sind? Darauf wird eine Antwort gegeben. NEUE MEDIEN, WISSENSCHAFT

UND

TECHNIK

IM

ZEITALTER

DER

GLOBALISIERUNG

Was heißt Globalisierung? – Die neuen Medien: Eine Kopernikanische Wende – Der Überlebensimperativ: Technologieentwicklung – Die Zukunft der deutsche Universitäten VERÄNDERTE KONSTELLATIONEN Globale Wirtschaft und globale Ordnung – Die Globalisierung der Finanzmärkte – Auf dem Weg zur virtuellen Organisation – Das Ende des klassischen Arbeitsmarktes – Zu einer neuen Wirtschaftspolitik EUROPA

IM

ZEITALTER

DER

GLOBALISIERUNG

Zur Ausgangssituation im Zentrumswettbewerb – Frankreichs Zentralismus: Grenzen der Marktwirtschaft – Das Überlebenssystem Italien – Großbritannien zwischen Tradition und Modernisierung – Deutsche Stärken im Umbruch – Neu gemischte Karten – Die Wiedergewinnung des Standorts Deutschland DIE EVOLUTION

DES

MITGLIEDSCHAFTSCODES

Zur Soziologie der Grenzziehungen – Gesellschaft, Organisation und Interaktion – Gesellschaftsinterne Globalisierung und Entwicklungstrends – Die globalisierte Gesellschaft

290 Seiten Frankfurter Allgemeine Zeitung/Gabler Edition, Frankfurt am Main 1998

Bookpublications

223

INTEGRIERTES MANAGEMENT Was kommt nach der Lean-Production? Gerhard Preyer/Jakob Schissler Es gibt heute im Weltmaßstab kein Unternehmen, das seine Umstrukturierung nicht an den Bestandteilen des japanischen Modells der schlanken Produktion und des schlanken Managements, der Lean-Modelle, orientiert. Die Lean-Modelle sind zu einem integrierten Management fortzuentwickeln: dem über Vernetzung integrierten segmentierten Unternehmen, mit durchlässigen, ständig wechselnden Grenzen zwischen Unternehmen, Lieferanten und Kunden. Integriertes Management ist das Unternehmensmodell, das eine erhöhte Qualifizierung der Mitarbeiter und die Fähigkeit, in sich selbst steuernden Teams mitzuarbeiten, erfordert. Das bedeutet aber auch die Entwicklung eines neuen Führungsstils. Mit integriertem Management kann jedes Unternehmen sofort anfangen. Dazu wird in diesem Buch ein Weg gezeigt. Es wird eine wichtige Funktion als Impuls für weitere Entwicklungen haben. Inhalt DIE NEUEN HERAUSFORDERUNGEN 1. Vom japanischen und amerikanischen Management lernen? 2. Information, Wissen und Entscheidungen KREATIVITÄT – ORGANISATION – FÜHRUNG 1. Managementfähigkeiten 2. Innovative Organisation 3. Management und Führungsstile STRATEGIE – MARKT – KULTUR 1. Unternehmensstrategie 2. Controlling 3. Marketing 4. Zur Unternehmenskultur STICHWORT Was kommt nach der Lean Production? Frankfurter Allgemeine Zeitung, Verlagsbereich Wirtschaftsbücher, BlickbuchWirtschaft, Frankfurt am Main 1996, 164 Seiten.

224

Bookpublications

ÜBER DAS KÄMPFEN Zum Phänomen des Kampfes in Sport und Gesellschaft Axel Binhack Kämpfen ist stets als ein Beziehungsphänomen, gleichsam als “Fortsetzung der Kommunikation mit anderen Mitteln” zu verstehen. Anhand von sechs Strukturmerkmalen wird eine “Prototypik des Kampfes” entwickelt. Ihre zusammenhängende Dynamik ist gesellschafts- und kulturunabhängig in vielfältigen Konfliktformen zu erkennen. In dem vorgelegten interdisziplinären Anschnitt wird die Struktur der Kampfbeziehung von vergleichbaren Sozialphänomenen abgegrenzt und am Beispiel des Kampfsports in Form des “Zweikampfes” analysiert. 1. Merkmale einer Formalstruktur des Kampfes Der duale Beziehungsaspekt: Antagonismus und Ambivalenz – Der duale Trägeraspekt: Entscheidungsorientiertheit und tendenzielle energetische Totalität – Der duale Inhaltsaspekt: Zweckgerichtetheit und riskante Offenheit 2. Die Strukturmerkmale des Kampfes im Vergleich mit Arbeit, Spiel und den Beziehungsphänomenen Tausch, Flirt und Macht Der “antagonistische” Unterschied – Freiheit und Zweckgerichtetheit – “Zeitenthobene” innere Unendlichkeit und “zeitbeschleunigende” Entscheidungsorientiertheit – Die formalstrukturelle “Folgenhaftigkeit” des Kampfes 3. Zum Kampfsport Kampfsport als Radikalisierung des sportlichen Kampfes – Der Kampfsport als kulturelles Phänomen – Der Kampfdiskurs als Symbol männlichen Agierens – Ästhetische Perspektiven 4. Die Kampfsportart Karate-Do als Beispiel Historische Entwicklungslinien des Karate-Do – Der Schwertkampf und die “Leere Hand”: Zum karatespezifischen Aspekt der Waffenlosigkeit – Die Kultivierung des Kampfphänomens im Karate-Do – Der Kampf als Lehrmeister im Sinne des Zen Campus Verlag, Campus Forschung Bd. 768, Frankfurt am Main 1998

