The Metaphorical Problem:

Realism and Anti-Realism in the Philosophy of Metaphor

Andrew James Joseph McGonigal Department of Philosophy University of Glasgow September 2004 PhD Thesis

Andrew McGonigal 2004

Abstract
This thesis is concerned with the meaning of metaphors. In particular, it examines a contemporary dispute in the philosophy of language, primarily comprising critical responses to Donald Davidsons seminal work in the area, which focuses on the question of whether metaphorical utterances, qua metaphors, ought to receive distinctive semantic evaluations. I treat this debate as an instance of a more general form of philosophical dispute, which has been explored in some detail in recent work on the nature of realism and anti-realism. The thesis has five chapters. In the first chapter, I outline, motivate and evaluate two contrasting approaches to realism, proposed by Michael Devitt and Crispin Wright. I argue that neither is wholly satisfactory, but that a modified version of Wrights approach is likely to be most fruitful in the philosophy of metaphor. In the second chapter, I examine the character of Davidsons anti-realism, concluding that he is best thought of as an error-theorist about metaphorical meaning. I go on to set out a unified Davidsonian argument for semantic and pragmatic anti-realism about metaphor, and offer a sustained discussion and partial defence of the six premises that such an argument proceeds from. My third chapter outlines a series of common objections to Davidsons views, and argues that error-theorists have the resources to address many of these criticisms in a fairly plausible manner. In the fourth chapter, I go on to investigate the realist standing of metaphorical meaning in more detail. I examine the open-endedness of metaphor in the light of Wrights response-dependent theory of intention, and argue that this approach offers a novel response to certain anti-realist concerns. The fifth chapter concerns the relationship between metaphor and non-conceptual content. I argue that thinking of metaphorical meanings as non-conceptual entails that the non-propositional and limitless character of metaphor does not pose a fatal objection to a pragmatic realist account, contra Davidson. I apply my suggested account to two test cases: metaphors that describe ones emotional state, and religious metaphors, and

argue that in each case, thinking of the metaphors as expressing non-conceptual contents is potentially suggestive and helpful. In that chapter, I also examine the possibility of an robustly realist approach to metaphorical meaning, modelled on the epistemicist approach to vagueness set out in recent work by Timothy Williamson. I demonstrate how the dominant objection to this account can be partially defused, and go on to examine the final standing of the dispute between realist and anti-realist.

Abstract ..................................................................................................................................... 2 Acknowledgements ................................................................................................................... 6 Preface ....................................................................................................................................... 7 1. Truth, Realism and Balance ................................................................................................. 9 1.1 Introduction ............................................................................................................... 9 1.2 Realism, Existence and Independence .................................................................... 10 1.3 Extending Devitts Account? .................................................................................. 12 1.4 Wright on Truth....................................................................................................... 15 1.5 Pluralism, Truth and Realism.................................................................................. 18 1.6 Evidence Transcendence ......................................................................................... 19 1.7 The Order-of-Determination Test ........................................................................... 25 1.8 Cognitive Command ............................................................................................... 28 1.9 Width of Cosmological Role................................................................................... 38 1.10 Costs and benefits of Wrights account................................................................... 41 2. Davidson, Meaning and Metaphor ..................................................................................... 49 2.1 Introduction ............................................................................................................. 49 2.2 The Character of Davidsons Anti-Realism ............................................................ 50 2.3 Reductionism and Expressivism ............................................................................. 52 2.4 The Nature of Davidsons Error-theoretic Approach.............................................. 63 2.5 Davidsons Master Argument Against Metaphorical Meanings............................. 66 2.6 Word-meaning and Metaphor ................................................................................. 71 2.7 Sentence-meaning and Metaphor ............................................................................ 80 2.8 Metaphor, System and Error ................................................................................... 83 2.9 Two Marks of Metaphor ......................................................................................... 93 2.10 Conclusion............................................................................................................. 100 3. Replies, Objections and Suggestions ............................................................................... 103 3.1 Introduction ........................................................................................................... 103 3.2 Complaints about the Causal Theory.................................................................. 104 3.3 Metaphor and Context ........................................................................................... 109 3.4 Metaphor and Other Linguistic Phenomena.......................................................... 115 3.5 Metaphor and Reduction ....................................................................................... 133 3.6 Conclusion............................................................................................................. 144 4. Metaphor and the Tests for Realism.......................................................................... 146 4.1 Introduction ........................................................................................................... 146 4.2 Wrights Argument against Error theories ............................................................ 147 4.3 Metaphor and Evidence Transcendence................................................................ 154 4.4 Metaphor and Cognitive Command 1 ................................................................... 157 4.5 Metaphor and Cognitive Command 2 ................................................................... 161 4.6 Metaphor and Intention ......................................................................................... 169 4.7 Metaphor and Order-of-Determination ................................................................. 177 4.8 Conclusion............................................................................................................. 183 5. Metaphor and Non-Conceptual Content .......................................................................... 184 5.1 Introduction ........................................................................................................... 184 5.2 Metaphor and Emotional Content ......................................................................... 186 5.3 Metaphor, Emotion and Non-conceptual Content................................................. 189 5.4 Further Thoughts on the Limitlessness of Metaphor ............................................ 194 5.5 Metaphor and Religious Language 1..................................................................... 196 5.6 Metaphor and Religious Language 2..................................................................... 203 4

5.7 Metaphor and Vagueness ...................................................................................... 209 5.8 Conclusion............................................................................................................. 215 Conclusion............................................................................................................................. 217 Bibliography.......................................................................................................................... 221

Acknowledgements

This book has been a long time in the making; far longer than ought to have been the case. During the last six years of thinking about metaphor, I have been helped, inspired and provoked by a great number of philosophers and friends. Special thanks are due to my supervisor, Gary Kemp, whose insightfulness, patience and support has been constant throughout my graduate study. Thanks are also due to Jim Edwards, Bob Hale, Philip Percival, Jimmy Lenman, Chris Lindsay and Nick Zangwill at Glasgow; to John Divers, Bryan Frances, Joseph Melia, Matthew Kieran, and, especially, Roger White at Leeds; to Jos Bermudez, Alan Miller, Katherine Hawley, Ted Sider and Alex Miller. I should also mention the philosophical inspiration and personal encouragement I received from being exposed at an early stage in my academic development to the intellectual example of Crispin Wright. Without the help of these people my work and my life would be a lot worse than they are. Finally, I would like to thank my family, and particularly my parents, for their faith, encouragement and practical support. This thesis is dedicated to my wife Stephanie. Nothing I did would have been any good without her.

Preface

In this essay, I examine some of the ways in which a particular debate in the philosophy of metaphor a debate which concerns, to put it crudely, the question of whether metaphors have meaning can inform, and be informed by, contemporary reflections on realism and anti-realism. This choice of subject matter has the distinct disadvantage that neither topic is particularly pre-theoretically gripping. On the one hand, reflection on the core elements shared by, say, moral realism, scientific realism and mathematical Platonism is an especially abstruse variety of philosophical activity, one where obtaining the requisite alpine clarity of overview requires getting accustomed to thinner air than that which surrounds the more pressing, felt, substantive, first-order philosophical problems. The interest of undergraduates or non-philosophers is more easily sparked when considering the nature of scientific success, or the problems of reconciling the claims of justice and welfare, than when identifying the best formulation of cognitivism, or the relationship between metaphysical and semantic accounts of realism. Metaphor might seem a juicier topic of discussion than realism, but the particular question of metaphorical meaning is rarely felt to be worthy of serious consideration. Blank stares from the folk are hardly unsurprising in this area. If by meaning we mean the property that ordinary speakers of English tend to attribute in their talk about meaning and meaningfulness, it can scarcely be denied that metaphors are meaningful. The debate I am concerned with, unsurprisingly, appeals to a more technical notion of metaphorical meaning; otherwise this would be a very short book. Once again, however, the price of precision and theoretical interest is a certain dislocation from first order issues. It is easy to feel, when reading even the best work by analytic philosophers on the topic of metaphor that something has been lost; that the original motivation for engaging in talk of metaphorical meaning involved some quite separate impulse, perhaps misplaced or ill-conceived, but in any case somewhat orthogonal to the debates that we find ourselves engaging in.

I do not in any way intend these remarks to signal a lack of engagement with, or enthusiasm for, the question of realism, or the methodology of analytic philosophy of language. My sympathies lie wholly with that tradition. The point is rather simply to signal explicitly that this book, like those of many analytic philosophers, may deliver results that are less well integrated with efforts in other arts and humanities subjects, and with the wider practical concerns of those whose work or life involves grappling with language, than might have been hoped. I would have liked to have written a book that had some consequences for the way real metaphors are actually thought about and interpreted. I might yet, but I rather fear this is not it.

1. Truth, Realism and Balance

1.1

Introduction

The aim of this book is to examine the way in which recent work on realism might help advance attempts to give a satisfactory answer the following question: do metaphors, qua metaphors, express distinctive linguistic meanings? In this chapter, I begin to set out what I take to be the most constructive way of characterizing the nature of the debate between realists and anti-realists in general, considered in abstraction from any given subject matter. My hope is that regarding the disagreement between those theorists who think that metaphors are typically associated with characteristic meanings, and those who do not, as a special instance of such a general form of debate about realism, will cast some welcome light on some difficult issues in the philosophy of metaphor. I begin this more challenging work in later chapters. For the present, my concern is primarily elucidatory. I introduce the topic, in Sections 1.2-1.5, by elaborating two influential contemporary conceptions of that debate, presented in recent work by Michael Devitt and Crispin Wright. 1 I argue that the generality and topic-neutrality of Wrights conception offers a more attractive approach for our purposes, and that there are in any case some reasons to remain suspicious of Devitts presentation of the issues. Since much of the interest of Wrights approach relates to the cruces, or tests for the realist standing of a given type of fact, that he outlines, I go on to examine and discuss a selection of such criteria in some detail in the next four sections. However, I do not embrace Wrights approach uncritically. In the final section, I raise some worries which relate to the putatively unassuming character of a minimalist approach to truth, and, relatedly, to the suggestion that an anti-realist construal of a given area of thought ought to be the default dialectical position. I conclude by briefly
1

See Devitt (1997) and Wright (1992)

outlining a conception of the realist/anti-realist debate that seems to me to respect many of Wrights key insights, but which seems to avoid some of the more unwelcome implications of his position.

1.2

Realism, Existence and Independence

Michael Devitt has argued that we ought to characterise realism about the external world in the following terms: Tokens of most current common-sense and scientific physical types objectively exist independently of the mental. 2 The various elements of this definition require some elucidation. Devitt tells us that for a token of some common-sense type (chairs, tables, mountains) or scientific type (electrons, quarks) to exist objectively is to for the object not to be
constituted by our knowledge, by our epistemic values, by our capacity to refer to it, by the synthesizing power of the mind, by our imposition of concepts, theories, or languages. 3

Devitts Realism

It is not wholly clear what Devitt means by an object being constituted by our knowledge, or by our epistemic values, but the thrust of the thought about objective existence is, I hope, clear. His common-sense realist thinks of the world as being as it is regardless of our epistemic and semantic access to it. Such a conception of objectivity might seem to make the qualification regarding independence from the mental redundant. Devitt, recognises this possibility, but argues that it is worth including it to forestall the possibility that a sophisticated anti-realist might posit unknowable or unconceptualized entities, that were nevertheless somehow dependent on minds for their existence. Perhaps certain past, non-actualised permanent possibilities of sensation might have this status, for example. Such an anti-realist might
2 3

See Devitt (1997) Ch. 2, passim. Ibid, p. 15

10

argue that she too could provide an account of our common-sense intuitions regarding the objective existence, in the above sense, of the external world. In any case, Devitt is surely right to endorse the idea that realism about an external world commits one to this kind of modesty to use Wrights terminology in characterising the degree to which the existence and nature of the external world depends on human minds and cognitive abilities. It is also common ground that such realism involves a certain epistemic presumption, marked in the case at hand by Devitts inclusion of ontological commitment to many of the entities postulated by contemporary scientific theory. Let us grant, finally, that anti-realism about the external world can be characterised in general terms as the rejection of either such modesty or presumption. 4 I have deliberately laid stress upon the sense in which Devitt and Wright agree on a core general characterisation of realism, and, ipso facto, anti-realism. The key elements independence and existence, modesty and presumption - comprise a shared and plausible background theory, even if differences of emphasis remain. Issues become more contentious, and more immediately pertinent to the project at hand, with the following questions 1. Is it unproblematic to extend this kind of general picture of realism to debates between realists and anti-realists in other areas of discourse? 2. How might actual debates between realists and anti-realists be profitably prosecuted? 3. Given the above intuitive characterisations give a satisfactory account of realism and anti-realism in general, which particular species of each genus are tenable, and which most plausible, in particular disputes? Much of this book will examine potential answers to the third question with reference to the special case of realism about metaphorical meaning. I will concentrate for the

See Wright (1992) pp. 1-3

11

moment, therefore, on briefly examining the divergent answers that Devitt and Wright offer to the first two questions.

1.3

Extending Devitts Account?

Devitt is explicitly concerned with realism about the external world, committed to the existence and mind-independence of both common-sense and scientific entities. He does, however, extend his account beyond this domain. For example, he argues for a certain realist attitudes towards semantics, which he call the factual perspective.5 He is also defends a certain kind of value realism, with respect to epistemic value. 6 These further realisms are defended, in part, by appeal to external world realism, but it should be clear that they are logically independent from it. (For example, Davidsonians reject the factual perspective on semantics, and many philosophers would feel suspicious of the claim that empirical procedures can establish normative claims.) Each of these further realist theories, however, fits more or less into the model Devitt has set out as constitutive of external world realism. In each case, instances of semantic and epistemic properties are thought of as existing objectively, and doing so in a manner which is not constitutively dependent on human cognitive responses. 7 Moreover, each of the realist accounts that Devitt offers are in accord with a certain set of methodological maxims that he offers: In considering realism, distinguish the constitutive and evidential issues.

Maxim 1

See e.g. Devitt (1997) p. 190: From [the factual] perspective the semantic properties of symbols are

explanatory in theories of mind and language. Thus, on the one hand, their existence is supported by a wide spectrum of evidence, and, on the other hand, their nature is not determined by that evidence.
6

Ibid p. 78 I see no reason to doubt that most questions of the goodness of [epistemic] procedures are

concerned with objective matters of fact. Though this normative task is outside psychology, it is not outside scienceIt is an empirical question which procedures are good.
7

Except perhaps per accidens, as when all the instantiated semantic properties are properties of creatures

with minds. It is not clear what Devitt should say about this kind of case.

12

Maxim 2

Distinguish the metaphysical (ontological) issues of realism from any semantic issue. Settle the realism issue before any epistemic or semantic issue. In considering the semantic issue, dont take truth for granted. Distinguish the issue of correspondence truth from any epistemic issue.

Maxim 3

Maxim 4

Maxim 5

It is not wholly clear what the status of such maxims are. There is clearly something right about the idea that e.g. epistemological, metaphysical and semantic issues can often be profitably distinguished from one another. On the other hand, there is something uncomfortable at least it feels so to me about taking such maxims to be simply an expression of methodological inclination. I shall try and give more substance to this intuition in my discussion of the self-reflective worries for Wrights minimalism in Section 1.10 below. For the moment, I simply want to note that, even if we follow such maxims, there is some reason to wonder whether Devitts model of external world realism can be extended straightforwardly to cover other realist/anti-realist disputes. Bernhard Weiss, for example, claims that if
we thought of realism as a view about a certain range of entities then we would miss the potential similarities between realism about, say, other minds and realism about the past (where no range of entities seems to be under discussion). 8

Weisss latter example is, perhaps, poorly chosen. At least one realist/anti-realist debate about the past that conducted between eternalists and presentists precisely turns on the existence or non-existence of a range of entities, with the eternalist holding, and the presentists denying, that past and future objects and events exist in just the same sense as present ones do. I think his general point is well taken, though. There are cases where it is difficult to think of realism as consisting in the combination of existence and
8

Weiss (2002) p. 51

13

independence that Devitt appeals to in the case of common-sense objects. One example might be the debate about the existence of qualia. Neither friends nor foes of qualia need think of them as existing wholly independently of human subjective responses; quite the opposite. Both realists about qualia and their opponents can accept that the existence of qualia depends on their being perceived. This seems straightforwardly inconsistent with Devitts characterisation of independence. 9 Perhaps some adjustment could be made to that theory, to allow us to state the similarities and differences between external world realism and qualia realism perspicuously. Surely, however, it is methodologically desirable, when engaging in reflection on philosophical debates and positions, to be able to operate at a level of abstraction that enables us to draw analogies directly with similar debates in other areas. It seems that Devitts account of external world realism does not generalise straightforwardly to other debates in this way. Another example might be the debate between genuine modal realists and actualist realists. 10 Both groups of theorists can hold that possible worlds exist, and that such possible worlds are mind-independent. There is a real sense, however, in which the actualist defends a less realist position than the genuine modal realist. (That is why actualist realism has proved to be a more attractive option in the debates about modality). A useful overview of the realism issue should allow us to draw finer grained distinctions than Devitts account, as it stands, allows us to do. Again, it might be the case that Devitts account could be extended or modified so as to include the kind of generality and topic-neutrality that it currently seems to lack. 11 The maxims that I outlined earlier might seem to offer a natural starting point for such a expanded project. For the moment, I merely want to insist that Devitts treatment of external world realism is not straightforwardly extensible to other, seemingly directly analogous areas of thought, and that therefore, as it stands, gives us only limited insight into how actual debates between realists and anti-realists in other contexts might actually
9

See Devitt (1997) p. 16 The realist rejects esse est percipi for the objects he believes in. No object that I follow Divers (2002) in using actualist realism to refer to the position that Lewis (1986) terms ersatz One way would be to avoid talking of modal realism simpliciter, and instead distinguish e.g. realism

is tied to perception for its very existence has the required independence.
10

realism, and Stalnaker (1976) terms moderate realism.

11

about possible worlds, from realism about possibilia.

14

proceed. Rather than exploring possible ways in which Devitts account might be developed, however, I now want to turn my attention to a different approach, set out in recent work by Crispin Wright, which seems to build in from the start exactly the kind of generality, fine-grainedness and dialectical sensitivity required for the project at hand.

1.4

Wright on Truth

Wrights approach to the question of realism proceeds via the notion of truth. He is a truth minimalist in two senses. Firstly, he is a minimalist about truth. In this context, a minimalist about truth is someone who holds that it is necessary and sufficient for a particular linguistic expression to qualify as a truth predicate that its use accord with a particular set of interlinked platitudes; putatively a priori principles that chime with our ordinary thinking about truth. 12 These platitudes include, for example to assert, doubt, fear, that p is to assert, doubt, fear, that p is true. aptitude for truth is preserved under embedding in particular, truth-apt propositions have negations, conjunctions, disjunctions, etc., that are themselves truth-apt. for a proposition to be true is for it to reflect reality, accurately reflect how matters stand, tell it like it is, etc. a proposition can be true without being justified and vice versa. 13

Transparency

Embedding

Correspondence

Contrast

The platitudes are intended to capture the minimal set of commitments incurred by any theorist who is concerned with truth in the relevant sense at all. They are common
12

Wright (1999) Truth: A traditional debate reviewed, p. 226. The canonical statement of Wrights Wright (1999) p. 227. Other platitudes relate to the absoluteness and timelessness of truth, as well as

minimalism is given in his (1992). See also his (1998) for a useful summary of the position.
13

the opacity of truth this latter embracing a variety of principles to the effect that a particular truth might be outside a particular speakers or communitys cognitive reach at a particular time, or at any actual time, or for any possible time.

15

ground between different substantive conceptions of the nature of truth, and between proponents of substantive theories of truth and their deflationary opponents. For example, both correspondence and coherence theorists of truth are committed to the truth of the correspondence platitude, although only the former attempts to base a theory of truth on a philosophically rich conception of what that relation amounts to. In effect, the platitudes serve to implicitly define the notion of truth, but remain silent whether it has an essential nature, or if so, what that essence might consist in. The second sense in which Wright is a minimalist concerns truth-aptitude. A utterance is truth-apt in the relevant sense if it can be (semantically) correct to evaluate it as true or false. 14 A discourse about a particular subject matter morality, say, or colour - is truth apt if a suitable range of the utterances which go to make up the discourse are truth apt. This second aspect of minimalism itself involves two sub-components. Wright argues that it will be necessary and sufficient for a particular discourse to be apt for the application of a truth-predicate that the discourse in question involves assertoric content. This, I think, ought to be uncontroversial. For a sentence of a language to possess assertoric content is simply for it to be capable of being used, in a suitable context, to say something about how things are. Once we have such a notion of a sentence saying how things are, it ought to be a short step to characterising a sentence as true if things are indeed how it says they are, and false otherwise. 15

14

One might say: if it can have the property of being true or false, were it not for the deflationists denial

that the truth predicate genuinely attributes a property. I should note here that, for ease of exposition, I blur over important differences between describing utterances, sentences, sentences in a language, etc, as truth apt.
15

For reasons of space, I here will rely on appeal to this intuitive link between possession of assertoric

content and truth aptitude. For an explicit demonstration that any sentence that meets the constraints of syntax and warrant will be apt for the application of a predicate that accords with the set of platitudes referred to above, see the first two chapters of Wright (1992). The basic idea is that if, as outlined below, it suffices for a sentence to possess assertoric content that it a) has a certain syntactic form and b) is associated with a suitable set of conditions under which we count it as epistemically warranted, then it ought to make sense of it being capable of meeting those conditions, and of it continuing to do so under arbitrarily close scrutiny of the warrant with which we hold it, and arbitrarily large improvements of our informational state. Wright calls sentences that meet such conditions superassertible, and argues that superassertibility counts as a truth predicate, in that it satisfies all the relevant platitudes. I discuss superassertibility further in Section 1.6.

16

The second, more controversial and distinctive claim of the minimalist about truthaptitude, is that it suffices for a discourse to possess assertoric content that it meet certain constraints regarding firstly, syntax and secondly, discipline. 16 The syntactic criteria - roughly, that the relevant sentences should embed appropriately within the right logical and propositional attitude contexts - suffice to ensure that an utterance is assertoric in character, syntactically suitable to act as a conventional vehicle for a saying that p. Sayings should allow for the expression of belief, for the drawing of inferences on the basis of the information that they carry, and for disagreement about their accuracy, and the relevant syntactic embeddings ensure that these demands can be met in ways that are syntactically well-formed. The second constraint insists that there should be public norms which determine the conditions under which particular claims within the discourse in question ought to be asserted or denied. 17 Wright thinks of these as being cashed out in epistemic terms, claiming that the use of the sentences in question must be governed by agreed standards of warrant. The demand that a discourse display this kind of discipline is intended to ensure that we are dealing with something genuinely worth regarding as contentinvolving, serving to carry information about how the speaker takes the world to be. Meeting the constraint establishes the type of minimal normative standard which is often taken to be necessary for linguistic meaning, since it allows a distinction to be drawn between cases in which a sentence or term is correctly used, and those in which it is misapplied. It is an important component of Wrights claim that such standards must include prescriptive and descriptive norms; there must be something like rules governing particular assertoric moves within the discourse, and participants in the discourse must to some degree also actually respect such rules. In summary, then Wrights key claims are as follows;
16

So again, we have two sub-components. The structure is as follows: Wrights truth minimalism consists

of two claims, about truth and truth-aptitude. The latter comprises two sub-claims a) that possession of truth aptitude is co-extensional with possession of assertoric content and b) that it suffices for an utterance or discourse to be truth apt that it meet two constraints regarding i) syntax and ii) discipline.
17

See Wright (1998) p. 185, and, for a fuller account of the nature and motivation of the background

conception of warranted assertibility, Wright (1993) pp. 35-40, 403-433.

17

1. it suffices for a sentence (which comprises part of a linguistic practice) to possess assertoric content that it meet minimal constraints relating to syntactic form and public standards of epistemic warrant 18 2. any such sentence which possesses assertoric content will be suitable for the application of a certain predicate T that accords with a certain set of platitudes 3. any predicate that accords with such platitudes will be a truth predicate.

1.5

Pluralism, Truth and Realism

As well as subscribing to minimalism about truth and truth-aptitude, Wright endorses a pluralism about truth. This should not be thought of as involving an ambiguity in the meaning of the predicate __ is true. That term can be thought of as being wholly (implicitly) defined by the set of platitudes that serve to identify the characteristic marks of truth. Rather, truth-pluralism comprises the claim that what constitutes truth might vary from discourse to discourse:
The kind of plurality thats envisaged may be brought out by a comparison with identity. Minimally, identity can be characterised as that relation which is universally reflexive and a congruence for an arbitrary property. To that extent, the concept of identity is uniform across varying kinds of object. But that uniformity had better be consistent with our recognising that what constitutes identity is subject to considerable variation depending on the kinds of objects concerned. The identity of material objects is constituted by spatial and temporal continuity; for cardinal numbers, according to Freges famous proposal, identity is constituted by the one to one correspondence of an associated pair of concepts; for the directions of a pair of straight lines, identity is constituted by those lines being parallel; and for persons, identity is constituted by well, its notoriously difficult to say, but the case is different from each of the preceding. Identity, one might thus say, is formally uniform, but may vary in constitution as we

18

I prescind here from the interesting a pressing question of how we can genuinely talk in terms of

epistemology and warrant in a way that doesnt presuppose the notion of truth, contrary to the direction of Wrights derivation.

18

consider different potential identicals. Clearly there is space for a similar contention about truth 19

Wrights central idea in Truth and Objectivity is that we can profitably approach the question of realism by consideration of the different ways in which truth might be constituted in different areas of thought and language. We have already seen that Wright and Devitt converge fairly sharply on an intuitive construal of realism about the external world. Whereas Devitt tackles the issue in rigorously anti-semantic terms, however witness his second and third methodological maxims Wright attempts to approach it via an inquiry into the nature of truth as we apply it to talk about observable, everyday objects (and their scientific cousins). The inquiry takes the form of the examination of a number of cruces; distinctive properties which it is possible for truth-predicates in given areas of discourse to display, and which can serve to focus the debate between the realist and anti-realist, in a satisfyingly wide number of philosophical contests. Wright discusses four such key properties in Truth and Objectivity: Evidence Transcendence, the Order-ofDetermination Test, Cognitive Command and Width of Cosmological Role. Since I agree that each of these marks an important area of focus for disputes about realism, and they shall play an important role in later chapters, it is worth examining each constraint in some detail here.

1.6

Evidence Transcendence

Michael Dummett has famously, or infamously, argued that realism about a given area of fact consists in the view that the language in which we describe such facts includes sentences with evidence-transcendent truth conditions, grasp of which constitutes mastery of such sentences. 20 For example, realism about the external world is the view that understanding certain sentences (about e.g. inaccessible planets) involves grasping truth-conditions that in a certain sense outrun our current evidence; realism about
19 20

Wright (1998) p. 186 Classic statements can be found in Dummett (1978) and Dummett (1991).

19

mathematics is the view that understanding certain sentences (about e.g. Goldbachs conjecture) involves grasping truth-conditions that similarly transcend our current evidence; etc. More precisely, Dummetts Realism consists of the following claim The literal content of realism about a given area of fact A consists in semantic realism about the sentences that concern A where semantic realism is characterised as follows Our understanding of undecidable sentences about an area of fact A consists in grasp of their truth conditions, where an undecidable sentence is one that meets the following two conditions a) we currently have no evidence that bears on its truth or its falsity b) we do not know a procedure which, if correctly implemented, is guaranteed after finitely many steps to put us in a position where we have evidence that it is either true or false. 21 There are some important issues which arise with respect to this formulation regarding, for example, the question of whether the relevant notion of evidence is one which essentially involves a certain transparency, so that one who had such evidence would know that she did so; the extent to which we are permitted to idealize away from some of our contingent human limitations, etc. It is more important for our immediate purposes, however, to get clearer on the motivation for Dummetts view, and the sense in which elements of such a Dummettian approach are carried over into Wrights Evidence Transcendence test. Few contemporary philosophers are convinced that debates between realists and antirealists can only be literally construed in terms of semantic realism, as (DR) claims. Certainly, the claim requires much more support than Dummett, or any other

(DR)

(SR)

21

Here and elsewhere in this section I have benefited from Alexander Millers discussion of theses issues

in Miller (forthcoming, a) and Miller (forthcoming, b). My formulation of semantic realism draws directly on Miller (forthcoming, a), Sections 3 and 4.

20

philosopher, has given it to date, if it is to deserve acceptance. A more modest claim, however, while hardly uncontroversial, has more widespread support: (At least) one important element of an intuitive realism about a given area of fact A, in the presence of certain fairly plausible background assumptions, entails semantic realism about A-sentences. Why ought we to believe (DR*)? Firstly, lets get clearer on the nature of some background assumptions that might be relied upon in establishing it. These include the following: If a subject S understands a statement T, then he knows what it states. If a statement T states that P, and a subject S knows what T states, then he knows that T states that P. P3) If S knows that T states that P, then, by virtue of that very fact, S knows that (things are as T states they are if and only if P) P4) If, in virtue of the fact that S knows that T states that P, S knows that (things are as T states they are if and only if P), then S knows that the truth condition of T is P. These seemingly platitudinous assumptions can be employed in an argument to the effect that understanding a given statement entails grasping its truth conditions. 1) 2) 3) 4) 5) 6) 7) S understands statement T. T states that P. S knows what T states. S knows that T states that P are as T states they are if and only if P). S knows that the truth condition of T is P truth condition of T is P. Conditional Proof Assumption. Assumption By 1, P1), MPP By 2, 3, P2), MPP

(DR*)

P1) P2)

In virtue of the fact that S knows that T states that P, S knows that (things By 4, P3, MPP By 5, P4, MPP

If S understands statement T, that states that P, then he knows that the By 1, 2, 6,

21

I think that the premises of the above argument are true, for one important sense of linguistic understanding, and that the argument is valid. 22 Be that as it may, it must be granted that the premises are at least prima facie plausible and the line of thought prima facie coherent. 23 Moreover, the argument seems to provide initial support, perhaps via an inference to the best explanation, for the following thesis Understanding an assertoric utterance of a declarative sentence consists in grasp of its truth conditions. Now, recall that one element of an intuitive realism about e.g. the external world identified by both Devitt and Wright comprised a certain modesty about humanitys relationship to the universe; that the worlds existence and nature does not depend in any way on our
knowledgeour epistemic valuesour capacity to refer to itthe synthesizing power of the mindour imposition of concepts, theories, or languages.
24

(TC)

Given the absence of such dependence, it ought to be possible, by the realists lights, for the world to be a given way say, for conditions C to obtain in the heart of a black hole - even though we have no evidence that bears on the matter, and have no idea how we might even begin to instigate a search for such evidence. The case is similar for the mathematical Platonist, the realist about the past, the genuine modal realist, etc. Thus, it
22

In this I diverge from both Devitt and Miller, who are sympathetic to the thought that P2 involves an

illegitimate substitution into an opaque knows-what context. I think this is simply an error. Some knows-what contexts do not allow certain inferential transitions. For example, if I know what gold is, and gold is Johns favourite element, it does not follow that I know what Johns favourite element is, or that it is gold, at least on one disambiguation of those claims. However, some other such contexts clearly do allow such transitions: if I know what your favourite colour is, and your favourite colour is red, then it follows that I know that your favourite colour is red. The above inference is, I believe, of the latter, harmless kind. See Devitt (1991) pp. 270-271, and Miller (forthcoming, b), Section 7 for discussion.
23

There is no doubt a perfectly good sense of understanding in which one can understand a declarative

sentence without knowing what it states for example, if you dont know the semantic value of indexical or demonstrative expressions it contains. Nevertheless, theres also a perfectly good sense in which if you dont know such facts, you dont really understand the statement.
24

Devitt (1997) p. 15

22

seems that a central element of realism about certain areas entails a commitment to the possibility of there being undecidable sentences in that area, in the sense that I outlined above. Now, given a truth-conditional conception of meaning, as characterised in (TC), we can argue to (SR) as follows 1) A central element of an intuitive realism about an area of fact A is the commitment to a modest conception of the relationship between human beings and the existence and nature of A-facts. 2) In many areas where realism is at issue, a modest conception of the relationship between human beings and the existence and nature of A-facts entails the possibility of there being an undecidable A-sentence such that a) we currently have no evidence that bears on its truth or its falsity and b) we do not know a procedure which, if correctly implemented, is guaranteed after finitely many steps to put us in a position where we have evidence that it is either true or false. 3) Given the above two premises, it follows that, in many areas where realism is at issue, a central element of an intuitive realist view of an area of fact A commits one to the possible existence of undecidable A-sentences. 4) 5) By (TC), understanding an assertoric utterance of a declarative sentence consists in grasp of its truth conditions. By 3) and 4), it follows that, in many areas where realism is at issue, a central element of an intuitive realist view of an area of fact A commits one to the possible existence of undecidable A-sentences, where our understanding of assertoric utterances of such sentences would consist in grasp of their truth conditions. 6) By (SR), semantic realism is the thesis that our understanding of undecidable sentences about an area of fact A consists in grasp of their truth conditions Conclusion: In many areas where realism is at issue, a central element of an intuitive realism about an area of fact A commits one to the possibility of semantic realism about A-sentences. The above argument, then, seems to give us grounds to subscribe to (DR*) 23

(DR*)

(At least) one important element of an intuitive realism about a given area of fact A, in the presence of certain fairly plausible background assumptions, entails semantic realism about A-sentences.

Such grounds are, of course, defeasible, and the line of argument is no doubt vulnerable to attack at many points. Nevertheless, given (DR*), we can begin to see why Wright identifies the question of whether we can make sense of A-facts obtaining in an evidence-transcendent manner as an important crux for realism disputes. The connection runs in two directions. Firstly, evidence that semantic realism is somehow incoherent, or in any case implausible in some strong sense, will, given (TC), be evidence that an important element of realism is implausible. Dummett and Wrights reflections on e.g. how semantic competence is acquired and manifested, the rulefollowing considerations, etc., are of course intended to provide exactly such evidence against semantic realism. 25 Secondly, a defence of the claim that there can be undecidable sentences of a given discourse counts against the claim that truth in that area can be constituted by superassertibility. A sentence is superassertible, remember, if we are capable of gaining evidence that would epistemically justify its assertion, given our current state of information, and we would remain so justified given any way that that state of information might be enlarged or improved. 26 An undecidable sentence is one whose assertion cannot be so justified, and thus one whose truth cannot consist merely in being superassertible. Thus, given that a construal of truth as superassertibility seems to favour an anti-realist construal of the nature of the contested discourse, with truth being built out of human epistemic concerns and standards, a demonstration of the possibility of undecidable sentences in a given discourse blocks one important anti-realist version
25

See e.g. Wright (1993) pp. 13-29, Hale (1997) passim, Miller (forthcoming a and b), Miller (2002) for a See Wright (1992) pp. 47-48. The formal definition he gives there runs as follows: a sentence is

detailed overview and appraisal of the strategy.

26

superassertible if and only if it is, or can be, warranted, and some warrant for it would survive arbitrarily close scrutiny of its pedigree and arbitrarily extensive increments to or other forms of improvement of our information. The notion is introduced and elaborated in Ch. 13 of Wright (1993).

24

of the truth-constituting property. Providing a defence of the possibility of undecidable A-sentences, therefore allows the realist to foreclose on an anti-realist account of A-facts which presents them as constructed from human assertoric practices and values. Much more could and should be said here, but my purpose is to provide an outline of why evidence transcendence matters to realism disputes, not to extend or evaluate those disputes. Suffice it for the moment to say that the following test Does the discourse at issue include sentences with regard to which a) we 27 have no evidence that bears on their truth value, and b) we lack a conception of any way in which we might come to get such evidence? seems prima facie to provides one genuine and important focus for critical attention when determining whether a realist construal of that discourse is appropriate.

Evidence Transcendence

1.7

The Order-of-Determination Test

The second focus for debates about realism that Wright identifies concerns the Euthyphro dilemma, or Order-of-Determination test. Wright is interested in different ways of interpreting what he calls Provisional Equations, which take the following general form For a set of optimal conditions C, a state of affairs P, and a subject S: If C holds, then (It would be the case that P iff S would judge that P). There are two different ways, Wright thinks, that we could understand the case where (PE) holds true for a particular C, P and S. We could understand the C-conditions as being such as to allow S to successfully track an independently obtaining fact that P. That is, we could understand the biconditional as indicating that S judges that P because
27

(PE)

Or, perhaps, some suitably idealised counterparts of us.

25

P is the case. For example, we might understand the instance of (PE) that told us that, under relevant C-conditions, S judged that x was square when and only when it was square, as holding because (i) x was in fact square, and (ii) under conditions C, S is a competent judge of squareness. Call this the extension-reflecting reading. On the other hand, we could understand the biconditional as indicating that it is Ss best opinion that constitutes the fact that P. In this case, we understand it as telling us that P is the case because S judges that P. For example, we might take the instance of (PE) that told us that, under relevant C-conditions, S judged that x was funny when and only when it was funny, as holding because the facts about funniness depend on our best judgements about whats funny. Call this the extension-determining reading. 28 It is clear that the extension-determining reading of the biconditional is far more conducive to an anti-realist view of the state of affairs P, and that the extension-reflecting reading is similarly conducive to realist intuitions about the mind-independence of the relevant species of fact. How are we to tell which way we should read the Provisional Equation in any given case? Wrights idea is that it should be read as extension determining just in case it meets a set of further constraints. First, the C-conditions must be specified in philosophically substantial terms, not by means of a whatever it takes ceteris paribus clause. Secondly, they must be a priori true. Third, whether the C-conditions are satisfied must be logically independent of facts about P. And finally, our case for reading it as extension determining must be extremal: there must be no better explanation of why the first three conditions are satisfied than the claim that Ss best judgements constitute the fact that P. What is the motivation behind these constraints? Wright wants to test whether or not a particular biconditional is extension determining or extension reflecting by examining whether or not there is a merely accidental, a posteriori, contingent link between our best judgements and whether or not the fact that P holds. That kind of link is what we

28

Since the PE only tells us about what is happening in optimal conditions, Wright normally describes our

optimal judgement of whether P as at best partially determining the facts about P. I have often blurred this distinction here.

26

would expect if the realm of fact in question had the sort of constitutive independence from human responses that realism attributes to it. It is clear, however, that if we specify the C-conditions by means of a whatever it takes clause, the biconditional will hold trivially true, and thus prevent us from examining whether the link between judgement and fact holds merely contingently. 29 So we must specify the C-conditions in more detail, without appeal to ceteris paribus clauses. If we do so, and the biconditional holds a priori, then that will be a sign that the facts of the matter cannot come apart from our best judgements of the matter, and thus that we should construe the Equation in the extension-determining sense. But such a sign will only be an accurate guide if the further two conditions hold. The independence condition is required in order that we can allow echoing, making use of the very concepts that we are concerned with in specifying the optimality conditions, without running the risk that we might be jeopardising the idea that the a prioricity of the PE can be a test of whether the relevant concepts are extension-reflecting or determining. By making sure that the concepts only occur, if at all, in contexts governed by intensional operators, we ensure that there is no 'hidden reference' to the extension built into the optimality conditions. We avoid the charge that in specifying the conditions under which, for example, we can best judge whether something is red, we have implicitly appealed to an response-independent property of redness, thereby rendering our proposed test valueless. The extremal condition, that there must be no better explanation of why the first three conditions are satisfied than that S's best opinions constitute the fact that P, is intended to leave room for the idea that the a priori co-extensiveness of judgement and fact might be a result of our infallibility about a particular type of fact, for example, and not an indication that our judgements constitute the relevant facts. In effect, the condition ensures that if we are to be justified in claiming such infallibility we must be able to give a pretty detailed story of why and how we can be infallible about this particular type of

29

I blur here the important distinction between necessity and a prioricity. In the case of extension

determining judgements, Wright seems to see the facts about our judgements as being the truth makers for e.g. the facts about what we meant. The a prioricity of the biconditional is interpreted as a sign that judgement and fact are non-contingently linked.

27

fact. In the absence of such a story, we are entitled to assume that our best judgements determine what the facts are. Given the fit between the extension-determining reading and anti-realism, it again seems that we have grounds to hold, prima facie at least, that Wright has identified another important question for disputants to focus on Given that we can identify a true Provisional Equation involving the discourse at issue, do the best judgements of ideally placed subjects constitute the relevant facts, or merely track independently constituted facts?

Order-of-Determination

1.8

Cognitive Command

The third test suggested by Wright attempts to establish whether thought and talk about the area in question is richly representational, in the sense that we might expect if a realist construal of the relevant discourse was appropriate. If the states of affairs in question really do have the kind of independence and autonomy that the realist attributes to them, then when we come to successfully represent them in thought and talk, we interact with the world in a way which is cognitive in a rich sense. On the contrary, if the states of affairs are essentially best thought of as shadowy projections of human thought or normative practice, as the anti-realist maintains, then talk of substantive cognitive achievement is misplaced, and such representation is a thin relation, secured merely by a certain sensitivity to the internal norms of the language game in question. Wrights discussion of this crux for realism is rather confusing. He appeals to the following incontestable principle:

28

Convergence/Representation Platitude

If two devices each function to produce representations, then if conditions are suitable, and they function properly, they will produce divergent output if and only if presented with divergent input. 30

Given such a platitude about representation, Wright thinks, we can generalise to a third test for realism Does the discourse display Cognitive Command?

Cognitive Command

when that notion is defined as follows A discourse displays Cognitive Command if and only if it is a priori that differences of opinion arising within a given discourse can be satisfactorily explained only in terms of divergent input (that is, the disputants working on the basis of different information, and hence guilty of ignorance or error, depending on the status of that information), or unsuitable conditions (resulting in inattention or distraction, and so in inferential error, or oversight of data, etc.), or malfunction (for example, prejudicial assessment of data, upwards or downwards, or dogma, or failings in other categories already listed). 31 As Edwards has pointed out, however, it is rather difficult to see the route that Wright discerns here. 32 If the Convergence/Representation Platitude is really a platitude, then it seems difficult to see how it can serve to differentiate genuine, thick
30 31

(CC)

Wright (1992) p. 91 Wright (1992) pp 92-93. See also Ch. 4, passim, where he defines the notion as follows (p. 144): a

discourse meets Cognitive Command if and only if it is a priori that differences of opinion formulated within the discourse, unless excusable as a result of vagueness in a disputed statement, or in the standards of acceptability, or variation in personal evidence thresholds, so to speak, will involve something that may properly be regarded as a cognitive shortcoming.
32

See Edwards (1994) pp. 66-69

29

representation from its minimalist simulacrum. This reflection, of course, doesnt invalidate the proposed test, merely the considerations that Wright appeals to in order to motivate it. It can still mark a discourse as apt for a realist understanding that it stand in relations of representation that are cognitive in a thick sense, and Cognitive Command, correctly understood, may provide a useful first step towards a characterisation of just such a sense. What if somebody insists that mere ignorance of the fact in question whether a particular shade is red, for example, or whether a particular metaphor means that P inevitably does involve something properly thought of as a cognitive shortcoming? That is, what if somebody insists that the real effect of the Convergence/Representation Platitude is to ensure that the mere truth-aptness of a discourse, even those for which only a minimalist treatment is appropriate, entails that any disagreement about a given statement signals a lack of grasp - on the side of at least one of the participants - of the way things genuinely are, although perhaps only resulting from ignorance of that very fact? For this type of objector, the price of the notion of representational content that the minimalist employs is the surrender of the possibility of cognitively faultless disagreement about whether the world fits a given representation or not. Wright has replied in various ways to this worry, the effect of which is to trivialise the (Cognitive Command) constraint by ensuring that any disagreement, whether apparently relating to genuine objective fact, or reflecting mere differences of inclination, is going to involve one of the disputants getting something wrong, namely, the truth value of the very proposition that they are disagreeing about. His response in Truth and Objectivity is to claim that the burden of proof is always on the side of the party who wants to argue that, in a particular discourse, there can be culpable cognitive dysfunction which consists solely of our ignorance of the very proposition whose realist standing is at issue. 33 Consider a case where have a disagreement about say, whether a certain speaker had a given communicative intention. In order for such discussion to display Cognitive Command, we must assume that the parties agree on all questions of a non-intentional
33

And e.g. disjunctions involving such propositions, etc.