225

Bookpublications

System der Rechte, demokratischer Rechtsstaat und Diskurstheorie des Rechts nach Jürgen Habermas hrsg. von Werner Krawietz / Gerhard Preyer Editorial: Zivilgesellschaftliche Assoziationen und spätmoderner Rechtssstaat (Werner Krawietz)

I JURISTISCHE ENTSCHEIDUNG, BEGRÜNDUNG

UND

DISKURSIVER

RECHTFERTIGUNG

DES

RECHTS

IN DER

PERSPEKTIVE

REFLEXIONSTHEORIE

Enrique P. Huba: Standortbestimmung in der zeitgenössischen Rechtstheorie - Rawls, Dworkin, Habermas, Alexy und andere Mitglieder der modernen ‘Heiligen (Rede-) Familie’, Thomas McCarthy: Legitimacy and Diversity: Dialectical Reflexions on Analytical Distinctions, Gerhard Preyer: Rechtsgeltung – Argumentation – Entscheidung II ORDNUNG, RECHT

UND

RATIONALITÄT

IM JURISTISCHEN

DISKURS

Karl-Heinz Ladeur: Rechtliche Ordnungsbildung unter Ungewißheitsbedingungen und intersubjektive Rationalität, Ulfrid Neumann: Zur Interpretation des forensischen Diskurses in der Rechtsphilosophie von Jürgen Habermas, Ota Weinberger: Diskursive Demokratie ohne Diskurstheorie III KONSENSUSTHEORIE

SUBJEKTPHILOSOPHIE? VORAUSSETZUNGEN UND FOLGEN DISKURSES IM DEMOKRATISCHEN RECHTSTAAT

VERSUS

JURISTISCHEN

DES

Michael Pawlik: Die Verdrängung des Subjekts und ihre Folgen. Begründungsdefizite in Habermas’ “System der Rechte”, Uwe Steinhoff: Probleme der Legitimation des demokratischen Rechtsstaats, William Rehg: The Place of Consensus in Democratic Legitimation: A Recommendation

Rechtstheorie Zeitschrift für Logik, Methodenlehre, Normentheorie und Soziologie des Rechts 3 1996, Habermas-Sonderheft (erschienen 1998) Duncker und Humblot, Berlin

226

Bookpublications

Wittgensteins Spätphilosophie Analysen und Probleme

Wulf Kellerwessel, Thomas Peuker (Hrsg.) Die Spätphilosophie Ludwig Wittgensteins entfaltet nicht nur in der gegenwärtigen Sprachphilosophie ihre Wirkung, sondern auch zunehmend über den Bereich der Sprachphilosophie hinaus. Zwar macht dieser einen Schwerpunkt des Wittgensteinschen Denkens aus, aber dennoch verdanken die Philosophie des Geistes, die Erkenntnistheorie und die Philosophie der Religion der Wittgensteinschen Sprachphilosophie wichtige Impulse. Diese Schwerpunktsetzung zugunsten der Sprachphilosophie wie auch ihre Zusammenhänge mit den genanten anderen Disziplinen der Philosophie zeichnet der vorliegende Band aus.