30

character, (since otherwise the parties would be working on the basis of different information, contra the first clause of the constraint). The issue at hand is whether or not divergent opinions regarding, say, a particular intention, must betray ignorance on one side or the other, given that none of the other explanations noted in (CC) are available. Wrights claim in Truth and Objectivity is that the burden of proof always lies on the part of the theorist who holds that, in any disagreement about the truth value of an assertion even one in which all other relevant facts are agreed upon - there is always a cognitive shortcoming. His idea is that talk of a cognitive shortcoming betrays a commitment to an object, fact or state of affairs with which we cognitively interact in some sense, even if merely by representing it as obtaining. Now, either the existence of such an entity will be in principle detectable or it will not. If not, then we need an account of how we can acquire and manifest grasp of the relevant concept. For example, in the case of intention, we need to be shown how to make sense of the notion of actiondirecting psychological states that can outrun all possibility of human cognitive contact. Such a demonstration may perhaps be given, but if so, it will surely be unsurprising that the discourse deserves a realist construal, since the relevant facts will clearly have the kind of autonomy from human cognitive affairs that realism maintains. The interesting case, therefore, is when the theorist holds that we can come to have warranted beliefs about such facts. In that case, Wright thinks, it is incumbent on the theorist who holds that e.g. talk about intentions meets Cognitive Command to identify an epistemological route to externally constituted facts about intentions. The theorist must, that is, provide details of when and how we might be justified in holding that such a fact obtains, bearing in mind that all other relevant facts are agreed upon, and that vagueness, varying standards of evidence, etc., have been excluded by hypothesis. Only if this type of route is provided has the intended sense of cognitive been respected. Otherwise, in the absence of a substantive epistemology, we are entitled to conclude that the sense of cognition in question is merely a minimal one, that consists in no more than adherence to the relevant practiceinternal standards of the discourse.

31

Such an epistemological route will presumably be either inferential, as in e.g. our knowledge of mathematical theorems or direct, as in perception and memory. If direct, then the case for acceptance of a perceptual or otherwise intuitional epistemology needs to be made. If indirect, it seems that the parties who disagree about whether P must disagree about the inferential transitions that link that fact to others, since, by hypothesis, there is agreement on matters that dont make mention of the particular fact that P, and on norms and degrees of justification, etc. In the example at hand, there must be disagreement about the principles linking the possession of a given intention with e.g. uncontested behavioural and dispositional facts. For example, I might assert, and you reject, the claim that a particular pattern of behaviour is constitutively tied to possession of the intention in question, so that if the subject behaves in such a way, she possesses the intention. Now, the status of these (conditional) principles has to be investigated. Since Cognitive Command is claimed to be being met trivially, these conditionals should also display it. Otherwise, the standards of acceptability that govern the discourse would permit differing opinions, since they would count both adherence and non-adherence to such principles to be acceptable. In that case, the discourse would not meet Cognitive Command. 34 So now we either need an intuitional epistemology for these, or help in understanding how their truth can outrun our cognitive powers, or another conditional, containing the original as consequent, for which the same problems will arise. Eventually, the thought is, the trivialising theorist is going to have to provide some substantive intuitional epistemology for the facts in question, or accept that the relevant, robust sense of cognitive achievement is simply not in play here. In effect, Wrights insistence on the provision of an epistemological route equates to the claim that cognitive shortcoming, in the intended sense, entails some failure of substantive epistemological process, rather than being signalled merely by disagreement about whether a given representational content fits the world. Cognitive Command is passed when it makes sense to think of us coming to grasp a given fact by means of some substantive epistemological procedure, tracing back ultimately to the fact in

34

See note 31, above for the formulation that includes standards of acceptability.

32

question, rather than merely being sensitive to the assertoric norms of a given linguistic practice. In more recent work, Wright has clarified how he thinks the Cognitive Command constraint should be defended in the light of a more telling formulation of the problem. The overall aim, remember, is to defend the constraint from the trivialising response, that is, the claim that mere ignorance of the purported fact in question can count as a cognitive failing, and thus that all discourses in which talk of assertoric content is appropriate will trivially meet Cognitive Command. We can take the response as crystallised in what Wright calls the Simple Deduction. Thus, where P is any claim that is at most minimally true, held true by a thinker A and held false by a thinker B: 1 2 3 4 2, 4 2, 3 1, 2 1) A accepts P 2) B accepts P 3) As and Bs disagreement involve no cognitive shortcoming. 4) P 5) B is guilty of a mistake, hence of cognitive shortcoming. 6) P 8) Not-(3) -A -A -A -A - 2, 4 - 4, 5, 3, RAA - 1, 6 - 3, 3, 7 RAA I canvassed above Wrights idea that enforcing a distinction between the sense of cognitive shortcoming as failure of process, and as failure of agreement, might help in avoiding the trivialising objection. The distinction was there left intentionally vague, but is fleshed out by Wright in his analogy between genuinely representational discourses and taking a photograph
two cameras that produce different conflicting representations of the same scene must, one or both, have functioned less than perfectly, not merely in the sense that one (or both) gives out an inaccurate snapshot but in the sense that there must be some independently noticeable defect, or limitation, in the process whereby the snapshot was produced. So too, it might be suggested, with Cognitive Command: the motivated requirement is that differences of opinion should involve imperfections of pedigree:

1, 2, 3 7) A is guilty of a mistake, hence of a cognitive shortcoming.

33

shortcomings in the manner in which, one or both, they were arrived at, of a kind that might be appreciated independently of any imperfection in the result. 35

Unfortunately for that plausible thought, it looks like an amended version of the Simple Deduction can be run for the process sense of cognitive shortcoming. 36 Consider a case where we are dealing with a subject matter, where if a particular fact holds, we must be able to come to know that it does by implementing a humanly feasible process. Thus, the following Evidential Constraint (EC) holds for the subject matter in question: (EC) P -> it is feasible to know that P. We can run an argument structurally parallel to the Simple Deduction, with A, B, P as defined there, as follows: 1 2 2 1,2 1,2 1 1 1 1 (1) A believes P, B believes P, and neither has a cognitive shortcoming.- A (2) P (3) It is feasible to know that P (5) B has a cognitive shortcoming (6) P (7) It is feasible to know that P (9) A has a cognitive shortcoming (10) Not-(1) -A - 2, EC -4 - 2,1,5 RAA - 6, EC -8 - 1, 1, 9 RAA.

(4) B believes the negation of something feasibly knowable - 1, 3

(8) A believes the negation of something feasibly knowable - 1,7

Since the cognitive shortcoming in this case involves error about something which is within human cognitive grasp, there must be an error in the application of procedure and not just concerning the particular fact at issue. Thus, the line of defence proposed by Wright in Truth and Objectivity, where he suggests that the burden of proof should be placed on the side of the realist about the fact at issue to make good on an explanation of the relevant epistemology, looks beside the point. For here we have case where the

35 36

Wright (2001) p. 57-58 See Shapiro and Taschek (1996), and Wright (2001) 59-62

34

epistemology of the matter is, by hypothesis, not in question, since EC holds. But if the discourse is one in which we want to hold that a substantive realism is not an option say, talk about the twee, or smug - then it seems difficult to see how we can defend such a position. For if EC holds, then it seems that there can be no room for intelligible, blameless, differences of opinion about whether e.g. Clive James is smug. One party is always in the wrong. Given that cognitive blamelessness seems to be an essential feature of such disputes of inclination, and one that in part motivates an intuitive antirealism, it seems that Cognitive Command fails to adequately capture the key distinction in the area. Since Cognitive Command plays an important role in Alison Denhams defence of a non-reductive cognitivism about metaphor, which I discuss in Section 3.5, I have included a fairly substantive introduction to it here. I ought to briefly note, however, that Wrights current position is that his previous conception of what a dispute of inclination fundamentally consists in takes us down the wrong path. For he thinks that leads us to think of cognitively blameless situations as a kind of third option, so that we know that either (i) A knows that P (ii) B knows that Not-P, or (iii) we know that neither P nor P is knowable. In cases where EC holds, this picture clearly cannot work, at least as long as LEM is assertible. For then it looks like we are holding the following inconsistent set true: P or P P -> K(P) K (K(P) & K(P))

(LEM) (EC) (*)

In conjunction with LEM, EC lets us derive [K(P) or K(P)]. Another application of EC tells us that this itself is feasibly knowable. But then, with (*), we effectively have the claim that propositions of the form (R & S) and (R or S) can be simultaneously feasible to know. On the plausible assumption that the relevant sense of feasible knowability involves knowledge of actual fact, this is contradictory. Wrights suggestion is that instead of construing disputes of inclination as cases where we can come to know that neither side is at fault, we should instead think of them as

35

cases where we cannot, in some suitably epistemic sense, identify which side the fault is on, and, moreover, that we have no idea whether it even is metaphysically possible to come to identify such fault. In this case, Wright describes us as being in a Quandary with respect to the truth value of the disputed proposition; we do not know whether or not P, or any means or method of coming to know whether or not P; we cannot produce a reason for thinking there is any way of knowing whether or not P and do not even know whether it can be known that P, (in some strongly modal sense of can). In cases where we find ourselves in a quandary, Wright holds, we should reject LEM, while holding onto its double negation. 37 Since P or P P -> K(P) K (K (P or P)) 38

(LEM) (EC) (Q)

is an inconsistent set, 39 while (P or P) P -> K(P) K (K (P or P))

(LEM) (EC) (Q)

is consistent, the way is open for Wright to claim that an acceptable anti-realist resolution of the aporia is to reject LEM in cases where EC and Quandary hold. This provides him with the resources to finesse the Simple Deduction and its EC counterpart. Since the claim that neither A nor B has a cognitive shortcoming has the form of a
37

The background logic is assumed to be neutral as to the question of whether LEM holds, and thus to be

shared ground between classical and intuitionist logicians. This is required if the realists preferred resolution of the inconsistency, the rejection of EC for the discourse, is to avoid the charge of question begging against Wrights suggestion that in disputes of inclination we should adopt only the intuitionistically acceptable correlates of LEM.
38

Wright gives the conditions for being in a quandary as a conjunction. Since the first three conjuncts Assuming LEM as a substitution instance in EC, we could come to know that it was possible to know P

follow from the last, I have only considered it as my (Q).

39

or P, contra Q.

36

negative existential statement, the attempted reductio only takes us to a doubly negated existential conclusion. Since this is not intuitionistically equivalent to the existential claim that a cognitive shortcoming exists, Wright thinks the bullet can be safely bitten without incurring unwanted realist commitment to the existence of an apparently unknowable state of affairs settling the dispute beyond our ken. (Of course, he is also committed to denying that such a state of affairs fails to exist, but given intuitionistic logic, these claims are compatible). We may seem to have come a long way from metaphorical meaning. But the relevance of our current discussion actually isnt hard to see. We will want to know whether we can safeguard the intuition that there can be genuine disputes of inclination about metaphorical content. This involves giving an account of how such a dispute can genuinely be a dispute (contra e.g. the Indexical Relativist, and the Expressivist) and genuinely one of inclination (contra the Realist and Error theorist, both of whom are committed to holding that the facts are holding us to account in some sense, and thus that inclination can only spring from ignorance of one kind or another). The promise of Wrights account was that minimalism about truth could guarantee the first, and failure of Cognitive Command account for the second. The Simple Deduction forced a clarification of the sense in which inclination holds sway in the kind of cases were interested in. The key result turned out to be the claim that, in cases where EC holds, ignorance is involved in disputes of inclination, but not in the way that either the Realist or the Error Theorist claimed. Rather, Wright claimed that while the assumption that there was no cognitive shortcoming led to a contradiction, this did not automatically supply us with a warrant to assert that such a cognitive shortcoming must exist in every case. While the Simple Deduction threatened to trivialise the Cognitive Command constraint by reducing to absurdity the idea that the world could be somehow neutral on the issue of whether P or not P, Wrights argument trades on the fact that our state of information warranting assertions can be so neutral, even when we have a warrant to assert that not-(P or Not-P) does not hold. If Wrights suggestion can be made good, then we have the promise of a stable, if revisionary, account of how a genuine dispute about metaphorical meaning can involve no attribution of fault to either side, even in principle. Of course, this only evens the score with e.g. a determined realist, who already claims to have such an account, at least 37

if attribution of fault is heard as implying something like in principle being able to tell which side is at fault. Unlike the realist, however, Wright is not committed to claiming that such a fault exists. He may be a hairs breadth away, given his acceptance of the claim that it isnt the case that no such fault exists, but perhaps we should accustom ourselves to hair-splitting in philosophy. Nevertheless, I shall tend to confine my later discussion to Wrights original treatment of the Cognitive Command test. There are a number of reasons for this. First, I want to avoid prejudging the question of whether EC holds for meaning, as much as possible. Though I offer an extended discussion of Davidson in the next chapter, I explicitly aim not to appeal to doctrines, such as the transparency of meaning, that might not be common ground between rival philosophical treatments of semantics. Second, most of what I say elsewhere about Cognitive Command can be translated fairly straightforwardly into the new approach, and it would grate a little to do so explicitly on every occasion. Third, no sensible realist worth his salt is likely to accept the kind of revision of logic that Wright recommends. Given the reservations about the default status of anti-realism I raise in Section 1.10, and the holistic methodology I recommend in its place, it seems less likely that such revision will turn out to be acceptable. Fourth, my concern is often with e.g. Denhams and Hopkins treatment of the cognitive status of metaphor, which predated Wrights recent discussion. 40 Finally, the difficulties raised by Wrights admirable efforts to clarify the relevant sense of genuinely cognitive should not lead us too quickly to reject this notion as unclear or misguided. It may well prove more useful to take the notion as primitive, with Cognitive Command having the status of an intuitive introduction, rather than an explicit definition. In any case, it is of independent interest to grant such an intuitive sense, and explore whether ascriptions of metaphorical meaning might meet it.

1.9

Width of Cosmological Role

The last of the cruces that Wright thinks realism debates should focus on is Width of Cosmological Role. This has received least attention in the literature; perhaps
40

I discuss Denham in Chapter 2 and 3, and Hopkins in Chapter 4

38

surprisingly, given that Wright sees it as developing and generalising a widely discussed view of Gilbert Harmans, which contrasts the explanatory efficacy of physical facts and moral facts. 41 Wright holds that the most interesting element of the contrast concerns not the nature of the explanation involved, but rather the nature and diversity of the explananda. Any discourse which fulfils the minimal conditions requisite for talk of assertoric content to be appropriate, will, according to Wright, allow certain types of explanation. Appeal to the property of being funny, to take a stock example of a merely minimally truth apt discourse, will allow us perfectly respectable explanations of e.g. peoples television viewing behaviour, or the contrasting sales of certain novels. Harmans insight, according to Wright, is not best captured by the thought that cases where anti-realism is the appropriate stance are cases where talk of explanation is wholly inappropriate, but rather that there are certain extra explanatory liaisons that come into play in realist discourses. One of the most interesting things about Wrights account of realism is his insight that there are a range of quite distinct and separable ways in which the realist might attempt to give substance to his intuition that the facts in question are somehow constitutively independent from human interests and responses. Investigation of the variety of ways that such facts can enter into quite different species of explanation provides another such focus for investigation. One of the notable things about the type of explanatory role played by the comic properties of television programmes and novels, for example, is that it involves a route that goes through a certain kind of human response. The comic properties of things, when they explain the features of a certain situation, explain it in virtue of their having certain effects on certain human beings, suitably equipped with senses of humour. In contrast, explanations that appeal to facts which we tend to think of in realist terms the fact that a certain table has a certain shape, for example while they may well involve certain human responses in certain cases, need not do so in all. That the table has a certain shape can enter into explanations of why I choose it as an example in a physics tutorial, but also why it remains stable under the Earths gravitational pull, why the dog likes to lie under it, why it wont fit through the door. These latter explanations are not mediated by human responses, judgements or values,

41

See Harman (1997) Chapter 1

39

and their existence thus helps to give substance to realist claims about the independence of facts about the shape of the table from human subjectivity. Another key test for establishing the realist standing of a discourse, then is Width of Cosmological Role Can the putative realm of fact enter into explanations in a way that isnt secured merely by discourse about it being apt for minimal truth? It isnt wholly clear what explanatory role is secured merely by possession of the relevant kind of assertoric content. As noted above, Wrights discussion suggests that one key element is the capacity to enter into explanations that arent mediated by the fully conceptualised attitudes of participants in the discourse. Something clearly seems right about this; if citing the existence of the relevant fact to explain why a subject holds a given attitude is, in a sense, merely a long winded way of noting that the subject is confident that the epistemic standards for a syntactically assertoric discourse have been met, then one would not expect that fact to be able to explain the behaviour of entities that are completely insensitive to such standards. On the other hand, there are complicating factors. Imagine that the world is Cartesian, and that concept users have non-material Cartesian egos associated with their bodies. Wouldnt it be possible that such egos played an explanatory role only in ways that were mediated by e.g. goals and beliefs? What about those who endorse David Lewiss view of possible worlds; surely realists, if anybody is? Does citing other possible worlds help explain what happens in this one, in a way that isnt mediated by our modal sense? It is of course wholly possible for talk of e.g. Cartesian egos, or other possible worlds to meet other of the tests, and yet fail this one, while still deserving to be thought of as realist. But its not clear that the suggested cases genuinely pose a problem for Width of Cosmological Role properly understood. For one thing, it ought to be enough, perhaps, that the facts in question can potentially enter into a wider range of explanations than that secured by minimalism. For example, even if only certain kinds of causal relations actually do hold between minds and bodies in the Cartesian world, it ought to have been possible that a broader set of causal relations held between the material and immaterial elements of the world. If a Cartesian ego can raise my arm, then it ought to be possible 40

for a vase to be caused to break just by having such an ego think a certain thought, for example. Secondly, although the Cartesian case essentially involves the presence of conceptualised attitudes, its not clear that it involves them in the right way. In the case where talk of explanation by a given fact is really just a matter of harmless paraphrase, it is essential that conceptualised attitudes mediate because these are required for sensitivity to the relevant communal epistemic standards. The Cartesian case makes conceptualised attitudes essential, but only as an accompanying, dependent feature of a metaphysical substratum, not as the genuine source of the facts in question. The case of possible worlds might invite a more direct reply. Perhaps the existence and nature of close possible worlds can partly explain e.g. why an animals perceptual capacities deliver information about its environment, or why non-actually-dissolved salt is soluble. In any case, the primary purpose of this chapter is, as noted, to outline Wrights tests for realism, and not to pursue specific debates further. Since the case of metaphorical meaning seems unlikely to enter into the explanatory relationships in question, we neednt concern ourselves over much with the best formulation of the constraint.

1.10 Costs and benefits of Wrights account

At the end of section 1.2, when discussing Devitts account of realism about the external world, I asked the following questions: 1. Is it unproblematic to extend this kind of general picture of realism to debates between realists and anti-realists in other areas of discourse? 2. How might actual debates between realists and anti-realists be profitably prosecuted? I suggested that Devitts theory did not invite an unproblematic extension to wide range of disputes, and did not suggest a framework in which such disputes could proceed fruitfully. Wrights account, in contrast, offers a welcome topic-neutrality, and the various tests, schematic as they are, have seemed to many philosophers to provide a 41

focussed dialectical structure in which longstanding debates might advance. Moreover, the general tactic, of establishing key constraints or tests, which serve to signal the appropriate metaphysical attitude one ought to take to a given discourse, is one that obviously can be profitably extended beyond the set of cruces that Wright himself identifies. For example, in recent work Christopher Peacocke has suggested the following indicators, each of which is relevant for the case of metaphorical meaning: a) Do true statements in the area have an a priori source? b) Is some role in causal explanation essential either to the truth of statements in the area in question, or to our having concepts of that area? c) Are statements in the problematic area predications of a property which also features in predications outside the problematic area? If so, does grasp of this identity of properties play some role in understanding statements in the problematic area? 42 This generalised, pluralist, fine-grained, diagnostic approach to disputes about the realist standing of a wide variety of discourses is one I find conducive, and will comprise the background methodology to my project here. I do, however, want to distance myself from certain other elements of Wrights recommended framework for the execution of disputes in particular areas. There is of course room for local disagreement about e.g. the exact formulation of the relevant cruces. I have some sympathy, for example, with a complaint that Williamson and Blackburn make about the inclusion of a prioricity in the formulation of e.g. Cognitive Command, and the Order-of-Determination test. 43 It might well seem that this inclusion counts against the generality and topic-neutrality of the approach, and that, given that there are philosophers who reject the claim that we can know anything a priori, a more neutral formulation should be given rather in terms of e.g. best overall explanation, with the putative a prioricity being a special case of the generic requirement, to be made out on its own terms by those sympathetic to that species of

42 43

See Peacocke (1999) Section 2.4 See Williamson (1996b) p 905-907, Blackburn (1998) p. 172

42

knowledge. My divergence from Wrights framework concerns a more fundamental structural element, however. He writes
A truth predicate, I have argued is one which satisfies a small set of basic principles most centrally, certain platitudes linking truth with assertion and negation. The characteristics possessed by any satisfier of these principles are the only characteristics essential to truth. Moreover, they are insufficient to motivate an intuitive realism about a discourse in which such a predicate applies. But a particular satisfier may, of course, have other characteristics as well. A basic anti-realism minimalism about a discourse contends that nothing further is true of the local truth predicate which can serve somehow to fill out and substantiate an intuitively realist view of its subject matter. Because of its unassuming character, this minimalism, I suggested, should always be viewed as the default stance, from which we have to be shown that we ought to moveIt is realism which must try to make good its case, by showing that minimalism about the relevant discourse is wrong showing that the minimal platitudes leave out features of the local truth predicate which substantially justify the rhetoric of independence, autonomy and full-fledged cognitive interaction by which realism pretheoretically defines itself. 44

Wright thus takes it that the unassuming character of minimalism shifts the burden of proof onto the realist in any given debate. I have several worries about this move. Firstly, it is not clear that this way of presenting the ground rules of the debate actually corresponds to the way in which Wright actually proceeds. For example, in the case of the Order-of-Determination test, the burden of proof seems to be on the anti-realist. It is she who has to demonstrate that the coextension of fact and best opinion holds a priori, can be specified in philosophically substantive terms, etc. Presumably, by Wrights lights, until this demonstration is carried out successfully, we have no warrant to shift from an extension-reflecting reading of the relevant biconditional. But that seems to signal that an intuitive realism, rather than anti-realism, comprises the default position in considering whether this test is met. 45

44 45

Wright (1992) p. 174 I owe this point to John Divers.

43

A second worry is that it is simply not clear why considerations of modesty alone should determine something as dialectically important as burden of proof. Why shouldnt other theoretical virtues parsimony, simplicity, coherence with other established theories, prior plausibility, elegance also be taken into account? Take the question of realism about the external world. Very few philosophers would want to insist that scepticism of a particular kind is the default position, even if, in a particular case, the sceptics claims assume much less than our own. A standing worry is that an overly modest starting position can itself serve to rob a realist of the resources that she requires to secure his case. Wright might, with some justification, complain that such a line of argument misplaces the sense in which minimalism is unassuming. It is not merely that the minimalist makes claims that are more ontologically parsimonious, or epistemically cautious, as the sceptic claims to do, but that she can explain features of our thought and practice in a given area in terms that respect - but dont exceed our pre-theoretic intuitions. (Contrast the revisionary implications of scepticism). In a sense, the minimalists demand is only that the realist justify his rhetoric, and what could be intrinsically objectionable about that? Nevertheless, even if Wright is correct to hold that many of the platitudes we associate with the notion of objectivity are in themselves inadequate to motivate a realist construal, I think it is still fair to ask why this consideration should always be considered forceful enough to land the realist with the burden of proof. In cases where we pre-theoretically dont associate a robust sense of objectivity with mere philosophical hyperbole say, in the case of the past, or the existence of other minds why should the failure of the realist to identify the precise features of the discourse that justify such a stance automatically entail that we ought to concede that the anti-realist is, so far, vindicated? The revisionary implications of such a position seems to me rather to demand, at the minimum, that the anti-realist give us good grounds for abandoning our intuitive realism, rather than merely taking the travails of the realist as motivation enough for persisting with a minimalist view. In my discussion so far I have been conceding, for the sake of argument, that the minimalist position is unassuming. Actually, however, this is far less clear than Wright makes it sound. I have already noted that several of the dialectical cruces that Wright identifies are formulated in terms of a notion of a prioricity that simply isnt common 44

ground among participants in the debate. Similarly, Timothy Williamson has pointed out that Wrights characterisation of warranted assertibility assumes that warrant for assertion demands something less than knowledge. 46 Wright defends this claim as uncontroversial given its prevalence in a certain tradition in epistemology, 47 but this seems to me to be less than is required if minimalism is genuinely to appeal only to platitudinous common sense. (Actually, even if Wrights reply were sufficient in this context, an unsympathetic reader might point to the reliance on the deliverances of common sense as itself manifesting a philosophically substantive commitment). Another group of critics have worried over the issue of whether Wrights position makes non-platitudinous demands in the philosophy of mind. 48 Given the extent of this kind of disagreement, we might wonder whether staking a claim to a philosophically neutral default position is really worth the candle. To determined realists about, say, the past, Wrights characterisation of the burden of proof is bound to appear simply as special pleading on behalf of the anti-realist. In any case, I shall not appeal to this element of Wrights framework in my discussion of realism about metaphorical meaning. Rather, I hold that the debate ought to proceed in a more even handed and holistic manner as it inevitably actually will, even for philosophers who are sympathetic to Wrights claims about burden of proof. The challenge to each participant is to reconcile the data, in the form of the results of the suggested tests, with their own preferred epistemology, semantics and metaphysics for the area in question. 49 Of course, as in many cases, the proposed data may end up being reinterpreted, reformulated or even rejected, and there may well be more than one defensible overall position at any given stage of the process. I thus preserve the admirable generality, topic-neutrality and dialectical focus of Wrights account, but

46

Williamson (1996b) p.907-908. Williamson provides a sustained defence of the opposing claim in his Wright (1996) p. 934-935 See Smith (1994a), Jackson, Oppy and Smith (1994) and Blackburn (1998) for the prosecution; Divers I do not mean to exclude e.g. their philosophy of mind, science, logic, ethics, religion, etc., but rather to

(1996a)
47 48

and Miller (1994) and (1995), Wright (1999) for the defence.
49

assimilate these under the broader headings outlined above.

45

substitute a general rush for coherence for Wrights advance from a default minimalism. 50 It might be difficult to see how to reconcile such a methodology with Wrights idea that degree of realism is a matter of the relevant features of the local truth predicate. After all, doesnt that commit us to a semantics first approach which is inconsistent with the kind of dialectical free-for-all endorsed above? Im not wholly clear on whether there is any genuine difficulty here, but to the extent that there is, I would sacrifice the relevant semanticism. None of the tests that I have outlined seem to me resistant to rephrasing in terms that omit the notion of truth. The question of whether the proposed objects, facts or states of affairs might obtain without our having any evidence for them so doing, for example, seems to be one that only appeals to metaphysical and epistemological notions. If each such test can be rephrased in terms that alethiphobes can endorse, then no objectionable semanticism seems to be being presupposed. A more substantive worry concerns a kind of self-reflective worry that arises for the philosophical treatment of realism. What happens if the philosophical facts, the facts that philosophers interested in realism about a given area attempt to establish, themselves are only apt for a minimalist treatment? After all, such facts seem to admit of permissible disagreement, at least in the sense that there can be long-lasting, entrenched disputes between sincere and willing participants, that we dont know how to resolve. Evidence that we can be in a quandary with regard to e.g. moral facts will surely carry over to the philosophical case. Philosophical facts do not seem to display Wide Cosmological Role, and it seems to me to be an open question whether we can make sense of their outrunning the potential evidence that we might have for them, or whether a tracking epistemology is appropriate for them. If such facts deserve a less than maximally realist treatment, however, how can any discourse deserve a more robust one? The claim that e.g. talk about the external world deserves a maximally realist construal will itself only be less than maximally realist. If it is e.g. cognitively permissible to reject such a claim, then how seriously can we take the supposed realist

50

This methodological position is hardly novel, and bears a fairly close resemblance to those outlined in

Peacocke (1999) and Miller (forthcoming a and b). Blackburn (1984) pp. 3-7 is an early advocate of a broadly coherentist methodology.

46

result which claimed to have established it? Doesnt the reflection that an intuitive antirealism was equally cognitively respectable itself serve to undercut the idea that the world is really out there, wholly independently of any decision we might have made on the matter? It seems that the degree to which one is anti-realist about the claims of philosophy to establish degrees of realism sets a limit to the degree to which one can be a realist in any other area, threatening a much wider anti-realist victory. It is a familiar phenomenon in epistemology that a dialectical draw favours the sceptic; here we seem to have the beginnings of a line of thought that would establish a similar outcome for anti-realism more generally. To use a phrase of Wrights drawn from a different but related context that of Wittgensteins rule-following considerations the worry has to be that any form of restricted realism about the outcome of philosophical theorising sets an upper bound on the robustness of the realism which is available anywhere. 51 I think this is a serious worry, and it is not one I have an answer to. For the purposes of this project, I intend to ignore it. It is, of course, fairly common methodological practice to ignore deep and dangerous problems while pursuing more limited concerns. In the case of metaphorical meaning, however, where many of the most plausible positions involve a substantially anti-realist component, the move is perhaps slightly more principled than normal. Even if the upper bound to realism does turn out to be determined by the status of philosophy itself, there is still room to examine the interesting range of local debates which fall short of that limit. The case of discourse about metaphorical meaning is an example of this latter class, in which one of the best argued positions that presented by Donald Davidson maintains that such talk falls short even of the modest standards that are requisite for minimal truth. It is to elaboration of this position that I now turn.

51

Wright (1992) p. 212

47

48

2. Davidson, Meaning and Metaphor

2.1

Introduction

Davidson is an anti-realist about metaphorical meaning. That thesis is true, but the sense or rather senses in which it is true have not always been easy to grasp. The motivation for his position has not always been clearly brought out, for all the wealth of commentary that his first, seminal, paper on the topic attracted. 52 Nor, relatedly, has the connection between Davidsons anti-realism and broader philosophical concerns with the nature of meaning been satisfactorily elucidated. There are, of course, honourable exceptions. 53 But the insights and blindspots, costs and benefits of Davidsons position have yet to receive a fully satisfactory treatment. My aim here is to make a pass at it from within the framework of a broadly Wrightian conception of the realism debate, in the hope that a new angle of illumination might reveal its contours more clearly. My aim, however, is more substantive than exegetical. While I will aim to characterize and interpret Davidsons position in a manner that is broadly consistent with his compressed remarks on metaphor, my primary concern will be to present his arguments in what I take to be their strongest and most interesting form. I will begin by examining the sense in which we should think of Davidson as an antirealist about metaphorical meaning, and then go on to set out an argument for his position. The argument as presented is not wholly explicit in Davidsons work, but draws upon recognisably Davidsonian themes and principles. I will then go on to clarify, motivate and evaluate the status of the premises of that argument.

52 53

Davidson (1984a) The best discussions of the position that I know are those included in Stern (2000), White (1996) and

especially White (unpublished). I am very grateful to Roger White for allowing me access to this latter work, from which I have benefited immensely.

49

2.2

The Character of Davidsons Anti-Realism

Davidson describes the central thrust of his paper in the following terms:
This paper is concerned with that metaphors mean, and its thesis is that metaphors mean what the words, in their most literal interpretation, mean, and nothing more. Since this thesis flies in the face of contemporary views with which I am familiar, much of what I have to say is critical. But I think the picture of metaphor that emerges when error and confusion are cleared away makes metaphor a more, not less interesting phenomenon. The central mistake against which I will be inveighing is the idea that metaphor has, in addition to its literal sense or meaning, another sense or meaningif I am right, a metaphor doesnt say anything, beyond its literal meaning (nor does its maker say anything, in using the metaphor, beyond the literal). This is not, of course, to deny that metaphor has a point, nor that that point can be brought out using further words. 54

He summarizes his contrasting view of how metaphor works its wonders as follows
I depend on the distinction between what words mean and what they are used to do. I think metaphor belongs exclusively to the domain of use. It is something brought off by the imaginative employment of words and sentences and depends entirely on the ordinary meaning of those words and hence on the ordinary meanings of the sentences they comprise. 55

And later he writes

The central error about metaphor is most easily attacked when it takes the form of a theory of metaphorical meaning, but behind that theory, and statable independently, is the thesis that associated with a metaphor is a definite cognitive content that its author wishes to convey and that the interpreter must grasp if he is to get the message. This theory is false as a full theory of metaphor, whether or not we call the purported cognitive content a meaning. 56
54 55 56

Davidson (1984a) p. 245-246 Ibid p. 247 Ibid p. 262

50

Three theses seem to be clearly endorsed in these passages: The words that are employed in metaphorical utterances do not thereby acquire new, distinctive, metaphorical senses, but rather play the same semantic role as they do in straightforwardly literal utterances. The sentences that are employed in metaphorical utterances do not thereby become associated with new, distinctive metaphorical contents, but rather continue to have their ordinary literal meanings, which are determined by the literal meanings of the words that make them up (together with the mode of composition of such words). Metaphorical utterances are not only, or typically, or characteristically, vehicles for the communication of a distinctive cognitive content. Theses (D1-D3) are clearly expressions of an anti-realist view regarding the existence of distinctive metaphorical meanings, whether these are thought of as associated with words, sentences or communicative acts. But what is the character of this anti-realism? In evaluating the strengths and weaknesses of a given realist or anti-realist stance with regard to a particular set of issues, it is often useful to begin by identifying the position as an instance of a generic kind. This allows us to clarify key aspects of the position; to identify promising analogies between the case at issue and other, structurally related, debates; and to look for shortcuts when developing a given dialectic. How, then, should we classify Davidsons anti-realism? Three broad interpretative paradigms suggest themselves Davidson is a reductionist about metaphorical meaning. He rejects the idea that there are sui generis metaphorical meanings by simply identifying metaphorical meanings of words or sentences with the corresponding literal meanings. (This paper is concerned with that metaphors mean, and its thesis is that metaphors mean what the words, in their most literal interpretation, mean, and nothing more). Davidsons position is 51

(D1)

(D2)

(D3)

Reductionism

analogous to e.g. a primary quality view of colour, or a typeidentity view of the mind. Davidson is a kind of expressivist, emotivist, non-cognitivist or non-factualist about metaphor. 57 He rejects the idea that there are sui generis metaphorical meanings, replacing the idea that metaphors typically work by communicating propositional contents with the idea that metaphor is a distinctive type of speech act. (I depend on the distinction between what words mean and what they are used to do). Davidsons position is analogous to Blackburns moral expressivism, or Wittgensteins view of selfascriptions of intentions. Davidson is an error theorist about metaphorical meaning. He holds that talk of sui generis metaphorical meaning makes sense, but attributions of it to utterances or their parts are globally false. (The picture of metaphor that emerges when error and confusion are cleared away makes metaphor a more, not less interesting phenomenon). His position is analogous to Mackies view of moral facts, or Fields view of mathematics. Each of these suggestions has something going for it, and I will endeavour to say something about each in turn. However, I think it is important, both in getting to grips with Davidsons arguments for his anti-realism, and for the more general project of understanding how the debate about metaphorical meaning relates to other realism debates, that we see that the third interpretation is the correct one.

Expressivism

Error theorist

2.3

Reductionism and Expressivism

There is clearly a sense in which the Reductionist reading of Davidson is right to insist that he holds that metaphorical meaning is simply literal meaning. After all,
57

For my purposes in the chapter, I will treat these terms as merely stylistic variations of one another.

52

metaphorical meaning, in one sense at least, can surely be equated with meaning possessed or expressed by the metaphor, and Davidson openly claims that the only meaning expressed by the metaphor is the literal meaning. This is rather misleading, however. Firstly, the sense in which Davidson is using metaphorical meaning, when e.g. he talks about attacking a theory of metaphorical meaning in the third quotation above, is not the sense which is appealed to by the Reductionist. The phrase doesnt simply equate to meaning expressed by the metaphor when understood in its intended sense. Rather, it is equivalent to something like result of a meaning-shift undergone by sub-sentential parts in virtue of being included in a metaphorical utterance. Understood as it ought to be, it is clear that Davidson is not identifying the metaphorical meaning of the utterance or its parts with the corresponding literal meaning, and that any suggestion to that effect at the beginning of What metaphors mean can only be the result of an ill-timed rhetorical flourish. Secondly, the Reductionist construal simply mischaracterizes the nature of the error Davidson takes his opponents to have committed. Consider the supposedly analogous positions in the philosophy of mind or colour. Someone who endorses the primary quality view of colour, or a type-identity view of mental states and properties, will typically hold that these physicalistically respectable properties have always been the intentional objects of our thought and talk. Pain was always C-fibre stimulation, or whatever; it is simply that we didnt recognise it as so, as a result of a subjective mode of presentation, or a misleading philosophical mythology. That thesis, in part, is what distinguishes the Reductionist position from its Error theoretic rival, which maintains that our pre-theoretic thought and talk about the area in question simply has nothing answering to it. Davidson isnt claiming, however, that we, and his opponents, have always been referring to the literal meaning of the words and sentences in question, when we invoke the notion of metaphorical meaning. On the contrary, his claim is that we have been conflating what the metaphor makes us notice, or calls to our attention, or what is brought to mind when interpreting the metaphor, with something correctly thought of as a meaning. Construing Davidson as a Reductionist, therefore, obscures, rather than

53

illuminates, the nature of his position, and also the analogies that hold between it and related standpoints in other philosophical debates. What then of the reading that thinks of Davidson as an Expressivist? This interpretation agrees with the Reductionist one insofar as they both have Davidson rejecting sui generis metaphorical meanings, but rather than seeing him as arguing that metaphorical meaning just is literal meaning, the Expressivist account has him denying that making metaphors is usefully assimilated to assertions of any type of content. Rather, some other linguistic act, some distinctive use of language is involved. An early sketch of a nave Expressivist position is provided by Beardsley
according to the Emotive theory, a word has meaning only if there is some way of confirming its applicability to a given situation roughly, only if it has a clear designation. For example, the sharpness of a knife can be tested by various means, so that the phrase sharp knife is meaningful. We may also suppose that sharp has some negative emotive import, deriving from our experience with sharp things. Now, when we speak of a sharp razor or a sharp drill, the emotive import is not active, because these phrases are meaningful. But when we speak of a sharp wind, a sharp dealer, or a sharp tongue, the tests for sharpness cannot be applied, and therefore, though the individual words are meaningful, the combinations of them are not. In this way the emotive import of the adjective is released and intensified. 58

For this type of Expressivist, then, the metaphor maker is not dealing with any genuine content when she utters the metaphorical sentence, since there is just none there for her to be asserting. Rather, she must be doing something else with the sentence, such as trying to affect her interlocutors attitudes or emotions in a particular way. Numerous well known problems lurk here. Firstly, the notion of emotive import is unclear, and does not account for how the same word or phrase can be used with
58

Beardsley (1958) pp. 134-5, quoted in Soskice (1985) p. 26. Beardsley characterises the possession of

content along verificationist lines, but this is strictly inessential to Expressivism, nave or sophisticated. Some criterion for the possession of genuine content is presumably required, however. That the fact that the nave Expressivist cannot take successful embedding in logical and propositional attitude contexts as such a criterion may well constitute her greatest problem.

54

opposing emotive import, as, for example when we use the commendatory phrases a sharp mind or a sharp wit along with the denigratory ones mentioned in the passage above. 59 Secondly, we lack an explanation of the analogies between the inferential behaviour of genuinely contentful statements and that of merely emotive ones. Thirdly, it just seems plain phenomenologically wrong to claim that getting a metaphor is merely a matter of being prompted into a certain affective state. The nave Expressivist admittedly has some room to manoeuvre, especially with respect to this last problem. She may hold, for example, that what metaphor brings about need not be purely affective, but can include items which bear content. In this case, making a metaphor is similar to, say, uttering nonsense syllables in an attempt to make the spy tapping your phone believe that you have a secret code. Your utterance possesses no literal meaning in itself, but is intentionally produced to cause another to have thoughts which do have a content. Nevertheless, deep problems remain with the nave Expressivist account of the inferential behaviour of metaphors, given the thesis that they lack literal content. Even relatively extreme poetic metaphors seem to be able to embed in conditionals (if philosophy is just showing the fly the way out of the fly-bottle, then it does not reveal deep metaphysical truths about the structure of reality) and under negation (Dan Quayle was no Kennedy). Expressivists such as Blackburn and Gibbard have aimed to discharge the explanatory burden of accounting for such behaviour in the moral case, and it cannot be ruled out a priori that an account could be given in the metaphorical one. However, it is fair to say that we currently lack such an account and have little idea of how to get hold of one. 60 A simple minded Expressivism, then, will hold that truth conditional contents are not involved at all in metaphor, and thus incur the difficult burden of explaining why such expressions can embed in e.g. logical and propositional attitude contexts. A more plausible line for the Expressivist is available, however. Typically, this will involve
59 60

Beardsley makes this point in his discussion of the position. For an introduction to the debate see Blackburn (1984) Ch. 6, Blackburn (1988), Gibbard (1990). For

criticism of Blackburns project see Hale (1986) and (1990). A useful discussion and overview is provided in Miller (2003). I sketch a pseudo-fictionalist solution to the embedding problem in Chapter 3.

55

granting that the propositions in question possess a truth conditional content, which is apt for assertion, but denying that the characteristic purpose of making a metaphor is to assert something. A sophisticated Expressivist might draw an analogies between metaphor and linguistic phenomena such as performatives. In using a performative such as I hereby promise to pay you 10 , I arguably do not assert anything, that is, represent myself as believing that I am making a particular promise. Rather, I typically make the promise by uttering the words. This is not because I could not use the sentence to make an assertion: the sentence has a content, determined by the conventional rules of English, which I could sincerely present as true. It is this propositional content which enables the sentence to be embedded in logical and propositional attitude contexts, and which explains the relation between the promising and temporally related descriptions (I will promise her the money, If I promise her the money, then I will give her the money). In a similar way, according to the sophisticated Expressivist, the literal vehicle of a metaphor may well have a truth conditional content available for assertion, but our typical practice of making such claims is not best represented as assertoric. The sophisticated Expressivist position thus begins to address the problem of accounting for the inferential behaviour of metaphor by allowing that metaphors usually have a truth conditional content associated with them, which allows them to embed in conditionals, etc. However, in making metaphors, we do not assert this content, but use the sentence in question in some other way. This might be to perform a sui generis speech act (often referred to as metaphorizing), to perform some other distinctive illocutionary act (such as inviting or encouraging the interpretation of the metaphor), or simply to achieve some aim of the speakers (to draw attention to certain features of a situation, or to make us think of somebody in a certain way, for example). The story, applied to the case of metaphor, typically looks something like this. The maker of a metaphor utters a sentence, which typically possesses literal content, and is thus apt for assertion. However, the speaker does not assert the sentence, which in the

56

normal case will be obviously literally false. 61 Rather, the metaphor maker implicitly invites, via contextual clues, shared adherence to Gricean conversational maxims, etc, the interpreter of the metaphor to look for a certain set of similarities between two or more objects, events, situations etc. The metaphor maker will typically intend that the interpreter get hold of some central similarities or analogies, but need not, and can in any case intend that the interpreter have an element of free play in coming up with ways in which the metaphor is apt. It is this aspect of metaphor making, the implicit invitation to make the most of it, which accounts for the oft cited open-endedness of metaphor. Of course, the metaphor may also prompt various affective states, and be intended by its maker so to do. This position has proved attractive to a number of careful philosophers, both as an interpretation of Davidson, and as a plausible account of metaphor in its own right. Simon Blackburn, for example, writes
[Metaphors] are typically couched in indicative sentences, certainly governed by norms of appropriateness, found in complex embeddings, yet certainly not intended as straightforward cases of truths or falsehoods. This is how the expressivist says it is in more controversial examples, such as commitment to conditional, moral or modal claims. 62

Of course, supporters of such a position accept that much still has to be filled in; the nature of the contextual clues, the nature of the actual mechanism by which we get hold of the various relevant similarities, the scope for invention, etc. As far as the semantics and pragmatics of metaphor go they argue, however, the broad picture is tolerably clear. Metaphor is a way of using a pre-existing, truth apt content to prompt, encourage or invite contemplation of or active investigation into some open-ended set of similarities. I am in agreement with much of what this type of position has to say, both exegetically and substantively. It is certainly true, for example that Davidson argues against the idea that the typical job of metaphorical speech is to convey a coded message, and
61

Or perhaps nonsensical: see White (1996) pp. 204-226. I argue against the nonsensical view of Blackburn (1998) p. 159.

metaphor in Chapter 3.
62

57

compares it instead with distinctive uses of language such as joking, promising, hinting, criticizing and, in an extended passage, lying
Now, let me raise a somewhat Platonic issue by comparing the making of a metaphor with telling a lie. The comparison is apt because lying, like making a metaphor, concerns not the meaning of words but their useThe parallel between making a metaphor and telling a lie is emphasized by the fact that the same sentence can be used, with meaning unchanged, for either purpose. So a woman who believed in witches but did not think her neighbour a witch might say Shes a witch, meaning it metaphorically; the same woman still believing the same of witches and her neighbour but intending to deceive, might use the same words to very different effectWhat makes the difference between a lie and a metaphor is not a difference in the words used or what they mean (in any strict sense of meaning) but in how the words are used. Using a sentence to tell a lie and using it to make a metaphor are, of course, totally different uses, so different that they do not interfere with one another, as say, acting and lying do. In lying, one must make an assertion so as to represent oneself as believing what one does not; in acting, assertion is excluded. Metaphor is careless of the difference. 63

The Expressivist reading of Davidson, then is correct to attribute something like the following thesis to him A metaphorical utterance involves a distinctive use of a segment of literal language, which does not alter in meaning by being so used, typically in order that a speaker might intentionally provoke or invite a certain view of the subject matter of the metaphor, (although it is also possible e.g. that the purpose of such an utterance might be to make us appreciate some fact, and that the metaphor itself may prompt or inspire us in ways that the speaker does not actively intend.) 64 I believe we will go astray, however, if we conflate (i) Davidsons position with respect to the question of realism about metaphorical meaning with (ii) his positive account of how metaphor works its wonders. The latter, but not the former, is broadly analogous
63 64

(E)

Davidson (1984a) pp. 258-259 For support for the idea that Davidson subscribes to (E) see his (1984a) p 261-264. (The quoted phrases

in my formulation of that thesis can also be found there.)