INHALT W. Kellerwessel, T. Peuker: Einleitung, A. Ofsti: Methodischer Solipsismus, Metasprachen(problem) und Dexis. Einige Überlegungen zum Privatsprachenproblem, T. Peuker: Die faktische Öffentlichkeit der Sprache. Zu Wittgensteins Privatsprachenargument, P. Niesen: Gemeinschaft, Normativität, Praxis: Zur Debatte über Wittgensteins Regelbegriff, A. Berndzen: Einer Regel entsprechen – einer Regel folgen. Zu einer Kontroverse in der Interpretation von Wittgensteins ”Philosophischen Untersuchungen”, G. Preyer: Sprachbedeutung ohne Regelbefolgung, C. Stetter: Sprache und Schrift bei Wittgenstein, H.J. Schneider: Mentale Zustände als metaphorische Schöpfungen, W. Kellerwessel: Zum Begriff der Gewißheit in Wittgensteins ”Über Gewißheit”. Ein Kommentar, M. Kroß: “Glaube Du! Es schadet nicht”. Ludwig Wittgensteins Vermischte Schriften zur Religion - Anhang. 302 Seiten

Verlag Königshausen & Neumann, Würzburg

Bookpublications

Donald Davidsons Philosophie Von der radikalen Interpretation zum radikalen Kontextualismus Gerhard Preyer Für Bruce Aune EINLEITUNG I RADIKALE INTERPRETATION, LOGISCHE FORM UND EREIGNISSE 1. Donald Davidsons Philosophie: Ein Überblick 2. Wahrheit, Bedeutung und radikale Interpretation 2.1 Radikale Interpretation als ein radikaler Externalismus (1) Der Zirkel zwischen Überzeugung und Bedeutung: Die Asymmetrie von RI (2) Die Strukur von Inhaltssätzen: sagen, daß ... (3) Der Grundsatz der Nachsicht (4) Die Autonomie der Bedeutung (5) Radikaler Externalismus: Der innertheoretische Schritt (6) Die Ontologie von RI 2.2 Von der Idiolekttheorie zum dritten Dogma des Empirismus (1) Ausgangs- und Übergangstheorien der Interpretation (2) Die Demontage eines Mythos (3) Die epistemischen Beschränkungen von Verstehen 3. Die logische Form von Handlungssätzen und die singuläre Kausalaussage 3.1 Logische Form und adverbiale Modifikation 3.2 Kausale Beziehungen 4. Wahrheit und Überzeugung 4.1 Zum Haupteinwand 4.2 Zur empirischen Offensichtlichkeit von Einstellungszuschreibung 4.3 Rationalität als normativer Begriff II KÖRPERBEWEGUNGEN UND HANDLUNGEN 1. Das logische Verknüpfungsargument 2. Primäre Gründe und die Identitätsthese: Die synkategormatische Fassung von Absichten 3. Basisakte 4. Flucht vor den Körperbewegungen 4.1 Thalbergs Handlungstheorie und der “Akkordeon-Effekt” 4.2 A.I. Goldmans Kritik an der Identitätsthese 4.3 H.L.A. Hart: Zuschreibungen 4.4 Körperbewegungen als Bestandteil von Handlungen III EINE RADIKALE THEORIE DES HANDELNS 1. Der Begriff der Einstellungsrationalität 1.1 Die Homogenität der Interpretation

227

228

Bookpublications

1.2 Bewertende Einstellungen 2. Handlungsbeschreibungen und Handlungsverursachung 3. Eine Handlungserklärung 3.1 Überzeugungen, Absichten, Situationen 3.2 Praktische Gedanken 3.3 Handlungsgründe 3.4 Zu Hempels Ansatz 4. Praktische Schlüsse 4.1 Zur Gültigkeit praktischen Schließens 4.2 Entscheidungen und die Ausführung von Absichten 4.3 Überzeugungen und willentliche Handlungen 5. Radikaler Kontextualismus Literatur 290 Seiten

Verlag Humanties Online, Frankfurt am Main, 2002 www.humanities-online.de

229

Bookpublications

LANGUAGE, MIND

AND

EPISTEMOLOGY

On Donald Davidson’s Philosophy Gerhard Preyer, Frank Siebelt, Alexander Ulfig (eds.)