58

to a typical non-cognitivist account of the function of e.g. moral or modal language. The purpose of metaphor, like the purpose of moral language on an Expressivist account, is often to do something, not to say something. But that should not mislead us into characterizing Davidson as an Expressivist about metaphorical meaning. His position with respect to metaphorical meaning is structurally significantly different from Expressivism about moral or aesthetic facts, whether nave or sophisticated in form. Alison Denham is one contemporary writer who falls into the trap of mistaking surface similarity in the treatment of linguistic role for depth of correspondence of metaphysical stance. 65 In her Metaphor and Moral Experience, a work whose partial aim, like my own, is to approach metaphor from within the compass of a Wrightian approach to realism, she writes
Some claim that it is not possible to state the truth-conditions of moral discourse in nonevaluative terms; the parallel thought [in the case of metaphor] is that it is not possible to state the truth conditions of some metaphorical discourse in non-figurative terms. 66

Again, in her introduction to that volume, she writes

The parallels [between moral judgements and metaphors] run deep, as can be seen by reflecting for a moment on the kinds of questions which often arise with respect to each. For instance: 1) Are moral judgements candidates for the assignment of truth values, or are they actually covert expressions of sentiment, not in themselves either true or false? Compare: Are metaphorical expressions genuine, truth-apt judgements, or do they serve merely to evince emotion and the play of imagery? 2) Can moral concepts be analysed in wholly non-evaluative terms? Compare: can metaphorical expressions be reduced to literal paraphrase? 67
65

Other theorists of metaphor who sail close to the wind include Blackburn (1998), who I quoted above;

Soskice (1985) pp. 26-27, who calls Davidsons position non-cognitivist, and notes its similarities to Emotivism; and Moran (1997) p. 260, who draws comparisons between Davidson and moral noncognitivists. Unlike Denham, however, most of these theorists manage to refrain from sailing too close.
66 67

Denham (2000) p. 283. n. 5. Denham (2000) p. 2. She goes on there to note two further deep parallels concerning, firstly, the

epistemological immediacy of facts about morals and metaphorical meanings, and secondly, the phenomenon of persistent, conflicting responses in each case.

59

I want to argue that, in fact, no relevant parallel of the kind that Denham appeals to actually holds between morals and metaphors. A bad analogy has led her astray, and led her to ignore the genuine differences between dominant forms of anti-realism about metaphorical discourse on the one hand, and moral discourse on the other. To see where and why the putative parallel fails, let's think more closely at what a discourse is, in the sense in which it has come to be used in debates about realism. For Denham, as for Wright, a discourse is an essentially intentional entity; discourse is always discourse-about. Denham writes
a discourse is loosely defined as a linguistic practice governed by community norms. A particular discourse within a linguistic community will be circumscribed by some distinctive, or more or less distinctive, set of concepts: aesthetic discourse, for instance, is circumscribed by the exercise of concepts such as 'beauty', 'elegance', 'grace', 'lyricism', 'ugliness', while moral discourse features concepts such as 'right', 'evil', 'malice', 'kindness', 'deceit', 'duty' and so on. 68

Since to have a concept is to have a "way of conceiving of some object (including object kinds) or property, whereby one is able to form true and rationally warranted beliefs concerning it", 69 discourses are fundamentally world-directed, with discourses individuated by the particular 'realm of fact' that they concern. It should have made Denham question the degree to which significant parallels hold between moral discourse and 'metaphorical discourse', conceived of as the production of metaphorical utterances or assertions, that the latter, in absolute contrast with the former, is not concerned with any particular feature or cluster of features of the world. Metaphors can be about phenomenal experiences, the external world, aesthetic properties, moral properties, persons, numbers, actions, God, the weather - anything, in fact, that any other discourse can talk about literally (and, according to some theorists, a bit more besides).

68 69

Denham (2000) p. 44. For Wrights view, see his (1992) pp. 2-18 Denham (2000) p. 44.

60

It seems that Denham has been misled by an ambiguity in the phrase 'metaphorical discourse'. On one reading, indubitably the more natural and everyday, the phrase refers to discourse which is metaphorical in character. On this resolution of the ambiguity, it is easy to see why metaphorical discourse, as opposed to, say, moral discourse, has no distinct subject matter; the relation to metaphor consists in the way a given state of affairs is being talked about, not the nature of the state of affairs itself. On the second reading, however, the one Denham requires if her supposed parallel is to go through, metaphorical discourse is discourse about the nature and meaning of metaphorical utterances; discourse whose individuating concepts include 'metaphorical meaning', 'simile', 'literal', 'dead metaphor', etc. This kind of discourse, though clearly concerned with a feature of the world, is a metadiscourse, since the feature of the world it concerns is a particular subset of what we might call first-order discourses, which in turn directly concern non-linguistic features of the world. 70 Denhams confusion is made clear in her discussion of what she terms the Reductionist position, which holds that the meaning expressed by metaphors can be fully spelled out in literal terms. She writes that for the Reductionist
Metaphorical discourse, if it is meaningful at all, is not conceptually autonomous, but parasitic or conceptually supervenient: all intelligible disagreements about the truth of judgements expressed metaphorically will thus depend on disagreements about the truth of judgements expressed in non-figurative terms. (The Reductionists position with respect to metaphor is in this respect analogous to that of some moral cognitivists who win cognitive standing for evaluative discourse by proposing to provide definitions of the concepts distinctive of it in naturalistic terms). 71

Denham here simply overlooks the fact that the sense of discourse in which concepts are distinctive of particular discourses is not the one we usually intend when we talk
70

This account actually needs to be complicated slightly, given that we can e.g. use metaphor to describe

metaphor, or literal language to describe linguistic practice, so that not all metadiscourses need concern first order discourses.
71

Denham (1998) pp. 230-231. I will go on to discuss Denhams positive views on reduction and

metaphor further in Chapter 3 below.

61

about metaphorical discourse; the latter usage most often refers to a mode of expression, not a particular subject matter. How does getting clear on this ambiguity help us elucidate the character of Davidsons anti-realism? For a start, it helps us to see that Davidson is no Expressivist, although there are surface similarities. The moral expressivist holds that (a) moral discourse is not typically used to state facts and (b) there are no moral facts. Davidson holds that (a) metaphors are not typically used to state facts and (b) there are no facts about metaphorical meaning. Convergence in doctrine is more real than apparent, however. Expressivists hold that moral discourse is not typically used to state e.g. the moral facts because there simply arent any such facts around to be stated. In the case of metaphor, however, there are plenty of facts around that metaphor could be used to state in a pragmatic sense, rather than a semantic one since metaphor can be used in the characterization of any subject matter. Davidson isnt disputing that metaphor could be used in such a way, only that it typically is. Generically, Expressivism combines an anti-realism about a particular kind of fact with the view that the discourse that we (or folk semantics, or philosophical prejudice) might mistakenly classify as representative of such facts actually has some quite different linguistic function. Davidson is an anti-realist about metaphorical meaning all right, but he does not hold that talk about metaphorical meaning actually has some nonrepresentational linguistic function. The aim of his paper isnt to reveal that Richards, Empson, Black, Goodman, and the rest were really expressing some non-cognitive attitude when they told us that sub-sentential parts of metaphors acquired new meanings in metaphorical contexts. On the contrary, the attitudes were cognitive enough, and aimed at delimiting certain linguistic facts; Davidsons complaint is simply that they got the relevant facts wrong. When Davidson characterizes Henles theory as saying that in its metaphorical role the word applies to everything that it applies to in its literal role, and then some, 72 he is attributing a theory to Henle, not some non-representational attitude. The analogies between Expressivism and Davidson on metaphor, then, flatter to deceive; they are real enough to mislead, but represent only surface similarities rather than continuities of metaphysical and semantic doctrine.
72

Davidson (1984a) p. 250

62

2.4

The Nature of Davidsons Error-theoretic Approach

Davidson is an error theorist about metaphorical meaning. He holds that talk of sui generis metaphorical meaning makes sense, and is to be construed in straightforwardly representational terms, but attributions of it to utterances or their parts are globally false. Error theorists in other areas, such as John Mackie in the case of morality and aesthetics, or Hartry Field in philosophy of mathematics, likewise claim that the relevant class of statements are genuinely assertoric in character, as a face-value construal would suggest (thus diverging from an Expressivist account) but that we are systematically mistaken about their truth values (thus diverging from e.g. non-reductive realists, reductivists, quasi-realists, indexical relativists and minimalists, each of whom hold that the statements in question are often true). It is to the fairly detailed and advanced critical discussion of such positions that we should look when seeking outside aid in evaluating structural features of Davidsons position. It is important here to recall that we are here concerned with facts about metaphorical content, in the sense of commitment to the truth or falsity of claims like Philosophys proper aim is to show the fly the way out of the fly bottle means that philosophys proper aim is to dissolve persistent and frustratingly complex conceptual problems that impede our progress through life, and prevent our living in a fulfilling manner. rather than facts about metaphor production, as when we quarrel about the type of linguistic function metaphors characteristically or typically perform. An Error-theorist can endorse the idea that metaphors role is typically not to state facts or make assertions; his point is that positive ascriptions of meaning to metaphors are uniformly false. We can distinguish two degrees of Error theory about metaphorical meanings. Firstly, one might hold that there are no strictly semantic facts which make claims about

(M)

63

metaphorical meaning true. 73 Of course, this has to go along with the claim that we thought that there were, in order to count as an Error theory. But perhaps this isnt totally implausible; lots of philosophers, literary critics, theologians and linguists did (and do) write as if metaphorical meanings were semantic in this sense, and we might be able to tell some story about epistemological deference to experts that convicted our linguistic community more generally. Lets call somebody who holds this kind of view a semantic error theorist. It will be useful to distinguish two varieties of semantic error theory; a folk version that held that the mistake was ingrained in the practices of the linguistic community at large, and an expert version that locates the mistake rather in the application of a semi-technical term. The phenomenon of epistemological deference mentioned above no doubt renders this a slightly artificial distinction, but I hope it will serve to mark out a difference in emphasis nevertheless. (For example, Mackies discussion suggests that he endorses the folk version of semantic error theory about morals; we are all error theorists in the expert sense about phlogiston.) A more charitable view holds that semantics offers a bad model of what talk about metaphorical meanings really involves, which is fundamentally pragmatic in nature perhaps involving something like Gricean speaker-meanings, although this is clearly not mandatory. The pragmatic error theorist holds that even this more charitable account is mistaken; there are just no facts about metaphorical meaning, whether this is construed in the narrow semantic sense or in a broader pragmatic sense, which could make sentences like (M) true. 74 The reason that we assert or deny such sentences is that we are mistaken about what kind of facts there are, for whatever reason. The explanatory burden the Error theorist of either pragmatic or semantic stripe must discharge is, therefore, threefold. Their aim ought to be to (1) give us a plausible account of how the facts actually stand; (2) explain how we could have been led into such massive error concerning them and (3) elucidate what the upshot of our mistake finally amounts to.
73

My aim is to remain relatively neutral on what counts as e.g. strictly semantic; it should be heard here There is logical space for an error theory that holds that, while the semantics of ascriptions of

in a broad and intuitive sense.

74

metaphorical meaning involve e.g. speaker meaning, these are all in fact false, and some other, strictly semantic theory of metaphorical meaning is in fact correct. I dont know of any theorist who defends this rather unattractive position, and shall in any case ignore it here; my pragmatic error theorist holds that metaphors dont have meanings in either a semantic or a pragmatic sense of meaning.

64

Typically, an Error theoretic account will go most smoothly if the effect of our mistake is fairly minimal. Field, for example, does not counsel us to immediately abandon all scientific theory that relies on our metaphysically suspect beliefs about the existence of numbers, but rather tries to provide nominalistically respectable translations of such science. We might want to make some further distinctions in this area about how such translations are construed. Let an eliminativist be an Error theorist who holds that our talk in a particular area is mistaken because, taken literally, it is straightforwardly false. A hard eliminativist counsels us to abandon the discourse altogether; a soft eliminativist provides us with a functionally equivalent replacement discourse. I take Davidsons position, in a nutshell, to be the following: he is a pragmatic error theorist, of the expert variety, and a soft eliminativist. 75 As noted above, I take him to subscribe to the following theses: The words that are employed in metaphorical utterances do not thereby acquire new, distinctive, metaphorical senses, but rather play the same semantic role as they do in straightforwardly literal utterances. The sentences that are employed in metaphorical utterances do not thereby become associated with new, distinctive metaphorical semantic contents, but rather continue to have their ordinary literal meanings, which are determined by the literal meanings of the words that make them up (together with the mode of composition of such words).

(D1)

(D2)

75

I thus disagree with a recent commentator, Garry Hagberg, (2001) p 289 who says that Davidsons

position is by no means an eliminative onehe allows that there is such a thing as metaphorical truth. Hagberg justifies this interpretation with reference to the following remark of Davidsons: metaphor does lead us to notice what might not otherwise be noticed, and there is no reason, I suppose, not to say that these thoughts and feelings inspired by the metaphor are true or false. I suspect Hagberg is here confusing the functional replacement that Davidson proposes with the kind of distinctively metaphorical truth that he explicitly rejects. For Davidson, the metaphor isnt true in anything like the way that ordinary senses are true. In the sense he allows, we could equally well say that e.g. hints, jokes or thought experiments were true or false.

65

(D3)

Metaphorical utterances are not only, or typically, or characteristically, vehicles for the communication of a distinctive cognitive content, even when such communication is pragmatically construed. Previous influential theories of metaphor falsely subscribed to doctrines that were inconsistent with (D1-D3) above. However, much of what such theorists wanted to capture about metaphor can be best upheld by dropping the idea that a given metaphor is a vehicle for the communication of a distinctive cognitive content, and instead thinking in terms of what it intimates, brings to our attention, or leads us to see.

(D4)

That Davidson subscribes to such doctrines is perhaps relatively uncontroversial. I hope that in the preceding I have made the case that we should see his adherence to such theses as revelatory of an Error-theoretic position, with certain distinctive features. In the following sections, I will go on to examine the nature of his arguments for (D1-D3). This will clear the way for a discussion of Davidsons suggested functional replacement of talk of metaphorical meaning in the next chapter.

2.5

Davidsons Master Argument Against Metaphorical Meanings

Davidsons What Metaphors Mean contains several direct and indirect arguments for (D1)-(D3). However, these arguments need not be thought of bringing wholly independent considerations to bear on the matter, although that is how they are sometimes presented. 76 In this section, I will outline how I think the arguments are intended to work together in a unified way to establish the desired conclusions. I will then go on to evaluate these arguments in the following four sections. Early discussions of metaphor which appealed to the existence of the metaphorical meanings of words or sentences often left the nature of these supposed senses inexplicit, to put it mildly. Nevertheless, we might hope that they would agree with the following principle:
76

E.g. in Nogales (1999) pp. 75-121

66

(A1)

The parts of metaphorical utterances only acquire distinctive metaphorical senses if they thereby become associated with something that deserves to be called their metaphorical meaning. Nothing deserves to be called a metaphorical meaning unless (a) there is something genuinely meaning-like about it and (b) there is something about it that distinctively relates to the metaphorical nature of the utterances it is putatively associated with.

Such a principle is, I hope, unexceptionable. Realism about metaphorical meanings is not secured by demonstrating the existence of elephants, eggplants or electrons, since those entities dont count as metaphorical meanings. The principle would perhaps seem dubious if it was read as reifying meanings, perhaps implicitly subscribing to the idea that meanings must be objects of some special kind, that words serve to name or denote. This is not its intended import however. (A1) merely demands that whatever the supposed metaphorical meaning of words consists in whether associated patterns of use, ways of thinking about the world, intentional properties, inferential roles, satisfaction conditions or whatever that entity, however pleonastically construed, must be the kind of thing that invites us to think of it as a metaphorical meaning. I also take it that the two conditions outlined in (A1), regarding meaning and relevance to metaphor, ought to be uncontroversial. Imagine that we had adopted the convention of printing (all and only) written metaphors in red ink, perhaps to alert the unwary reader to their metaphorical character. The ink colour thus associated with metaphors would clearly not thereby be correctly describable as the metaphorical meaning (henceforth, mmeaning) of the words, even though it was distinctively related to their occurrence in metaphors. The colour of tokens of the words would simply not be sufficiently closely related to the words semantic role to count as a metaphorical meaning. The fact that we feel no temptation at all to think of the ink colour as metaphorical meaning in this case, even though condition (b) is met, seems to indicate our implicit acceptance of something like condition (a). Similarly, imagine that we spoke a language in which the extension of every predicate systematically shifted in certain linguistic contexts. Perhaps certain forms of words were taken to indicate that the extension of the predicate was to be restricted to entities 67

that normally fall under this predicate and are owned by the speaker, for example. It would clearly not be correct to call this restricted extension the m-meaning of a given predicate, since there is simply nothing about it that relates in any distinctive way to metaphor. Such a systematic shift could occur in a linguistic community where nobody ever spoke metaphorically. Again, our reluctance to describe of this kind of case in terms of metaphorical meaning appears to indicate that some condition similar to (b) accurately captures a minimal constraint on what metaphorical meanings must be like. I will also appeal to two auxiliary principles that are required for Davidsons master argument to go through, which again, seems like it ought to be common ground between himself and his opponents (A2) If a proposed assignment of a metaphorical content to a utterance results solely from an unacceptable theory of metaphor, then the utterance does not express that content. If a sentence is associated with a metaphorical content solely in virtue of being a compositional product of its component semantic parts, then some of those parts must themselves have (or determine) metaphorical meanings. So much, I hope, ought to be common ground between Davidson and his opponents. 77 The body of Davidsons article, however, comprises arguments for three premises that have proved to be more controversial: The words employed in a metaphorical utterance do not come to be associated with anything that is both (a) genuinely meaning-like and (b) distinctively related to the metaphorical nature of the utterances in question. The m-meaning conveyed by a sentence used metaphorically would be identical with either (a) what is said by the utterance, which results solely from the
77

(A3)

(A4)

(A5)

(A3) may well seem less obviously neutral territory than (A2). For exegetical reasons, I delay my

argument for (A3)s independent plausibility until Section 2.7. Basically, I argue that rejecting (A3) entails rejecting compositional principles governing genuine word and sentence meaning that simply arent up for grabs, given (A1).

68

meanings of its component parts, together with their mode of composition, or (b) the extra propositional content communicated by the author of the metaphor, over and above what it says, as determined by a linguistic theory of pragmatics. An account of m-meaning which identifies it with the extra propositional content communicated by the author of the metaphor, over and above what it says, as determined by a linguistic theory of pragmatics, provides an unacceptably incomplete and misleading account of our practices of producing, using and reflecting on metaphors. In what follows, I will assume that (A1)-(A3) are either uncontroversial, (or at least, that they can be made so by more careful reformulation, in a way that does not substantively affect the rest of the argument). In a moment, we will look more closely at Davidsons arguments for (A4)-(A6). First, though, let us briefly assure ourselves that, if true, they would serve to establish Davidsons negative theses regarding metaphorical meaning. Let s be a sentence uttered metaphorically, w be a semantically significant part of s, and the following abbreviations hold
W = w has a metaphorical meaning M1 A = M1 is meaning-like B = M1 is distinctively related to the metaphorical nature of the relevant utterance S = s has a metaphorical meaning S1 C = S1is an m-meaning that results solely from composing the meanings of the component parts of s. D = S1 is an m-meaning that is determined pragmatically. U = Attribution of S1 to s results from appeal to an unacceptable theory of metaphor

(A6)

Then the above assumptions correspond to the following schemas

W -> A & B U -> S C -> W (W & A & B) S -> C v D D -> U

(A1) (A2) (A3) (A4) (A5) (A6)

Davidsons argument can then be taken to run roughly as follows:

69

A1 2 A1, 2 A1, 2 A4 A5 8 8, A5 10 A3

1) W -> A & B 2) W 3) A & B 4) W & A & B 5) (W & A & B) 7) S -> C v D 8) S 9) C v D 10) C 11) C -> W 13) W &W 14) (D -> U) & (U -> S) 15) D 16) S 17) S & S 18) (C v D) 19) S

(A1) Assumption, for reductio 1,2 MPP 2, 3, &I A4 2, 4, 5, RAA A5 Assumption, for reductio 7, 8, MPP Assumption, for v-E A3 10, 11, MPP 6, 12, &I A6, A2 Assumption, for v-E 14, 15, &E, MPPx2 8, 16, &I 9, 10, 13, 15, 17, v-E 9, 18, RAA

A1, A4 6) W

10, A3 12) W A1, A3, A4, 10 A6, A2 15 15, A6, A2 8, 15, A6, A2 8, A1-A6 A1-A6

The outline of Davidsons master-argument is thus, I hope, clear. By (A4), nothing meets the conditions outlined in (A1), so by (A1), anti-realism about m-meanings is the correct position. Thus, if we begin to understand a metaphor by coming to understand what its parts mean, and how they are structured, as we clearly do, the meaning in question can only be literal. That establishes (D1). By (A5), a sentence could only express a distinctive metaphorical content if it was either determined by the meanings and arrangement of its parts, or by the relevant pragmatic features involved in its utterance. If the former, then given (A3), and that (D1) has been established, the meaning in question would have to be the literal meaning of the sentence, violating (A1). If the latter, then, by (A6), we would be left with an unacceptable theory of metaphor, so by (A2), no such content is thus associated with the sentence. So sentences used metaphorically are not associated with distinctive cognitive contents, which is (D2). If metaphors were typically, or characteristically, vehicles for the communication of distinctive, pragmatically determined contents, then a pragmatic theory of metaphor

70

would be acceptable, so if (A6) is true, then no such contents are characteristically expressed, which is (D3). I therefore take myself to have established that Davidsons argument, as presented is valid, and relies on at least three assumptions which seem relatively uncontroversial. Our task in the next few sections, then, is to examine the prima facie plausibility of the three remaining assumptions: The words employed in a metaphorical utterance do not come to be associated with anything that is both a) genuinely meaning-like and b) distinctively related to the metaphorical nature of the utterances in question. The m-meaning conveyed by a sentence used metaphorically would be identical with either a) what is said by the utterance, which results solely from the meanings of its component words, together with their mode of composition, or b) the extra propositional content communicated by the author of the metaphor, over and above what it says, as determined by a linguistic theory of pragmatics. An account of m-meaning which identifies it with the extra propositional content communicated by the author of the metaphor, over and above what it says, as determined by a linguistic theory of pragmatics, provides an unacceptably incomplete and misleading account of our practices of producing, using and reflecting on metaphors. I shall take these one at a time.

(A4)

(A5)

(A6)

2.6

Word-meaning and Metaphor

What would it be for the words in a metaphorical utterance to be associated with something meaning like? In a recent paper, discussing his support for Quines thesis of the inscrutability of reference, Davidson tells us:

71

individual words dont have meanings. They have a role in determining the truth conditions of sentences.
78

In Chapter 1, Section 6, I gave an argument to the effect that understanding the meaning of an declarative sentence (in one important sense of that phrase) consists in grasping its truth conditions. However, for the moment, I wish to stay as neutral as possible about whether this is true. It would be disappointing if Davidsons arguments relied heavily on a background theory of meaning that his opponents rejected. That would entail that the dispute between them didnt really turn on the nature of metaphor per se, but rather on another dispute in the philosophy of language. The interest of Davidsons arguments would be increased if it turned out that this was not the case; that they turned out to be compatible with quite different conceptions of the nature of semantics. I will, therefore, try not to trade on features of Davidsons account that his opponents might reasonably reject e.g. the identification of meaning with truth conditions, 79 the inscrutability of reference, etc. Nevertheless, the notion that the meaning of an individual word is exhausted by its semantic role, and that the semantic role of a word is the way it makes a systematic contribution to the meaning of sentences in which it occurs, is not, I think, up for grabs. 80 That would lead to tension with the following plausible principles: (Compositionality) The meaning of complex linguistic expressions is wholly determined by their syntax, together with the meanings of their parts (in context). For example, the meaning of dogs bark

78 79

Davidson (2001) p. 79 Or elimination of meaning in favour of something related to truth conditions, which is probably a more I say involves making a systematic contribution rather than is the systematic contribution, since on

accurate way of stating Davidsons views.

80

some theories, the meaning of a word relates not only to the sentences it can compose, but also to other, related words. For example, full grasp of the meaning of healthy1, as predicated of an diet, may involve understanding a family of closely related predicates (e.g. the homonyms healthy2 healthy3 as predicted of a man, and his complexion). This goes beyond the minimal capacity to grasp the compositional role of the predicate.

72

depends on the meaning of dogs and barks, and the manner in which they are syntactically composed. Languages that can express a given propositional content composed via compositional route F out of sub-sentential parts W1-Wn can also express any other content which can be constructed from W1-Wn using F. For example, a language that can express dogs bark and cows moo can also express dogs moo and cows bark. The meaning of constituent expressions (in context) supervene on the complexes of which they are parts. For example, the meaning of dogs and bark couldnt change without the meaning of dogs bark changing. Such principles, however, are widely subscribed to, 81 and play a familiar and currently indispensable role in explaining the productive and systematic character of human mastery of language. They are not peculiar to truth-conditional theories of meaning, and it is often held to be a minimal condition on any satisfactory theory of meaning that it entail them. I will assume therefore, that Davidson is entitled, pro tem, to rely on the notion that the meaning of an individual word is its semantic role, and that the semantic role of a word involves making a systematic contribution to the meaning of sentences in which it occurs. 82 What significance does thinking of the meaning of a word in such a way have for Davidsons argument against metaphorical meaning? It provides the relevant way of explicating the similarity condition (a) that (A1) places upon any prospective such meaning:
81

(Systematicity)

(Reverse Compositionality)

See Szab (2004) for a note of caution, however. The three principles above could be weakened further

by reading them as generic statement, rather than universal claims, without the force of the point being seriously weakened.
82

For further defence of the importance and nature of compositionality et al, see Fodor and Lepore (2002).

I will return to discuss the issues briefly when I discuss the radical pragmatist objection to Davidsons position.

73

A1)

The parts of metaphorical utterances only acquire distinctive metaphorical senses if they thereby become associated with something that deserves to be called their metaphorical meaning. Nothing deserves to be called a metaphorical meaning unless (a) there is something genuinely meaning-like about it and (b) there is something about it that distinctively relates to the metaphorical nature of the utterances it is putatively associated with.

Nothing is genuinely worth calling the metaphorical meaning of a word that isnt similar to the literal meaning of a word in this respect: it makes a systematic contribution to the meaning of sentences to which it occurs. For it is precisely this feature of words that theorists of language posit word-meanings for: if it wasnt for this feature, it is far from clear what theoretical profit there would be in talking about word meanings at all. Word meanings are like phlogiston, the humours, and the ether; if they cant do the theoretical job that they were designed to do, then theres absolutely no point in having them around. So if theorists like Richards, Empson, Black and Ricoeur are to defend the claim that words acquire distinctive new senses when used metaphorically, then they ought to ensure that those senses are playing the right kind of theoretical role. Otherwise, its a bit like categorising afterimages as a distinctive new variety of pains (one that doesnt feel bad) or cabbages as distinctive instances of kings (ones that dont have a lot of power, or move about much). The problem for such theorists is that they precisely introduced such distinctive senses in a way that seems to prohibit the kind of systematic semantic role that (A1) rightfully proscribes. For example, Richards writes:
In the simplest formulation, when we use a metaphor we have two thoughts of different things active together and supported by a single word, or phrase, whose meaning is a result of their interaction. 83
83

Richards (1965) p. 93? check page ref. The idea that metaphors involved two distinct word or phrasal

meanings was widely shared before Davidsons article was first published in 1979. Beardsley (1976) p. 219 describes the conventional wisdom about metaphor as maintaining that in the usual case, to recognize a controlled sentence as metaphorical involves discerning two senses of the predicate term, in one of which the sentence is false.

74

This conception of the process which determines the metaphorical meaning of a word, however, completely rules out the possibility of the kind of systematic role mandated by (A1). Given that the m-meaning of the relevant phrase is wholly resultant from the local linguistic context, any shift to a new context leaves it open that a completely different sense might arise. And in fact, unsurprisingly, that is what seems to happen. Consider the role the word bronze plays in the following two metaphors:
1) In cities you build a language of circumspection and tact, a thousand little intimations, the nuance that has a shimmer of rubbed bronze. (DeLillo) 84 2) Pythagoras planned it. Why did the people stare? His numbers, though they moved, or seemed to move In marble or in bronze, lacked character. 85 (Yeats)

How do we understand the role the word bronze is playing in these two utterances? In the first, the two thoughts of different things active together involve the metal and the dark, warm, elegant and unyielding mores of the city. In the second, the comparison is between the stark, unchanging abstract world of Pythagorean numbers, and the chilled, balanced perfection of the art that they made possible. 86 Two things seem clear. Firstly, we clearly do not approach either utterance with a prior grasp of a preordained metaphorical class of entities which are picked out by the metaphorical sense of the word, wholly independently of the immediately local linguistic context in which it appears. When asked for things that are metaphorically bronze, I am confident that very few people would include a certain restricted class of nuances, and Pythagorean numbers. Secondly, we only come to grasp the supposed metaphorical sense by first coming to understand the utterance in which it plays a role. One simply cannot
84 85 86

DeLillo (1997) p. 446 Yeats (1962) p. 173 I am assuming that the richness of Yeats image results in part from our being encouraged by the

grammatical structure of the lines to think firstly of the numbers as things which metaphorically move in bronze, and then to reverse direction, conceiving of the statues as containing the numbers that measure their form. Nothing hangs on whether this is the right way to read the lines. Yeats says in On the Boiler p. 37 There are moments when I am certain that art must once again accept those Greek proportions which carry into plastic art the Pythagorean numbers, those faces which are divine because all there is empty and measured.

75

determine what the relevant sense is supposed to be, without first coming to understand the utterance as a whole; for it is only by so coming to understand it that we can identify what precisely the two thoughts of different things are, and how precisely they interact so as to determine the meaning of the relevant phrase. Compare and contrast the following literal examples:
3) John bought a bronze statue yesterday, which was very reasonably priced. 4) My favourite plate is bronze.

5) Nobody much visited the bronze factory in Africa any more, where the workers were
reputedly ill paid and married very early.

In the first two examples, the linguistic meaning we attribute to the word bronze can potentially do some genuine explanatory work. We can explain how we come to understand the utterances by adverting (a) to the fact that we approach them with a grasp of which kind of objects counts as the bronze ones, and (b) that we know enough about semantic roles of the other constituent words in the sentences, together with the way in which they are syntactically related, to be able to determine the content of the whole sentence as a function of the meaning and arrangement of its parts. By appealing to our systematic mastery of a finite lexicon and set of compositional rules, we can explain how we are able to instantly recognize the meaning of two sentences that we have never come across before. Moreover, such an explanation admits of extension to cases where the meaning varies systematically. A bronze factory is not related to bronze in the sense that a bronze statue is, but nevertheless, mastery of what it is for an object to be bronze can interact with systematic linguistic processes to explain how we can understand a cluster of related uses of the word. 87 But no such explanation is possible in the case of the two metaphors. In that case, we do not compute the semantic value of the sentence based on our knowledge of the semantics and structure of its parts. It is only by first determining what the metaphor is supposed to get across that we can then begin to determine what the supposed metaphorical sense of one of its component words might be.

87

Perhaps via grasping the various thematic roles of the subjects qualified by the term bronze a bronze

factory being a source of bronze, a bronze sheen being the characteristic effect of it, etc.

76

Similarly, Max Black claims that in the famous opening sentence of Davidsons article, Metaphor is the dreamwork of language, dreamwork is used metaphorically, via Davidsons attaching an altered sense to the words he is using in context, allowing him thereby to say something distinctively different from the meaning of the words taken literally 88 . But it is not through mastery of this altered sense that we come to understand the utterance, as in the case of sub-sentential parts with genuine meanings. How would the nature of such a sense become evident? Rather, all that can happen in this case is that we understand or endeavour to begin to understand what Davidson is trying to get across by his remark, and then assign a corresponding meaning to the relevant parts of it. But this last step seems wholly redundant. Since we only want word meanings to explain how we compute utterance-meanings, etc, then it is perfectly useless to assign distinctive word-meanings that depend wholly on our first understanding the meanings of the metaphors that they occur in. Of course, it is fair enough to engage in such a practice as part of the process of learning unfamiliar words. If, lacking a dictionary or knowledgeable companion, I read Manley Hopkins lines
Glory be to God for dappled things For skies of couple-colour as a brinded cow

and wonder what brinded means, then I only have the evidence of the surrounding linguistic context, together with my extra-linguistic knowledge (e.g. of skies, cows and the stylistic habits of Hopkins) to go on. But this is quite different from the case of the type of metaphorical meaning that Black adverts to. Having got as far as working out that brinded means something like streaked or patched with colour, I have a chance of using my new found knowledge to help work out the meaning of e.g.
(6) The brinded cat wound its way through the streets.

(7) The outside of the house was ill-kempt and brinded, not recently painted and polished as
shed expected.

88

Black (1978) pp. 185-187

77

But the type of metaphorical meaning assigned to words by Richards and Black is intended to be far more radically dependent on immediate linguistic context than the above type of case. To borrow a nice example of Roger Whites, try using pheasant in the same metaphorical sense in (9) as is most naturally assigned to it in (8)
(8) A poem is a pheasant. Shakespeares Juliet is a pheasant, not a dove or a hawk.

(9)

or metaphysician the same metaphorical sense in the following two metaphors (10)
The actor is A metaphysician in the dark, twanging An instrument, twanging a wiry string that gives Sounds passing through sudden rightnesses, wholly Containing the mind, below which it cannot descend, Beyond which it has no will to rise. (Stevens)

(11)

[Dylan] was re-incarnated as a travelling musical salesman in the spirit of James Brown or B.B. King: one hundred and twenty shows a year, the hardest working metaphysician in showbiz. (Mick Gold)

In the case where we use linguistic and contextual clues to master a genuine word-sense, we can go on to explain how that sense can contribute to an indefinite number of quite different and novel sentences. But in the case of the putatively metaphorical sense, we find this difficult to do, especially in the most interesting and effective metaphors. Typically, understanding the role that a word plays in a metaphor is quite unlike learning the meaning of an unknown word by examining its use in partially understood sentences. No doubt this isnt always the case. Perhaps when we come to grasp what it would be for Juliet to be a pheasant in the relevant sense, we can apply that term in just the same sense to other fictional characters, acquaintances of ours, or film stars, for example. But thus securing condition a) of (A1) that anything worth calling a metaphorical meaning should be genuinely meaning like comes at the cost of surrendering condition b), which insists that it should also be distinctively related to the metaphorical nature of the

78

utterances in question. For it now seems clear that this new sense of pheasant which we can stipulate is to serve as the meaning of a new coinage, m-pheasant admits of mastery by careful observers of our practice who dont know what pheasants are. Once they cotton onto the type of properties of people that we apply m-pheasant to, they seem to know all they need to accord with our application of the term. They dont have to think of the people as pheasants, through a prior literal sense of pheasant; they simply need to master the sense of m-pheasant. And that they can presumably do in the same way that they can learn other words of our language, by observation, imitation, guesswork, empathy, simulation and theory building. But that means that theres nothing distinctively metaphorical about the new meaning that has been introduced, except for its origin. Since many words that we now take to be associated with uncontroversially literal meanings similarly have etymologies involving metaphor, it seems that origin simply isnt enough to establish the distinctive relation that condition b) demands. It seems then, that Davidson can offer a fairly convincing defence of (A4) by offering a dilemma to his opponents. Is the supposed metaphorical meaning of the relevant subsentential part straightforwardly extensible to a wide range of novel linguistic contexts or not? If not, then it doesnt bear a close enough relation to the standard theoretical role of word-meanings to deserve the name. If so, then we have so far been given no reason to think of such a sense as distinctively metaphorical, depending for its existence and efficacy on our taking one thing as another, rather than as a new literal sense; possibly introduced, as many such senses have been, by the use of metaphor, but now bearing no essential or distinctive relationship to metaphor at all. There is no doubt much more to be said. For example, it is often claimed that apparently stone-dead metaphors, that dont demand our thinking of one thing as another, can come to life again given the right context. Such contexts often include cases where we mix dead metaphors to jarring effect:
(12) She was hot, but cold.

Moreover, a case might be made that it may not in fact be straightforward for observers who do not determine the new sense of pheasant via a metaphorical process to know 79

how it is to be extended to new cases. It is a familiar theme of a certain type of Wittgensteinian approach to the application of concepts that one might need to be inducted into a communal practice in order to extend them reliably to new cases, and there seems to be no obvious reason a priori why that might not include going at the thing via the metaphorical process. In any case, I have left the nature of the distinctive relation that (A1) demands at such an intuitive level that it might seem unfair to rule out certain kinds of etymological relations as potential satisfiers. So there is definitely much more that needs to be said to secure Davidsons defence of (A4). Nevertheless, I intend to rest content with the claim that there are at least prima facie good grounds for thinking that something suitably similar to (A4) might be a safe principle for Davidson to rely on.

2.7

Sentence-meaning and Metaphor

As we have seen, my reconstruction of Davidsons master argument attributes to him the following thesis The m-meaning conveyed by a sentence used metaphorically would be identical with either (a) what is said by the utterance, which results solely from the meanings of its component words, together with their mode of composition, or (b) the extra propositional content communicated by the author of the metaphor, over and above what it says, as determined by a linguistic theory of pragmatics. Call the putative distinctive metaphorical content expressed by the utterance the semantic sentential m-meaning, and that communicated by the author the pragmatic sentential m-meaning. 89 What reason is there to think that the m-meaning of sentences used metaphorically must be either semantic or pragmatic in the relevant sense? We have already met independently plausible principles that seem to support the idea that defending semantic sentential m-meaning entails defending the claim that the parts of sentences have metaphorical meanings. Firstly, we can note that (A1)s demand that
89

(A5)

Henceforth, the semantic m-meaning and the pragmatic m-meaning.

80

m-meanings are both meaning like, and suitably related to the metaphorical character of the utterance, applies equally to the sentential case. Secondly, we can note that e.g. the following principles tell us a lot about the character of genuine sentential meanings (Compositionality) The meaning of complex linguistic expressions is wholly determined by their syntax, together with the meanings of their parts (in context). For example, the meaning of dogs bark depends on the meaning of dogs and barks, and the manner in which they are syntactically composed. The meaning of constituent expressions (in context) supervene on the complexes of which they are parts. For example, the meaning of dogs and bark couldnt change without the meaning of dogs bark changing. It seems reasonable to demand that putative semantic m-meanings accord with these central principles governing genuine sentential meanings. But that seems sufficient for us to conclude that any such metaphorical sentence meanings would have to be determined by the metaphorical meanings of words. Otherwise, (Compositionality) and its sister principle would seem to fail. We can argue as follows. 1. Assume that a metaphorical utterance U expresses a distinct propositional content M that is distinct from its literal content L. 2. Assume further, for reductio, that M does not result from a compositional process on sub-sentential parts that themselves have distinctive m-meanings. 3. Since the sub-sentential parts of U have only their literal meanings, their compositional product will be L. 4. By hypothesis, L and M are distinct, so the semantic m-meaning of isnt determined by Us syntax, together with the meanings of its parts (in context). So (Compositionality) fails. 5. Since M is distinct from L, and not determined by the compositional process that fixed the meaning of L, and since the component words of U have only their literal meaning, there seems to be no obvious reason why M might not remain

(Reverse Compositionality)

81

the content of U while the meaning of Us parts changed. So (Reverse Compositionality) seems to be under pressure too. 6. Since giving up those two principles robs m-meanings of the right to be thought of as meanings, the assumption in line 2 must be rejected. So sentential mmeanings require sub-sentential m-meanings. 90 Given that the good standing of (A3) is secured by the above line of thought, however, anybody who (i) wants to defend the existence of metaphorical sentence-meanings, and (ii) has some sympathy with the arguments of the previous section against metaphorical word-meanings, will no doubt want to take a different approach. (After all, (A3) simply tells us the sentence m-meanings require word m-meanings, so if its secure, and you dont want the latter, youd better not take the former). The metaphorical content is not expressed by the utterance, such theorists might hold, but arises in some other way. However, given (A1), room for manoeuvre is limited somewhat. Its no good finding some apt propositional content, and simply proclaiming it the meaning of the metaphorical utterance. To be a meaning is to play a role which is similar enough to things deservedly regarded as meanings. Its no good just recalling some thought that the metaphor inspires, and proclaiming it the meaning of the metaphor. Such a content needs to link up in the right way to the linguistic knowledge that governs the communication of representational contents by means of natural language or other relevant representations. For metaphorical meanings to be of distinctive interest to linguistics and the philosophy of language, as their proponents typically take them to be, they need to resemble in relevant ways linguistic meanings. Since linguistic meaning is fundamentally what is communicated in successful instances of linguistic communication, metaphorical meaning had better be what is communicated by the making and taking of metaphors. If that communication does not take place in virtue of shared mastery of the m-meanings of words, then it had better be explicable in terms of some other distinctively linguistic capacity. Call any theory which aims to explicate the nature of such non-semantic but still linguistic communication a pragmatic theory. It

90

This is the proof of the independent good standing of (A3) promised in Section 5 above. I take it that

(A1) and the two principles governing compositionality ought to be common ground between any serious participants in the debate about metaphorical meanings.

82

seems then, that if metaphors genuinely have meanings, and those meanings arent explicable in semantic terms, then they will have to be explicable in pragmatic ones, if we can explain them at all. Bracketing appeal to mystery, it seems that semantics and pragmatics offer the exclusive options for a genuine theory of metaphorical meaning. But thats just what (A5) asserts. If the above line of thought is correct, then it seems that (A5) can in a sense be seen as an elucidation of (A1); metaphorical contents had better be meaning-like, so if they are not explicable in terms of word-meanings and compositional rules, they had better be suitably related to some more general theory of communication. However, it is worth keeping the two principles separate. Since (A5) claims more about what being meaninglike demands, it is correspondingly more open to challenge than the seemingly platitudinous (A1). Its notable how little we have needed to say in the above discussion about the shape of a linguistic pragmatics. We have not committed Davidson to the eventual success of a broadly Gricean model, for example, over a Relevance-Theoretic rival. 91 Nevertheless, I hope that I have offered some prima facie reason to think that (A5) defensible, and Davidson is entitled to rely on it, or a close counterpart, in his master argument against metaphorical meanings.

2.8

Metaphor, System and Error

We are left, then with the task of examining and evaluating (A6) An account of m-meaning which identifies it with the extra propositional content communicated by the author of the metaphor, over and above what it says, as determined by a linguistic theory of pragmatics, provides an unacceptably incomplete and misleading account of our practices of producing, using and reflecting on metaphors.

(A6)

91

See Sperber and Wilson (1986)

83

Davidsons remarks in defence of this thesis are compressed in the extreme, and it is fair to say that it seems at first face to be the most controversial of the principles that his argument relies on. Many philosophers have been attracted by the idea that metaphor makers use the literally false content of one sentence to get across some quite distinct content. Although this model of metaphor has been widely criticised in contemporary work on the topic, it remains a widespread and popular view among philosophers, and probably deserves to be thought of as the default position. An early and influential statement of the essence of the view comes from John Searle
The problem of explaining how metaphors work is a special case of the general problem of explaining how speaker meaning and sentence or word meaning come apart. It is a special case, that is, of the problem of how it is possible to say one thing and mean something else, where one succeeds in communicating what one means even though both the speaker and the hearer know that the meanings of the words uttered by the speaker do not exactly and literally express what the speaker meant. Some other instances of the break between speakers utterance meaning and literal sentence meaning are irony and indirect speech acts. In each of these cases, what the speaker means is not identical with what the sentence means, and yet what he means is in various ways dependent on what the sentence means.
92

We have seen in Section 2.3 above that Davidson agrees with theorists like Searle that metaphor is a matter of how words are used, rather than what they mean. Why then, does he think that holding that they are typically used by the speaker to communicate some other proposition leads to an unacceptable account of metaphor? Some philosophers have thought that a certain kind of systematicity, or rather the lack of it, is the issue. 93 Thus Joseph Stern says that the view that no speakers meaning is conveyed is

92 93

Searle (1993) p. 84 Systematicity in this sense is clearly different from the type of systematic linguistic creativity adverted

to above. It concerns, rather, the idea that the interpretation of metaphors does not depend on a codifiable set of strictly linguistic rules and abilities.

84

not peculiar to Davidsons discussion of metaphor. For in general he expresses doubt about the possibility of codifying the abilities and skills involved in so-called speakers or utterance meaning in the form of principles ( la Searle) or maxims ( la Grice), pre-established or conventional rules which would be either specific to language or linguistic activities like conversation. Instead the kinds of inferences and reasoning these activities involve require only, he says, the cleverness, intuition, luck and skill which are necessary for any rational activity or for devising a new theory in any field.
94

Similarly, William Lycan treats Davidsons claim that

There are no instructions for devising metaphors; there is no manual for determining what a metaphor means or says; there is no test for metaphor that does not call for taste.
95

as an argument against the Pragmatic View, commenting

As if directly inspired by that passage, Searle produced quite a number of such rules, and so far as they go they are plausible. 96

Although the aim of this chapter is not textual exegesis of Davidson, I think that this interpretation is neither very plausible as account of what actually drives Davidson to endorse something like (A6), nor the most charitable way of defending that principle. The suggested motivation runs roughly as follows: (1) Anything worth regarding as a meaning relates constitutively to distinctively linguistic knowledge, abilities, rules or conventions. (2) The type of interchange of propositional content with the interpreter coming to recognise what the speaker is trying to get across which is involved in metaphor, though genuine enough, is not constitutively related to distinctively

94 95 96

Stern (1991) p. 16 Davidson (1984a) p. 245 Lycan (2000) p. 221

85

linguistic knowledge, abilities, rules or conventions. Rather, it draws on quite general epistemic and practical knowledge and abilities. (3) Therefore, the type of interchange of propositional content which is involved in metaphor, though genuine enough, is not worth regarding as the meaning of the metaphor, whether construed in semantic or pragmatic terms. Firstly, there is little textual evidence in What Metaphors Mean that this line of thought is what is motivating Davidson there. Stern quotes from a different essay, Communication and Convention, 97 which does not discuss the case of metaphor, and which is primarily concerned with the question of whether appeal to public conventions is explanatorily essential in constructing a theory of communication. 98 The Davidsonian quotation that Lycan appeals to is immediately preceded by a remark that makes it clear that Davidson means it to speak to semantic accounts of metaphor, rather than the kind of pragmatic account adverted to here. 99 His point seems to be that there are no additional compositional rules that serve to determine metaphorical meanings on the basis of literal ones. I am not claiming that Davidson would wholly reject the spirit of the premises of the above argument. The idea that a theory of meaning, properly so-called, deals with only a very limited pattern of predictable behaviour, teased out painstakingly from the unruly

97 98

Davidson (1984b) Stern also cites remarks made by Davidson in an introductory summary to that essay. Davidson writes:

It is always an open question how well the theory an interpreter brings to a linguistic encounter will cope. In practice an interpreter keeps the conversation going by adjusting his theory on the spot. The principles of such inventive accommodation are not themselves reducible to theory, involving as they do nothing less than all our skills at theory construction. Since the kind of pragmatic account discussed by Searle explicitly does not involve revision of the semantic theory which assigns satisfaction conditions to the parts of the language, this remark does not seem to speak to it. See Stern (1991) p. 46 and Davidson (1984) p.xix. Davidson explicitly distinguishes between metaphor and malapropism, where such theory adjustment does occur, in Davidson (1986).
99

all communication by speech assumes the interplay of inventive construction and inventive

construal. What metaphor adds to the ordinary is an achievement that uses no semantic resources beyond the resources on which the ordinary depend. There are no instructions for devising metaphors. Davidson (1984a) p 246. The remark comes at the very beginning of Davidsons essay, whereas the discussion of pragmatic communication comes at the very end.

86

flux of linguistic interaction, is definitely a recurrent theme in his work. On the other hand, the idea that he hangs anything very much on the distinction adverted to in the second premise is simply not faithful to his stated view
[Grice] has shown why it is essential to distinguish between the literal meaningof words and what is often implied (or implicated) by someone who uses those words. He has explored the general principles behind our ability to figure out such implicatures, and these principles must, of course, be known by speakers who expect to be taken up on them. Whether knowledge of these principles ought to be included in the description of linguistic competence may not have to be settled: on the one hand they are things a clever person could often figure out without previous training or exposure, and they are things we could get along without. On the other hand they represent a kind of skill we expect of an interpreter and without which communication would be greatly impoverished.
100

Davidson is wise to profess neutrality on the question of whether we should accept the two premises of the above argument. Linguists and philosophers of language have long found it plausible and profitable to suppose that some theory of implicature can play an explanatory role in understanding communication. Such work is no doubt provisional and in need of development. Furthermore, it is surely too much to hope that the type of theory that emerges, even in the most favourable case, is likely to allow for the type of productive and predictive formalization available in syntax and semantics. Nevertheless, its not clear why we are to accept the claim of the first premise that anything worth regarding as a meaning must be constitutively related to distinctively linguistic knowledge and abilities. Why shouldnt it be sufficient that a pragmatics show how speaker meanings characteristically relate to a fairly typical, delimited set of skills and assumptions? Its all too easy to get slightly hysterical about such demarcatory issues. Given that working linguists typically find speaker meaning a useful explanatory posit, philosophers ought not to get too excited about what is in effect a terminological decision, made on the basis of a fluid and shifting pattern of use. Neither textual exegesis, nor a spirit of interpretative charity, then, lead us to conclude that Davidsons pragmatic error theory is motivated by a general scepticism about the
100

Davidson (1986) p. 437

87

place of pragmatics in linguistic explanation. Rather, I will suggest, the intended supporting argument for (A6) runs somewhat as follows: a) Anything worth regarding as a tolerably complete and informative account of our practices of producing, using and reflecting on metaphors must account for all of the distinctive marks of metaphor. b) An account of metaphor that, like Searles, characterizes such practices as being in essence a special case of saying one thing and meaning another, cannot account for all of the distinctive marks of metaphor. c) Therefore, such pragmatic accounts of metaphor provide at best an incomplete and misleading account of our practices of producing, using and reflecting on metaphors. Several comments on this argument are immediately required. Firstly, by marks of metaphor, I mean the distinctive features of our practice of composing, employing and engaging with metaphor, that give it the role in our thought and practice that it possesses. Secondly, it ought to be clear that the above argument does not aim to demonstrate that metaphor-as-implication never takes place, nor that a theory like Searles might comprise a part of a more general theory of metaphorical talk. Davidson has often been attacked for holding that such communication never takes place, but this may be to attack a straw man. He writes of the thesis that metaphors are associated with a distinctive cognitive content that the speaker intends to convey to an interpreter
This theory is false as a full account of metaphor, whether or not we call the purported cognitive content a meaning.
101

Similarly, Davidson treats Donnellans case of Smiths murderer as a case where Jones has said something true by using a sentence that is false, adding
This is done intentionally all the time, for example in irony or metaphor.
101

102

Davidson (1984a) p 262, my italics. The italicised section did not appear in the original version of the

article, which might either suggest a change of heart, or, more likely, the correction of a rhetorical flourish.
102

Davidson (1986) p. 440

88

Admittedly, Davidson does include the previously quoted remark

if I am right, a metaphor doesnt say anything, beyond its literal meaning (nor does its maker say anything, in using the metaphor, beyond the literal). 103

There are, however, a number of ways to interpret this charitably. We might see it as an ill-timed rhetorical flourish, a deliberately eye-catching overstatement of the actual thesis defended. We might take it as a sign that Davidson is here speaking of metaphor in generic terms, rather than making a universally quantified claim. We might take its inclusion as an oversight, a hangover from an early, less plausible version of the position. Later work suggests that the latter interpretation is probably the most likely. In Locating Literary Language, Davidson distinguishes three species of intention that he holds must be present in any speech act. 104 Firstly, there are ulterior intentions which aim at the achievement of some extra-linguistic end; a goal or purpose that could, at least in principle, be achieved in other ways. For example, one might want to call somebodys attention to similarities between philosophers and flies trapped in flybottles, or get them to think of the former as the latter. Secondly, there are pragmatic intentions to utter a sentence with a given illocutionary force. A given remark is intended to be taken as a command, a request, an invitation, etc. 105 It is unclear whether Davidson thinks that the intention to speak metaphorically fits in at this level, partly because metaphor seems to relate more closely to the manner of what is said, than the force with which it is put forward. Finally, there are the strictly semantic intentions, where one intends ones words to have certain meanings that will be taken as such by ones interpreter. For example, Wittgenstein intended the words Mendelssohn is not a peak, but a plateau 106 to mean that Mendelssohn is not a peak, but a plateau. (Had he
103

Davidson (1984a) p. 246, my italics. The remark is included both the original version of the article, Davidson (1993) p. 298-299 Davidson accepts the possibilities of borderline cases as when see you in July is half promise and Wittgenstein (1998) p. 4e

and the revised version which I have relied on in this thesis.

104 105

half prediction but only when such a mixed force is intended by the speaker.
106

89

mistaken the meaning of plateau for that of platypus, his intention would not have been fulfilled.) Davidson uses the label first meaning for the meanings of words picked out by the latter type of intentions. He writes
The usefulness of the concept of first meaning emerges when we consider cases where what is stated or implied differs from what the words mean. Sometimes too hot the eye of heaven shines means that the sun sometimes shines too brightly. But the first meaning of the eye of heaven purports to refer to the one and only eye of heaven. We can tell this because Shakespeare (we assume) intended to use words that would be recognised by a reader to refer to the one and only eye of heaven (if such a thing existed) in order to prompt the reader to understand that he meant the sun. We may wish to use the word meaning for both the first meaning and what the metaphor carries us to, but only the first meaning has a systematic place in the language of the author. 107

As if to eliminate any doubt, in a footnote to the above he adds

In my essay What Metaphors Mean I was foolishly stubborn about the word meaning when all I cared about was the primacy of first meaning.

In any case, I will assume that the best defence of the type of pragmatic error theory canvassed by Davidson in his original article involves granting that a speaker-meaning account might be applicable on occasion, but deny that it provides a satisfactory theory of metaphor. 108 Such a concession, while seemingly inevitable given the widely acknowledged datum that we sometimes do, or at the very least could, convey a cognitive content when speaking metaphorically, entails an immediate weakening of Davidsons argument as presented above. Compare (A6) and (A2)

107 108

Davidson (1993) p. 300 For stylistic convenience, I will continue to refer to the proponent of such a position as Davidson and

the position itself as Davidsonian, etc. This should not be understood as an implicit commitment to the actual Donald Davidson being willing to endorse them in letter or spirit.

90

(A2)

If a proposed assignment of a metaphorical content to a utterance results solely from an unacceptable theory of metaphor, then the utterance does not express that content.

(A6)

An account of m-meaning which identifies it with the extra propositional content communicated by the author of the metaphor, over and above what it says, as determined by a linguistic theory of pragmatics, provides an unacceptably incomplete and misleading account of our practices of producing, using and reflecting on metaphors.

The surface plausibility of these two principles seems to result from two quite different conceptions of an unacceptable theory of metaphor. Imagine that the only putative warrant we could get, even in relatively idealized conditions, for assigning a certain mmeaning to an utterance is a crazy theory of metaphor. Then, it seems, the utterance does not express that meaning. Thats the intended sense of (A2), and its the sense in which it ought to be common ground between Davidson and his opponents. Such a principle can be seen as resting on one incontrovertible principle, and one widely shared assumption about what metaphorical meaning must be like

(Warrant)

If a theory is crazy, then it doesnt give warrant to beliefs formed on the basis of it.

(Epistemic Access)

If a metaphor m-means that P, then (in somewhat idealized conditions) we can get some warrant for believing that it m-means that P.

Assume that the only putative warrant that we could get for an ascription of m-meaning was a crazy theory. By (Warrant), no warrant for believing that the metaphor did have such a content would spring from that theory. So we would have no warrant at all for accepting the ascription. Modus tollens on (Epistemic Access) gives us that the

91

metaphor does not have that content, and conditional proof gives us the plausible version of (A2).

However, (A6) is only plausible if we apply quite a different sense of unacceptable theory. Grant Davidson that any acceptable theory must account for all the marks of metaphor, and that a speaker-meaning account does not. Even so, if there are grounds for thinking that such an account is only unacceptable in the sense it is incomplete, partial or narrow in application, then that is a quite different thing than being the type of crazy theory that (Warrant) involves. Similarly, if the theory does not mislead us with respect to its proper objects, but only beguiles us into ignoring more complex cases of metaphor than straightforward implicatures. In those type of cases, an unacceptable theory may well be capable of giving warrant to beliefs, just as Newtonian mechanics gave warrant despite incompleteness and misleadingness of a similar kind. So it seems that Davidson currently lacks a unitary notion of unacceptable theory of metaphor upon which all of the principles he implicitly relies upon come out true.

It seems, then, that even before we begin to evaluate Davidsons claim regarding the explanatory power of a pragmatic theory of metaphorical content, his argument runs into trouble. Even if we grant that a pragmatic theory is unacceptably incomplete in the way that he suggests, that does not warrant the application of (A2), in the sense in which that principle is neutral territory.

The only answer available to a Davidsonian seems to be to restrict the scope of the argument. Let m be a metaphorical utterance that the pragmatic theory really does give a crazily unsatisfactory account of. Then (A2) and (A6) hold true of metaphors like m. The Davidsonian had better hope that a significant proportion of our metaphors are more like m, than the type of communicative metaphors adverted to by Searle et al. Otherwise, the pragmatic error theory seems to be a considerably less revisionary, and interesting, position than it originally seemed.

92

Imagine that Davidson is right to hold that a significant proportion of metaphors are not best explained in terms of pragmatic communication, but that some are. Does even this limited concession undercut the motivation for, or substance of, a pragmatic error theory? After all, we cant be making all that much of an error if we are right about metaphor expressing cognitive contents in a number of familiar cases.

It is instructive to consider analogous cases from other areas of philosophy. Hartry Field, for example, does not hold that every mathematical sentence is false, just that a substantial and interesting class are. Negative existentials that are counted as true by realists about mathematics there is no greatest prime number, for example will also be counted true by Field. That case, admittedly, might be thought to turn on a peculiar feature of the case. But there is comfort elsewhere. Take a Nietzschean view on everyday moral talk, for example. Nietzsche rejects the idea that the majority of claims about moral worth are literally true, holding instead that people generally are radically self-deceived about e.g. the value of giving money to famine relief out of a sense of pity for the poor and suffering. He does not try to reconstruct the semantics for such talk in terms more acceptable to an anti-realist worldview, however. It is thus appropriate to regard him as an error theorist about everyday morality. But that does not entail that he rejects, or ought to reject, every claim that everyday morality would make. Folk morality and Nietzschean master-morality may well converge on their appraisals of e.g. the moral worth of character traits like creativity, self-sufficiency, inner strength and dedication. If were content to continue to regard Nietzsche as the moral revisionist and error-theorist par excellence, then it seems that the Davidson has some room for limited concession.

2.9

Two Marks of Metaphor

What marks of metaphor, then, might a speaker-meaning account be thought to give a radically unsatisfactory account of? Davidson suggests two potential lines in What Metaphors Mean. Though neither is developed in any great detail there, both have proved suggestive. Firstly, there is the idea that what a metaphor brings to our attention is in a certain sense limitless. 93

Stanley Cavell mentions the fact that most attempt at paraphrase [of metaphors] end with and so on and refers to Empsons remark that metaphors are pregnant. But Cavell doesnt explain the endlessness of paraphrase as I do, as can be learned from the fact that he thinks it distinguishes metaphor from some (but perhaps not all) literal discourse. I hold that the endless character of what we call the paraphrase of a metaphor springs from the fact that it attempts to spell out what the metaphor makes us notice, and to this there is no clear end. I would say the same for any use of language. 109

Secondly, there is the idea that this limitlessness has a partial source in the nonpropositional character of what metaphor guides us to

Its not only that we cant provide an exhaustive catalogue of what has been attended to when we are led to see something in a new light; the difficulty is more fundamental. What we notice or see is not, in general, propositional in character. Of course it may be, and when it is, it usually may be stated in fairly plain words. But if I show you Wittgensteins duck-rabbit and I say, Its a duck, then with luck you see it as a duck; if I say, Its a rabbit, you see it as a rabbit. But no proposition expresses what I have led you to seeSeeing as is not seeing that. Metaphor makes us see one thing as another by making some literal statement that inspires or prompts the insight. Since in most cases what the metaphor prompts or inspires is not entirely, or even at all, recognition of some truth or fact, the attempt to give literal expression to the content of the metaphor is simply misguided. 110

There are two issues that become pressing here. Firstly, there is the question of how we should best understand the character of the suggested marks of metaphor. Secondly, there is the question of whether they are wholly inadequately explained by a pragmatic theory of metaphor. Since the last two chapters of this thesis will be concerned with addressing both topics in a degree of depth, I will offer a fairly compressed discussion at
109 110

Davidson (1984a) p. 263 Davidson (1984a) p. 263. It is worth drawing attention to the fact that Davidson here claims that in

most cases, what the metaphor prompts is not propositional, giving textual support to the suggested interpretation offered above.

94

this stage. Nevertheless, it is worth dealing with a few issues here. I shall proceed in reverse order.

A pragmatic account of metaphor typically distinguishes between what is communicated, or implicated, by the use of a sentence, and what that sentence literally says. Admittedly, pragmatic accounts need not limit themselves to such models. More radical pragmatist accounts challenge the idea that what a sentence means is wholly independent of the linguistic and social context in which it is uttered, even when all e.g. indexical and demonstrative elements of the sentence are taken into account. 111 Nevertheless, even such radical theories agree that pragmatic processes result in something that can be specified using a that clause. If John, holding a cigarette, asks Susan if she has any matches, we can take it that he is implicating, inter alia, that the matches are suitable for lighting the cigarette. Even if Sperber, Wilson and Carston are right to hold that processes of pragmatic enrichment determine that saying John has had breakfast typically states, rather than merely implicates, that John has had breakfast this morning, the output of the pragmatic process is still a propositional content. 112

Such convergence is unsurprising. The aim of a pragmatic theory isnt wholly clear, but often in practice amounts to the systematic investigation of what Gazdar has called meaning minus truth-conditions. 113 Since what is meant or said by using a sentence can typically be expressed using S said that P or U means that P constructions, pragmatics is typically concerned with assigning propositional contents to utterances. Moreover, since pragmatics aspires to the status of a well-confirmed linguistic theory, and linguistics aspires to be a (special) science, such attributions are typically literal statements of what is said by an utterance. 114

111 112 113

See e.g. Searle (1978) and Travis (1997) for statements of a view similar to this. See Recanati (1993) p. 261 for a useful discussion of the case. Gazdar (1979) p. 2. Levinson (1983) pp 12-32 suggests Pragmatics is the study of all those aspects of I do not mean to suggest that metaphor has no place in science. It clearly does. However, the

meaning not captured in a semantic theory as the most promising definition.

114

systematic attribution of contents to utterances by linguistic theories are typically not cases in point.

95

It is thus a serious problem for a pragmatic theory if what a metaphor gets across is nonfinitely specifiable, and even worse if it is wholly non-propositional. We can argue as follows

1) If a pragmatic theory of metaphor is to be successful, then it ought to issue in attributions like Speaker S communicates that p by uttering metaphor M, or Metaphor M states that p. 2) The limitlessness and non-propositional character of most metaphors means that no such attributions will be forthcoming by a pragmatic theory. 3) A pragmatic theory of metaphor will not be successful.

The first premise seems to be secured by the general aims of a linguistic pragmatics, while the second seems relatively secure, given then methods by which such a theory aims to match utterances and contents. It seems then, that if Davidson is right about the characteristic marks of metaphor, he has the chance to establish that pragmatic theories of metaphor are unacceptable in just the way that a unified reading of (A2) and (A6) demands.

The matter turns, then, on the status of the two marks of metaphor. The first, limitlessness, has often been associated with a demand for paraphrasability. In specifying the pragmatic content of an indexical utterance, we may have to use a different but related sentence, that expresses a suitably related content. For example, if I ask you if you can pass the salt, implicating that I would like you to pass the salt to me, then a theoretical specification of that content might replace the two indexicals with suitable proper names. 115 Similarly, the thought goes, a demand for a theoretical specification of the pragmatic content of a metaphor will involve the identification of a suitably related proposition.

115

Perhaps with an indication that the utterance was originally de se in form.

96

Thus Nogales reconstructs Davidsons argument as follows

(1) Metaphorical utterances typically do not admit of paraphrasing. (2) Any cognitive content can be expressed in at least two different ways (i.e. it can be paraphrased) (3) Being (easily) paraphrasable is a test of whether the cognitive content of a sentence captures its cognitive effect. (4) Therefore, the cognitive effect of a metaphorical utterance does not lie in the cognitive content of its terms. (5) Metaphorical meaning is defined so as to capture the cognitive effect of the utterance through cognitive content. (6) Therefore, there is no such thing as metaphorical meaning as defined. 116

The idea expressed in line 2 that Davidson thinks that anything can be said in two ways has gained a strangely wide currency in recent work. Thus Goodman notes that paraphrase of many literal sentences is also exceedingly difficult, and indeed we may seriously question whether any sentence can be translated exactly into other words in the same or other language, 117 an argument echoed approvingly by Nogales. Similarly, Denham suggests that the claim of non-paraphrasability, if it is to provide a distinction between utterances that express pragmatic contents and those that do not, must become one of two claims
Either (1) metaphorical contents are non-replaceable because the truth-conditions of metaphorical sentences, unlike those of literal sentences, cannot be stated independently of the sentence itself, or (2) metaphorical contents are non-replaceable because they do not exist; unlike literal sentences, metaphorical sentences have no truth conditions (apart
116

Nogales (1999) p. 75. I have set out Nogales reconstruction verbatim. In a footnote commenting on

line (3), she says Underlying this argument is Davidsons conception of what it means to be paraphrasable, which seems to involve not only the cognitive content of an utterance, as evidenced by its truth conditions, but also its effect, which seems to include feelings as well as thoughts we are led to contemplate.
117

Goodman (1981) p. 176.

97

from those attaching to them as literally read) and hence no cognitive content (apart from their literal content. It is not open to Davidson to opt for (1) because he has already rejected the view that the truth-conditions of structurally simple literal sentences can be stated independently of the sentences themselves which leaves them on a level with metaphors. He thus opts for (2). 118

All these views seem to me to miss some fairly obvious disanalogies between the case of metaphors and those of other utterances. Firstly, Davidson is arguing against the idea that a pragmatic theory might ascribe speaker-meanings, or other pragmatically determined propositional contents, to metaphors. Since, in the case where theorist and speaker share a language, such a theory can straightforwardly give a literal specification of what is being said by a literal utterance by using that very sentence, or, in the case of indexicals, etc, a suitably related one, the issue of paraphrasability does not arise. When John utters assertorically the cat is on the mat, he typically aims to communicate, inter alia, that the contextually salient cat is on the contextually salient mat. There is simply no need, contra Nogales, for Davidson to commit himself to the idea that everything can be said in two ways; nor, unsurprisingly, does he do so. Goodmans point is thus an ignoratio elenchi, based on a misreading of Davidsons line of argument. Denhams diagnosis of what motivates Davidsons concern with paraphrase is similarly misguided. 119 Davidson clearly does hold that the truth conditions of structurally simple sentences can be stated independently of those sentences, since e.g. (a) radical interpretation in the non-homophonic case would otherwise be impossible and (b) even in the homophonic case, indexical transformations require that quite different sentences be used to specify truth conditions. So Davidson could just as easily opt for (1), for all Denham has shown. (Denham is right to wonder whether the pragmatic theorist might profitably surrender the idea that a linguistic theory ought to issue specifications of content in only literal terms. I will discuss this idea in the next chapter).

118 119

Denham (2000) p. 259 Rather unsurprisingly, given the wholly misleading and uncharitable construal of his position that

precedes it, during which she accuses Davidson of muddling meaning with perlocutionary effect (!) and, in an oversight of convenience, having neglected the sense/force distinction.

98

The paraphrasability of literal sentences, and related red herrings relating to the nonpreservation of Fregean tone, ought then to be put aside. The question is whether it is reasonable to believe that a systematic pragmatics is likely to be able to assign suitable truth-conditional equivalents to metaphorical utterances. Unfortunately, however, it is at this point that the debate tends to collapse into appeal to theoretically loaded intuitions. Realists about metaphorical meaning, confident in the idea that metaphors can be true or truth-apt, tend to characterise Davidsons scepticism as mistaking difficulty for impossibility. Sometimes they even go so far as to provide putative paraphrases, often for relatively stale, straightforward or one-dimensional metaphors, and then extrapolate to the general case. Anti-realist attention is divided between two contrasting rejections. In the case of superficial and glib summary, they are wracked with the uneasy apprehension that more than mode-of-presentation is omitted from such construals. In the (rare) case where a substantive elucidation of a metaphor is presented, they become sceptical of the idea that the speaker could really have intended to convey all that by her metaphor. Even for a single theorist, it is easy to vacillate between rejecting the phenomenology of content (surely Wittgenstein was saying something about Mendelssohn, that we might reject, endorse as true, or ask for proof of?) and the phenomenology of paraphrase (surely its true that we cant say what a given metaphor conveys in non-metaphorical terms isnt that partly why we appeal to metaphor?) An account that attempted to reconcile both intuitions is clearly desirable, and has been much attempted. Since I will revisit the question of the possibility of paraphrase in some depth in the final two chapters, and offer a sustained discussion of the putatively nonpropositional character of metaphor in Chapter 5, I shall draw my discussion to a close here. Let us return to (A6). An account of m-meaning which identifies it with the extra propositional content communicated by the author of the metaphor, over and above what it says, as determined by a linguistic theory of pragmatics, provides an unacceptably incomplete and misleading account of our practices of producing, using and reflecting on metaphors I have argued that this principle, and with it much of the surface plausibility of Davidsons case for a pragmatic error theory, depends upon (i) the way in which the 99

(A6)

putative distinguishing marks of metaphor are developed and elaborated, (ii) the case being made that no suitable pragmatic theory can account for them and (iii) a significant class of metaphors actually displaying the supposed distinguishing marks. Each of these issues requires a good deal of work to be made good, and the principle is correspondingly far less plausible, prima facie, than the others adverted to in Davidsons argument. Nevertheless, we have not yet identified an unambiguous reason to reject it outright.

2.10 Conclusion
I have argued that we should think of Davidson as an error theorist about metaphorical meaning, whether construed in semantic or pragmatic terms. In the latter half of this chapter, I have examined the negative aspect of that position. Davidsons argument against metaphorical meaning relied on the following six fundamental principles. The parts of metaphorical utterances only acquire distinctive metaphorical senses if they thereby become associated with something that deserves to be called their metaphorical meaning. Nothing deserves to be called a metaphorical meaning unless (a) there is something genuinely meaning-like about it and (b) there is something about it that distinctively relates to the metaphorical nature of the utterances it is putatively associated with.

(A1)

(A2) solely

If a proposed assignment of a metaphorical content to a utterance results from an unacceptable theory of metaphor, then the utterance does not express that content.

(A3)

If a sentence is associated with a metaphorical content solely in virtue of being a compositional product of its component semantic parts, then some of those parts must themselves have (or determine) metaphorical meanings. 100

(A4)

The words employed in a metaphorical utterance do not come to be associated with anything that is both a) genuinely meaning-like and b) distinctively related to the metaphorical nature of the utterances in question.

(A5)

The m-meaning conveyed by a sentence used metaphorically would be identical with either a) what is said by the utterance, which results solely from the meanings of its component parts, together with their mode of composition, or b) the extra propositional content communicated by the author of the metaphor, over and above what it says, as determined by a linguistic theory of pragmatics. An account of m-meaning which identifies it with the extra propositional content communicated by the author of the metaphor, over and above what it says, as determined by a linguistic theory of pragmatics, provides an unacceptably incomplete and misleading account of our practices of producing, using and reflecting on metaphors.

(A6)

The first of these I took to be a priori defensible. The second, (A2), was derivable from two more fundamental principles, one a priori, and the other a substantive but widely endorsed anti-realist principle about meaning of any kind: If a theory is crazy, then it doesnt give warrant to beliefs formed on the basis of it.

(Warrant)

(Epistemic Access)

If a metaphor m-means that P, then (in somewhat idealized conditions) we can get some warrant for believing that it m-means that P.

101

The third principle was defended on the grounds that its rejection entailed the rejection of independently plausible principles governing the relationship between the meanings of sentences and those of their parts, such as (Compositionality) The meaning of complex linguistic expressions is wholly determined by their syntax, together with the meanings of their parts (in context). For example, the meaning of dogs bark depends on the meaning of dogs and barks, and the manner in which they are syntactically composed. The fourth was defended by a dilemma. Either supposed metaphorical senses of words are compositional or they are not. If not, then they are not genuinely meaning-like. If so, then their distinctively metaphorical character has been covertly abandoned. The defence of (A5) had a similar structure: metaphorical meanings had better be meaninglike, so if they are not explicable in terms of word-meanings and composition, they had better be suitably related to some more general theory of communication. I took it that the arguments offered here gave us some reason to take these principles seriously. Of course, prima facie plausibility is not truth, and each of those principles needs considerably more in the way of defence than I have offered here. Some of this work will begin in the next chapter. Nevertheless, the majority of the remainder of this thesis will relate to the most controversial principle, (A6), and I will resume discussion of that principle in the final two chapters. For the moment, however, I turn to objections to the Davidsonian account outlined here.

102

3. Replies, Objections and Suggestions

3.1

Introduction

Davidsons error theory has rightly commanded a lot of critical attention over the last twenty five years. Each component of that theory the negative case for semantic and pragmatic anti-realism about metaphor, the diagnosis of the mistakes that led theorists to posit metaphorical meanings, the suggested functional replacement has been examined, reformulated and criticised. The evaluation of the theory has been far from uniformly negative. It is widely recognized, even by realists about metaphorical meaning, that the conventional wisdom about discerning two senses of the predicate term that Beardsley had adverted to three years earlier, was shown to be misguided by the considerations that Davidson brought to bear. 120 Contemporary recognition of the importance of elucidating the dependence of metaphorical language upon its literal base, and upon its context of utterance, can also be seen to have resulted from sustained critical engagement with Davidsons article. Nevertheless, and unsurprisingly, Davidsons position as a whole has not commanded widespread acceptance. An error theory of whatever stripe is typically a revisionary enterprise, so it is predictable that philosophers have been interested in investigating whether a more conservative position, consistent with the genuine insights of Davidsons argument, might enable us to preserve our pre-theoretic practice of ascribing truth and meaning to metaphors. In this chapter, I will outline and evaluate some influential objections to Davidsons theses. I do not pretend to completeness; there are no doubt many objections, good, bad and indifferent, that I do not consider here. My purpose is to rather complement and reinforce the discussion of the previous chapter by addressing a selection of the concerns that the Davidsonian approach regularly provokes. In particular, I will concentrate most of my attention on worries relating to the adequacy of an error theoretic functional replacement for talk of metaphorical meaning.
120

See Beardsley (1976)

103

3.2

Complaints about the Causal Theory

Davidsons error theoretic approach fails to provide an acceptable functional replacement for our pre-theoretic talk about the workings of metaphor. His Causal Theory suggests that metaphors work by brute causation, nudging us into noticing similarities in a way that might just as well be brought about by a pill, or a bump on the head. But thats hopeless. In particular:

1. Such an account ignores the fact that metaphors are produced and interpreted within the space of reasons. Metaphors can be misinterpreted, supported by evidence, produced for reasons, rejected as unwarranted or false, etc. (Lycan, Nogales) 121

2. Such an account wantonly ignores the fact that we know very little about each others cognitive architecture and subjective associations, and thus can hardly be expected to predict the causal effects of producing a given metaphor. 3. Such an account cannot explain why the syntax of metaphors might be important, and thus cannot account for the clear metaphorical difference between e.g. Surgeons are butchers and Butchers are surgeons (Stern, Nogales). 122 Such an account also blurs the difference between metaphors and strings of nonsense syllables, or word-salads, which might equally well causally bring about some recognition of similarities, etc. (Lycan). 123 Davidson is often presented as endorsing a Causal Theory of the way in which metaphor works its wonders, albeit one that is only sketchily developed. This is slightly misleading. In many ways, it would be more accurate to say that Davidsons scattered

121 122 123

See Lycan (1999) p. 212, Nogales (1999) p. 121 Stern (2000) pp. 47-48, Nogales (1999) p. 118 Lycan (1999) p. 211

104

remarks on the subject are an attempt to indicate why we should not expect to have any interesting explanatory theory of the workings of metaphor. It is true, however, that, like many error theorists, he does see the need to explain how what is genuinely valuable about our practices can outlast the discovery of the falsity of much of our theorizing about them. The nature of this account has, however, been widely misunderstood. We began, in the last chapter, to investigate the source of the error that Davidson takes us to have fallen into in our pre-theoretic thought about metaphor. A misunderstanding of the nature of meaning in general in particular, a failure to give proper weight to the a priori connections between word-meaning on the one hand and principles such as (Compositionality) and (Warrant) on the other has led us to make a useless theoretical posit, a cog that turns out to be driven by what it was intended to drive. Getting clearer on meaning and metaphor is supposed to let us see that to the extent that systematic enquiry into the latter is possible, its place is external to linguistic theory properly so-called, being concerned instead with the realm of extra-linguistic goals and perlocutionary effects. To help us get clearer on Davidsons line of thought here, let us consider two analogous cases. Firstly, take the case of warnings. A warning is something that can be achieved without recourse to natural language. For example, one might draw a picture of a bull, and leave it attached to a suitable fence. Nevertheless, it is clear that warnings often are expressed linguistically, and that when they are, they make use of e.g. the syntactic and semantic properties of language. Consider the case of warning somebody that a bull is in the field, by uttering the sentence theres a bull in that field. Such a speech act can clearly be done for reasons, often very good ones in fact. It can be morally and practically evaluated, and one can be held to account for performing it e.g. on the basis of limited or irrelevant evidence. It can be misinterpreted by the unwary; as a dare, for example. Ones purposes in so acting can go unrecognised, remain unfulfilled or be challenged. Nobody thinks that the words bull or field change their meaning in the utterance, and implications made by the utterance that the bull is dangerous, for example do not take on the status of a special warning meaning of the utterance. Consider also the case of speaking pedantically. When one speaks pedantically actually, you havent shown that asserting the consequent is invalid, just invalid in first 105

order logic with identity - one may aim to bring about quite definite effects. For example, one may wish to prompt ones student into being careful not to confuse logic as a whole with the type of logic we teach at an elementary level. Those aims, again, may be well- or ill-advised, pursued with wilful disregard for pedagogical evidence, etc. Nobody thinks that the words logic or identity change their meaning in such utterances, and nobody confuses the aim of the utterance with what is conversationally implicated by one who makes it. Nor does anybody think that it is a particularly sensible project to investigate the mechanism by which warnings and pedantic reminders about logic work their wonders. (Apart from anything else, the holism of evidence means that there are simply too many ways in which we can come to believe that e.g. somebody is warning us of something). A Davidsonian thinks of metaphor in similar ways. In making a metaphor, one typically has some goal or purpose in mind. Often, the point of a metaphorical utterance is to get somebody to see similarities and analogies between two things, or two situations. Often, it is to get somebody to see or think of one thing as another; or to put it less conventionally, if just as obliquely, to think of one thing through another. This goal is typically extra-linguistic in the sense that it is often something that, in principle, one could achieve without using natural language. The purpose with which a metaphor is made can be evaluated, rejected or endorsed, and the means chosen to achieve that purpose criticised or applauded. People may fail to take up the metaphor in the intended sense, or may respond with ridiculously far-fetched comments and elucidations. This does not imply, however, that we should assimilate speaking metaphorically to a distinctive kind of speech act. Speaking metaphorically is most naturally contrasted with speaking literally, and speaking literally is not a distinctive speech act. To say something literally is a way of saying something, not an alternative to saying something. The same is true of making metaphors, although if Davidson is correct, our purpose in making metaphors only involves communicating a particular proposition in relatively atypical cases. Inviting someone to see philosophers as flies trapped in fly-bottles is something that could be done in a number of ways, only some of them linguistic; speaking metaphorically is one effective way to do it. In this respect, speaking metaphorically is more like the practice of speaking pedantically than that of giving warnings. 106

Nevertheless, it ought to be clear that such acts typically take place within the space of reasons, on any reasonable understanding of that enigmatic phrase. An invitation is something that can be offered for good reasons or bad; its nature and purpose can be mistaken and misconstrued in a variety of familiar ways; it can be based on mistaken beliefs, or comprise an ineffective route to its goal. A Davidsonian simply need not think of metaphors as brute causal prompts, not amenable to intentional explanation, and to think that she must is to misread Davidsons remarks on the topic. A worry might remain about the case of falsehood. Surely, we can often reject metaphors, deny them, etc, in ways that go beyond the clash of purposes here expressed. I will deal with this in Section 3.3 below. For the moment, my concern has been to defuse a more general worry; that a causal account of metaphor assimilates it to brute psychological promptings, which do not admit of explication in terms of reasons, and whose effects it unreasonable to expect a normal producer of metaphor to predict. I take this charge to be misguided. 124 It seems to be this misunderstanding of the Davidsonian account that underlies the final objection outlined above. 125 Stern claims:
in point of fact, Davidsons own explanation of how metaphor works does not appeal to more than the separate literal meanings of the individual words in the sentence, ignoring any contribution made by the string syntactically or semantically structured as a sentence. For Davidson, there is no difference between a metaphor and a poem like T. S. Eliots The Hippopotamus: both are devices that alert us to aspects of the world by
124

This misreading of Davidson seems to have been fostered by the construal of his account offered by

Richard Rorty. See in particular Rorty (1991) p. 167, where he compares metaphors with anomalous non-linguistic phenomena like platypuses and pulsars that can prompt new insights in a wholly unpredictable manner. Much of what Rorty says in that article is strictly correct, if expressed in way which is likely to mislead, but some of it seems to me to be straightforwardly incorrect. For example: live metaphors can justify belief only in the same metaphorical sense in which one may justify a belief not by citing another belief, but by using a non-sentence to stimulate ones interlocutors sense organs p. 169. This is about as plausible as holding that my recognition of the fact that somebody is warning me that a bull is in the field cannot justify my belief that a dangerous bull is in the field.
125

Or at least, if it is not this, the objections seems to me to be wholly unmotivated.

107

inviting us to make comparisons. But Eliots poem works simply by the alternating presentation or display the brute juxtaposition, as it were of stanzas or clauses referring to hippopotami and the Church. Likewise, Davidson would have us believe that metaphor works simply by way of the linear sequence of literal meanings of the individual words of the utterance, regardless of its sentential syntax. 126

Like many commentaries on Davidsons position, this is uncharitable and implausible in the extreme. A disaffected teenager who asserts that all politicians are the same is not thereby credited with believing that they resemble each other in every respect. Similarly, even if Davidson had said there is no difference between the way metaphors work and the way that poems like Eliots work which he didnt - he might be reasonably be taken to mean, given the immediate context, that they are both things that can alert us to similarities in the world without requiring special meaning-shifts. 127 He would not have thereby committed himself to regarding them as equivalent in every respect. In particular, he need not hold that it is simply the brute juxtaposition of the terms in the metaphorical sentence that brutely causes the perception of certain similarities to spring to our minds. When dealing with the complexity and efficacy of our norm-governed practices, he can easily afford to be far less brutish than that, without giving up on his basic claim, that we have mistakenly assigned to metaphorical meaning what belongs to the realm of perlocutionary effect. If Davidsons account relied on treating the sentences used in metaphor as unstructured strings of words or sounds, which then served to prompt comparisons between two subjects, then he would be vulnerable to the Surgeons/Butchers objection, and to the worry that he cannot distinguish the case of metaphor from that of nonsensical strings. But it doesnt. For all Stern has shown, a Davidsonian can perfectly well hold that quite complex syntactic, semantic and pragmatic properties of sentences play a practically

126 127

Stern (2000) p. 47-48 Davidson (1984a) p. 256 comments on the poem as follows. Here we are neither told that the Church

resembles a hippopotamus (as in simile) or bullied into making this comparison (as in metaphor), but there can be no doubt the words are being used to direct our attention to similarities between the two. Nor should there be much inclination, in this case, to posit figurative meanings, for in what words or sentences would we lodge them?

108

essential role in bringing about the intended perlocutionary effects. 128 I take Davidson to be a Davidsonian in this regard.