Introduction: On Donald Davidson’s Philosophy Alexander Ulfig)

(Gerhard Preyer, Frank Siebelt,

PHILOSOPHY OF LANGUAGE Jerry Fodor, Ernie Lepore (New Brunswick, USA): Meaning, Holism, and the Problem of Extensionality; Olav Gjelsvik (Oslo, Norway): Davidson’s Use of Truth in Accounting for Meaning; Wilhelm K. Essler (Frankfurt/Main, Germany): Was ist Wahrheit? Arend Kulenkampff (Frankfurt/Main, Germany): Eigennamen und Kennzeichnungen EPISTEMOLOGY Roger F. Gibson (St. Louis, USA): Quine and Davidson: Two Naturalized Epistemologists; Eva Picardi (Bolongna, Italy): Davidson and Quine on Observation Sentences; Ralf Naumann (Düsseldorf, Germany): Events and Externalism; Dorit Bar-On (Chapel Hill, USA): Conceptual Relativism and Translation; David K. Henderson (Memphis, USA): Conceptual Schemes after Davidson; Frank Siebelt (Frankfurt/Main, Germany): Singular Causal Sentences and two Relational Views PHILOSOPHY OF MIND AND THEORY OF ACTION Louise M. Antony (Raleigh, USA): The Inadequacy of Anomalous Monism as a Realist Theory of Mind, Louise Röska-Hardy (Darmstadt, Germany): Internalism, Externalism and Davidsons Conception of the Mental, Marcia Cavell (Berkeley, USA): Dividing the Self, Ralf Stoecker (Bielefeld, Germany): Willensschwäche - Wie ist das nur möglich? Klaus Puhl (Graz, Austria): Davidson on intentional Content and Self-Knowledge, Johannes Brandl (Salzburg, Austria): Sharing Beliefs and the Myth of the Subjective, Kirk Ludwig (Gainesville, USA): First Person Knowledge and Authority, Gerhard Preyer (Frankfurt/Main, Germany): Rationalität: Absichten, Primärgründe und praktisches Denken. Donald Davidson: Dialectic and Dialogue 445 Seiten Kluwer Academic Publishers, Synthese Library, Dordrecht

230

Bookpublications

Reality and Humean Supervenience Essays on the Philosophy of David Lewis Gerhard Preyer, Frank Siebelt (eds.) “This is a very fine collection of essays about David Lewis’s work. Centered around such central Lewisian doctrines as Humean supervenience, modal realism, the counterfactual analysis of causation, and physicalism, the collection is a fitting tribute to his outstanding work and to its influence.” – Jeremy Butterfield, University of Cambridge “This collection gathers papers on David Lewis’s metaphysics. Since the editors have assembled a superb group of authors, the result is at the cutting edge of the best work in contemporary metaphysics. Together with Lewis’s writings it would make for an outstanding seminar in metaphysics.” – Ernest Sosa, Brown University “David Lewis’s work occupies a central place in contemporary metaphysics, and in this volume some of our leading metaphysicians comment, interpret, and critically appraise Lewis’s important and influential claims and arguments on issues like modality and possible worlds, Humean supervenience, endurance and perdurance, counterfactuals and time, causation, and the mind-body problem. This is a timely and highly useful collection indispensable to students of Lewis’s work in metaphysics and philosophy of mind.” – Jaegwon Kim, Brown University Preface Gerhard Preyer, Frank Siebelt Reality and Humean Supervenience - Some Reflections on David Lewis’ Philosophy MODAL REALISM Phillip Bricker: Island Universes and the Analysis of Modality, John Bigelow: Time Travel Fictions, Peter Forrest: Counting the Cost of Modal Realism, Paul Teller: Lewis’s Defence of Counterpart Theory, Harold W. Noonan: The Case for Perdurance PHYSICALISM, CAUSATION AND CONDITIONALS Daniel Bonevac: Naturalism for the Faint of Heart, D. M. Armstrong: Going throug the Open Door again: Counterfactual vs. Singularists Theories of Causation, Jonathan Bennett: On Forward and Backward Counterfactual Conditionals REDUCTION

OF

MIND

Terence Horgan: Multiple Reference, Multiple Realization and the Reduction of Mind, Michael Tye: Knowing what it is like: The Ability Hypothesis and the Knowledge Argument

Rowman and Littlefield Publishers, Lanham, USA, 2001

Bookpublications

231

Wilhelm K. Essler „Unser die Welt“ Sprachphilosophische Grundlegungen der Erkenntnistheorie Ausgewählte Artikel herausgegeben von Gerhard Preyer

Einleitung: Zur Struktur von Erfahrung 1. Fundamentals of a Semi-Kantian Metaphysics of Knowledge 2. Kant und kein Ende 3. Tarski on Language and Truth 4. Am Anfang war die Tat – Semantische und epistemologische Anmerkungen zur Sprachstufung 5. Was ist Wahrheit? 6. Gorgias hat recht! 7. Was ist und zu welchem Ende treibt man Metaphysik 8. Das logische Aufbauen von Welten 9. Unsere Welt – trotz alledem 10. Selbst das Selbst ist nicht selbst 11. „Erkenne dich selbst!“ Der Versuch von Hinweisen auf die Weisheit des Sokrates und auf die des Buddha Schakyamuni 12. Offenes Philosophieren