3.3

Metaphor and Context

1. Davidson argues that the context-dependence of metaphor is inconsistent with semantic or pragmatic realism about m-meanings. But literal language is equally context dependent. Therefore, Davidson is either confused, inconsistent, or making an ad hoc ideological exception. (Kittay, Denham) 129

2. The suggested defence of (A3) and (A4) ignore the possibility of treating metaphors as analogous to indexical and demonstrative utterances. (Stern) 130

3. The suggested defence of (A3) and (A4) ignore the possibility of metaphorical meaning resulting from primary pragmatic processes (Sag, Reanati) 131,132 Davidsons famously compressed style of argument arguably reaches its apotheosis in the following much-cited passage
Once we understand a metaphor we can call what we grasp the metaphorical truth and (up to a point) say what the metaphorical meaning is. But simply to lodge this meaning in metaphor is like explaining why a pill puts you to sleep by saying it has a dormative power. Literal meaning and literal truth-conditions can be assigned to words and
128

The above quotation from Sterns immediately follows discussion of an objection of Whites and

Margalits to the effect that metaphors taken literally are often nonsensical. Stern may think that the suggested feature of Davidsons account enables him to sidestep this latter objection, but if so, its clearly a poisoned chalice. I discuss the White/Margalit objection below, in section 3.3
129 130 131 132

Denham (2000) pp. 262-263 Stern (2000) passim. Reanati (1993) p. 263-264 Sag (1981) p. 264

109

sentences apart from particular contexts of use. This is why adverting to them has genuine explanatory power.
133

I have already made some suggestions regarding how I think we should understand Davidsons point here (in Section 2.6. above), which I will not recapitulate here. However, given that misunderstandings of the sense in which Davidson views the literal as essentially context-independent are still fairly commonplace, it may be worth making some brief remarks. Thankfully, a useful set of distinctions has recently been articulated by Josef Stern, that serve to address the Kittay/Denham mistake succinctly. Let us distinguish three types of context dependence; presemantic, semantic and post-semantic. Pre-semantic context dependence springs from the fact that in interpreting a linguistic act as involving the utterance of some sentence-type, we are forced to draw on a substantial number of rich and detailed contextual cues and clues. Stern gives an apposite example:
I hear the sound pattern . Even knowing that the speaker is speaking English, I must decide whether what I heard was the first person indexical I or the common noun eye or the affirmative aye or the groan ai. In making this judgement, we rely on all sorts of contextual cues the appropriateness of the alternative types within the immediate string and then within the larger discourse, our beliefs about the speaker and his intentions, and so on.
134

The fact that both metaphorical and literal speech are context dependent in this sense does not, clearly, invalidate Davidsons argument. For it is surely common ground that understanding what is being said in this sense requires contextual information. A sentence-type S is semantically context-dependent just in case an utterance of S only expresses a propositional content after its indexical and demonstrative parts are assigned determinate semantic values by context. Thus That is a hat is intuitively semantically context-dependent, and The largest terrestrial sea is wet is not. Stern accommodates the context-dependence of metaphor by assimilating it to a form of semantic contextdependence. But Kittay and Denham do not endorse such a view, and this is not the
133 134

Davidson (1984a) p. 247 Stern (2000) p. 42

110

significance of their claim that the metaphorical is on a par with the literal with respect to degree of context-dependence. The relevant conception required for Davidsons point is post-semantic contextdependence. Imagine that we have assigned a propositional content to S, by resolving pre-semantic contextual alternatives in order to identify literal word-meanings and mode of composition, and then filling in any indexical gaps with relevant semantic values. Call this the first meaning of the utterance. Then
that utterance may then be used for an indefinite number of extra-linguistic ulterior purposes or intentions: to warn, promise, deceive or threaten. Which of these secondary intentions is realized also depends on the context the speakers and interpreters mutual beliefs, intentions and expectations. Yet, whichever further intention is attributed to the speaker, and however the utterance is used, its first meaning remains, indifferently , as the first of the means to these ends. Hence the first meaning is the meaning it has on all its uses or, more accurately, regardless of how it is so used. 135

Literal meaning and metaphorical meaning, as conceived of by e.g. Richards and Beardsley, clearly differ with respect to this property of indifference. M-meaning, thought of as a new sense of the predicate term, varied from context to context, unpredictably, and in a way that could only be read back into the metaphor by somebody who had already grasped its impact. In the case of literal word meaning, in contrast, a predictable and systematic contribution is made to a whole range of utterances, across a whole range of contexts. Davidsons point and this ought surely to have been clear simply could not have been that literal sentences were in no sense dependent on context. That would have been crazy. Rather, it was that m-meanings were sensitive to both local linguistic context and broader, post-semantic purpose, while meanings proper are not. That said, more sophisticated theorists, like Stern, Recanati and Sag, have argued for a variety of conceptions of the dependence of metaphor on context that makes more trouble for the Davidsonian. For Stern, sentences like Juliet is the sun when used metaphorically, have a hidden indexical element, making them semantically context135

Stern (2000) p. 43

111

dependent. For rich pragmatists, the proposition directly expressed by a sentence is not its first meaning, as defined above, but rather some related proposition, which the pragmatic context of utterance helps determine. 136 I do not have space to investigate either of these theories in depth here, and they must remain as standing and sophisticated challenges to any error-theoretic project. However, I will make the following few remarks with regard to Sterns theory. Sterns idea is that e.g. a sentence like a) Grey trees whose lungs had filled up with winter suddenly exhaled a breath of leaves does not have the semantic and syntactic structure that it appears to, but rather something like b) Grey trees whose Mthat <lungs had filled up> with winter suddenly Mthat <exhaled> a breath of leaves. Sterns outline and defence of the Mthat operator is fairly complicated, and I cannot explore its intricacies here. Suffice it to say that it is an operator, somewhat analogous to Kaplans Dthat, 137 which converts a literal expression into a metaphorical expression, whose contribution to the truth-conditions of the whole sentence varies from context to context. Thus, while the context of (b) assigns Mthat <exhaled> one set of properties, the treatment of (c)-(e) as metaphorical expressions will assign it different ones.

136

As well as the reference cited above, see e.g. Bezuidenhout (2001) for a defence of this type of See Stern (2000), passim, for a thorough introduction to and defence of the mysteries of the Mthat

position.
137

operator. Some significant differences from Kaplan include (1) the fact that whereas Dthat attaches to descriptive expressions to form directly referential terms, Mthats paradigmatic use involves attaching to predicates to form new, context sensitive predicative terms (2) that whereas the character of a Dthat expression is a function from contexts to individuals, the character of an Mthat expression is a function from sets of pragmatic presuppositions to sets of properties (3) that whereas Dthat requires some form of demonstration, Mthat does not, being parametric, like genuine indexicals (I, now, etc).

112

c) d) e)

The city languidly exhaled the vapour that it had breathed in from the clouds. Wittgenstein exhaled philosophy. This was Clinton; he exhaled, while never apparently inhaling.

In each of (b)-(e), some distinct set of properties is determined as the semantic value of the metaphorical component Mthat<exhaled>. These properties are determined, as in the case of pure indexicals like I and now, by a general semantic rule, namely, that the properties in question are those m-associated with exhaled in the relevant context. How a property gets to be m-associated with a given expression in context is a pragmatic matter, rather unclear in detail, but Stern makes some more general suggestions. The key point is that Stern takes himself to have offered a broadly semantic model, similar to those applied in the case of indexicals and demonstratives, which can be used to model competence with metaphor. When one understands a metaphor, one (a) maps the literal string onto a string containing a Mthat operator (b) identifies the set of properties that the expression within the scope of the operator is pragmatically presupposed to be associated with in context and (c) substitutes those properties for those of the corresponding literal predicate. 138 Of course, to the sceptic, it will look as if all Stern has done is to defer all the difficulties attendant on explaining metaphors relationship to context to a pragmatic theory, thereby allowing himself a simple semantics. Nevertheless, the position is a serious challenge to the Davidsonian, and deserves to be taken seriously. I will content myself with the following observation. Stern apparently thinks of Mthat as expressing an tacitly known rule, which is psychologically real in the sense of being represented in some mental lexicon. 139 He explicitly defends the idea that nobody who failed to master such a rule could count as genuinely understanding or making metaphors. He says:
Metaphorical competence involves mastery of a general skill that one can apply to arbitrary expressions across the language. More theoretically, metaphorical competence consists in knowledge of a schematic rule that applies to all expressions that admit a metaphorical interpretation. This schematic rule governs the characters of all
138 139

Stern gives a different but related account for nominative metaphors, etc. See Stern (2000) p. 225-229 See Stern (2000) p. 198-205

113

metaphorical expressions of the type Mthat [], for each substitution instance of . The metaphorically competent speaker knows how to generate metaphorical expressions given her knowledge of the expressions to be interpreted metaphorically.Either one knows this schematic rule and has the general skill or one doesnt. It makes no sense to say that one could know this rule, or have this skill, for some expressions but not for others. 140

There are two key points here. Firstly, it only manages to seem even remotely plausible that a single rule is involved in interpreting all metaphors because Stern has kicked all the difficult work of saying what determines which properties are m-associated with each given , and how we recognise this into the pragmatics. Secondly, Sterns claim seems to run into trouble with a test for semantic commitment that Kripke famously suggested we apply in the case of Donnellan-type objections to Russellian theories of definite descriptions. 141 Sterns theory clearly differs from the Davidsonian account with regard to the question of whether English contains a special indexical/demonstrative rule of metaphor. How are we to decide who is correct? Following Kripkes suggestion, we might consider a hypothetical language, which is as like English as possible, except that it is stipulated not to contain a psychologically real Mthat operator. In such a language, would people still formulate and understand metaphors in much the way that they actually do? If so, then there seems to be little need to posit such an operator, since it seems designed to explain the emergence of a phenomenon that would equally arise in its absence. As Kripke asks, why posit a semantic ambiguity that is insufficient in general, and superfluous for the special case it seeks to explain? 142 It is rather difficult to apply this test with any certainty, given the obscurities about what determines which properties are m-associated with expressions, but the overwhelming tendency is surely to affirm that the phenomena would arise in any case. Even if we had not mastered a rule that mapped literal predicates onto sets of metaphorically apt properties, we could surely still come to see what somebody was getting at by means of a peculiar form of words. Indeed, given that there doesnt seem to be anything
140 141 142

Stern (2000) p. 198 Kripke (1979) pp. 85-91 Ibid p. 88

114

distinctively linguistic about thinking of one thing as another I can have a sudden sense of the childish or animalistic character of a companion, without employing any expressions of natural language it would be wholly surprising if the absence of such a rule prohibited metaphoric thought. Stern does tell us that the properties which are massociated with an expression in a context C are pragmatically presupposed to be associated with in C, in the following sense: Speaker S presupposes a proposition p in a context c by uttering the sentence s iff (1) S represents herself as believing that p; (2) S represents herself as believing that the other members of c represent themselves as believing that p; and (3) S represents herself as believing that the other members of c recognize that she, S, represents herself as believing that p from her utterance of s. 143 But since Mthat only works when such presuppositions are identified, it seems that Stern can give no account of metaphorical thought in the absence of conversation. This is bad enough, but devastating in the context of Kripkes test. For since, in a community that lacked mastery of Mthat, speakers could presumably still articulate such thought, and hearers latch on to the fact that they were doing so, it seems that metaphorical practice could carry on much as it actually does. And that simply signals the uselessness of the appeal to Mthat in getting to the heart of metaphorical competence.

(P1*)

3.4

Metaphor and Other Linguistic Phenomena

Davidsons error-theoretical approach makes a mystery of some incontrovertible linguistic data related to metaphor. This undermines the claim to have provided a suitable functional replacement for ascriptions of metaphorical meaning. In particular

143

On the basis of this root definition, Stern goes on to define what it is for a sentence and an utterance to

presuppose a proposition, and defines Mthat with respect to the latter. But this makes no difference to the point made above. See Stern (2000) p. 122-123 for details.

115

(1)

Error theory about metaphorical meaning cannot account for the behaviour of metaphors which are embedded in non-assertoric contexts. (Cohen, Stern, Moran). 144

(2)

Syntactic facts about metaphor demand the positing of metaphorical meanings which are independent of speaker intentions. (Stern). 145

(3)

Such an account is committed to treating metaphorical sentences as false when they are taken literally, when in fact many are neither true or false, but rather semantically anomalous (White, Margalit and Goldblum, Stern).
146, 147 , 148

The Frege-Geach problem the demand for an explanation of how utterances that lack cognitive content can embed in non-assertoric linguistic contexts, such as negated or conditional sentences has been a deep and painful thorn in the side of expressivist accounts of e.g. morality, modality and aesthetics. At first glance, it is difficult to see why anybody would think that it was a problem for error-theorists, however. After all, the error-theorist diverges from the expressivist precisely in allowing the kind of straightforwardly truth-conditional treatment of such contexts that the realist herself proposes. Things are interestingly more complicated than that in the case of metaphor, and realism about meaning more generally. An error theorist about meaning in general cannot, it would seem, help herself to the above explanation of the semantic nature of nonassertoric contexts, for the simple reason that by she will, by definition, reject such explanations as erroneous, given the plausible assumption that a sentence cannot be apt

144 145 146 147 148

Cohen (1993) passim, Stern (2000) p 69-71, Moran (1997) p. 260 Stern (2000) p. 69 White (1996) pp. 204-226 Stern (2000) p. 47 Margalit and Goldblum (1994) pp. 234-237

116

for truth if it does not say anything. 149 Assume that a theorist holds that we are wrong to assert any claims like (*) Snow is white says that snow is white. perhaps because they see semantic talk as a regrettable hangover of a scientifically dubious folk theory. 150 Given that all sentences like (*) are false, it seems that snow is white does not say anything, and is thus neither true or false, given the assumption that truth-aptitude requires content. We therefore cannot use the standard truth-conditional explanation of the way in which the meaning of (**) If snow is white then it is coloured. depends on the meaning of its parts, since that explanation relies on the antecedent sentence having a truth value. 151 An error theorist about the metadiscourse, who holds most of our claims about semantic talk is false, may thus also end up struggling to account for linguistic practice in first-order discourses. 152 A similar phenomenon is alleged to occur in the case of metaphorical meaning. Stipulate that metaphors dont express (distinctively metaphorical) propositional contents, whether semantically or pragmatically construed. The question then becomes how we should understand linguistic contexts like
149

I do not mean to suggest that every form or degree of anti-realism about meaning entails rejection of a

truth-conditional explanation of negation and conditionals. A minimalist about meaning might give such a truth-conditional account, but characterise truth as superassertibility. Wrights response-dependent account of meaning can also endorse a truth-conditional explanation. (See e.g. Wright (1987), (1989a-c)). I discuss Wrights account, and its relevance to the case of metaphor, in the next chapter.
150

Such a theory would be analogous to the eliminativism about folk psychology notoriously defended I do not mean to suggest that such a semantic error theorist is likely to see this failing as causing any This type of phenomenon provides another reason to substitute the type of rush for coherence adopted

by the Churchlands.
151

further problems not already inherent in the position.

152

in chapter 1 for Wrights burden of proof on the realist approach. The a priori links between different discourses where realism is an issue means that it is implausible to treat them as a series of isolated debates, with realism always on the back foot.

117

(M1) Its just not the case that, as the Wittgensteinian quietists seem to think, philosophy is just getting flies out of fly bottles. (M2) If the garden was a slum of bloom, then it could hardly have been winter that Stevens was writing about in Banal Sojourn It is clear, firstly, that the immediate problem faced by the error theory about meaning in general can be avoided. Because metaphorical utterances are typically straightforwardly false, and thus truth-valued, when taken literally, there is space for a truth-conditional account of non-assertoric contexts. Unlike the crude expressivist about moral sentences, or the global irrealist about meaning, the Davidsonian can appeal to the literal content of metaphors to explain why they can imbed in more complex sentences at all. The Davidsonian isnt quite off the hook, however. The deeper challenge is to give an account of what is going on when we react to metaphors by apparently negating them, or making conditional inferences from them, or reporting beliefs concerning them. The realist about metaphorical meaning has a relatively straightforward account. For the semantic realist, the metaphorical sentence comes to express a new proposition, distinct from the literal, which can then be negated, conditionalised, feature in belief reports, etc. For the pragmatic realist, the implicated proposition is available to play a similar role. Assume that someone asserts (a) Susan has two children

thereby implicating (b) Susan has exactly two children

The following seem perfectly normal responses to (a) (c) (d) No, she has three children Well if she has two children, well only need two presents.

118

(e)

You may believe that she has two children, but I think youre wrong Im sure she has three.

In each case, however, it is the implicated proposition, and not the proposition which is literally expressed, that is negated, conditionalized, etc. (Or at least, if there is a more general problem here about how merely implicated propositions can apparently be negated, etc, the speaker-meaning theory of metaphor has only one problem to solve. The Davidsonian apparently has two: how negation can govern implied rather than expressed propositions, and how negation can govern metaphors when no proposition is implicated.) The problem is not one that the Davidsonian can duck. Responses like (M1) and (M2) are completely central and everyday aspects of our dealings with metaphor. Any attempt to write them off as the result of an unfortunate mythology of meaning would entail that the error-theoretic position was radically revisionary of our practice, so that it would require significantly more in the way of detailed argument than has been offered to date. Let us return to the case of warnings. If, in response to your warning about the bull in the field, I reply with any of the following (i) (ii) (iii) If the bulls in the field, we had better walk around the long way It might be in the field, or it might be penned in there at the back. Yes, I had thought that it would be.

the force of your utterance is not preserved in the new contexts. Similarly, although, on the Davidsonian account, Wittgenstein and Stevens may have put forward their metaphors as a means of encouraging us to see the end of philosophy as escape, or flowers as slum-dwellers, the illocutionary force of such speech acts is presumably not preserved in (M1) and (M2). So what is going on in those cases? The problem is a difficult one, and it is not an issue that I have resolved to my own satisfaction. But there is perhaps room for Davidsonian to take something like the following line. Although the illocutionary force of a warning, or an invitation, might not survive into non-asserted contexts, pedantry and literality certainly can 119

(iv)

John hasnt submitted a list of his research publications, but rather a list of his publications along with those of his articles that have been accepted for publication but have not yet appeared.

(v)

If John hadnt submitted a list of his research publications, but rather a list of his publications along with those of his articles that have been accepted for publication but have not yet appeared, then we must point it out to him.

(vi) (vii)

John - and Im speaking literally here researches the sex life of pot plants. She had a sudden insight into university life when she realised that John and Im speaking literally here researches the sex life of pot plants.

Illocutionary force might not hold up well under embedding, but ways of speaking certainly do. In a similar vein, I dont think that it is unreasonable to construe (M1) Its just not the case that, as the Wittgensteinian quietists seem to think, philosophy is just getting flies out of fly bottles. (M2) If the garden was a slum of bloom, then it could hardly have been winter that Stevens was writing about in Banal Sojourn as (M3) Its just not the case that, as the Wittgensteinian quietists seem to think, philosophy is just to put it metaphorically - getting flies out of fly bottles. (M4) If the garden was indeed as in Stevens metaphor - a slum of bloom, then it could hardly have been winter that Stevens was writing about in Banal Sojourn How do this help? The basic problem, remember, was not to account for how metaphorical utterances can embed in logical and propositional attitude contexts, but to give an account of what is going on when we do so. The first step that might be made here is to hold that such a treatment of (M1) and (M2), taking them to be saying implicitly what their counterpart utterances say explicitly, makes space for thinking of their component sentences as governed by some kind of modifying operator, qualifier or 120

marker. 153 Our mutual recognition that such an element was, in some sense, modifying the component sentences, might help to explain why we react to such utterances in a fairly structured and predictable way. The problem now is to see how appeal to such an modifier can avoid surrendering the error-theoretic character of Davidsons account. For example, imagine that, with Stern, we think of the relevant sentence as governed by an Mthat operator. Such an operator, remember, converts a literal expression getting flies out of fly bottles into a metaphorical expression, whose contribution to the truthconditions of the whole sentence varies from context to context, in somewhat the way that Kaplans Dthat operator does. Then we have e.g. (M1) and (M3) modelled by (M5) Its just not the case that, as the Wittgensteinian quietists seem to think, philosophy is just Mthat (getting flies out of fly bottles). Since Mthat (getting flies out of fly bottles) is alleged to determine the same truthconditional contribution, in our present context, as e.g. dissolving persistent and frustratingly complex conceptual problems that impede our progress through life, and prevent our living in a fulfilling manner, we seem to have surrendered our anti-realism about metaphorical meaning completely. Employing an operator like Mthat which is essentially just a way of mapping old expressions onto new metaphorical senses, in a context-dependent manner entails adopting realism about metaphorical meanings. So a Davidsonian has to steer a careful course. He has to give an account of what is going on in when we embed metaphors, while preserving a suitably anti-realist account. What is at issue, in effect, is Davidsons right to subscribe to the final of his key theses, discussion of which I delayed to this chapter: Previous influential theories of metaphor falsely subscribed to doctrines thatwere inconsistent with (D1-D3) above. However, much of what such theorists wanted to capture about metaphor can be best upheld by dropping the idea that a given
153

(D4)

I do not mean to suggest that such an operator must be psychologically real, or syntactically

represented. The hope is that some theory of asides will allow such modifiers to guide interpretation, without thereby becoming genuine parts of the string. They should rather be thought of as metalinguistic self-interruptions, intended to ease interpretation. In cases where the need for such interruption is left implicit, their separability from the sentence is not impugned.

121

metaphor is a vehicle for the communication of a distinctive cognitive content, and instead thinking in terms of what it intimates, brings to our attention, or leads us to see. The time has come to examine how a Davidsonian might earn the right to (D4), in the light of the phenomenon of metaphors embedding in non-assertoric contexts. To help identify one potential such course, I suggest we return to Davidsons positive characterization of metaphorical practice, slender though that is:
if I show you Wittgensteins duck-rabbit and I say, Its a duck, then with luck you see it as a duck; if I say, Its a rabbit, you see it as a rabbit. But no proposition expresses what I have led you to seeSeeing as is not seeing that. Metaphor makes us see one thing as another by making some literal statement that inspires or prompts the insight. 154

Let us set aside bizarrely literalistic interpretations of this passage, that take Davidson to be proposing an account of metaphor which relies wholly on the phenomenon of perceptual aspect perception. 155 Davidsons thought is clearly that, via the making some wholly literal statement, we are brought to some sort of imaginative insight into a situation. The metaphor presents us with a kind of lens or prism through which we think of one object, event or situation in the light of another. It is difficult to spell out what such a thought amounts to, although it seems fairly clear that such a process is (i) not intrinsically linguistic (ii) often passive, in the sense that a given metaphoric point of view can force itself upon us, as when I am suddenly struck by the sheer childishness of a colleague (iii) often draws on the rich resources of the imagination, in both propositional and experiential ways. Perhaps despite its vagueness, we might appeal to such a conception of the effects metaphor can have upon us, opening up the possibility that we might think of embedded sentential contexts as somehow relating not to the content of the metaphor, but to the content of such an imaginative engagement. We might take the qualifying marker metaphorically speaking as a pragmatic cue that such an shift in interpretive approach is required. 156
154 155 156

Davidson (1984a) p. 263 Kemp (1991) rightfully takes such theories to task. Similarly, we might take Im guessing here to defeat the normal presupposition of informed

inference in If John comes then Im guessing here Jane will leave.

122

A preliminary and crude attempt at this might be to take: Metaphorical utterance SM (in part) metaphorically-means* that P iff according to an apt act of imaginative insight I that SM, in context, itself prompts or suggests, P. Consider an example. Wittgensteins metaphor encourages me to imaginatively construe philosophers as flies trapped in fly-bottles. Such an activity might, inter alia, lead me to think of philosophical problems as essentially dissoluble puzzles that prevent us living in a fulfilling manner. Such a course of thought is, arguably, mandated by the metaphor itself in context, given Wittgensteins view of philosophical endeavour. At the very least, such a construal is interpretively appropriate. So (F) warrants my claim that the famous metaphor metaphorically-means* that philosophical problems as essentially dissoluble puzzles that prevent us living in a fulfilling manner. But that claim is consistent with my holding that strictly speaking, there are no metaphorical meanings. It is true that expressions will become associated with metaphorical-meanings*, but thats simply a useful fiction; a convenient way of relating metaphors to the content of the imaginative acts they prompt. Nothing deserves to be called a metaphorical meaning, since nothing is suitably meaning-like, and suitably related to the metaphorical character of the relevant utterance. But we might mock up a simulacrum of our talk about metaphorical meaning, projecting a certain kind of perlocutionary effect of the metaphorical utterance back onto the metaphor itself. In this way, we can hope to avoid the charge that we confuse what the metaphor makes us notice with its meaning, while hoping to explain the inferential linkages that embedded contexts reflect.157

(F)

157

This move may seem very non-Davidsonian, in the light of the marks of metaphor discussed in Section

2.9 above. After all, doesnt Davidson explicitly distance himself from such a suggestion? He says, remember: If what the metaphor makes us notice were finite in scope and propositional in naturewe would simply project the content the metaphor brought to mind onto the metaphor (Davidson (1984a) p. 263). So it might well seem that he takes the kind of fictionalist projection suggested above to be ruled out by the limitless, non-propositional character of metaphor. There is much in this criticism, and I will address the relevance of such marks in later chapters. For the moment, my concern is less Davidson exegesis, than an examination of whether an error theorist who rejects metaphorical meaning when speaking strictu sensu can nevertheless give some structured account of embedding. But it is worth noting

123

What is it for a proposition to hold according to an apt act of imaginative insight? This is hard to spell out, although I hope the intuitive idea is clear enough. Basically, it is for that proposition to comprise part of the content of the imagining; the way the imagining presents things in general as being. This intuitive characterization allows for representations drawing on expressive resources from both the imagined world and the real world to enter into the propositional content of the imagining. Thus, we can say both that Wittgenstein characterizes philosophical talk as being a useless, frustrated, buzzing, and that he thinks that it is not a worthwhile intellectual pursuit. The former type of propositions will often be presented in sentences that are themselves seemingly metaphorical in character. Those who believe in the eventual dispensability of metaphorical modes of speech will be able to hold that the content of the imagining will eventually be able to be fleshed out in wholly literal terms. Those who do not may be reassured that the non-reductive element of Davidsons account is preserved by such a fictionalist addendum. 158 Given such a translation from the imagined world, the Davidsonian might try to develop a suitable semantics. I shall not try to outline such a semantics in detail here, but rather simply continue to sketch an approach which is avowedly rough, and no doubt flawed in detail, to illustrate how one might begin on such a task. Imagine that we divide the sentences of our language up into literal sentences like:
that the following remark can be taken in a way that accords fairly well with the suggested strategy. Once we understand a metaphor we can call what we grasp the metaphorical truth and (up to a point) say what the metaphorical meaning is. But simply to lodge this meaning in metaphor is like explaining why a pill puts you to sleep by saying it has a dormative power. Davidson (1984a) p. 247. This suggests that we do grasp something when we understand a metaphor, but that this is the effect, rather than the explanation, of the metaphors success.
158

Why not take the metaphors to be direct expressions of such imaginative states, rather than the more

roundabout route canvassed above? While there is no doubt room for an expressivist or quasi-realist development of such a theory, I will not explore it here. It is worth noting that one prima facie advantage of the current proposal over such an account is that the above account does not require that anybody had such thoughts prior to reflection on the metaphor itself, whereas the expressivist account would tend to see the metaphor as in some sense the result of some prior imaginative state. But there is no doubt room for skirmishing, and for compromise solutions. My primary concern here is with the standing of an errortheoretic account in the light of worries about embedding.

124

(1)

Juliet is the sun

and metaphorical sentences like (2) Juliet metaphorically speaking is the sun.

Metaphorical sentences arent a syntactic kind, obviously. Rather, they contain expressions that are correctly recognisable as having been uttered metaphorically. Sentences like (1) are typically straightforwardly false, whereas metaphorical sentences can be said (in an extended, fictional sense) to be true, when things are as a related apt act of imaginative insight I suggests they are. In more detail Metaphorical sentence SM (in part) metaphorically-means* that P iff according to an apt act of imaginative insight I that SM, in context, itself prompts or suggests, P. Now let *[SM, P1-Pn] abbreviate S: SM metaphorically-means* that P1, & SM metaphorically-means* that P2, & SM metaphorically-means* that Pn Such a notation simply lets us conjoin the various particular propositions that (F) allows us to ascribe to the metaphor. Then we have SM is true with respect to Pi iff *[ SM, Pi] & Pi SM is true simpliciter iff SM is true with respect to each member Pj of some suitably weighted subset of {P1, P2,Pn} and then the standard semantics for non-atomic sentences 125

(F)

(1) (2)

(3) (4) (5)

A & B is true iff A is true and B is true not-A is true iff A is not true. If A then B is true iff not-A is true or B is true.

and so on. Such a semantics clearly doesnt commit the Davidsonian to realism about metaphorical meanings, since we are distinguishing them from the merely fictional metaphorical-meanings*. However, it seems that such a semantics can help the errortheorist set out a realist-style semantics for embedded contexts in a straightforward way. Take, for example, (M1) Its just not the case that, as the Wittgensteinian quietists seem to think, philosophy is just getting flies out of fly bottles. Contextual clues help us identify this as equivalent to (M3) Its just not the case that, as the Wittgensteinian quietists seem to think, philosophy is just to put it metaphorically - getting flies out of fly bottles. which (somehow) combines two propositions: (a) (b) Its not the case that philosophy is just to put it metaphorically - getting flies out of fly bottles Wittgensteinian quietists seem to think philosophy is just to put it metaphorically - getting flies out of fly bottles. For ease of presentation, I shall ignore the latter, and concentrate on the former. It has the form not-SN, where SN = Philosophy is just to put it metaphorically - getting flies out of fly bottles Now we know that

126

Metaphorical sentence SM (in part) metaphorically-means* that P iff according to an apt act of imaginative insight I (that SM, in context, itself suggests), P. Imagine that the relevant act of imaginative insight IN has as its content: philosophys proper aim is to dissolve persistent and frustratingly complex conceptual problems that impede our progress through life, and prevent our living in a fulfilling manner. Then by semantic axiom (1) we have SN is true iff (according to IN, philosophys proper aim is to dissolve persistent and frustratingly complex conceptual problems that impede our progress through life, and prevent our living in a fulfilling manner), & (philosophys proper aim is to dissolve persistent and frustratingly complex conceptual problems that impede our progress through life, and prevent our living in a fulfilling manner) So by employing semantic axiom (4), we can derive not-SN is true iff its not the case that [(according to IN, philosophys proper aim is to dissolve persistent and frustratingly complex conceptual problems that impede our progress through life, and prevent our living in a fulfilling manner), & (philosophys proper aim is to dissolve persistent and frustratingly complex conceptual problems that impede our progress through life, and prevent our living in a fulfilling manner)] Since, by hypothesis, first conjunct of the right-hand-side is true, the above biconditional is equivalent to not-SN is true iff Its not the case that (philosophys proper aim is to dissolve persistent and frustratingly complex conceptual problems that impede our progress through life, and prevent our living in a fulfilling manner) and since 127

SN = Philosophy is just to put it metaphorically - getting flies out of fly bottles we have more or less the desired result. The metaphorical sentence is correctly rejected just in case things arent the way the imaginative viewpoint they prompt presents them as being. Similarly, from (M2) If the garden was a slum of bloom, then it could hardly have been winter that Stevens was writing about in Banal Sojourn we move to (M4) If the garden was indeed as in Stevens metaphor - a slum of bloom, then it could hardly have been winter that Stevens was writing about in Banal Sojourn which has the form If SP then P, where the obvious abbreviations are made. Axiom (5) tells us that this is true iff not-SP is true or P is true. And this, modulo the standard worries about the paradoxes of the material conditional, is exactly what we would expect. The conditional comes out as false when things are as the metaphor prompts us to see them, and yet the interpretation presented in the consequent does not fit with them. So we have some explanation of the use of the conditional, and its place in our talk and thought about metaphor. In rejecting metaphors, we reject the imagined view of the world that they invite. In extending them via conditional thought, we implicitly move within such a worldview, and when we interpret them, we examine the actual world through the intellectual and experiential lens that they provide. It seems then, that in tackling metaphors which occur in e.g. negated contexts, mixed conditionals, belief-attributions, etc, the Davidsonian has some room to manoeuvre. Perhaps the above account, sketchy and provisional as it is, will ultimately prove to be unsatisfactory. Nevertheless, the general approach should surely seem attractive. The great appeal of such a fictionalist strategy is that it seems to provide the error theorist

128

with a means of appropriating realist-style semantics, while rejecting realist ontology. 159 Such an approach makes it far easier to provide functionally equivalent replacements for rejected discourses than would otherwise be the case. Of course, success is far from guaranteed, and must be earned, via the provision of a plausible translation scheme in and out of the fiction. A fully defensible version of fictionalism about metaphorical meaning can scarcely be attempted here. Nevertheless, I hope to have indicated one promising strategy for the Davidsonian, and given the beginnings of an outline of how such an account might go. 160 Such an approach might also be employed to deal with an interesting problem recently raised by Stern. 161 Stern notices that though we might agree that both (1) (2) Achilles is the sun Juliet is the sun

where the relevant imaginative viewpoints are very different in each case, there is something strange about the sentence (3) while (4) Sol is the sun, and Juliet is, too. Juliet is the sun, and Achilles is, too.

is even worse. Stern holds that this shows that metaphorical meanings are necessary linguistic posits. He appeals to an analogous case, arguing that the semantic distinctness
159

I do not mean to suggest that the semantics sketched above works exactly like a standard fictionalism,

say in modality or morality. The point is merely that the introduction of pseudo-meaning promises the best strategy for a Davidsonian who wishes to give some account of embedded contexts.
160

I do not pretend originality in linking metaphorical content and imaginative construal. The above

account has obvious affinities to ideas set out in e.g. Walton (1990), (1993) and Yablo (1998). However, Yablo takes his account to underwrite a realism about metaphorical content, rather than the type of pseudo-fictionalism mooted here.
161

Stern (2000) p. 70-71

129

of the may of possibility and that of permission is established by reflection that while in a given use of (5) John may leave tomorrow

may can be assigned either interpretation, in the following (6) John may leave tomorrow, and Harry, too

grammar demands that we assign the same interpretation to antecedent and anaphoric term. Explaining such an external constraint on what speakers can manage to mean by (6) is just, thinks Stern, what we want word and sentence meaning for. And since a seemingly similar phenomenon occurs in (3) and (4), it might seem that metaphorical meaning is required there too. However, it is clear that the account sketched earlier gives the Davidsonian some chance of addressing this worry. Firstly, reading (2) as 2*. Juliet metaphorically speaking is the sun

itself seems to might serve to explain why we find (4) problematic. After all, (2*) invites an imaginative construal of Juliet as the sun, which the literal statement frustrates. Davidson might argue that 4* Sol is the sun, and Juliet is metaphorically speaking - too. is actually semantically and syntactically fine, but that the intrusive presence of the literal identity claim simply frustrates the ability of the metaphor to conjure up a suitable imaginative viewpoint. Similarly 3*: Juliet is- metaphorically speaking the sun, and Achilles is metaphorically speaking too.

130

is actually true, but simply makes a demand on imaginative focus that we find it hard to discharge. After all, Sterns case is surely overstated if the claim is supposed to be that we simply cant hear the sentences in their intended sense, as happens in the case of may. The following sentence seems fine to my ears: 7. Juliet is (metaphorically speaking) the sun, and Achilles is (metaphorically speaking) too; but the sense in which the former characterization is apt is very different from that of the latter. whereas no such move seems available in (6). Again, such a prospective reply needs to be examined and defended in greater detail than I can spare here. My aim is only to examine whether it is obvious that a Davidsonian must retreat in the light of such linguistically based objections. And, although the matter is a delicate one, it doesnt seem to me that the case has yet been established. What then, about the charge that Davidson cannot give a plausible account of clearly anomalous sentences? That worry, remember, was expressed as follows:

(3)

Such an account is committed to treating metaphorical sentences as false when they are taken literally, when in fact many are neither true or false, but rather semantically anomolous (White, Margalit and Goldblum, Stern) 162, 163 , 164

Several philosophers have taken issue with Davidsons remark that most metaphorical sentences are patently falseon occasion patent truth will do as well. 165 Even philosophers like White and Stern, who are in general sympathetic to Davidsons position, or at least willing to accept that he is right about words retaining their usual literal meaning in metaphor, resist the idea that the sentences composed out of those words need have any literal meaning at all. Such philosophers are willing to give up the idea that a syntactically well-formed sentence composed of significant parts must also be
162 163 164 165

White (1996) pp. 204-226 Margalit and Goldblum (1994) pp. 234-237 Stern (2000) p. 47 Davidson (1984a) p. 258. Davidson make related remarks elsewhere.

131

significant; insisting instead that metaphors like they donated his face to the wildlife fund are semantically anomalous, or nonsensical. Stern suggests that Davidson might argue that the suggested distinction between the literally meaningless and the false rests on a discredited analytic-synthetic distinction. He envisages a persistent objector who replies that the rejection of analyticity is primarily a rejection of its epistemological use to ground an a priori / a posteriori distinction which is simply not what is at issue here. 166 But actually, leaving aside his actual motivation for assimilating the nonsensical to the false, it is not clear that Davidson need appeal to that principle at all. A more straightforward procedure would be to argue as follows. Let W be a thing that it is supposedly nonsensical to imagine donated to a wildlife fund my face, say - and let sentence S be instance of the form W is donated to the wildlife fund. Then Davidson can run through instances of the following schematic argument (1) It is a priori that for all x, if x is donated to the wildlife fund, then x is the kind of thing that can be donated to wildlife funds. (Assumption) (2) It is a priori that, for all x, if x is not the kind of thing that can be donated to wildlife funds, then it is not the case that x is donated to the wildlife fund. (From 1, Modus tollens) (3) It is a priori that W is not the kind of thing that can be donated to wildlife funds. (Assumption) (4) So, it is a priori that it is not the case that W is donated to the wildlife fund. (From 2 and 3, Universal Quantifier rules and Modus ponens) (5) It is not the case that W is donated to the wildlife fund. (From 4) (6) Line 5 follows from a priori true statements by truth-preserving steps, and is thus true. (A prioricity of 1, 3) (7) Line 5 is the sentential negation of S (Definition of S) (8) For all sentences X, if X is the sentential negation of Y, and Y is true, then X is false. (Definition of classical negation) (9) S is false (from 6, 7, 8, &-I, Universal Quantifier rules and Modus ponens) (10) If S is false, then it is not nonsensical. (Assumption)
166

Stern (2000) p. 328

132

(11) S is not nonsensical. (From 9, 10, Modus ponens) Thus it seems that Davidson has an argument that proceeds from relatively uncontroversial steps, from plausible or shared assumptions, to the conclusion that metaphorical sentences are false when taken literally. The issue of analyticity neednt arise. 167 This section has been a long one, so let me recap briefly. As in the case of many other error-theories, a lot turns on whether a suitable functional replacement for our existing practices of ascribing m-meanings can be provided. I have argued that, unlike many error-theorists, the Davidsonian faces a difficult variant of the Frege-Geach problem. I have attempted to sketch the beginnings of a reply to this worry on behalf of Davidson, introducing a way of talking as if the content of the imaginative viewpoints that metaphor provokes really comprised the meaning of the metaphorical utterance. On the basis of such a fiction, I have attempted to show how worries about e.g. metaphors appearance in negated or conditional contexts might be addressed. I do not pretend to have set out anything detailed enough to be regarded as a theory; rather, my aim was to set out a strategy for the Davidsonian, hoping to provide just enough illustrative detail for its contours to be discerned. I then speculated whether the appeal to imaginative viewpoints might help finesse Sterns problem regarding metaphor and anaphora, and concluded by rejecting a distinct set of worries related to semantic anomaly.

3.5

Metaphor and Reduction

(1) Davidsons objection to simile theories of metaphor illegitimately characterises such theories as committed to the idea that metaphors be reduced to explicit literal similes. (Fogelin) 168

(2) Davidsons account illegitimately demands that a linguistic theory ought to issue specifications of content in only literal terms (Denham) 169
167 168

Thanks are due to Gary Kemp for the argument presented here (inter alia!). Fogelin (1988) Ch 2.

133

Davidson makes a series of objections to simile theories of metaphor. Famously, he argues that everything is like everything else, so if metaphors were literal comparisons, they would be trivially true, instead of interestingly significant, as the supporters of metaphorical meanings had held. Recently, Fogelin has argued that this reply overlooks the obvious; the simile-theorist should identify the meaning of the metaphor with the figurative meaning of the corresponding simile. Fogelin sketches an account of the difference between literal similitude and its figurative counterpart in terms of varying standards of salience, modes of relevance and evaluation governing the likeness claim. 170 The reply masks a deeper worry, which is the sheer underdevelopment of the simile theory, whether literal or figurative. We are simply not shown that every metaphor can be parsed into a corresponding simile. Davidsons jokey paraphrase of Virginia Woolfs definition of a highbrow is designed to highlight the point. Woolfs

(a) A highbrow is a man or woman of thoroughbred intelligence who rides his mind at a gallop across country in pursuit of an idea. becomes

(b) A highbrow is a man or woman whose intelligence is like a thoroughbred horse and who persists in thinking about an idea like a rider galloping across a country in pursuit of well, something. The attempt is manful, and the joke light-hearted enough, but the philosophical point is there, and ought to have been addressed by Fogelin. We simply havent been shown how to turn e.g.

(c)

The garden was a slum of bloom

169 170

Denham (2000) p. 259. See pp. 318-328 for an extended discussion. Fogelin (1988) pp. 75-76

134

into simile form. Perhaps: The garden was like a slum of bloom? But whats a slum of bloom? It seems that we have to understand the metaphor in order to make the comparison, so rephrasing as a comparison scarcely casts light upon the metaphor. Similarly with e.g. (d) (e)

The Dean gloated into the office. The Dean did something into the office that was like gloating.

Grammar and insightful analysis tend to come under strain, if we take Fogelins approach seriously. A more interesting question is whether Davidson implicitly appeals to reductivist presuppositions that can be challenged even by non-simile theorists. Perhaps the idea that a systematic linguistic pragmatics requires the possibility of literal paraphrase can be challenged. If I understand her correctly, that is what Alison Denham aims to do in her Metaphor and Moral Experience and elsewhere. 171 There, Denham is interested in what she terms the conceptual autonomy of the metaphorical. She gives two different glosses on the nature of conceptual autonomy as applied to metaphors. The first seems to coincide with the condition that Wright has called disputational purity; (DP) A discourse concerning a particular realm of fact - moral discourse, modal discourse, etc - is disputationally pure just in case there is no other discourse such that "the rational intelligibility of differences of opinion expressible in the former will depend on the existence of differences expressible in the latter". 172 Denham, citing Wright, gives the following definition of conceptual autonomy

171 172

See Denham (1998) and (2000), passim. Wright (1994) p. 155.

135

(W)

A discourse is conceptually autonomous if and only if it is not possible to state the truth conditions of assertoric utterances within it solely in terms of concepts extraneous to it. 173, 174

She goes on to assert the equivalence of this account with Wright's notion, as expressed in (DP)
An alternative way of thinking about conceptual autonomy is this: A discourse fails to be conceptually autonomous if and only if all intelligible disagreements about the truth of judgements expressible in the discourse finally depend on disagreements about the truth of judgements extraneous to it. If it fails in this way, we may say that it is conceptually supervenient.
175

I do not disagree with this equivalence claim, but since it may not seem immediately obvious how the intended equivalence with Wrights original formulation runs, it is worth examining a little further. Let us take, for example, assertoric utterances of the form Metaphorical utterance m in context C metaphorically means that p. 176

(M)

If it is possible to satisfactorily state the truth conditions of claims like (M) in terms which do not mention metaphorical meaning or the like, then, if two subjects should disagree about whether or not (M) is true for a given m, C and p, this should ramify into some disagreement statable in those other terms. Intelligible disagreement is going to presuppose that either some aspect of the subjects reasoning is incorrect, or that they disagree about whether or not the truth conditions of (M)-sentences have been met. Moreover, if these truth conditions can be stated in other terms, then the two subjects ought, if they are thinking rationally, to disagree about whether some fact expressible in

173

Denham actually uses 'if' on both occasions where I have used 'if and only if'; since this weaker claim Denham (2000) p. 283. n. 5. A similar definition is given in Denham (1998) p. 228 n. 3. Denham (2000) p. 283. n. 5. We don't actually say this kind of thing too much , so (M) has to be heard as a useful simplification.

seems unmotivated, and since Wright makes the stronger, I have used the latter.
174 175 176

136

these other terms obtains, since their disagreement should carry over into the new idiom. Call the way of thinking about conceptual autonomy expressed by (W) the Wright way. The second elucidation of conceptual autonomy that Denham gives, however, seems rather different. On this account, metaphorical utterances are conceptually autonomous if and only if "the contents they express are not reducible to literal paraphrase" 177 or "metaphorical assertions admit of no non-circular analysis in non-figurative terms". 178 This is not an entirely happy characterisation, since Denham wants to hold that theorists like Davidson and Rorty, who are anti-realists about metaphorical content, agree that metaphors are conceptually autonomous. Let us amend the definition, then, to Metaphorical utterances are conceptually autonomous if and only if either they do not express propositional contents, or the contents they express are not reducible to literal paraphrase or admit of no non-circular analysis in nonfigurative terms. Call this way of thinking about conceptual autonomy the Denham way. Denham clearly thinks that (D) is relevantly equivalent to (W). Thus, in the same footnote as she gives the two definitions cited above, she writes
Thus some claim that it is not possible to state the truth-conditions of moral discourse in non-evaluative terms; the parallel thought here is that it is not possible to state the truth conditions of some metaphorical discourse in non-figurative terms. 179

(D)

Again, in her introduction she writes

The parallels [between moral judgements and metaphors] run deep, as can be seen by reflecting for a moment on the kinds of questions which often arise with respect to each.

177 178 179

Denham (1998) p. 228. Denham (2000) p. 307. Denham (2000) p. 283. n. 5.

137

For instancecan moral concepts be analysed in wholly non-evaluative terms? Compare: can metaphorical expressions be reduced to literal paraphrase? 180

I have argued above (Section 2.3) that distinguishing between discourses and metadiscourses help us see that this has to be the wrong way to think about the relationship between metaphors and morals. I will not repeat that discussion here. 181 My interest is rather in the second elucidation of conceptual autonomy that Denham gives. Metaphorical utterances are conceptually autonomous if and only if either they do not express propositional contents, or the contents they express are not reducible to literal paraphrase or admit of no non-circular analysis in nonfigurative terms. Since Denhams aim is to show that certain metaphors which meet this condition do express propositional contents, she clearly embraces the second disjunct of the righthand-side of (D). The category of metaphors Denham is concerned with are those she terms phenomenological metaphors. These, she holds, have the following properties (a) They function catachretically, in that they introduce a new use for an old concept, to fill a lexical gap.
180 181

(D)

Denham (2000) p. 2. It is perhaps worth noting that the two definitions offered by Denham, (W) and (D), seem non-

equivalent. Say we grant that metaphors express propositions that don't admit of literal expression. Still, it seems at least consistent to hold that the truth about whether a given metaphor expresses such a proposition p is analysable into some statement about whether its author intended to say that p, or about some pattern of use of p, or the like, as long as p remains non-literally expressible. Thus disagreements stated in terms of metaphorical meanings might be held to depend on disagreements about speaker intentions, or about use properties. Hence, an issue of whether m metaphorically means that p or not might be conceptually autonomous is Denham's sense and not in Wright's. Conversely, if all metaphors were amenable to literal paraphrase, it might still come out that disagreement about metaphorical meaning was compatible with complete agreement on facts concerning speaker intentions, use, context and the like. Just as disagreement about whether some particular incident is funny might be consistent with shared knowledge of all other relevant facts of the matter (to take Wright's example of a disputationally pure discourse), so metaphorical meaning might be more intimately related to individual taste and preference than we tend to think.

138

(b) They take on a representational role which would otherwise be fulfilled by a response-dependent concept, i.e., a role which would be filled by a response dependent concept were one available in the speakers lexicon. 182 (c) They are minimally truth-apt. (d) They are conceptually autonomous, in the sense expressed by (D). Moreover, given that she holds that a metaphor implicitly proposes some relation of similarity to obtain between its topic and vehicle, she suggests (e) If any metaphors are of this kind, then their truth or assertibility conditions will consist in the obtaining of some relation of similarity that is i) capable of yielding determinate truth assignments ii) cannot be expressed in non-figurative terms and iii) is a property that would otherwise be identified by a response dependent concept. Denham introduces the notion of austere similarity, to describe such a relation, and argues that it is best thought of as being constituted by the responses of suitably situated subjects. She draws an analogy with the perceptual case. Defenders of response dependent accounts of colour hold that the similarity between two shades of e.g. pink is not reducible to any deeper metaphysical feature that the two shades have in common, but consists merely in the fact that epistemically well-placed subjects judge that they look the same. In the same way, argues Denham, the relations of austere similarity expressed by phenomenological metaphors does not pick out a set of common features, that we might in principle enumerate. Rather, such relations obtain in virtue of the fact that certain external states of affairs, picked out by the vehicle of the metaphor, strike us as brutely similar to certain of our experiential states. We may not be capable, even in principle, of setting out any exhaustive list of common features underwriting such judgements, but only of to appealing to a limited and provisional set of seemingly relevant features. In extreme cases, we may even be debarred from
doing much more than recording the proposed similarity in the very same terms as the metaphor, merely making explicit that the relation is one of similarity by inserting the
182

Denham (1998) p. 236.

139

requisite like or as. Even if the reader is able to proffer some non-circular elucidation, that elucidation may fail to be given in wholly non-figurative terms it may be that his best prospects lie in offering further metaphors and further figurative similarity claims as elaborations of the target one. 183

Denhams theory thus bears some resemblance to the Fogelins theory outlined above. 184 Fogelins is a non-reductive simile theory, since it does not attempt to assimilate metaphors and similes to cases of literal comparison. 185 Denhams theory of phenomenological metaphors is doubly non-reductive, however. Not only does she claim that the simile obtained from the metaphorical utterance is an (implicit) non-literal comparison, but she casts doubt on the idea that either explicit or implicit figurative comparisons can be fully spelled out in literal terms. Fogelin takes similarity to consist in a shared set of salient features, and differentiates between literal and non-literal comparisons by postulating a contextually variable parameter for salience. Denham, in contrast, holds that the facts about similarity that constitute the metaphors truth condition need not hold in virtue of any shared set of features, except, trivially, their striking us as similar. To summarise, Denham seems committed to the truth of the following schema Phenomenological metaphor q means that phenomenological state X is like Z

(P1)

where Z is a placeholder for some actual or possible type of state of affairs picked out by the vehicle of the metaphor q, and where the relevant concept of likeness is one that can be fully captured via the following (a priori and necessary) provisional equation For all subjects S in ideal epistemic circumstances, (Phenomenological state X is like state of affairs Z iff S judges that X is like Z) The truth conditions for the metaphor q can thus be given as follows

(P2)

183 184 185

Denham (1998) pp. 249-250. Fogelin (1988) See Tirrell (1991).

140

(P3)

q is true iff for some X and Z, q means that phenomenological state X is like Z, and all subjects S in ideal epistemic circumstances would judge that X is like Z.

Since the truth conditions essentially include a type of state of affairs picked out via a figurative comparison, (X is like Z) phenomenological metaphors are conceptually autonomous in the Denham sense. Is an analysis of this kind consistent? As we can see, Denhams preferred account of the truth conditions for phenomenological metaphors involve appeal to ideal subjects. The judgements of such subjects constitute the facts about whether a given experiential state is primitively similar to the state of affairs picked out by the vehicle of the metaphor. 186 Now, there are well known problems with explicating metaphorical meaning in terms of likeness, some deriving from the fact that s can be like s in irrelevant respects, and some from the fact that s can fail to be like s and yet be falsely believed to be so. For example, Wittgenstein is like Kim Cattrall in virtue, inter alia, of being human, but the metaphor Wittgenstein was a real Kim Cattrall is not thereby appropriate. Similarly, gorillas are shy, retiring creatures, but The bouncer was a gorilla does not convey that he was shy and retiring. If Denham is to avoid these problems carrying over her account, she needs some principled way of excluding the possibility that her supposedly ideal subjects might make their judgements as a result of false or irrelevant beliefs. 187 It might seem that Denhams appeal to austere similarities would circumvent this problem. After all, that appeal was supposed to allow Denham to reject the idea that whether a given metaphor is true ultimately depends on the obtaining of some set of similarities which might, in principle, be stated in literal terms. However, it is important to be clear that this move does not remove the need for an account of the ideal subject that explains why such subjects would not have false or irrelevant beliefs. Denham agrees with Wright that any judgement-dependent account of some putative realm of
186

The connection with the Order of Determination test outlined in Chapter 1 should be obvious. I Its not enough to simply rule this out by fiat, since that would render (P2) trivial, violating the

discuss the various tests for realism further in the next chapter.
187

substantiality constraint on provisional equations. I refer the reader back to Section 1.7 for further discussion of the constraints on such equations.

141

facts must give a substantive account of the nature of the subjects and conditions that allow constitutive judgements to be made. 188 Denham is, of course, aware of the old worries about similitude. She writes
As Goodman and Davidson are fond of pointing out, all similarity statements are trivially true: they cannot fail to be true, since everything is in some respects similar to everything elseIf I were concerned with absolute truth conditions attaching to sentences in null contexts this would be problematic, for it would suggest that the truth conditions of metaphors (viewed as implicit similes) were hopelessly underdetermined. However, the level of meaning at which I have located the determination of truth conditions is context relative (Grice's utterance type occasion meaning), and so that worry need not detain us: we can rely on our knowledge of the background discourse and situational context in which the metaphor occurs to constrain the range of possible similarities to a class of relevant ones. 189

The idea seems to be, then, that the ideal subjects whose judgements constitute the facts about whether X austerely resembles Z can be a priori guaranteed not to do so on irrelevant grounds, since their background linguistic knowledge will always constrain the range of similarities to relevant ones. As above, then, we can represent such putative background knowledge as I, and the linguistic knowledge that entitles us to constrain the 'range of possible similarities' as a conditional of the form I -> F, where F stands for the class of similarities relevant to a given utterance of, say q. But now an incoherence emerges in Denham's theory. For if a given phenomenological metaphor q is to be conceptually autonomous in the sense of being irreducible to literal paraphrase, then F must include either the corresponding simile s or some other irreducible figurative expressions f1-fn. Otherwise we would have identified some set of similarities which could provide a literal paraphrase of q, namely, the very set specified in the consequent of the relevant conditional I -> F. The class of similarities F, if it is not to provide exactly the type of literal paraphrase that Denham aims to avoid, must therefore itself include figurative elements.
188 189

See Wright (1994) p. 112, Denham (1998) pp. 45-49 Denham (2000) pp. 292-293.

142

Now, however, we are back where we started. For the problem of how we narrowed down the set of possible similarities, excluding irrelevant ones in a principled manner, was supposed to have been solved by citing our mastery of such conditionals. Since it seems that, in order to master such conditionals, we must already have identified the relevant set - given that the conditionals involve either the very figurative expression in question or some other for which the original problem arises - it seems that we have merely deferred the hard question; why those similarities. Moreover, this requirement seems inconsistent with one of the defining characteristics of phenomenological metaphors, namely, the fact they represent an instance of catachresis, plugging a lexical gap by introducing a new use for an old word or phrase. Denham writes
Phenomenological metaphors are used to represent phenomenally characterized states of affairs and so used because (to the knowledge of informed and competent speakers) no equally suitable word or expression occurs in the languagethey are one way of attempting to conceptualize phenomenology in contexts in which the character of an experience or its objects is too fine-grained or too idiosyncratic to be represented by our standard repertoire of linguistic concepts - our standard lexicon. 190

But if, ex hypothesi, a phenomenological metaphor is being introduced to make up for the lack of a linguistic concept in our standard repertoire, how could we have already mastered conditionals like I -> F, which use either the very concept that was supposed to be being introduced, or some other, for which analogous problems arise? To summarize, Denhams theory seems committed to both (a) grasping a conditional which includes some irreducibly metaphorical element q is part of, or entailed by, speaker competence with a given language L and

190

Denham (2000) pp. 283.

143

(b)

the metaphor q can be introduced into a language for the first time, to fill a lexical gap, by means of some linguistic rule that only draws on the existing conceptual resources of the language L.

But the two theses seem inconsistent. Since it seems that q (or some other figure for which a parallel problem arises) must be a part of the pre-existing linguistic rule that supposedly enables us to master its truth-conditions, (a) and (b) are mutually exclusive. I conclude that we have, so far, no reason to believe that phenomenological metaphors are both cognitive and conceptually autonomous, on either of the resolutions of that notion. So Denhams attempt to elaborate a theory of metaphor that allows metaphors both to express distinctive cognitive contents, and to be resistant to literal paraphrase, is, I believe, unsuccessful. That said, it is clear that rejecting an instance of a theory-type is clearly not to undermine the approach in general. In Chapter 5, I will outline an approach to metaphorical content that would provide the combination of propositional content and irreducibility that Denham aims to secure.

3.6

Conclusion

There are many objections to Davidsons theory of metaphor, and many rival theories have been constructed in the thirty or so years since What Metaphors Mean was published. I have not attempted to address all of them here, and nor have I aimed to treat those I did discuss in the depth that they deserve. However, I hope to have demonstrated that certain influential objections may be more tractable than is sometimes acknowledged, and that some key rival theories may face distinctive difficulties of their own. My central aim has been to elaborate the Davidsonian position in a way that complements the discussion of the previous chapter, and to demonstrate its prima facie defensibility in the light of popular criticism. I will now broaden my focus of attention, turning to the marks of realism outlined in Chapter 1. How do considerations of Truth-Minimalism, Evidence Transcendence, Cognitive Command and Order of Determination bear on my conclusions to date? 144

145

4.

Metaphor and the Tests for Realism

4.1

Introduction

In Chapter 1, I set out a generic approach to disputes involving the objective status of a given realm of fact. I elaborated Wrights minimalist and pluralist approach to truth, and outlined the range of tests that he suggests might serve to focus debate between the realist and anti-realist. In the next chapter, I went on to examine the nature of Davidsons well known anti-realism about metaphorical meaning, and to describe what I took to be his motivation for that position. The detailed character of Davidsons error theory emerged more clearly in the last chapter, in the course of correcting misplaced objections to it. Moreover, I there attempted to highlight the range of dialectical resources, defensive and offensive, that are seemingly open to the error-theorist in securing his position against more serious objections, and making trouble for his realist opponents. My stated aim in this work is to offer a preliminary examination of the significance of recent work on realism for the question of whether we should think of metaphors as expressing propositional linguistic contents. It ought to be clear at this stage that I face a strategic difficulty. On the one hand, a comprehensive review of theories of metaphor, and putative replies to Davidson, is far too ambitious a project to attempt here. I have endeavoured to develop error-theoretic rejoinders to central, widely endorsed and influential objections, and to point out some of the difficulties faced by Davidsons opponents, but no doubt many questions still remain, with many more being raised by my suggested replies. On the other hand, having outlined in some degree of detail the motivation and dialectical resources of error theory about m-meaning, and offered a cautious optimism about that positions standing in each case, it might seem that I have rather foreclosed on

146

a broader examination of the various cruces I set out in Chapter 1. As I will argue in Section 4.2 below, the fact that the error theorists fundamental motivation relies on appeal to norms which are internal to semantic discourse, broadly construed, seems to entail that such a discussion could only concern the realist standing of semantic talk in general, since that is what realists about m-meaning can be defensibly held to aspire to. Such a task is clearly also much too ambitious to be undertaken here. Moreover, given that such work would have served merely a prelude to pointing out the error-theoretic reasons for holding that, in the case of metaphor such talk failed of truth, whatever its realist pretensions, the overall effect would be rather bathetic. I intend to proceed as follows. From Section 4.3 onwards, I bring the detailed discussion of the character of Davidsons error theory to a close, and instead embark upon a slightly more neutral enquiry into the standing of his (A6). I shall try not to rely on the assumption that other aspects of a broadly Davidsonian theory have been demonstrated to be correct, since (a) there are significant rival theories that I have not (and will not) consider in depth, and (b) I have conceded that even a Davidsonian should accept that a speaker-meaning account of metaphorical meaning has its place. I will endeavour to relate the supposed limitlessness and non-conceptual character of metaphor to the tests for realism about m-meaning in a way that of interest to, and consistent with, a number of different theories of metaphor. However, I do think that the ontological parsimony and prima facie defensibility of an error-theoretic account places a significant burden of proof on Davidsons realist rivals. I begin, therefore, by considering one final argument against a generic error-theoretic position, an argument that, if good, and applicable to the case at hand, would successfully serve to undermine the motivation for Davidsons position.

4.2

Wrights Argument against Error theories

Wright, remember, is a minimalist and a pluralist about truth. His truth minimalism consists of two claims, relating to (1) truth and (2) truth-aptitude. The former amounts to the claim that it is sufficient for a given predicate T to meet the conceptual 147

requirements on truth that it accord with a series of platitudes. A richly metaphysical or epistemic account of truth is not secured merely by reflective appeal to such uncontroversial principles. Minimalism about truth aptitude comprises two sub-claims; (a) that possession of truth aptitude is co-extensional with possession of assertoric content and (b) that it suffices for an utterance or discourse to be assertoric that it meet two constraints regarding (i) syntax and (ii) discipline. Roughly, if a sentence embeds appropriately under negation, etc, and there are publicly shared standards which determine when it is correct and incorrect to employ it, then the sentence possesses assertoric content. Finally, Wrights pluralism results from his view that although truth is conceptually minimal, it may be metaphysically robust, in a variety of kinds and degrees, across different subject areas. Truth can be defined in minimal terms; but what falls under or realizes that definition in particular cases may be metaphysically substantial. Wright argues that such a conception of truth makes life difficult, if not impossible, for error theorists. Discussing the question of whether scope for error theory is simply closed off by a minimal theory of truth, he writes:
The position, I think, is that error-theoretic proposals remain theoretically feasible but that their development is interestingly constrained. Suppose a philosopher denies that anything or much of what we say within a given discourse is true but grants that its truth predicate admits interpretation as superassertibility by its standards of assertoric warrant. 191 Then he has to produce reason to deny that anything, or very much, of what we say is superassertible by the lights of those standards. And that will be a commitment to denying that any, or very many, of the statements of the discourse are even assertible by those standards.

Wright goes on to identify two possible strategies for the error-theorist who aims to show precisely this:

191

A sentence is superassertible, remember, if we are epistemically justified in asserting it, given our

current state of information, and we would remain so justified given any way that that state of information might be enlarged or improved.

148

One is to contend that while the standards in question are perfectly coherent, little of what we accept really complies with them; our error, on this view, will consist in a propensity to apply those standards erroneously. The other is to argue that the discourse is not governed by coherent standards of assertoric warrant, and consequently that nothing is genuinely warranted by those standardsThe great question, it seems to me, is not the coherence of error theory but its motivation; why insist on construing the discourse in terms of a notion of truth which has us in massive error when the alternative of superassertibility is prima facie available and would avoid the charge? 192

Wrights question is an important one, and pressing in the present instance. Davidsons theory of metaphor apparently represents everyday talk and thought as being in massive error. When we call metaphors true or false, or talk about what they say, or signify or mean, we are completely mistaken. How can this be the most charitable interpretation of our practice? Isnt the minimalist account - that represents us as being answerable to standards of warrant which are internal to the practice of making/interpreting metaphors, and which can, therefore, secure the defensibility of our standard ways of thinking about metaphorical meaning - by far the more plausible position? This seems especially worrying in the light of the mocked-up account of truth and falsity that I sketched on behalf of the Davidsonian in the last chapter. If some such account is available, then why not simply identify the norms governing our attributions of m-meaning with the norms of the alleged fiction? The error theorist seems compelled to explain why he takes the standards of correct use that govern our way with metaphors as having been set so high that we almost always fail to accord with them, when he himself is committed to the existence of a much lower standard, whose demands we manage regularly to respect. The idea that minimalist theories of metaphorical meaning have such a default status has been defended in the literature, notably by Denham. 193 She writes:
How does my account of phenomenological metaphor answer the objection that irreducible metaphors do not express truth-apt judgements? Can it show that phenomenological metaphors are held accountable to sufficiently stable standards of correctness in the absence of any non-circular analysis in literal terms? I aim to do that
192 193

Wright (1992) pp. 86-87 Denham (2000) Ch 7-9

149

somewhat indirectly, by showing how we might assess whether or not a subject understands a given metaphor; I take it that if we have to hand standards or norms which are sufficiently disciplined to determine attributions of understanding, we have also standards or norms which are sufficiently disciplined to confer the status truth-apt. 194

I actually agree with many of the suggestions Denham makes regarding the criteria for manifesting understanding of a metaphor. The capacity to elaborate a metaphor using a mixture of literal commentary and further figuration; the fact that one is able to pick out paradigm instances of objects, events, etc, that would fall under particular metaphorical characterizations; the ability to extend the metaphor in the natural way all these suggestions seem eminently sensible. So it may seem that Wrights minimalist challenge to the error-theorist can be reinforced by the development of a set of plausible proposals describing the nature of the more modest standards the minimalist appeals to. However, such an argument is not, I think, nearly as attractive in the case at hand as it is in e.g. that of moral, aesthetic or mathematical talk. Consider an analogy. It ought to be uncontroversial that error-theory about phlogiston, or the humours, is the appropriate position to take. When scientists posited the existence of such substances, it was hoped that they would thereby play a genuinely explanatory role in accounting for certain natural phenomena. When it turned out that such explanatory work could not be delivered in a way that was consistent with other of the commitments of natural science, the original motivation for believing in such entities simply lapsed. The good standing of assertions about phlogiston depended upon their being sanctioned by the existing, reflectively endorsed norms governing scientific practice and discussion. It would have been wholly inappropriate to think of retreating to a less demanding norm of correctness for e.g. phlogiston-assertions. The point of the standards that govern scientific discourse as a whole would simply have been undermined by such a move. The same, the Davidsonian should maintain, is true of the case of metaphorical meaning. Better founded, more fundamental commitments of our thought about meaning count seriously against us our taking talk of m-meaning seriously.

194

Denham (2000) p. 318-319

150

It is no doubt arguable that talk about meaning is less robustly objective than theoretical science, although this is scarcely uncontroversial. The issue can be pressed on a number of fronts. For example, in Section 2.8 above, I committed the Davidsonian to the following principle; If a metaphor m-means that P, then (in somewhat idealized conditions) we can get some warrant for believing that it m-means that P. Such a principle bears in an obvious way on the question of how the objective standing of the discourse is illuminated by the first of Wrights key tests: (Evidence Transcendence) Does the discourse at issue include sentences with regard to which a) we 195 have no evidence that bears on their truth value, and b) we lack a conception of any way in which we might come to get such evidence? Moreover, Davidsons well-known identification of amenability to radical interpretation as the constitutive mark of the semantic might lead us to think that meaning-talk is best construed in a response-dependent manner, with semantic facts being constituted rather than tracked by the judgements of a suitably informed radical interpreter. 196 So the Order of Determination test might also be taken to be resolved in the favour of the anti-realist about semantic facts, broadly construed; (Order-of-Determination) Given that we can identify a true Provisional Equation involving the discourse at issue, do the best judgements of ideally placed subjects constitute the relevant facts, or merely track independently constituted facts? Questions regarding Wide Cosmological Role seem similarly resoluble, since there seems little prima facie reason to suppose that semantic facts enter into a suitably wide
195 196

(Epistemic Access)

Or, perhaps, some suitably idealised counterparts of us. See (1984) passim, especially Essays 1,2, 9 and 10.

151

variety of explanations. Finally, it is unclear whether it is really a priori that every disagreement over a semantic fact results from a cognitive failing, in the way that Cognitive Command demands, since it is unclear exactly what e.g. Quinean-style semantic indeterminacy shows us about the cognitive standing of judgements of linguistic meaning. 197 So it might seem that minimalism about meaning in general is mandated by broadly Davidsonian commitments in other areas of philosophy of language, providing a disanalogy with the scientific case alluded to above. 198 This would be to mistake the point of the analogy with outdated scientific theory, however. The point is rather that whatever the realist standing of the constitutive norms governing a given discourse, it must always be a live possibility that assertions which appeared to be warranted by such norms might turn out not to be. So much is written into the notion of a genuine norm or standard. A standard that could not, even in principle, fail to be lived up to, could provide no conceivable influence on action or belief, and thus would hardly be playing a normative role at all. One cannot sincerely intend to bring ones actions into accord with something that will sanction them no matter what. 199 Susceptibility to error is built into the notion of respecting a norm. So even if semantic discourse is accountable to merely minimal standards of correctness which is itself, of course, hotly debatable there ought still to be an open possibility that we have mistakenly taken certain of our judgements to be warranted by those standards, when in fact they ought to have been rejected. That, in effect, is the contention of the Davidsonian position I have been sketching in earlier chapters. Such a position does not accept that discourse about the meaning of metaphors is independent of semantic discourse more widely construed. Rather, it holds that constitutive elements of that wider discourse Compositionality, Epistemic Access, the constitutive goals of a pragmatic theory, et al provide the resources for a demonstration that no suitable account of m-meaning is genuinely forthcoming even by the potentially minimal lights of that very discourse. Just as in the scientific cases, principles and data that we count as better grounded, or at least more fundamental, are
197 198

See e.g. Quine (1960) Ch 2, Evans (1975) for discussion of such semantic indeterminacy. Wittgensteinian and Kripkean worries about rule-following might also serve to establish a meaningSee Petitt (1990) for an extended defence of this thought.

minimalism. For relevant discussion of the stability of such a position, see Wright (1992) Ch 6.
199

152

alleged to overthrow the apparent warrant for committing to a given theoretical entity. This does not seem a wholly implausible position to take. Accord with principles like Compositionality, or something very in their neighbourhood, seems simply mandatory for anything worth regarding a properly semantic discourse at all. And if discourse about m-meaning realism about distinctive propositional contents conveyed by metaphors is not a form of semantic discourse, it becomes very unclear what is. In a sense, then, the Davidsonian error theorist takes the first of the two dialectical strategies that Wright offers:
to contend that while the standards in question are perfectly coherent, little of what we accept really complies with them; our error, on this view, will consist in a propensity to apply those standards erroneously.

However, this has to be understood in a qualified manner. The error theorist will typically concede that much of semantic discourse is well-founded in the light of its proper standards, rejecting only those parts of it that mistakenly extended to the story of how metaphor works its wonders. The intended lesson of the phlogiston example was that the debate regarding the existence of metaphorical meaning is a debate which is internal to semantics. Theorists like Denham, who aim to give a self-standing minimalist account of the metaphor, simply fail to give sufficient regard to the internal connections between superficially different discourses. Such a mistake is unsurprising, and tempting, in the light of the sketchy and intuitive notion of discourse that Wright appeals to in the course of setting out his framework for debates about realism. But it is a mistake nevertheless. Metaphorical meaning, if it lives at all, lives in the shadow of meaning proper, and inconsistency with key commitments of that grounding discourse spells doom for realism about metaphorical content. Mere surface discipline in insufficient in this kind of case. 200

200

Similar arguments could be adduced against e.g. minimalism about the analytic/synthetic distinction. If

it turns out that the notion of analyticity is thrown into confusion by central facts about meaning in general, then adverting to the surface discipline of ascriptions of analyticity could not pace minimalist descendants of Grice and Strawson suffice to rehabilitate it.

153

The foregoing line of thought no doubt requires much more in the way of elaboration and qualification. Like many other aspects of this work, it marks more of a direction of travel than a route map. If it can be made defensible, however, then it marks one, admittedly modest, area in which debates about the nature of realism can be clarified by examination of the case of metaphor. It is easy to overlook how much is built in to Wrights notion of a discourse, and equally easy to assume that change of focus marks change of subject. Do moral talk, aesthetic evaluation and epistemology count as three different forms of discourse, or as three sub-branches of a more general language-game involving the making and justifying of normative ascriptions? Its difficult to tell, but a lot hangs on the answer. If it turned out, for example, that deep, constitutive epistemic commitments undercut a certain species of moral or aesthetic judgement, Wrights plea for charitable construal of everyday talk would carry quite different weight, depending on how we individuated discourses. 201 I do not pretend that Wright is unaware of this, nor that the plea for charity of construal might not be reformulated in a way that would help address it. But as the case of Denham has shown, it is easy for even sophisticated theorists, familiar with the details of Wrights work, to overlook the possibility of appeal to internal clash of standards. Even if the dialectical strategy that I have offered to the error theorist fails in a given case, or even in every case, it at least ought to be considered seriously and argued against. The case of metaphorical meaning, I believe, brings this simple moral out quite nicely. 202

4.3

Metaphor and Evidence Transcendence

In Section 2.5, I outlined what I took to be the best master-argument for Davidsons key theses about metaphor. A central plank of that argument was principle (A2): If a proposed assignment of a metaphorical content to a utterance

(A2)
201

Similarly, transcendental arguments for the existence of epistemic norms would bear in quite a different Moreover, it gives further support to the broad church, holistic methodology that I defended in Chapter

way on the metaphysical possibility of moral norms.

202

1. If we arent sure exactly what a discourse is, it can hardly be wise to rely on such a semantic notion to the exclusion of other equally well-founded epistemic and metaphysical commitments.

154

results solely from an unacceptable theory of metaphor, then the utterance does not express that content. I expressed the hope there that (A2) would be common ground between realists and antirealists about metaphor. Such an expression can often turn out to be overly optimistic. In Section 2.8, in a slightly different context, I did set out one obvious defence of (A2), were one to be demanded, as follows. Grant the following two principles: If a theory is crazy, then it doesnt give warrant to beliefs formed on the basis of it. If a metaphor m-means that P, then (in somewhat idealized conditions) we can get some warrant for believing that it m-means that P. Now, assume that the only putative warrant that we could get for an ascription of mmeaning was a crazy theory. By (Warrant), no warrant for believing that the metaphor did have such a content would spring from that theory. So we would have no warrant at all for accepting the ascription. Modus tollens on (Epistemic Access) gives us that the metaphor does not have that content, and conditional proof gives us the plausible version of (A2). I noted above that the latter principle, (Epistemic Access), seems, if good, to resolve the first of Wrights cruces: (Evidence Transcendence) Does the discourse at issue include sentences with regard to which a) we 203 have no evidence that bears on their truth value, and b) we lack a conception of any way in which we might come to get such evidence? 204

(Warrant)

(Epistemic Access)

203 204

Or, perhaps, some suitably idealized counterparts of us. I will abbreviate these principles to (EA) and (ET) for ease of reference in this section.

155

If we can always get some warrant for the true belief that a metaphor means that P, then it seems harmless to think of some moderately idealized counterpart of us as actually possessing such evidence. Moreover, our grounds for accepting something like (EA) presumably depends upon our having some conception of how we might go about getting some evidence relating to what the metaphor means. Either way, commitment to (ET) seems assured. Most extant realists are likely to agree with such a position. The general strategy embarked on by a realist theory of metaphor is to show how some well-founded linguistic theory justifies us in ascribing meaning to metaphors. Admittedly, presumption of global knowability is rarely explicitly stated, but that is in general best explained by the fact that it is widely adhered to. 205 Take, for example, any theory of metaphor that gives a special determinative role to speakers intentions. 206 We typically assume that we can come to possess warrant for beliefs about what people intend, and that such an assumption is not merely a reflection of wholly contingent epistemic luck, but something that reflects on what it is to form an intention.. There are, of course, complications. We might be skeptical about whether intentions are always really warrantable from the third person point of view, or might have theoretical or dispositional reasons for ascribing widespread self-deception regarding our actual goals and purposes. Nevertheless, first person access is a perfectly respectable way of getting access about ones plans, and even the most extreme Freudian has some idea about how we might go about getting some evidence regarding what somebody intends to do. So it seems that adherents of the default, speakers meaning position ought to sign up to an anti-realist reply to (ET). More generally, contextualists like Recanati, Bezuidenhout or Stern are likely to include the intentions of the speaker as one relevant factor determining the nature of the context in question. 207 Since sensitivity to such contextual factors is generally a feature of such
205

It is easy to see why this might be so. Wittgensteins insistence that meaning be manifested in use, and For example, Grice (1975), Searle (1993), Martinich (1984), etc. See e.g. Sterns account of presupposition, that plays a central role in the semantics of the Mthat

Quines rejection of the myth of the museum, have had lasting influence in the philosophy of language.
206 207

operator, and that relies on recognition of what a speaker is representing themselves as doing. Stern

156

contextualists accounts of linguistic mastery, commitment to the epistemic tractability of intention seems secured. A similar argument can be given for sophisticated comparison theorists such as Fogelin, who make essential play with shifting contextual standards in giving their non-reductive account of figurative likeness. Similarly, theorists who make essential appeal to linguistic convention, like Denham, seem committed to our knowing, perhaps tacitly, what such conventions demand of us. Nevertheless, I shall, in Chapter 5, sketch a robustly realist account of limitlessness, that entails that (ET) fails. As far as I know, no contemporary realist endorses such a position, but it does admit of some independent motivation. Since it turns on assimilating the open-endedness of metaphor to a kind of vagueness, however, the most interesting construal of (ET) is not threatened. For the distinction between e.g. the domain of mathematical truth, which we can conceive of as containing interestingly unknowable elements, and judgements of the relative painfulness of our experiential state, is surely not wholly undermined by granting that it is conceivable that we could (essentially) lack warrant in certain borderline cases of the latter. I provisionally conclude, then, that (ET) ought to be answered in an anti-realist spirit, in the case of metaphorical meaning. The case of metaphor, like that of colour, morals, beauty and phenomenology, seems fated to be fought within a more closely epistemically demarcated area than e.g. realist/anti-realist disputes about mathematics, fundamental physics, or the past.

4.4

Metaphor and Cognitive Command 1

The case of (Cognitive Command) is more difficult to appraise, even putting aside Wrights broader worries about that supposed test. 208 A discourse is held by Wright to
accepts that his account of presuppositions relies on the idea that they are somehowmanifest in the utterance. Stern (2000) p 122. That seems enough to secure the suggested answer to (ET).
208

For ease of discussion, I will assume that (Cognitive Command) has somehow been improved so that

it can test for questions relating to the sense in which talk about metaphorical meaning is robustly, cognitively representational. If this simplifying assumption turns out to be false, the Quandary account of the difference between disputes of inclination and matters of fact is presumably an available fallback

157

be robustly, cognitively, representational, remember, just in case some defensible version of the following holds; (CC) It is a priori that differences of opinion arising within a given discourse can be satisfactorily explained only in terms of divergent input (that is, the disputants working on the basis of different information, and hence guilty of ignorance or error, depending on the status of that information), or unsuitable conditions (resulting in inattention or distraction, and so in inferential error, or oversight of data, etc.), or malfunction (for example, prejudicial assessment of data, upwards or downwards, or dogma, or failings in other categories already listed). 209 On the one hand, it is difficult not to agree with the letter of Davidsons claim when he claims that there is no test for metaphor that does not call for taste. 210 On the other, it is difficult to get clear on what precise significance this is supposed to have for the debate about cognitive standing, particularly when the broader context of his remarks make it clear that his primary concern is to reject the idea that metaphor production, recognition and interpretation is rule-governed in any mechanical sense of that term. Such a position is perfectly compatible with a certain kind of particularist, intuitional epistemology of metaphor. Perhaps, then we ought to apply Wrights forced march for the cognitivist about metaphorical meaning. One who holds that metaphorical meaning accords with some suitable revised cousin of (CC) must make the case by identifying a cognitive route to the relevant facts. Such a cognitive route will be either inferential or direct. If direct, then the intuitional epistemology needs to be defended. If inferential, then the nature of
position. If that also fails and I suggest later that it can be put under pressure then a fallback to brute intuitions regarding cognitive status seems inescapable.
209

Wright (1992) pp 92-93. See also Ch. 4, passim, where he defines the notion as follows (p. 144): a

discourse meets Cognitive Command if and only if it is a priori that differences of opinion formulated within the discourse, unless excusable as a result of vagueness in a disputed statement, or in the standards of acceptability, or variation in personal evidence thresholds, so to speak, will involve something that may properly be regarded as a cognitive shortcoming.
210

Davidson (1984a) p 245

158

our robust warrant for the conditionals that capture such inferences must be explained. Pending provision of such a substantive epistemology, we are entitled to construe talk of cognition in a merely minimal manner, a mere reflection of the platitudes requisite for minimal truth. I cannot examine every such potential epistemology here, (although later in this chapter I will examine an under-explored account which is amenable, inter alia, to the speakermeaning theory). Rather, I will content myself with brief consideration of two broader issues. The first is, that in the light of my earlier methodological scepticism regarding burden of proof, it is not wholly clear that Wrights demand is a fair one. Imagine a theorist who thinks that metaphorical meaning is determined via a two stage process;

We display a sensitivity to a rich variety of relevant features of context, somehow integrating them into our unconscious linguistic processing In the case of many metaphors, such pragmatic knowledge intrudes into the derivation of the proposition expressed by a given utterance. This might happen e.g. (i) because, prior to availability to consciousness, the semantically determined proposition is delivered to a pragmatics module which applies some function from propositions to propositions or (ii) because, prior to availability to consciousness, the semantic derivation of the relevant proposition is coeval with a pragmatic derivation, so that the eventual proposition expressed is already partially pragmatically conditioned. Now, imagine that it proves difficult to deliver a detailed account of the nature of such a complex contextual sensitivity. Ought we to conclude that such a putative realist explanation is thereby disfavoured, and anti-realism wins by default? Such a verdict seems to me to be excessively harsh. Perhaps in discourses about mathematics or modality, where we it is difficult to see how we could even stand in the right kind of relations underwriting such a sensitivity, appeal to intuitional epistemology carries a heavy burden of proof. But in the case of e.g. sensitivity to ones context, or the intentions of others, or the tacitly recognized demands of linguistic convention, or the

159

contents of ones imaginings, 211 it seems an unfairly strong constraint on the realist that a full epistemology is to be provided. Surely, we might think, we can begin to understand how a causal sensitivity to contextual factors might exist, even if we are unable to give a theoretical account of it. If thats right, then the lesson of the failure of Wrights suggested forced-march is unclear. Wright does give us more general guidance on when appeal to intuitionism is permissible;
We ought not to associate a special faculty with a particular region of discourse, a faculty, that is, apt for the production of non-inferentially justified beliefs essentially involving its distinctive vocabulary, unless the best explanation of our practice of the discourse, and especially the phenomenon of non-collusive assent about opinions expressed therein, has to invoke the idea that such a faculty is at work. 212

The problem here, as so often, is to see how such a principle can properly serve to constrain debates about realist standing. For one important contributory factor to an explanations good standing is its coherence with prior probability and other theoretical commitments. Somebody who, like Nogales, Stern, or Yablo, simply finds it overwhelming pre-theoretically plausible that metaphors typically express propositions, is unlikely to accept any theory as best that doesnt deliver that result. 213 If the only way to secure such a result is to appeal to a broadly intuitional epistemology, based on e.g. an under-specified causal sensitivity to contextual factors, then that appeal will be
211

I do not mean to suggest that the kind of ersatz account of metaphorical meaning sketched in the last

chapter is a realist account. But it is an interesting question whether every disagreement about the ersatz meaning of a metaphor must depend on some cognitive disagreement. It might, if e.g. the only apt act of imaginative insight is that originally engaged in, or intended, by the maker of the metaphor. I attempt to make such an account slightly less implausible than it would otherwise be later in this chapter. On the other hand, if a plurality of apt such construals are available, some sort of relativism about metaphorical meaning might be defended. I cannot pursue these matters here, except to note that decisions relating to the logic of the according to operator will play a large role in this case, as in that of more straightforward fictionalisms.
212 213

Wright (1992) p. 152 Stern (2000), Yablo (1998), Nogales (1999). I pick these theorists merely for illustrative purposes;

many other philosophers and linguists share their intuitions.

160

made. Again, if we had it written into our methodology that the anti-realist position wins by default, the spectre of a dialectical stand-off need not be particularly worrisome. But I have distanced myself from such a position on a number of occasions now. The difficulty of establishing a neutral framework for the resolution of divergent intuitions about cognitive standing, then is the first of the broad issues that I wanted to deal with. This difficulty might lead us to look not for general constraints, but on local argumentation, to resolve the question. The second issue I will discuss in this context concerns the potential of one seemingly promising such attempt.

4.5

Metaphor and Cognitive Command 2

In recent work, Robert Hopkins has developed and defended Kants claim that aesthetic judgements are autonomous, in a sense that I will describe below. 214 He argues that a straightforward line of argument seems to show that a discourse that mainly comprises such autonomous judgements cannot display Cognitive Command. Since it also seems defensible to ascribe autonomy to ascriptions of metaphorical meaning, it might seem that Hopkins argument provides a more direct and effective route to settling their cognitive status than Wrights forced march approach. I will set out the suggested argument below, and go on to argue that while interesting, it does not provide an obviously new route to determining whether discourse about m-meaning is cognitive in the relevant sense. In introducing the notion of autonomy, Hopkins compares aesthetic and colour judgements. Imagine that you are looking at a surface that you judge is blue, on the basis that it looks phenomenally blue to you. Now imagine that some larger group of people, whom you have no reason to believe are in any worse a position to judge the colour of an object on the basis of the way it looks, and whom are apparently careful and sincere, dissent from your judgement. Lets say that they insist that the surface is clearly red. What is it rational for you to do? Intuitively, you ought to withdraw your
214

See Hopkins (2000) and, especially, Hopkins (2001)

161

judgement. Even in the case where the surface genuinely is blue, and your judgement had previously amounted to knowledge, it seems that the presence of so much apparently informed disagreement acts as an epistemic defeater. What might be underwriting such an epistemic defeat? Hopkins suggests, plausibly enough, that it may have something to do with the fact that colour judgements are cognitive in something like the following sense; Cognitive judgements are genuinely representational judgements. In particular, (i) it seems a priori that when two subjects make contradictory cognitive judgements, the warrant for the judgement of at least one can be rationally criticised and (ii) if two subjects make contradictory cognitive judgements, prescinding from vagueness, etc, exactly one of them must be correct. 215 Given such a characteristic feature of colour discourse, it seems that the following line of thought the fault-allocating argument - is open to a rational subject who finds herself in the above situation: 1) I and my opponents disagree about whether the surface is blue. 2) One of us is at fault 3) They outnumber me, in general I and they are equally competent in matters of this sort, and weve all tried to access the facts in the same way. 4) So it is likely that I am fault. 5) So it is likely that the surface is not blue, as my opponent says. The first and third premises simply reflect the situation as described, while the second reflects the apparently cognitive status of colour judgements. Hopkins thus holds that commitment to Cognitives application to colour, plus the supposition that the two parties are of equal competence but different size, suffices to secure the transition from

Cognitive:

215

Hopkins explicitly draws on Wrights work on Cognitive Command in his discussion here. I hope the

analogy is obvious.

162

(1)-(5). The availability of such a pattern of reasoning is what explains the fact that I ought, rationally, to withdraw my intuitive colour judgement in such circumstances. Hopkins contrasts that example with the case of beauty. In this case, he argues, it is intuitively not always rational to withdraw ones judgement in the light of disagreement. If you find, say, a piece of music very beautiful, you may be surprised and baffled to find that those who you had previously taken to reliably agree with your considered aesthetic judgements now disagree. You may return to the music, asking them for reasons to support their judgements. But if your disagreement persists through these procedures, it is far less clear than it is in the colour example that you are rationally compelled to withdraw your judgement, although perhaps you should place less confidence in it. Thus, Hopkins takes judgements of beauty to be autonomous in the following sense; Autonomy: When one party finds herself disagreeing with (several) others who share a view then; (a) for ordinary empirical matters this is sometimes reason enough for her to adopt their view, but (i) this happens much more rarely in the case of beauty, and (ii) the opposing view does not act as a defeater to my judgement to the same extent as it does in the empirical case. Instead (b) she should place less confidence in her view; and (c) she should, if possible, test the issue by re-examining the disputed item. Now, imagine that judgements of beauty were cognitive. In that case, it would seem, we ought to be able to run through exactly the same kind of fault-allocating argument as we could in the colour case. That argument would move from the stipulated fact of my disagreeing with a greater number, as apparently well-placed as I, together with observation that, by Cognitive, one of us must be at fault, to the conclusion that it is 163

likely that I am that the one in the wrong. But the availability of that transition seems to be in irrevocable tension with Autonomy. So it seems that we are forced to give up either the idea that aesthetic judgements of this kind are cognitive, or that they are autonomous. We seem to have some kind of test for genuine cognitive character. What about the case of metaphorical meaning? Consider the famous conclusion of Fitzgeralds The Great Gatsby
And as I sat there brooding on the old, unknown world, I thought of Gatsbys wonder when he first picked out the green light at the end of Daisys dock. He had come a long way to this blue lawn, and his dream must have seemed so close that he could hardly fail to grasp it. He did not know that it was already behind him, somewhere back in that vast obscurity beyond the city, where the dark fields of the republic rolled on under the night. Gatsby believed in the green light, the orgiastic future that year by year recedes before us. It eluded us then, but thats no matter tomorrow we will run faster, stretch out our arms furtherAnd one fine morning So we beat on, boats against the current, carried back ceaselessly into the past.

Imagine I interpret the final line along the following lines: our individual projects and concerns, that imbue our lives with meaning, involve a constant, pointless struggle to achieve an impossible objective that is both determined and frustrated by the contingencies of our past histories, personal and social. And imagine that friends or critics that I respect agree on interpreting the line rather differently: the mention of boats alludes back to the Dutch sailors mentioned a few paragraphs earlier, who first gaze on the fresh, green breast of the new world, that had pandered in whispers to the last and greatest of all human dreams. The line expresses, according to them, a fundamentally American theme; the inevitable corruption of the seemingly new world by European decadence, symbolized by capitalism, consumerism and elitism. I see the line as making a universal claim about what it is to be human, they see it as making a claim about what it is to live in a certain kind of American society.

164

Surprised by their disagreement, I re-draw their attention to the phrase the last and greatest of all human dreams, and to the recalled advice of Nick Carraways father, which opens the novel. 216 They respond by reminding me in turn of the role that the contrast between the cosmopolitan East and the moral Midwest plays in the novel, and point to the thematic significance of Oxford education and World War I, not to mention that stock Modernist image of corrupt European capitalism, the cigar smoking, thicktongued, bulbous-fingered, Jewish gangster Meyer Wolfsheim, of The Swastika Holding Company, fixer of the World Series. Imagine that things go on like this, without an obvious resolution of the issue. 217 Do the mere facts that I have no reason to think that they are insincere or limited in capacity, and that they are numerically superior, give me good grounds to think it is I who am in the wrong? It is surely at least defensible to hold that they do not, and that judgements of metaphorical meaning are autonomous. It is quite normal for us to feel that metaphors are precisely open to interpretation in a way that literal language is not. 218 It is in any case interesting to explore the question of whether establishment of such autonomy would provide the antirealist about metaphor with a telling argument against treating judgements of m-meaning as subject to (Cognitive Command). Given the worries about a stand-off I raised earlier, in the context of Wrights forcedmarch argument, it would be reassuring to be able to identify some such focus of discussion. Nevertheless, I intend to argue that it is not obvious that there is an inconsistency in holding that judgements are both autonomous and cognitive. The faultallocating argument may be valid. But it seems that there are enough apparently cognitive discourses where something like autonomy seems permissible, for us to be justifiably concerned about whether it genuinely can be relied upon.