Digitale Publikation: Verlag Humanities Online, Frankfurt am Main, 2001 www.humanities-online.de

232

Published Volumes

Subscription / Digital Version I subscribe PROTOSOCIOLOGY – pdf-file – 12,- Euro each issue, abroad: $ 25 (including courtage); payable by cheque or transfer. Date, signature ................... Separate order I order PROTOSOCIOLOGY volume Nr. ....... Single issue: 15 Euro, abroad: $ 27 (including courtage); payable by cheque or transfer. Email-posting with keyword. Date, signature ............ Your address: Name ....................

Address ....................

E-Mail ................... Send the order to: PROTOSOCIOLOGY ISSN 0940-4147 Editorial office: Stephan-Heise-Str. 56, D-60488 Frankfurt am Main, E-Mail: [email protected], Tel. 069/769461

Editor: Gerhard Preyer, Johann Wolfgang Goethe-Universität Frankfurt am Main, Department of Social Sciences. Editorial staff: Georg Peter. www.protosociology.de

Published Volumes

233

ProtoSociology: Digital volumes available Vol. 11 is a free download. Visit our homepage under “Special Service”. Vol. 16 – Understanding the Social: New Perspectives from Epistemology Vol. 15 – On a Sociology of Borderlines: Social Process in Time of Globalization Vol. 14 – Folk Psychology, Mental Concepts and the Ascription of Attitudes Vol. 13 – Reasoning and Argumentation Vol. 12 – After the Received View – Developments in the Theory of Science Vol. 11 – Cognitive Semantics II – Externalism in Debate (free download!) Vol. 10 – Cognitive Semantics I – Conceptions of Meaning Vol. 8/9 – Rationality II &III (double volume)

Visit our hompepage: www.protosociology.de Or send order to: PROTOSOCIOLOGY, editorial office: Stephan-Heise-Str. 56, D-60488 Frankfurt am Main, Tel. 069/769461 Via E-Mail: [email protected] or [email protected]

Editor: Gerhard Preyer, Johann Wolfgang Goethe-Universität Frankfurt am Main FB 3: Department of Social Sciences. Editorial staff: Georg Peter. www.protosociology.de

Bookpublications

231

Wilhelm K. Essler „Unser die Welt“ Sprachphilosophische Grundlegungen der Erkenntnistheorie Ausgewählte Artikel herausgegeben von Gerhard Preyer

Einleitung: Zur Struktur von Erfahrung 1. Fundamentals of a Semi-Kantian Metaphysics of Knowledge 2. Kant und kein Ende 3. Tarski on Language and Truth 4. Am Anfang war die Tat – Semantische und epistemologische Anmerkungen zur Sprachstufung 5. Was ist Wahrheit? 6. Gorgias hat recht! 7. Was ist und zu welchem Ende treibt man Metaphysik 8. Das logische Aufbauen von Welten 9. Unsere Welt – trotz alledem 10. Selbst das Selbst ist nicht selbst 11. „Erkenne dich selbst!“ Der Versuch von Hinweisen auf die Weisheit des Sokrates und auf die des Buddha Schakyamuni 12. Offenes Philosophieren

Digitale Publikation: Verlag Humanities Online, Frankfurt am Main, 2001 www.humanities-online.de

234

Published Volumes

Über den Krieg · Pragmatismus und Faschismus Zeitgemäße Erziehung · Der Schatz im Acker Rationierung und Allokation im Gesundheitswesen · VELBRÜCK WISSENSCHAFT · Praxeologie der Moderne · Autopoietische Realität und konnotative Theorie · Der Sinn der Zeit · Die Grenzen des Verstehens · Neu im Herbst 2002 Herfried Münkler Über den Krieg Stationen der Kriegsgeschichte im Spiegel ihrer theoretischen Reflexion · 293 Seiten Broschiert, Fadenheftung ca. EUR (D) 35,– / sFr 59.– ISBN 3-934730-54-X

Peter Vogt Pragmatismus und Faschismus Kreativität und Kontingenz in der Moderne · 368 Seiten Broschiert, Fadenheftung ca. EUR (D) 35,– / sFr 59.– ISBN 3-934730-56-6

Walter Herzog Zeitgemäße Erziehung · Die Konstruktion pädagogischer Wirklichkeit · 684 Seiten Gebunden, Fadenheftung, Schutzumschlag ca. EUR (D) 50,– / sFr 81.– ISBN 3-934730-55-8