216

Whenever you feel like criticizing anyone, he told me, just remember that all the people in this world Assume for convenience of example that neither of us is convinced by the other partys positive case.

havent had the advantages youve had.

217

Agreed multiple meanings of metaphors raise interesting issues, but they are distinct from my concerns here.
218

Again, any fully satisfactory Davidsonian semantics for fictional metaphorical meanings would have

to make a decision about how to account for these. One would be to supervaluate on acceptable interpretations. Another would be to use a non-standard logic, that allowed for inconsistent claims to be held true of the same utterance. Another would be to relativize metaphorically-means* to interpretations.

165

To see why, we can begin by noting that it isnt quite right to suggest that granting premises (1)-(3) of the fault-locating argument immediately secures the transition to (4). Imagine that aesthetic judgements are robustly true or false, but in a way that completely escapes our epistemic powers. Then the mere fact of their being cognitive in the above sense, together with my being at a numerical disadvantage, in no way warrants the move from (3) to (4). (If an unseen ball is either red or blue, then a hundred people guessing blue in no way makes it more likely that it is blue!). We can see that when a certain kind of epistemic draw is secured in this case, a nil-nil draw Autonomy is not threatened by the mere fact of the discourse being cognitive. Endorsing Autonomy is fine, for all Hopkins has shown, as long as the move from (3) to (4) can be resisted. 219 Lets say we attempt to remedy this by appealing to an earlier commitment to epistemic transparency: If a metaphor m-means that P, then (in somewhat idealized conditions) we can get some warrant for believing that it m-means that P. This principle is, of course, independently plausible, at least on standard construals of metaphorical meaning, so a defender of Hopkins line ought to be able to take it for granted. However, its not clear that adding this as a premise will always help. Consider an analogy. Scientific judgements are often warranted via standard scientific enquiry. Now, imagine that I am a scientist who believes that there is intelligent humanoid life on other planets, and that a significant number of my friends and colleagues, on the basis of similar evidence and methodology, disagree with me. Does such disagreement give me grounds to conclude that it is likely that I am at fault, and ought to adopt their view? Not at all. For a start, I may have a different subjective probability set than they. Since the shared scientific methodology includes the commitment to favour theories that are e.g. simple, elegant and coherent with prior probabilities, it doesnt seem obvious that I ought simply to adopt my opponents view,
219

(Epistemic Access)

Of course, in the case where the facts are unknowable, clauses (b) and (c) of Autonomy will be

independently implausible. But that hardly distracts from the point.

166

or even to become agnostic. In this case, it seems perfectly defensible to maintain my existing belief, despite the paradigmatically cognitive nature of scientific discourse. 220 Mere disagreement does not warrant me in revising my judgement. Perhaps then, we ought to bracket differences that merely result from such subjective, non-cognitive differences: (Epistemic Access*) If a metaphor m-means that P, then (in somewhat idealized conditions) we can get some warrant for believing that it m-means that P, and such warrant does not rely even in part on subjective, non-cognitive differences. But the issue is delicate. In the first place, it is not clear whether the capacity to undergo certain kinds of creative imaginings, of the type that seems important for thinking of one thing as another, is relevantly non-cognitive. Surely, however, no account of metaphorical meaning will want to risk cutting it off from some fundamental notion of thinking-as. In the second place, it isnt wholly clear that it is irrational for me to persist in my judgement in the scientific case even if e.g. our prior probabilities are similar. It seems at least conceivable that I may simply come to a different view to you on the basis of similar evidence, and yet be rationally warranted in maintaining it in the face of opposition, simply because the case is one where we have so little evidence either way. Such a position certainly seems true of e.g. philosophical practice. Timothy Williamson is an epistemicist about vagueness, while many of his professional colleagues are not. Neither need be ignorant of any relevant a priori consideration, and may well start off their investigations with similar degrees of intuitive attachment to certain principles. But its extremely uncomfortable to think that the mere fact of such disagreement acts as a defeater for Williamsons philosophical beliefs. And its equally uncomfortable to think that e.g. philosophy of language, or metaphysics is not cognitive

220

Of course, one might respond to the fact that subjective probabilities play a role in scientific

methodology by becoming an anti-realist about science. My concern with this move is that it tends to impose a global anti-realism, whereas the most interesting forms of anti-realism seem contrastive with genuine matters of fact.

167

in the above sense, or depends ultimately on non-cognitive preferences. 221 So it seems that, even in the presence of (Epistemic Access*), the transition from (1)-3) to (4) is more complicated than Hopkins seems to suggest. Consider one final example. I am playing poker, and have been lucky enough to improve on the draw to the highest hand, a royal flush. Various parties are observing the game, including each players hand, and making side-bets on the eventual outcome. To my surprise, they dont bet heavily on me, but instead start backing my opponents. I have no reason to believe that their perceptual equipment, etc, is any better or worse than mine, nor that they are engaged in any great conspiracy or subtle strategy. I check again with my co-players what the rules are, and have my prior beliefs apparently confirmed. Is it rational for me to fold? After all, my belief that I have the best hand relies ultimately on causal sensitivities employed in perception, memory and elementary computation, just as in the colour case. But somehow I dont feel that I would fold in such a situation, nor that there is any great rational pressure on me to do so. The case would seem strange to me, but not one in which it would not seem obvious to me that I am at fault. So the matter seems autonomous, but also, surely cognitive. The issue is clearly a difficult one, but I suggest that the valid instances of the faultallocating arguments offered by Hopkins are really enthymematic, implicitly relying on like the following premise; 1) I and my opponents disagree about whether the surface is blue. 2) One of us is at fault 3) They outnumber me, in general I and they are equally competent in matters of this sort, and weve all tried to access the facts in the same way. 4) The best explanation of the disagreement in this scenario is that I am at fault. 5) So it is likely that I am fault. 6) So it is likely that the surface is not blue, as my opponent says.

221

Apart from anything else, such a view threatens the kind of globalizing argument canvassed in

Section 1.10.

168

Our new premise is made probable by (1)-(3) in e.g. the colour case, and provides a suitable abductive link to (5). It fails in e.g. the case where no evidence at all is forthcoming, and the scientific case, just as we would intuitively predict. (Similarly, the fact that philosophy is very difficult, and poker-type cases very rare and weird, seem to provide competing explanations of divergence, blocking the move to (4) in those cases). But its inclusion somewhat diminishes the interest of Hopkins argument for the question of whether discourse about m-meaning is cognitive. For the original promise of the argument related to its ability to circumvent a frustrating realist/anti-realist stand-off regarding best explanation. It is now far less clear whether a genuinely new route for profitable debate has really been identified.

4.6

Metaphor and Intention

So far this chapter, I have (i) provisionally concluded that even the realist ought to grant that talk about metaphorically meanings is epistemically constrained, and (ii) expressed some methodological worries about the application of Cognitive Command, or some improved descendent thereof, for settling questions of the cognitive status of such judgements. I now want to turn to the last of Wrights tests for realism that I will discuss here, the Order-of-Determination test. 222 I will approach this issue somewhat indirectly in this section, via an examination of the relationship that intention stands to metaphorical meaning. I will then conclude my discussion of Order-of-Determination in Section 4.7. Intention plainly has something to do with metaphor. We can choose to make a metaphor, and intend our utterance to be taken as such. We have a certain authority with respect to our utterances; we are entitled in many circumstances, to treat lack of respect for a prior intention as a cognitive failing. You've misunderstood me, we might say, I wasn't speaking literally, or I didn't intend that aspect of the metaphor to be emphasised. Admittedly, this need not be, and in all probability isn't always the case. When speaking
222

As I noted in Chapter 1, Wrights other test, Width of Cosmological Role seems unlikely to be passed

by metaphorical contents. However, I will suggest in the next chapter that the contents of some metaphors may play a role which isnt always mediated by the attitudes of concept users, thus defending a wider cosmological role for them than might have been expected.

169

to a psychoanalyst, or writing a novel, for example, one might implicitly or explicitly renounce a certain degree of authority over the interpretation of one's metaphors, along with some of one's literal utterances. Moreover, there might be many other types of cases where speakers happily allow their metaphorical utterance to be extended in a way that is new or surprising to them. We should not let consideration of these cases blur an important distinction, however. To take a somewhat analogous case, there is a big difference between being happy with a way a witty remark is picked up and elaborated upon, and claiming the embellishments as one's own. We need an account of the role of intention that allows us to make just this kind of distinction in the case of metaphorical utterance. This is important both for the type of Davidsonian position that I have outlined and defended in earlier chapters, and for the most plausible realist positions. Everybody ought to grant that metaphors effects, whether or not they are taken to involve the expression of propositional contents, are highly context dependent. But our best models of context represent them as highly sensitive to conversational purposes and goals, local and general. 223 Since these in turn are likely to depend on the aims and intentions of conversational participants, a satisfactory account of the metaphor and intention is desirable for all concerned. For the Davidsonian, such a theory will be important in (a) the cases where the speaker-intention model is appropriate and (b) developing the kind of pseudo-fictionalist semantics seemingly necessary for avoiding the Frege-Geach problem. For realists who hold the default, speaker-meaning view, the need for a plausible account of intention and metaphor will be inescapable, while for their strongest rivals, contextualists, the need to tell a story about what a context is should provide suitable motivation. On the other hand, there might seem to be a problem in reconciling this consideration with the peculiarly open-ended quality of metaphor. Many commentators, for example, have felt dissatisfied with a straightforward "speaker's meaning" account of metaphor, which in its crudest form holds that:

223

I say more about the nature of context in Chapter 5.

170

(I) S's metaphorical utterance M means that P iff S intends M to convey that P. 224 Such an account is clearly so over-simplistic as to allow straightforward disqualification as a serious theory of metaphorical meaning; the spiralling complexities of speech act theory have taught us that. Yet we might be sceptical of the prospects for any such account, however hedged and qualified, and not only for this crude precursor. Surely, we might want to say, this whole approach to metaphor is in danger of just misrepresenting the phenomenology of metaphorical utterance. We don't typically have a distinct content in mind, as it were, that we then express via the use of metaphor. Indeed, if we did, it would seem to render the motivation behind metaphorical utterance a little mysterious. If you intend to say that P, then why not just say it? We seem to be close to a view of metaphor that characterises it merely as amusing embellishment, or useful shorthand. And such a view has often been felt to be philosophically and phenomenologically unsatisfying. Of course, one response to this worry would be to deny that we have any privileged access to the content of our own intentions, to hold that we can have at best inductive evidence, resulting from a process of self-interpretation, for what we intend and mean, even in the most everyday cases. But this move, in turn, makes a mystery of the very authority we grant to speakers, which originally counted in favour of the speaker-intention model. We need a theory of metaphor that allows us to be faithful to the phenomenology of metaphor making, while also delivering a satisfying account of how we interact with metaphor makers. It may seem, however, that no speaker-meaning theory can deliver both desiderata. David Cooper has objected to the speaker's meaning view of metaphor along related, if not strictly analogous, lines. 225 His objection contends that such a view cannot account for the indeterminacy of metaphorical content. A metaphor is indeterminate in Cooper's sense iff it admits of more than one interpretation, none of which can be demonstrated as uniquely correct. Cooper holds that any successful account of metaphor must give us a story about metaphorical indeterminacy.

224 225

See Searle (1993) for one of the first accounts of this kind. Cooper (1986) pp. 71-77. Cooper characterises the speakers intention model of metaphor as the

standard view.

171

It is clear that Cooper's characterisation is unsatisfactory as it stands - it lets in cases where all interpretations are clearly incorrect, for example. 226 Rather than try to offer an improved version, however, I intend to take it merely as gesturing at some important aspects of our normal thought about metaphor: namely, that metaphor is apt for competing, independently satisfying interpretations of which there is seemingly no a priori guarantee, in the general case, that we will have reason to adopt one rather than another. The relation between this thought and our sense that attribution of meaning to metaphor is apparently autonomous, and that what metaphor leads us to is in some sense limitless, are, I hope, fairly obvious. 227 Cooper considers three possible ways in which the speaker-intention model could try to account for such indeterminacy. Firstly, the indeterminacy might be caused by our ignorance of what exactly it is that the speaker intended. Secondly, the speakers intention might be somehow open-ended or indeterminate. Cooper characterises this option, following Searle, as implying that when S utters some metaphor of the form A is B, he intends to mean or implicate a range of meanings; A is C1 and/or A is C2, C3 etc. Finally, we might take the indeterminacy to be a feature related to the fact that different possible speakers could use the same sentence to convey different contents. Cooper rejects all three of these proposed accounts. Cooper has a number of objections to the first proposal:

226

There are lots of other worries that could be raised against the definition. If Quine and Davidson are

right about the indeterminacy of translation, then all metaphors will count as trivially indeterminate, for example.
227

I do not intend to suggest that relations of simple entailment are involved here. Clearly, a metaphor

could be indeterminate in Coopers sense, and yet admit of a competing series of quite determinate and limited paraphrases. I take Cooper to be guided in part by the intuition that metaphor is limitless both horizontally and vertically as it were; that any one interpretation is open-ended, and that rival interpretations can be equally defensible. His definition of indeterminacy makes it seem as if it is the latter that is at issue, but the second of his suggested replies on behalf of the speaker-meaning theorist suggests that he is just as concerned with the former. The relevant notion of open-endedness is multidimensional and plural. I say some more to help disambiguate it slightly in Section 4.7, and in the next chapter.

172

(M1) A metaphorical content M is indeterminate iff we are ignorant of what an actual speaker S intended to convey by his corresponding metaphorical utterance P. He first considers the case were we have no knowledge of S at all, and concludes that this case collapses into a special case of the third proposal (M3, see below); with the indeterminacy being caused by the speculation about what different possible speakers might have meant. He then argues that even if we do know quite a lot about the identity and context of S, a limited amount of ignorance about his environment can still leave us completely in the dark about S's intention. He cites solely literary cases
In learning about writers such as Rimbaud, Mallarm, Marinetti and Ezra Pound, one soon learns that speculation as to what they intended to communicate by individual metaphors is pointless - in the dual sense of being a waste of time and beside the point. But this does not mean that it is pointless to try and interpret the metaphors, nor that any old interpretation will do. 228

Having argued that the interpretation of metaphor can be relatively determinate even when we know little about the speaker's intentions, Cooper goes on to argue the converse; that we can be sure of what the speaker's intentions were without determining the metaphorical content.
The speaker, poet or painter does not have exclusive rights to interpretation - and even if he did his interpretation would not have to mimic his intention at the time of composition. 229

So, he concludes, ignorance of an actual speaker's intention is neither necessary nor sufficient for a metaphor to have an indeterminate content. If the point here is solely that we can come up with satisfying interpretations of apparently metaphorical utterances that mention little about the actual speaker, then the point is well taken. But that is surely not what should be at issue here. We should be careful not to confuse the idea that we can interpret what a person knowingly said in
228 229

Cooper, (1986) p. 72 Cooper (1986) p. 73

173

making an utterance with the idea that we can interpret the sentence he uttered in a way that satisfies us in some respect. The whole idea of something being an unintentional double entendre, for example, relies on there being a gap between what a speaker actually said and the possibility of construing it in a satisfying way. We should similarly admit the existence of non-intentional metaphors in the sense of sentences or utterances that we can intelligibly or usefully treat as if they were metaphors which we can generate and interpret for particular purposes. Perhaps Burroughs' technique of cutting up newspapers and randomly assorting often suggestive sentences would be a clear example of how we can intentionally bring about such non-intentional metaphors. Nobody should claim that only intentional utterances can be interpreted as if they were metaphors, just as nobody should suggest that only intentional jokes or joke shaped utterances - are funny. But this should not lead us to play down the role of intention in the practice of joke telling, nor in metaphor. In addition, we must be careful not to assume that we can give a unified account of how we should best interpret metaphors, no matter in what context they may arise. The cases Cooper cites, involving Modernist literary writers, may well demand a different treatment than the everyday case, but this is not to say that we haven't identified an important feature of the everyday case when we link metaphor to an intended use of a sentence. Let us assume, in the interests of charity, that Cooper's target is only the 'nave' intention theorist, who thinks that M1 says all that needs to be said about metaphoric indeterminacy. His arguments against nave intentionalism then seem cogent; metaphoric determinacy and actual speaker intention can come apart. What about the second idea then, namely; (M2) A metaphorical content M is indeterminate iff the speaker's intention is open ended. Cooper admits that a speaker can, on occasion, intend for a metaphor to be taken in 'several ways at the same time'. 230 Perhaps when I say "John is a real giant among men",
230

Cooper (1986) p. 74

174

I can intend that his mother will think that I am praising him, while knowing that you will really know that I am drawing attention to his weight problem. But, Cooper argues, this is not happily characterised as intending to mean C1 or C2 by P, but rather C1 and C2. Moreover, it cannot suffice for a metaphor to be indeterminate that the relevant intended meaning be vague, since that would suggest not incommensurate interpretations but a single one that 'matches the speaker's intention in vagueness'. 231 Nor can S intend merely to try out a striking sounding sentence, for that would render the relevant speaker's meaning non-existent, not open-ended. Cooper claims that there is only one kind of case that
might happily be described as one of 'speaker's open ended meaning'. This is where the proposition meant by the speaker is of the open-ended form 'P or Q or '. [But] To say that a speaker might have meant P, or might have meant Q, is not equivalent to saying that he meant a disjunctive proposition P or Q or 232

Cooper rightly points out that whereas we might be happy to say of a notoriously ironic friend that we often don't know whether he is saying P or saying not-P, we never take him to be uttering the tautology P or not-P. All these arguments, I would suggest, are fine as far as they go. But Cooper is wrong, I think, to take the best construal of M2 as involving vagueness or disjunction. In the latter half of this essay, I want to outline what I take to be a better picture of how intention and indeterminacy are related. We can deal briefly with (M3) A metaphorical content M is indeterminate iff different possible speakers could mean different things by the corresponding utterance P. This is untenable, and Cooper rightly demolishes it. Firstly, there are just too many possible speakers. We have to narrow them down to the 'most reasonable ones', and
231 232

Cooper (1986) p.75 Cooper (1986) p.75

175

Cooper argues, we can do so precisely because we have a prior idea of what counts as a reasonable interpretation of the metaphor. Our grasp of the metaphoric content determines the relevant possible speakers, and not vice versa. I am slightly sceptical about the general effectiveness of this response to more plausible modifications of (M3), but do not intend to take issue with it here. Instead, let's go back to the problem that we started out with. There is a tension between two ways we might want to think about metaphor. On the one hand, we are pulled in the direction of saying that the speakers intention must in some sense constrain acceptable interpretation of metaphor. After all, how else are we to explain the common sense idea that the speaker has often selected a particular metaphor, that she has reasons for using the expression that she does, etc? On the other hand, there is definite substance to the intuition that lies behind Cooper's objection. The idea that in metaphorical utterance we have a definite and pre-existing content to convey, that the problem of selecting a metaphor is one of how best to dress up such a content in borrowed clothes, seems to completely misrepresent the phenomenology of what we do. Typically metaphors spring to mind with a rather vague feeling of aptness. We can often struggle to express or even elucidate in literal language what we meant by a metaphor. Yet typically we can recognise interpretations of a metaphor as being in or out of accord with the way we meant them to be understood. We ought not to envisage ourselves as throwing metaphors out into the world, semi-randomly as it were, to fare as they happen to be taken up and elaborated on, whether by ourselves or others. But similarly, it seems difficult to see how, after the metaphor has been correctly and fully interpreted, taken in the way we meant it, we could have had a prior intention to mean or implicate all of that. Spelling out the intended effects of a metaphor, even in the case where one has produced it oneself, can be a long and tortuous process. Plausible interpretations that are out of accord with ones intent need to be set aside; good suggestions about how to convey in literal terms what one was getting at need to be enthusiastically adopted, even when the idea of putting things that way would never have sprung to ones mind in a million years. The problem or at least, one important problem - is to see how any half-way motivated theory of speaker intention can account for first-person authority over metaphor, while not surrendering the datum that one rarely has some clearly delimited proposition in mind at the time when the metaphor is 176

made. And if my earlier discussion is correct, that seems to be a problem for realists and error theorists alike.

4.7

Metaphor and Order-of-Determination

I think we should see this tension as an instance of a more general issue about intentional states, discussed by Crispin Wright in a number of papers 233 and usefully summarised by Jim Edwards:
The problem is to reconcile the first person epistemology of such intentional states, the fact that we normally take a persons sincere avowals of his own intentional states to be authoritative, with the fact that an intentional state may also have to answer to future behaviour, behaviour which the subject need not have had in mind when he made the avowal. 234

In the case of metaphor, the intentional state in question is what the speaker intended to mean by the metaphor. The future behaviour is the responses and judgements made by the speaker concerning which interpretations and extensions of the metaphor are in accord with his original intention. Wrights attempted resolution of the problem involves taking the speakers considered belief about what he intended to mean to constitute the facts about what he meant, rather than seeing them as tracking an independent fact of the matter. What I intend by a metaphor can be open-ended precisely because for any given interpretation, extension or development of the metaphor, there is not an independently determined, pre-existing fact about whether it accords with my intention. Before we go on to consider the particular case of metaphor, lets remind ourselves Wrights general account is structured, since we have covered quite a lot of ground since my discussion in Chapter 1. He is interested in different ways of interpreting what he calls Provisional Equations, which take the following general form

233 234

See, for example, Wright (1987), (1989a), (1989b), (1989c) and (1992). Edwards (1992), p. 21.

177

(PE)

For a set of optimal conditions C, a state of affairs P, and a subject S: If C holds, then (it would be the case that P iff S would judge that P).

There are two contrasting ways, Wright thinks, that we could understand the case where a given provisional equation holds for a particular C, P and S. We could understand the C-conditions as being such as to allow S to successfully track an independently obtaining fact that P. That is, we could understand the biconditional as indicating that S judges that P because P is the case, and S is infallibly good at tracking that fact in those conditions. Call this the extension-reflecting sense. On the other hand, we could understand the biconditional as indicating that it is Ss best opinion that constitutes the fact that P. In this case, we understand it as telling us that for P to be the case consists in nothing more than that subjects like S judge that P in Cconditions. In this sense, P-facts would be metaphysically dependent on certain tendencies of human response. Call this the extension-determining sense. 235 We are to determine which way the Provisional Equation should be read by checking whether (1) the C-conditions have been specified substantially, not by means of a whatever it takes ceteris paribus clause (2) the equation holds true a priori. (3) whether the C-conditions are satisfied must be logically independent of facts about P and (4) that there is no better explanation of why the first three conditions are satisfied than the claim that Ss best judgements constitute the fact that P. 236 If the result of all four of these checks is positive, then the relevant instance of PE is to be read in the extension determining sense. If not, it is to be taken as extension-reflecting. How might such an account look with respect to first person self-ascriptions of intention? Consider the Provisional Equation for Intention (PEI) for any subject S and content P

235

As I noted in Chapter 1, since the PE only tells us about what is happening in optimal conditions,

Wright normally describes our optimal judgement of whether P as at best partially determining the facts about P. I continue to ignore this distinction here.
236

Wrights motivation for these constraints is explored in more detail in Section 1.7

178

(PEI) If conditions C hold then (S believes that S intends that P iff S intends that P). 237 Wright argues that in suitable conditions a subject's judgements about his own intentions constitute the fact that he has such and such intention. It is not at best a contingent matter whether or not we have access to our own intentions, he thinks. Rather, it is precisely the fact that our best opinion determines whether or not we have a particular intention that explains why we are effortlessly, non-inferentially and generally reliable about [our own] psychological states. 238 Of course, it is possible for us to be self-deceived about our own states of mind. Moreover, there seems to be no straightforward way of ruling out such self-deception in formulating the C-conditions in PEI without running into trouble with the substantiality condition. Nevertheless, Wright plausibly argues, the 'grammar' of intention is such that we are a priori entitled to presume that we are not deceived, unless we have actual evidence to the contrary. That is, although we cannot include a 'no self-deception' clause in the optimality conditions and still fulfil the substantiality condition, we can still be a priori justified in holding that any given instance of PEI is true. 239 Such justification is a priori but defeasible, since evidence that S was in fact selfdeceived would remove it. Wrights claim, therefore, is that the fact that PEI is a priori justified is, in the absence of a better explanation, enough to show that our best opinions about our own intentions are extension-determining rather than extension reflecting. 240 Since it is our best judgements about what we intend or intended that constitute the facts about our intentions, we can reconcile the idea that we have a definite authority with respect to our own intentions with the fact that there need be no propositional content
237

Statistically standard health, statistically standard external conditions, possession of relevant concepts, Wright (1989b) p. 289. I have drawn here on the useful outline of Wrights position given in Edwards That is, we are not a priori justified in holding that the universal closures are true, but we are in holding For a defence of the move from a priori truth to a priori justification see Divers and Miller (1994)

judging after a period of careful reflection, etc

238

(1992)
239

that any given instances are.

240

179

that we need have in mind, in the sense of being able to spontaneously or even reflectively identify. at the time of making a metaphor. 241 We now have the resources to make the case that in making a metaphor, the speaker's intention concerning how it should be understood can be essentially 'open-ended'. The suggested reply is that intentions concerning the communication of a metaphorical content are merely a special case of intending in general. In the correct conditions, my judgements about my own intentions are, as Wright has plausibly argued, constitutively linked to the facts about what those intentions are. Its simply difficult to make sense of a genuine intention that could float free of our sincere, undeceived self-ascriptions. 242 So there is apparently no need for the intention theorist to have to make the case that in intending to authoritatively convey a propositional content P by uttering a metaphor M, a speaker S must somehow have had the resources to specify or delimit P available to him when intending to utter M metaphorically. It may be a perfectly acceptable picture to conceive our access to the facts about what S intended by a particular metaphor as primarily relating to when, under optimal conditions, he would judge that a particular interpretation is in accord with what he intended. Cooper originally defined a metaphor as being indeterminate just in case it admitted of more than one interpretation, none of which can be demonstrated as uniquely correct. We have seen that we must distinguish the idea that there can be satisfying and productive interpretations of 'non-intentional metaphors' from the notion that a speaker's utterance has a metaphorical content that admits of more than one interpretation. So Cooper requires a narrower definition of what indeterminacy consists in. What I want to argue is that the intuition that metaphors are in some sense indeterminate because a
241

I have concentrated on the case of intention, but it seems highly likely that a similar account could be

devised for our first person authority over the nature of our imaginings. Experiential imagination is a puzzling case, however. In the case of intention, we generally do not check back, phenomenologically, with the content of our thought. In the case of imagination, we often have the sense that we do. Imagine a metal fork lying in clear light on a white tablecloth. Do your best to visualize it. Now, are the prongs pointing up or down? Theres a definite sense of checking back against a prior image, in the way that we might redirect perceptual attention to one aspect of a perceived scene. But it is clear our epistemic relationship to the merely intensional world of the imagination is very different from the perceptual world, that we causally engage with.
242

See Peacocke (1999) for a contrasting account of this phenomenon, however.

180

speaker simply couldn't have had that complicated and open-ended and yet determinate a content in his mind's eye, as it were, may be simply misplaced. The speaker typically does not have a independently determined content in mental view, which he then conveys with a greater or lesser degree of success. Nor does such a content live, wholly determinate yet inaccessible, in some intention box in the lower reaches of the speakers sub-personal psychology. 243 Nor is he merely in equal hermeneutic standing with his interlocutor, forced to interpret his own metaphors in just the same fashion as he interprets those of others, and as others interpret him. Rather, the speaker can inhabit an epistemically optimal position from which to arbitrate between different interpretations, making constitutive judgements as to which was in accord with his earlier intention. But this doesn't entail that indeterminacy in Cooper's sense is completely removed, that one can uniquely demonstrate that one interpretation is correct. Such a demonstration is not possible since the interpretation at hand is always defeasible, conditional not only on the possession of a continuing warrant to hold that the relevant epistemic conditions are ideal, but also on the speaker's future judgements and avowals relating to alternative or modified interpretations. Of course, all this still leaves plenty of room for a speaker to be mistaken about what he intended. If optimal conditions do not hold, or there is evidence that he is self-deceived, then PEI is silent about whether or not S's judgements constitute the facts about what he means. We might favour a subjunctive account, appealing to what the speaker would have meant. Alternatively, in many cases where the optimal conditions for PEI do not hold, the optimal conditions for radical interpretation still might, and we might take the facts to be constituted by the judgements of a actual or hypothetical radical interpreter. In other cases, including perhaps the case of the Modernist novelist and the psychoanalytic patient, I might be taken to have renounced my first person authoritative standpoint with respect to my intentions, even when I actually inhabit best conditions.244

243

As e.g. Sperber and Wilson seem to suggest at times. See their (1986a) passim, and (1986b) pp. 547We might characterise what is happening here in different ways. Is it a refusal to make the judgement

549
244

about my intentions, even if ideal conditions hold? Or a second order intention that my first order intention not be taken as authoritative? (I might of course still require that my second order intention be taken as such.)

181

But there will be a wide area of cases in which it is precisely my best judgements that are authoritative. 245 We have the beginnings of a potential solution, then, to the problem we started off with. In the everyday case, there is a firm link between speaker intention and metaphor. But we needn't imagine that this entails having the whole metaphorical content antecedently 'in mind' in any philosophically problematic way. This was the one aspect of a problematic picture that Cooper, in suggesting that metaphorical content is indeterminate, was no doubt reacting against. Instead, it is open to us to replace this picture with another which allows the nature of our knowledge of intention itself to be characteristically 'open'. Wright's theory has just this feature, since the existence of facts about what I intended are conditional on my best judgements, many of which, at any given time, I shall not have considered or made. Much more than this has to be said, of course, in order to differentiate metaphorical limitlessness from the general indeterminacy highlighted by Wright's account of intention. After all, that account applies as much to applying the rule for + as to reflecting on what Eliot meant by Midnight shakes the memory/As a madman shakes a dead geranium. There is room for tinkering with this proposal. Perhaps in some cases we are happier to admit that, at the time of utterance, we didn't quite know exactly what we meant to convey, while in the literal case we resist this much more strongly. Moreover, there will no doubt be many cases where the particular propositional content intended to be conveyed by the speaker is just not to the point: we like the metaphor because of its non-propositional effects on us, say the way the words sound together. Nevertheless, it seems clear that important aspects of the intuitive open-endedness or limitlessness of metaphor have been omitted. In the next two chapters, I turn some other ways in which they might be accommodated.

245

Or if not, the case has to be made. I take it that e.g. postmodernist critics aim to show that the

optimality conditions are never fulfilled - the lures of patriarchy, ideology or the unconscious mean I am always self-deceived. I leave open the question of whether this is a coherent thought.

182

4.8

Conclusion

I began this chapter by examining an argument of Wrights which, if good, would have served to undercut much of the prima facie motivation for the type of error theory canvassed in Chapters 2 and 3. I concluded that the continuity of ascriptions of metaphorical meaning with semantic ascriptions generally was likely to provide a convincing answer for the Davidsonian, and undercut Denhams attempt to provide a minimalist account of metaphorical meaning. I then broadened my focus slightly, endeavouring to examine the objective standing of metaphorical discourse in the light of Wrights test, in a way that did not prejudge more detailed issues relating to the good standing of particular realist theories. I concluded that most contemporary parties to the debate were likely to conclude that metaphorical meaning was not evidence transcendent, and had an at best limited Cosmological Role. I discussed two strategies for settling questions of the cognitive status of such judgements, due to Wright and Hopkins, but concluded that neither of them provided a promising general strategy for settling disputes about realism, contrary to first appearances. I then noted that all parties, including the Davidsonian, were likely to require some suitable account of the relationship between metaphorical meaning and speaker-intention. Having discussed and evaluated Coopers presentation of the issues, I suggested that a first-person, response-dependent account of speaker-intention might provide a suitable and underexplored model of that relationship. Such an account would clearly impose an antirealist response to the Order-of-Determination test. Nevertheless, I concluded, such an account could only hope to provide at best a partial explication of the key notion of limitlessness, appealed to by Davidson in his rejection of the speaker-meaning account. I will examine two further treatments of that notion in what follows.

183

5. Metaphor and Non-Conceptual Content

5.1

Introduction

Any satisfactory theory of metaphor ought to have an answer to the question: why do we use metaphors? This question is especially pressing for the realist about metaphorical meaning. Given that you want to communicate some proposition, why not just say it outright? There are of course straightforward answers; to be entertaining, to enliven, to be elegant and concise, to enhance rhetorical effect. These answers, however, despite their classical ancestry, have often seemed to miss something important out. Other purposes for metaphorical speech have been suggested, such as the cultivation of intimacy between speaker and interpreter, and there is no doubt much more that could be usefully said about this type of issue. 246 Moreover, the idea that intending to express some proposition means having it there in mind, ready to be suitably adorned in metaphorical garb, is clearly mythology, which the broadly Wittgensteinian account of intention offered in the last chapter can help us begin to see past. Nevertheless, these solutions, and dissolutions of the puzzle are apt to seem unsatisfying. Something seems still to be missing the sense that, when one employs a metaphor, one is often doing the best one can to get something across. 247 The error theorist about meaning has some kind of answer to our question: one uses metaphor to try to get us to think of one thing as another. 248 This is something that

246

See Cohen (1978). Cohens suggestion is discussed, elaborated and endorsed in Cooper (1986) pp. Yablo (in his (1998)) says that the cardinal rule of metaphor is: make the most of it. But often this is

153-178.
247

clearly not what we want at all. Interpretive fireworks are not at all to the point in many cases; were trying to convey something, just those words are the best we can do at formulating it, and we want our interlocutor to catch on, not take off on flights of associative creativity.
248

Thus Lamarque and Olsen (1994) p. 360. The constitutive aim, definitive of metaphorical utterance

per se, is simply this: to invite or encourage a hearer to think of, conceive of, reflect on, or imagine one

184

could be done in other ways; by use of simile, mimicry or juxtaposition, for example. But what is it to think of or imagine one thing as another? Why invite someone to do such a thing, and what good does it do? These are no doubt bad questions, although it is not always easy to think of which better ones should replace them. In any case, the sheer generality of the answer involves a loss as well as a gain. A real question seems in danger of being drowned out in the hubbub of a general theory of experiential and propositional imagining. So let us ask it again: in what sense, in employing metaphor, are we doing the best we can? In this chapter, I intend to argue that figurative language plays an important and distinctive role in helping us talk about aspects of the world that would otherwise be resistant to linguistic expression. Moreover, that role neatly coheres with the distinctive Davidsonian marks of metaphor outlined in section 2.9. This ought to be good news for the realist about metaphorical meaning, since it helps provide the resources for a principled rejection of (A6). However, if I was right to suggest previously that everybody, including the Davidsonian error-theorist, ought to sign up to at least the possibility of pragmatic communication via metaphor, such an account ought to be of interest to both sides in the realism debate. In particular, I will claim that we use metaphors to capture aspects of representational states which have a content that the attributing subject lacks the concepts to specify directly. When we lack the concepts to specify the character of certain states in straightforwardly literal terms, metaphor helps us employ the concepts we have to do so. In that sense, what metaphor draws our attention to is often non-conceptual and nonpropositional in character. Moreover, such specifications are limitlessness in a number of relevant senses. We naturally feel that such characterizations are often irreducible to literal paraphrase, that the process of interpretation is self-sustaining and constantly developing, and that the idea of finishing would have no clear application. 249 In each case, the non-conceptual character of the content in question helps us to provide an account of why this should be so.

thing (state of affairs, idea, etc) in terms associated with some other thing (state of affairs, etc) often of a quite different logical type.
249

Davidson (1984a) p. 263

185

I begin by outlining the model with attention to the case of the emotions. I argue that there are elements of our representations of our own emotional states which are nonconceptual in character, and suggest that metaphor and other figurative language often best captures the precise phenomenology of such states. I then extend the model to the case of religious metaphor, demonstrating in detail how it can be used to answer a challenge raised by William Alston to the coherence of irreducible metaphors in theology.

5.2

Metaphor and Emotional Content

It is an important and interesting feature of our normal experience of our emotions that we find it difficult to describe in literal terms the exact qualitative and experiential character of what we are feeling. In communicating with each other about our emotions, we often use metaphors and other figurative expressions to try and capture the finegrained phenomenology of our emotional states. Proper appreciation of such metaphors can often be an essential precursor to empathetic and imaginative identification with another persons emotional states, and to the development of explicit knowledge of the nature of our own emotions. A distinctive and valued aspect of literary skill concerns the ability to provide depth of characterisation making the character live - precisely by imaginatively recreating recognisable emotional states via such figurative use of language. Take, for example, Nick Carraways description of his distorted, emotionally charged dreams of West Egg:
I see it as a night scene by El Greco: a hundred houses, at once conventional and grotesque, crouching under a sullen, overhanging sky, and a lustreless moon. In the foreground four solemn men in dress suits are walking along the sidewalk with a stretcher on which lies a drunken woman in a white evening dress. Her hand, which dangles over the side, sparkles cold with jewels. Gravely the men turn in at a house the wrong house. But no-one knows the womans name, and no-one cares.

186

Or Jamess description of the coldly avaricious nature of Osmonds emotional attachment to Isabel Archer
Her mind was to be his attached to his own like a small garden plot to a deer park. He would rake the soil gently and water the flowers; he would weed the beds and gather an occasional nose-gay. It would be a pretty piece of property for a proprietor already farreaching. 250

Such metaphors evoke a quite precise sense of the specific emotional charge, intensity and hue of the emotional states inhabited by the characters, a sense that we would find difficult to capture in literal terms. I have chosen fictional examples for ease of reference, but I want to maintain that figurative language plays an exactly analogous role in our everyday expressions and descriptions of emotion; we capture best how we feel at times by sticking resolutely to the figurative. Its easy to recognize this, even in ones own case; we grasp for pictures and analogies to convey how we feel, or how others seemed. The problem is not to recognize the phenomenon, but to explain it. I have noted several times that well-founded worries about semantic treatments of metaphor have prompted many theorists to move to a pragmatic theory of metaphor, where metaphor is accounted for at the level of speaker-meaning rather than sentence meaning. On this type of account, the producer of a metaphor utters a sentence that literally means that q in order to convey or implicate some other proposition r. Metaphor is thus seen as structurally analogous to irony or sarcasm, where my utterance of e.g. that was really clever can serve, in the right circumstances, to implicate my belief in the negation of the proposition I literally expressed, or made as if to express. However, I have also noted that this type of account seems to run into difficulties in accounting for the marks of metaphor. One central problem relates to paraphrasability. In the typical case of sarcasm, I could have said literally what I instead implicated. It is this kind of ability that, in part, makes it appropriate to identify the implicated proposition as what I was really trying to get across. In the case of metaphorical utterance, on the other hand, it often feels simply impossible to say in literal terms what

250

Cited in Denham (2000) p. 327

187

a speaker got across by using a metaphor. In simple cases, with practically dead metaphors or near idioms, of course, this isnt difficult at all; I can utter he isnt the sharpest knife in the drawer, and convey my belief that he is stupid. With even moderately complex metaphors, however, paraphrases are often implausible and hard to identify. Since the type of pragmatic theory described seems to rely on the producer of the metaphor having some proposition in mind, as it were, which she intends to communicate obliquely by uttering some other sentence in the appropriate context, this constitutes a real problem for proponents of the theory. A natural temptation is to retreat to a generalized error-theory about all but the simplest and stalest metaphors, that admit of quite straightforward elucidation. Again, we seem to tick-tock between two conflicting inclinations; on the one hand the inclination to reject the wooden paraphrases, and the pragmatic theory that seems to demand them, and on the other, the sense that we clearly do often reach for quite complex metaphor when we most sincerely want to get something across about what its like to feel a certain way. I want to resolve the felt tension between these compelling lures of thought. Perhaps in purely literary cases, what a metaphor-maker means by his words is less important than what meanings can be made from them. In the everyday case, however, the use of metaphor is rarely a symptom of the fact that the speaker was really playing around with words; instigating a game where each participants aim is to come up with striking and original similarities and analogies based on the metaphor. Rather, using metaphor often involves doing the best we can to get something across that we cant convey in any other way. The emotional case is a useful illustration precisely because it is so clear that we may well be trying to communicate something tremendously significant by using metaphor. We neednt be trying merely to politely invite certain thoughts or attitudes in our interlocutor, nor to let her come up with her own inventive readings of what we said, since it seems difficult to see why should we care about them. Rather, were trying to communicate the way things are for us, something that (a) genuinely possesses content, and thus can stand in the right kind of normative relations to our attempt at putting it into metaphor, and yet that (b) we dont seem to be able to get across in other, more straightforward ways. A natural idea is to hold that the reason that these two conditions obtain is that were dealing with non-conceptual content.

188

5.3

Metaphor, Emotion and Non-conceptual Content

Lets take a simple case where I turn my attention inward onto my own emotions; lets say I realise that I am feeling more resentful and competitive towards a colleague than I had first realised. Now, it seems prima facie non-controversial that I am representing a particular aspect of my current psychological state. Of course, there have been philosophers who have denied this (thus conforming to the well confirmed universal generalisation that for all prima facie non-controversial claims C there exists a philosopher who has maintained not-C.) In particular, error-theorists about the emotions hold that most folk terms for emotion almost certainly lack an objective reference, so that there is nothing for such terms to represent, while avowalists have held that self ascriptions of mental states should not be thought of in representational terms at all. 251 Nevertheless, there are well known problems with these positions, especially relating to the motivation for adopting them. In particular, the latter seems committed to providing a plausible theory of representation which entails that our self-ascription of mental states come out as non-representational. It seems, however, that the opposite is true. A recent, and convincing set of necessary criteria for a given state to count as representational has the consequence - unsurprising no doubt - that my second order thoughts about the emotional states I am in are genuinely representational. They count a state as representational only if 1. It serves to explain behaviour in situations where the connections between sensory input and behavioural output cannot be plotted in a law-like manner (Parsimony) 2. It admits of cognitive integration; it connects up in the right way with other psychological states, both representational and motivational (Integration). 3. It is compositionally structured in such a way that their elements can be constituents of other representational states (Compositionality) 4. It admits the possibility of misrepresentation (Error).

251

Eliminativists about emotion may include Churchland (1981) and Griffiths (1997). Wittgenstein is

often read as the paradigmatic avowalist.

189

It is clear that these criteria (proposed by Jose Bermudez) are met in the case of my thought about my emotional states.252 There is no relevant causal story to be told about the way that e.g. my realisation that I am more resentful than I had thought affects my behaviour, and how it does so will be in part a function of psychological interaction between that state and other of my beliefs and desires. I can think other self-involving thoughts, and thoughts about other people's emotional states, so compositionality is respected. And I can be wrong about the character and quality of my emotional states; mistaking justified resentment for envy, or a short-lived romantic crush for genuine feelings of love, for example. The interesting thing about these criteria, however, as Bermudez points out, is that while
they are all satisfied paradigmatically by the conceptual contents that are the objects of folk psychological propositional attitudesit does not follow that conceptual propositional content is the only genuine form of content. 253

In particular, there is theoretical space here for the ascription of non-conceptual content to our experiences of emotional states, where a non-conceptual content of an experience is defined as follows (NC) An experiential state E of a subject S has a non-conceptual content iff it meets the criteria for being a representational state outlined above, and S could be in the state E even if S lacked the concepts requisite to specify the content of E. Notice that this definition does not rule out the possibility that a subject might indeed possess the concepts necessary to specify the content of the experiential state that they are in. Indeed, in examples like the one above, where I came to realise that I was angry, I do have certain of the relevant concepts. The central point is, of course, that I needn't have the concepts to be in the state; there is no intrinsic or essential connection between my being able to represent in experience e.g. my emotional state as being thus and so, and my possessing the concepts to do so.

252 253

See Bermudez (1998) Ch 3 and 4 for extensive discussion of these criteria. Bermudez (1998) p. 94

190

What might give us theoretical reason to decide that my experience of my emotional states possesses non-conceptual content? There are, I believe, several motivating factors, familiar from discussions of non-conceptual perceptual experience 1. My experience of my own emotions is more fine-grained than my conceptual repertoire. I can internally discriminate states that fall under the same conceptual descriptions. For example, I can distinguish the character of the longing boredom that I experience waiting for a bus, from that I undergo while awaiting a tiresome acquaintance's punch-line. Similarly, I can distinguish the feelings of liking or attachment I have towards a range of new workmates, even though I may not be able to verbalise or even express in thought precisely what the relevant difference comes to. 254 2. Uncontroversially limited concept users such as young children or higher mammals can nevertheless seem to manifest a relatively sophisticated emotional repertoire (boredom, self-satisfaction). Such attributions aren't obviously a result of misguided anthropomorphism. 255 3. It seems prima facie possible that I might recreate exactly in memory states of emotion that I underwent before acquiring the relevant concepts, such as a moment of resentment or euphoria undergone as a young child. Since such states would be by hypothesis phenomenologically identical before and after acquisition of the relevant concept, and would seem to meet the constraints on being genuinely representational outlined above, the most natural explanation is that the states possess non-conceptual content on both occasions, now and the remembered past. 256 If the above comprises a defensible set of theses, then I believe that we have the resources to explain the fact that we often seem forced to resort to metaphorical constructions, if we are to do justice to our experience of emotions. Moreover, we can simultaneously account, at least in part, for the non-paraphrasability of an important
254

See Peacocke (1993) Ch 3 and Peacocke (1994) for further elaboration of this line of thought applied in There is a sharply increasing body of research on the contrasts and inter-relations between human and See Martin (1992) for an analogous argument in the case of perception.

the case of perception.

255

animal emotions. For an illustrative discussion, see McNaughton (1989) Ch 11.

256

191

class of metaphors, without resorting to the abandonment of any notion of metaphorical meaning. The story we should tell goes roughly as follows. In experiencing our emotional states, we are presented with states which, while representational, are nonconceptual in character. Unsurprisingly, therefore, given the fine-grainedness noted above, we often lack the necessary concepts to communicate such states to others directly. We therefore use metaphor to tell each other about the states that we are in, using the concepts that we do possess to get across the way things are with us that we don't have concepts for. Nevertheless, since non-conceptual states are genuinely representational, they provide the kind of normative friction which is necessary to account for our practice of not always treating metaphorical utterances as just invitations to contemplate salient similarities, or useful devices for the promotion of lateral thinking, but rather sometimes as genuine attempts to get something, currently unspecifiable, but often proposition-like, across. 257 If thats right, then a realist has resources with which to take some of the sting out of Davidsons (A6): An account of m-meaning which identifies it with the extra propositional content communicated by the author of the metaphor, over and above what it says, as determined by a linguistic theory of pragmatics, provides an unacceptably incomplete and misleading account of our practices of producing, using and reflecting on metaphors.
257

(A6)

The content communicated by such a metaphor may correspond to what Peacocke (1992) p. 77 calls a

proto-proposition, composed of individuals and properties. See Bermudez (1998) pp. 96-100 for a useful introduction to this notion. Having such Russellian propositions in play opens up the possibility that metaphorical content may have a wider Cosmological Role than it is natural to expect. In a sense, the sensitivity of non-concept users to such representational states already serves to establish this. It is clear, however, that while non-concept users may be sensitive to the content expressible by metaphors, they are not thereby credited with such sensitivity via metaphor. The sketchiness of Wrights Wide Cosmological Role constraint makes it difficult to settle finally on a judgement as to the correct view to take on this matter. See Stern (2000) p. 188 for a related distinction between referential and purely conceptualized propositions. My conception of metaphor as apt for the expression of non-conceptual content was developed independently of Stern, and differs in detail but the two approaches share a common theme. It is of course a very old and intuitive idea to hold that metaphor helps us say what we could previously grasp, but not put into words.

192

That principle, remember, was supported by the following line of thought: a) Anything worth regarding as a tolerably complete and informative account of our practices of producing, using and reflecting on metaphors must account for all of the distinctive marks of metaphor. b) An account of metaphor that, like Searles, characterizes such practices as being in essence a special case of saying one thing and meaning another, cannot account for all of the distinctive marks of metaphor. In particular, it cannot give an account of the limitlessness and non-propositional character of what metaphor draws to our attention. c) Therefore, such pragmatic accounts of metaphor provide at best an incomplete and misleading account of our practices of producing, using and reflecting on metaphors. If the above sketch of an account is defensible, then the pragmatic theorist has established some room for manoeuvre. A form of limitlessness is established, since no literal paraphrase is likely to seem satisfying. Our existing vocabulary will be far too impoverished and clumsy to capture the complex fine-grainedness of our emotional lives, as tracked from the inside. Moreover, there is scope for the realist to argue that the relevant form of non-propositionality is also captured. After all, trying to convey to somebody how one is feeling, is unnaturally thought of as trying to communicate that such and such is the case. Davidsons motivating examples of non-propositionality perceptual and pictorial content share many relevant features with our experiences of our own emotional states. Pictures, perceptions and emotions are analogue in form, admitting of shading and degree. They are multi-dimensional, and experientially integrated in a hugely complex manner. They admit of gestalt shifts, as when one is forced to see ones black mood as suddenly self-indulgent and ridiculous. They are resistant to judgement: a tromp loeil picture can still seem strikingly real once the illusion has been unmasked, and anger can outlast the revelation of its unmotivated character. Such observations, however, do not seem to undercut the idea that pictures or

193

perceptions can be genuinely representational. In a broad sense of proposition, they possess propositional content. 258

5.4

Further Thoughts on the Limitlessness of Metaphor

The pragmatic realist can do even better in accounting for our intuitions of the limitlessness of metaphor. I have already distinguished two forms of this intuitive notion; one relating to the dependence of m-meaning on future judgement, and the latter on the unavailability of literal, conceptualised paraphrase. It is clear, however, that there are other strands to the original conception that might be profitably teased apart. One involves a kind of open-endedness, the capacity for continual elaboration and development. It is this notion which is often taken to motivate anti-realisms about metaphor. However, given the gestalt-type nature of our representation of our emotional state, it is clear that the realist can begin to account for why we might feel that metaphors are open-ended in this sense also. Lets go back to the case of Nick Carraways dream:
I see it as a night scene by El Greco: a hundred houses, at once conventional and grotesque, crouching under a sullen, overhanging sky, and a lustreless moon. In the foreground four solemn men in dress suits are walking along the sidewalk with a stretcher on which lies a drunken woman in a white evening dress. Her hand, which dangles over the side, sparkles cold with jewels. Gravely the men turn in at a house the wrong house. But no-one knows the womans name, and no-one cares.

It is clear that Nick does not intend us to take much of this literally; he does not literally see the dream scenario as a night scene by El Greco, and the sky is not literally presented as sullen. Nevertheless, the metaphors provide an effective, almost indispensable means of capturing and communicating the fine-grained representational content of the dream,
258

Moran (1997) p. 257-258 makes a similar move, noting that we might e.g. model propositions as sets of

possible worlds. As befits my attempt to remain fairly neutral about controversial semantic issues, I will not critically examine this possibility here. My purpose is merely to note that a workable model of propositional content may be open to the realist.

194

including the emotional charge and timbre that pervades it. So much was argued for above. We neednt, and shouldnt, rest content with a purely reflective account of emotional metaphor, however, metaphor as mirror to the soul. In employing the metaphors in reflecting upon his emotional states, Nick brings those states into a light that illuminates but also colours and contours. In employing the El Greco metaphor, Nick at once expresses and casts new light upon his emotional state. Aspects of the dream that originally provoked the figure may seem afterwards to be less important than other elements of it that the metaphor fits; the strange static quality of the movement within the dream-narrative, for example, or perhaps a quasi-religious sense of sin, abandonment and foreboding. The metaphor is not used up in its original expression, but establishes a new gestalt under which the significance of the dream can be re-examined and assimilated. Similarly, consider the following passage, from the fourth of Abelards personal letters to Heloise:
And so I ask you, sister, to accept patiently what mercifully befell us. This is a fathers rod, not a persecutors sword. The father strikes to correct, and to forestall the enemy who strikes to kill. By a wound he perceives death, he does not deal it; he thrusts in the steel to cut out disease. He wounds the body, and heals the soul; he makes to live what he should have destroyed, cuts out impurity to leave what is pure.

Abelard is almost certainly consciously punning on fatherly punishment as symbol of the justice of God, and as infamous component of his own Historica calamitatum. Nevertheless, we can easily imagine that it is not so; that the original metaphor, drawing on stock Christian imagery in the service of theodicy, comes to organize Abelards sense of his situation in a way that suddenly reveals to him the tortuous character of his relationship with his own sexual nature. Simply stating literally that God only inflicts evil upon us for our own good is clearly far less likely to provide such a moment of insight and reorientation. Metaphor, motivated by one set of insights, can come to reconfigure ones sense of things in a way that provides smoothly and fittingly for others. In this sense, the earlier picture of emotional metaphors can also seemingly lay

195

claim to a satisfying notion of open-endedness. Of course, it need not be non-conceptual content that provokes such an episodic reflective pattern. Nevertheless, it seems clear that the way in which metaphor serves to focus and elaborate ones sense of such emotional episodes, the very states that one lacks conceptual resources to tackle reflectively, entails that metaphor has a special role to play here.

5.5

Metaphor and Religious Language 1

I now want to turn to my second example of the role that non-conceptual content can play in the philosophy of metaphor, namely, religious metaphor. William Alston has presented an argument which, if sound, would prove the impossibility of irreducibly metaphorical talk - where such talk is irreducible iff it says something which cannot be said, even in part, in literal terms - about God. 259 In this section I examine two answers to Alston that might be given by a theist. The first develops a late suggestion of Alston's, which seems to undermine his main argument. I conclude that this reply, though logically consistent and independently motivated, is weaker than a second reply, which argues that Alston's argument has little force against a theist who subscribes to the idea described above, that such metaphors can express non-conceptual contents. Alston argues against the possibility of irreducibly metaphorical talk about God, where a metaphor is irreducible in the relevant sense if what it says cannot be said, even in part, in literal terms. It is clear that the denial of this type of strong irreducibility (SI) is perfectly consistent with maintaining that metaphor plays an important or even essential role in our theological talk and thought, and that such weak irreducibility might ultimately turn out to be the more puzzling and interesting variety. Moreover, there may well be room for doubt as to whether there are any good grounds for endorsing such strong irreducibility in the first place; Alston notes his independent disagreement with claims about God's transcendence and ineffability, and with the non-verifiability of religious language, each of which might potentially be employed in arguments which motivate an appeal to the strong form of the position. Nevertheless, since I take it that both the position and Alston's argument against it is are of some interest in their own
259

Alston (1989)

196

right, I propose to examine the merits of Alston's case against the possibility of strong irreducibility, and to offer two replies on behalf of the opposing position. Although it is not strictly implied by the position as stated, I will take it that the defender of SI is a theist, and moreover, I will assume that there is no special burden of proof placed on her by Alston simply in virtue of her theism. It is as well to be clear from the beginning how Alston thinks of metaphorical meaning. He holds to a pragmatic realist theory, where, as ought to be familiar by now, speakers use sentences with a given literal meaning to convey or implicate some propositional content distinct from that expressed by the sentence taken literally; this metaphorical content is equivalent to what the speaker is saying via his use of the metaphor. It is of course possible for speakers to use metaphor for purposes other than assertion - to invite a hearer to reflect on (or play with) a model or exemplar of a given situation, in order that she might identify striking similarities and analogies for herself, for example - but in many cases, the speaker will have some fairly definite set of resemblances 'in mind' which he wishes to convey. Thus Alston writes
[In] the typical metaphorical statement is "building on" the relevant meaning of his predicate term in two ways. First, he is presenting the kind of thing to which the term literally applies as a model of the subject. Secondly, he has in mind one or more resemblances between model and subject, and he extracts from these resemblances what he means to be attributing to the subject. 260

For the purposes of this section, I will allow Alston that this is a convincing theory of metaphor, although there are clearly many issues of substance and detail that might reasonably be challenged. How, then, given such a theory, does his argument for the impossibility of SI proceed? Let us concentrate on the example Alston gives: God is my rock. Alston's argument can then be represented as follows 1) Strong irreducibility is true. 2) Alston's model of metaphor is correct.

(Assumption, for reductio) (Assumption)

260

Ibid p. 23

197

3) God is my rock is an example of the type of irreducible metaphor SI is concerned with, and can be used to say something about God. 4) God is my rock is irreducible 5) God is my rock expresses a proposition expressible, even in part Call this property P. speaker must have cognitive access to P property P. (Assumption) 9) If it is possible for the speaker to form a concept of the property P, then other members of the language community can form a concept of P. (Assumption) 10) If other members of the language community can form a concept of P, then P can become the literal meaning of a predicate term in the language. (Assumption) 11) If P can become the literal meaning of a predicate term in the language, then God is my rock can be expressed in wholly literal terms. 13) Strong irreducibility is false. (Assumption) 12) God is my rock can be expressed in wholly literal terms, contra 4. (7, 8, 9, 10, 11) (By reductio, 1, 4, 12) (Assumption)

(by 1 and 3) (by 2 and 3).

6) God is my rock must attribute some property to God which isn't literally (by 4 and 5). (by 2 and 6) 261 7) Since the speaker is attributing the property P to God by means of the metaphor, the 8) If the speaker has cognitive access to P, then the speaker has a concept of the

As noted above, I intend to grant Alston the correctness of his account of metaphor and the appropriateness and typicality of the example he offers (assumptions 2 and 3). I will also accept that the argument as presented is valid, and that assumptions 10 and 11 are true. It is thus common ground between myself and Alston that the argument against SI is sound if 8 and 9 are true. For convenience, I will consider the question of their truth by examining two slightly more general principles, which I will assume provide the sole support for 8 and 9. If a subject has cognitive access to a property, then the subject has a concept of that property.

Conceptual Access

261

This step in the argument may seem less than obvious. Remember, however, that on Alston's account

of metaphor, a speaker must 'have in mind' the respects in which the model resembles the subject of the metaphor.

198

Mutual Access

If it is possible for the speaker of a natural language to form a concept of the property P, then other members of the language community can form a concept of P.

I suggest that the defender of SI can best resist Alston's argument by rejecting either one or both of the above principles. The challenge, of course, is for the SI-theorist to provide principled grounds for doing so, and so avoid the charge of a mere ad hoc refusal to accept Alston's reductio. I will begin with outlining a possible strategy for rejecting mutual access. The prima facie grounds for rejecting the Mutual Access principle appear bleak. After all, given plausible assumptions about the shared cognitive and conceptual capacities of natural language users - endorsed by empirical evidence as well as common sense - it seems to follow that, in the vast majority of cases, a species of conceptual mastery attainable by an arbitrary member of a natural language community can be duplicated by his peers. Of course, this need not always be the case, at least in practice. The externalist tradition in epistemology and semantics has stressed the widespread phenomenon of epistemic and linguistic division of labour. 262 Given finite capacities and ever more specific fields of human intellectual enquiry, it may well be practically impossible to acquire conceptual mastery of many theoretical concepts. Moreover, it is a familiar truth that experts can develop a higher degree of conceptual sensitivity than others possess with regard to the properties of wine, humans or song; and perhaps we can imagine certain of them making use of innate sensitivities simply not available to the rest of us. Your inborn capacity for colour discrimination, or gift for mathematical physics may, as a matter of fact, allow you to conceptualise elements of your experience which are literally unreachable by those of us whose sensitivities are naturally less apt for honing. Nevertheless, this can hardly be the kind of case that the SI-theorist would wish to rely on. Ineffable experience of and thought about God seems unlikely to be best represented as the unspeakable privilege of a genetically blessed few.

262

See e.g. Putnam (1975); Burge (1979).

199

A better tactic is for the defender of irreducibility to look for a principled way to differentiate between certain speakers and others, in a way that lets them effectively undermine mutuality. Unusually, Alston identifies precisely such a route in a footnote towards the end of his paper. In a footnote to his discussion of I.M. Crombie's position 263 Alston writes
Although I am reading Crombie as a panmetaphoricist, he is susceptible to an interestingly different reading according to which theological metaphors can be literally paraphrased, though not by mere mortals. That is, we might think of Christ, who, according to Crombie, guarantees that the models he provides for us are suitable models, as being able to spell out the crucial similarities in literal terms. This, then, would be an extension of the familiar situation in which a poet uses a metaphor with some definite intention in mind that he could express literally, but where none of his readers could do so, though some of them are "grasped" by the metaphor. 264

Alston is correct to note that this is an "interestingly different" way of interpreting or developing Crombie's position, but seems not to fully recognise the direct relevance of this type of account for his argument against strong irreducibility. It seems to me to speak directly against the mutual access principle, in cases where the linguistic community includes members who are more than "mere mortals". Let's avoid textual issues as to the correctness of this reading of Crombie, and simply take the defender of this type of account to be committed to the following claim: In cases where a linguistic community has members who are more than mere mortals, it is possible for certain speakers to form concepts of certain properties which other speakers cannot. Asymmetrical Access entails that the mutual access principle doesn't hold in cases where the linguistic community includes those who are more than mere mortals, and thus, if good, would block Alston's argument against SI. Moreover, it has, I believe, a
263 264

Asymmetrical Access

As outlined in Crombie (1955) and (1957) Alston (1989) 36 n. 23.

200

certain degree of independent plausibility, and thus provides a means for the SI-theorist to avoid the charge of ad hoc rejection of Alston's premises. After all, as noted above, we are familiar with the idea that possession of certain cognitive capacities can be essential for concept mastery, at least for basic possession of such concepts. Normally sighted people have psychologically and epistemically basic mastery of colour concepts, for example, since they have the ability to grasp colours under the mode of presentation afforded by visual perception. Other members of the linguistic community (for example, congenitally blind, or colour-blind members), if they count as possessing the concepts at all, do so in virtue of the existence of those who exercise basic mastery. Perhaps a person who couldn't visually distinguish between red and green might know that redness was the property of looking red to normally sighted people under standard conditions in the actual world, and might even have some other, reliable way of telling red things from green. (We can imagine that certain light wave detectors might be combined in such a way that she could discriminate objects on the basis of their colour). Nevertheless, if she counts as possessing colour concepts at all, it is partly in virtue of the existence of those who exercise basic mastery of such concepts, since it is the visually basic judgements of these members which determine e.g. what counts as being co-extensive with, or a reliable detector of, red things. 265 Moreover, we can imagine that trying to get across to such people what it was like to make such judgements on the basis of visual appearance alone would involve appeal to metaphor, analogy and image; scarlet being like the sound of trumpet, for example. Nor is this phenomenon restricted to sensational or perceptual qualities. We might offer an autistic person metaphors, models or exemplars to help them track the emotional and psychological states of others, for example. I claimed above that the existence of this type of asymmetry of biological capacity didn't by itself provide a very promising model for directly undermining Mutual Access in the theological case. However, it does serve to provide an interesting set of structural analogies which can be drawn on in defending Asymmetrical Access. Since we can reasonably expect that more than mortal members of natural language communities have distinctive cognitive capacities which mere mortals lack, we can expect that they can form concepts of certain properties - have ways of thinking about certain properties which are simply not open to their cognitively limited peers. It is open to the Christian
265

I assume here, for purposes of illustration, a response-dependent account of colour.

201

defender of SI to hold that only Christ can form basic concepts of the properties expressed by irreducible metaphors, and that any mastery of them that we might come to possess is bound to be partial, derivative and reliant on his use of them in practice and judgement. A natural policy, then, as we have seen above, is for Christ to make appeal to the concepts and capacities that we do have basic mastery of in order to elucidate those which we do not. Such an employment of Asymmetrical Access in the theological case has the advantage of blocking a potential reply that Alston might make against the claim that considerations concerning experts and basicality suffice for rejecting Mutual Access. Such a reply would hold that the modal claim which the latter claim makes - that other members of the language community can form a concept of P - has to be heard as 'can in principle form a concept of P'. It isn't clear that such a reply works even against those considerations; Kripkean style necessities regarding the essentiality of species and origin might mean that it was metaphysically impossible for certain members of linguistic community to acquire certain capacities possessed by others, in cases where the linguistic community involved several distinct kinds of creatures, for example. But in any case, it seems that the theistic defender of SI has good grounds for saying that even given such a strengthened reading of Mutual Access, there might be good grounds for rejecting it when the linguistic community involves both cognitively finite and non-finite members. The final position of the SI-theorist who rejects the mutual access principle by endorsing Asymmetrical Access, then, is independently motivated to the extent that it appeals to other cases which make plausible the claim that concept possession should be asymmetrically and externally individuated. Christ, unlike other members of his linguistic community, can form basic concepts of certain properties of God; the best way for him to convey the nature of these properties to creatures who lack such basic capacities may involve metaphor, myth and symbol, given to us in the form of parable. Nevertheless, we can count as having knowledge of God's nature to the extent that we stand in certain communal relationships to Christ. Just as some epistemological externalists argue that we can count as possessing theoretical knowledge, or mastery of the meaning of the terms of our language, in virtue of standing in certain deferential relationships to scientific experts and to the linguistic community as a whole, so we can 202

count as having knowledge of God and the meaning of metaphors that describe him to the extent that we intend to be accommodating our practice to Christ's. It is perhaps worth noting that this defence of SI is not dependent on Alston's rather surprising and counter-intuitive claim that this account "would be an extension of the familiar situation in which a poet uses a metaphor with some definite intention in mind that he could express literally, but where none of his readers could do so, though some of them are "grasped" by the metaphor". I find it difficult to see why Alston takes such a case to be familiar, given that the phenomenology of poetic composition seems to count fairly drastically against it, and also why, on his account, the readers fail to be able to identify the content of the relevant intention. If it is because the poem concerns some ineffable subjective experience, then it is difficult to see how it can comprise such a content, given Alston's views on the necessity and mutuality of concept possession. A natural alternative interpretation is that the readers are merely contingently ignorant; that had they been around to ask the poet the correct questions, he could have identified for them the proposition that he had decided to dress up in borrowed clothes. Such an account seems to me to distort the nature of artistic creation and interpretation, but more seriously, to drain much of the interest from his suggestion of how to read Crombie. For if it is only the fact that Christ chose to speak indirectly, or wasn't asked the right questions, which differentiates our knowledge of God from his, then, apart from the radical contingency of this account of God's ineffability, there seems to be no reason why we might not be able to stumble across the relevant concepts for ourselves, so that the moderate strengthening of Mutual Access mentioned above would suffice to rebut Crombie's suggestion.

5.6

Metaphor and Religious Language 2

Such a strategy seems to me to be philosophically defensible and logically consistent; it may even prove promising for some purposes. Nevertheless, I think there is some cause for concern about whether it alone can do the requisite explanatory work. In particular, the defensive strategy as outlined so far might be taken to rely overly much on Christ's knowledge of God as providing the epistemic link between us and God. Christian SItheorists might want to hold that, for example, metaphors employed by the pre-Christian 203

prophets and psalmists can express or serve as a model for thought about God, while non-Christian theists might not accept that natural linguistic communities have included members with qualitatively different cognitive powers and capacities. Alternatively, concern may be felt about whether a purely externalist model of our knowledge about God can do justice to the reflective component of much of our theological talk and practice. For these reasons, and for the independent interest of the question, it's worth examining whether an independent rejection of Conceptual Access (the claim, remember, that if a subject has cognitive access to a property, then the subject has a concept of that property) might be motivated. The best line here, I believe, is for the defender of SI to appeal to other cases where philosophers have wanted to appeal to non-conceptual contents of our propositional attitudes. For, again, drawing analogies with independently motivated philosophical position allows the SI-theorist to deflect the force of Alston's reductio, while avoiding the charge of mere ad hoc rejection of another premise. It seems too quick, though, just to saddle Alston with the general claim that cognitive access entails conceptual access - so that just any reason to posit non-conceptual content would serve to rebut his argument. Alston might justifiably claim that strictly he needn't defend the general claim, (although this is the claim that he would most naturally be read as making) but only the weaker thesis that, if a subject has access to a theological property, then she has a concept of that property. From now on, then, I will argue against this weaker claim. The defender of irreducibility will be on strongest ground if he can make the case that the contents of thought about God, according to him, cohere with a general theory of representation, and yet are available to creatures who lack any of the relevant kinds of concept. This latter condition seems to block certain of the arguments presented in the literature in support of the possibility of non-conceptual contents of thought - for example, the claim that our perceptual experience is more finely-grained than our

204

conceptual capacities. 266 Such arguments generally presuppose a certain level of conceptual mastery, (visual concepts, for example, along with grasp of an objective spatial material world) but argue that the subtle variations of, say, shade, tone and shadow that our visual experience presents us with outrun our genuinely conceptual discriminatory capacities. In the case of thought about God, however, certain types of SI theorist may well want to hold that we can't form any concept of God's nature, (although of course this isn't the only conceivable view, and certainly isn't entailed by the basic position). For those who are motivated by the thought of the complete alterity and transcendence of God, though, the kind of conceptual capacities presupposed by the fine-grainedness argument simply won't be present, so that the needed analogy won't genuinely hold. A better approach for the SI-theorist, I believe, is to rely on Bermudez model, that allows for wholly non-conceptual contents. It has been a while since I introduced his account, so let me briefly summarize it here. Bermudez outlines a plausible set of necessary conditions which a psychological state must meet if we are to count it as genuinely representational. He then argues that this general account of what it is for thought to have a content, to cognitively map aspects of the objective world, does not require that the subject of the thought should possess the concepts necessary to specify that content. Rather, the conceptual / non-conceptual distinction comprises a further division of representational states into those whose contents presuppose conceptual mastery on behalf of the relevant subject, and those which might be undergone by creatures who lacked those concepts completely. Bermudez four requirements on psychological states being genuinely representational are as follows 1) They should serve to explain behaviour in situations where the connections between sensory input and behavioural output cannot be plotted in a law-like manner.

266

Again, Peacocke (1992) and particularly his (1994), 419-429, provide a useful overview of such

arguments.

205

2) They should admit of cognitive integration, being able to connect up with other representational and motivational states. 3) They should be compositionally structured in such a way that their elements can be constituents of other representational states. 4) They should permit the possibility of misrepresentation. Given such a conception of a representational state, then, Bermudez offers us the following definition of when its content is non-conceptual (NC) The content of a psychological state T of a subject S is non-conceptual if the state meets the criteria for being genuinely representational outlined above, and S could be in the state T even if S lacked the concepts requisite to specify the content of T. Does the content of our thoughts about God meet Bermudez' four criteria for being genuinely representational? I believe that any theist is going to be committed to defending the claim that it does. Our thought about God, is going to be relevant to what we believe and how we act; and the prospects of explaining such action as prayer, reflection and self-sacrifice in stimulus-response terms seems wholly impossible, as the first condition demands. The connection with belief and action is going to ensure that such thought must be integrated with other cognitive states; and the possibility of speaking and thinking falsely about God, or recognising that his qualities are not shared by human beings, is going to ensure that any such thought must admit of the recombination of its parts, and the possibility error and misrepresentation. So each of Bermudez' criteria will be met. Given that a theist will typically be committed to the claim that thought about God is representational in Bermudez' sense, how might she establish that we could be in such states even if we lacked the relevant concepts? Well, she might hold that God's otherness, transcendence and infinity - however these notions are fleshed out - give us prima facie reason to believe that we couldn't conceptualise his properties, so that the burden of proof is rather on the proponent of the (modified) Conceptual Access claim. How does holding such a position point towards a positive account of the usefulness of strongly irreducible metaphor? Well, defensively, it blocks Alston's argument, so that 206

the SI-theorist is, for the moment at least, on stable ground. More positively, she might argue as follows. Attribution of non-conceptual content is standardly intended to help us explain apparently cognitive and purposive behaviour in creatures that arguably haven't mastered concepts - animals, newborn babies, etc. Attribution of such content 'piggy backs', as it were, on the conceptual capacities of the attributing theorist: it is because we have the concept of squirrels, and elm trees, and spatial locations, that we can specify the dog's mistaken thought as one that the squirrel is in the elm tree. In this way, we can get the explanatory benefits of appeal to intentional states, while avoiding having to attribute an implausibly sophisticated conceptual range to beasts and babes, or to play down the normative and holistic character of concept acquisition and mastery. It is open to the defender of SI to hold that a similar relationship holds between ourselves and God as holds between the dog and the theorist. We entertain genuinely representational, non-conceptual thoughts about God, which we express via metaphors. Thoughts with which contents? Those specified by God - since only something as powerful and cognitively unlimited as God could properly conceptualise his properties. We, not the dog, can specify the content of his thoughts about the elm tree; God, not us, can specify the content of our thought about God. The propositions that our metaphors express are thus cognitively within our grasp, and yet non-reducible, even in part, to literal language. Moreover, this second strategy, if good, seems to avoid the disadvantage noted with the first approach, since it is consistent with the claim that subjects from many different traditions and historical periods can think about God, and express that thought, no doubt with widely varying efficacy, in metaphor, model and parable. This type of reply, as with the first account, draws strength from analogies with independently motivated positions. States which plausibly involve non-conceptual content in Bermudez' sense - including perceptual, proprioceptive, emotional and aesthetic states - are ones where we find metaphors centrally important in describing and understanding our thoughts. In these cases, metaphors, analogies and models may indeed play a central role in the process of conceptualization by which we eventually come to be able to specify the states of thought that we are in. The idea that this need always be the case needs arguing, though, and there doesn't seem to me to be any reason why the SI-theorist should feel obliged to hold that it happens in the theological case.

207

There are no doubt many difficulties which might be raised with both approaches, evident on even this brief outline. I want to briefly consider two points which might be raised with reference to the rejection of Conceptual Access, however. The first is that, since God can specify the literal content of our metaphors, they are paraphrasable after all, and Strong Irreducibility is false. The defender of SI should concede that, in this sense, irreducible metaphors can be paraphrased, but hold that they were only ever committed to the weaker modal claim that they could not be paraphrased, even in part, and even in principle, by creatures such as us. Since God knows everything that can be known, he knows what our metaphors say, if they say anything, but this rebuttal of SI seems rather too quick to address the most interesting versions of the claim of irreducibility, and is definitely not in the spirit of Alston's argument as presented. The second objection which might be raised is that I have appealed to many nonmetaphorical qualities of God in setting up the above argument - for example, that he would assign particular contents to the irreducible metaphors we use to talk about him. Doesn't such appeal to our knowledge of God's nature undercut many of the motivations which might be given for endorsing SI in the first place? In particular, can such an account be squared with any theory which holds that God's nature is wholly ineffable? I would want to make two replies to such an objection. The first is the local point that, as presented in this paper, Alston's argument is presented as establishing the impossibility of strongly irreducible theological metaphors. Even if my argument doesn't show that such metaphors can play the role that the supporter of ineffability required of them, it would still show that such metaphors were possible in theological contexts, for all Alston's argument shows, given either certain beliefs about externalism and basicality, or, more interestingly for my purposes here, about non-conceptual content. Secondly, the ineffability theorist might make a certain reflexive move in response to the above challenge, and claim that the philosophical claims she makes about God in the course of defending SI are, ultimately speaking as it were, as metaphorical as talk about shepherds, kings and rock. Of course, since she is committed to such metaphors expressing non-conceptual contents, which can only be specified by God, she will hold that the philosophical remarks about God and metaphor she makes themselves are 208

metaphors whose content she can entertain, although only God can specify. But it isn't immediately obvious why this move should be taken to be viciously circular, or selfdefeating, at least in a context where the burden of proof isn't assumed always to fall on the theist. If she occupies states which genuinely represent aspects of God, and which are cognitively integrated with her other psychological states, then why shouldn't she employ these in reasoning about God? Such a result, even if consistent, may be considerably less than many defenders of ineffability might want to endorse; perhaps the idea that we can occupy any states which can genuinely represent the nature of God, conceptual or otherwise, simply fails to meet a certain type of sense of God's alterity. Nevertheless, such an account would, if stable, at least provide one elucidation of God's ineffability. That alone, it seems to me, provides it with sufficient interest to warrant more general consideration. I have argued that the defender of strong irreducibility has room to counter Alston's attempted reductio, and sketched two possible lines of thought which might allow her to do so. Both have costs in terms of philosophical commitments, but I have argued that for the most part, they are not particularly steep, and have struck philosophers as possessing some independent plausibility. I conclude that, pending further investigation at least, and for all Alston has shown us, theists are free to appeal to irreducible metaphors in theology.

5.7

Metaphor and Vagueness

I want to briefly discuss one final disambiguation of limitlessness; one highlighted by Davidsons evocative remark that any attempted paraphrase of metaphor is bound to fail because there is the idea of finishing would have no clear application. It is clear that Davidsons remark may be taken in a number of different ways, drawing attention to a number of salient features of the limitlessness of metaphor, including some discussed earlier. I want to draw attention to one way in which the realist might aim to explain why the idea of finishing had no clear application, namely, by claiming that __metaphorically means that__ is a vague term.

209

Take a standard example of metaphor: Romeos claim that Juliet is the sun, say. It seems clear that there are some things that this definitely doesnt mean; that Juliet is gassy, for example. Moreover, there may be some features that no satisfactory elucidation of the metaphor could omit; that Juliet is important to Romeo, for example. But between these to categories, it is clear that there is a large grey area, where we are unsure whether a given property is really part of the metaphorical content expressed. Thus Stern describes the metaphor as conveying that Juliet is exemplary and peerless, worthy of worship and adoration, one without whose nourishing attention another cannot live, one who awakens those in her presence from their slumbering, who brings light to darkness. 267 It is simply not clear to me that the metaphor genuinely mandates the claim that Juliet e.g. awakens those in her presence from their slumbering. So it seems that, in general terms, we can often identify three categories of metaphorical assignment; those properties that definitely seem part of the metaphorical meaning, those that definitely do not, and those for which it is neither definite that they do nor definite that they do not. Such a division clearly invites a treatment in terms of vagueness. Such an account would have a natural account of why there was no clear stopping place in paraphrasing metaphor we simply stray onto a wide, multi-dimensional borderline, where it is often far from clear whether a given property genuinely counts as part of the meaning of the metaphor. Since I am dealing with a broadly realist account of metaphorical meaning, I will examine how the most straightforwardly realist account of vague terms epistemicism might account for this phenomenon. Epistemicists about vagueness explain the linguistic behaviour of vague terms, including their capacity to give rise to the famous sorites paradoxes, in terms of our being ignorant about certain facts about the world. For an epistemicist about redness, there is a unknowable fact of the matter regarding which precise shade on the colour spectrum marks the transition from redness to non-redness. For the epistemicist about metaphorical meaning, similarly, there are complete and precise facts about whether a given paraphrase of a metaphor is correct or not, but we simply cannot know what they are. Since, as noted above, most contemporary philosophers of metaphor endorse an
267

Stern (2000) p. 9

210

anti-realist answer to the Evidence Transcendence test, epistemicism will no doubt seem to an extremely revisionary proposal. The epistemicist thus has some important explanatory debts to discharge. They ought to tell us: (1) why we should believe in sharp cut-offs in paraphrases of metaphor. (2) why we are unable to know the positions of such cut-offs. (3) how the answers to the above two questions can be integrated with our pretheoretic conception of epistemology and semantics. I do not have space here for a satisfactory treatment of these issues. I can however, sketch the beginnings of a line that such a theorist might take. Firstly, the epistemicist ought to motivate his position by noting (a) his ability to give a straightforward resolution of a number of versions of the sorites paradoxes which (b) accords well with a minimal statement of the phenomenon of vagueness while (c) maintaining classical logic, an intuitive and clearly understood system and (d) avoiding the counter-intuitive results that seem to plague its rivals. 268 That begins to address (1). Secondly, the epistemicist ought to tell something like Williamsons story regarding our ignorance of such cut-offs. 269 Arguably, knowledge demands something like modal safety: it is a priori that, if S knows that P, then it couldnt easily have been the case that S believed that P and P was false. The problem with gaining knowledge of exactly where a given colour cuts off, or paraphrase ends, is that it is just too easy to form false beliefs in the area. Even if one happens to stumble by chance on the correct shade, or statement of the metaphors meaning, one does not count as knowing that one has done so, since we lack the relevant powers of discrimination to clearly distinguish such choices from their incorrect, closely overlapping neighbours. So one is typically necessarily ignorant of where redness cuts off, and what the uniquely correct paraphrase of a metaphor really is. The real puzzle for the epistemicist about m-meaning is in addressing the latter half of explanatory debt (3). What aspect of our linguistic thought and practice could possibly serve to determine that metaphor meant one perfectly precise paraphrase rather than
268

For a detailed defence of (a), (c) and (d), see Williamson (1994) and (1997), and Sainsbury and

Williamson (1997). For a defence of the claim that epistemicism coheres best with a minimal description of the phenomenon of vaguness, see Greenough (2003).
269

See Williamson (1994) Ch 7 and (2000) Chs. 5 and 7

211

another, closely related one, in a way that transcended all possibility of successful human enquiry? Surely, one might feel, this is wholly to surrender the Wittgensteinian insight that meaning has an internal connection to the way words and sentences are used? If this is the cost of finessing limitlessness, many will no doubt feel that it is hardly worth the candle. Nevertheless, it may be too soon to hope that such an extreme realist about metaphorical meaning will be forced into capitulation. The following answer may be open to him. Call each potential paraphrase of a metaphor a candidate-meaning of the metaphor. Think of such meanings as conjunctions of claims M1-Mn relating to the metaphor, a little like Sterns paraphrase of Juliet is the sun, only spelled out in some language which is infinitely more complicated than our own, suitable for expressing literally the fine-grained non-conceptual contents that English can only gesture at figuratively. Acceptable candidate-meanings will include each claim Mk that definitely belongs to the paraphrase of a given metaphor, and none of the claims that definitely do not belong to such a paraphrase. The vagueness of metaphor consists, according to our epistemicist, in our being unable to tell which of the acceptable candidate-meanings of the metaphor is the correct one. Even if we could formulate such paraphrases in such a language, and happened to pick the right one out by chance, we could never count as knowing that we had done so, since we cannot, we are supposing, reliably discriminate the correct candidate meaning from its incorrect, partially overlapping neighbours. So we can never have sufficient warrant for thinking that we have correctly paraphrased a metaphor, and a form of limitlessness is accounted for. The proposed epistemicist will think of a given metaphor as being ambiguous between its acceptable candidate-meanings. 270 Our linguistic practice, together with the way the world is, fails to determine a unique satisfier for the definite description correct paraphrase of the metaphor. Now, arguably, it does not make sense to ask after the unrelativized truth value of an ambiguous utterance. Take the claim that many bats can fly. Is that true? Well, its true of vampire bats, but not of baseball bats. The question

270

I owe this thought to Sider and Braun (2004). Those authors conclude that no sentence containing a

vague term has a truth value, and no argument involving one is valid. My epistemicist concedes that may be true of sentences, but denies it of utterances.

212

of the truth value of the sentence is only properly resoluble once the ambiguity in the string has been eliminated. A similar feature holds true of arguments. Is: something is a flying bat, so something is a bat valid? Yes, on some resolutions of the ambiguous terms; no on others. Since sentences involving vague terms are treated as if they were ambiguous in the more traditional sense, the same story goes for them. On this view, Frege was right to hold that vagueness in natural language must be eliminated before evaluation in terms of truth, validity, etc, can take place. Such a view is obviously radically revisionary. We normally think of claims like: (a) (b) Post-boxes are red Men with no hair are bald

as being straightforwardly true. Since, on the mooted view, each such vague sentence is ambiguous between a number of equally acceptable precisifications, it is strictly speaking incorrect to think of any of them being apt for truth or falsity. Similarly, intuitively valid arguments (All red things are coloured, this thing is red, so this thing is coloured) are actually not apt for such appraisal after all. 271 My epistemicist attempts to sweeten the pill with the following fable. Although such sentences are non-truth apt, since they do not express unambiguous propositional contents, utterances of such sentences in context do express such unique meanings. How can this miracle occur? Think of a linguistic context as a device for maximizing the intelligibility of conversational participants. 272 Now, conversational participants normally aim to speak truly, and reason validly, not just approximate to each. Moreover, they expect the same of other interlocutors. We ought to predict that context will thus shift in a way that allows them to achieve their aims. After all, thats exactly what happens when I tell you to put all the beer in the fridge, or describe my desk as flat. Contextual standards shift appropriately in response to the aims of conversational participants. How exactly this happens is no doubt a puzzling matter, which we dont
271

Sider and Braun accept this conclusion, but argue that approximate truth is all we need and should Compare Lewis (1979) p 420. conversational score does tend to evolve in such a way as is required

care about.
272

in order to make whatever occurs count as correct play.

213

yet understand fully, although we can model some aspects of it. Nevertheless, poor grasp of a phenomenon is not a reason to reject it, especially when we rely on appeal to it to do important explanatory work. How might context shift to let us achieve our aims? Since, by hypothesis, there are a selection of equally good candidate meanings, it cannot shift so as to pick an independently favoured one. In a sense, context is in the position of Buridans Ass. Both have a goal to eat straw, to maximise intelligibility which they will fail to achieve if they remain static in the face of equally good options. The obvious solution for the Ass in this case is not to think about things too hard; to pick an arbitrary bundle at random, rather than settling for hunger in the face of equally tempting alternatives. Similarly, the best result for context is to simply assign an arbitrary precisification to every vague term in the utterance. That way, intelligibility is maximized. If context refuses to assign a value in the case of a draw, then our utterances end up failing of truth and validity, contrary to our communicative intentions. And thats exactly what we think context-shifts tend to prevent. Now, since, by hypothesis, such a shift by context assigns an arbitrary precisification to vague terms, we cannot hope to know which semantic value it has assigned. Safety is violated, since it is just to easy for us to form false beliefs about which of the permissible sharpenings of e.g. red or bald has been assigned to a given utterance. So the basic epistemicist story about vagueness holds. Nevertheless, on the suggested account, the epistemicist can seem to circumvent their greatest problem; of identifying what it is about our pre-theoretic linguistic practice that could possibly determine sharp cut-offs. The answer turns out to be relatively simple: our shared expectation that the correct interpretation of ones thought and talk is the one that maximizes intelligibility. Our epistemicist simply puts a characteristic spin on this plausible thought, by denying that we can always in principle come to know what such a correct interpretation is. A similar story goes for metaphor. Metaphors, according to our hero, are ambiguous between their rival candidate-meanings. In cases where we want to get something fairly definite across, to speak the truth, in an extended sense, context shifts to assign an arbitrary such candidate-meaning to the metaphor. The search for paraphrase is clearly doomed, on this account, since we would never be able to know that we had lighted on 214

the correct restatement even if we had. So a form of limitlessness is established. We can identify some propositions that lie clearly on either side of a large, grey area, but beyond that point, we are out of our epistemic depth.

5.8

Conclusion

In the early part of this thesis, I outlined and offered some preliminary defence of a broadly Davidsonian error-theory about metaphorical meaning. In the last two chapters, I have aimed to examine some key issues relating to realism about metaphorical meaning in a broader, and more neutral manner. In Chapter 4, I discussed such meaning in the light of Wrights cruces, and began to examine whether they might help highlight previously unconsidered resources for accounting for the distinctive marks of metaphor. In this chapter, I continued that project. I argued that thinking of metaphors as serving to express non-conceptual contents provided some resources for a realist account of limitlessness and non-propositionality. I then described two ways in which such a model could be independently insightful and significant, in the philosophy of emotion and the philosophy of religion. Finally, I sketched a novel, robustly realist treatment of non-paraphrasability, in terms of a certain kind of context-dependent resolution of vagueness. It seems, then, that Davidson has not marshalled conclusive arguments, based on the marks of metaphor, that would serve to establish that a broadly pragmatic realist model is hopeless. Since, as I have already concluded, Davidson himself should agree that metaphors can be used to implicate propositions, this is not wholly bad news for the error-theorist. Nevertheless, if the defence of Davidsons (A6) depends maintaining on the vast majority of metaphors failing to convey such propositions, at the supposed cost of surrendering the marks of metaphor, then such a defence seems to fail. Of course, my discussion here has been provisional, sketchy and carried out at an intuitive level. Nevertheless, I hope to have established several strategies by which the realist about metaphorical meaning might hope to give an account of the limitlessness and nonpropositionality of metaphor, without surrendering central doctrines. In the absence of compelling reason for thinking that such strategies must fail, the pragmatic realist account of metaphorical content seems to remain as an important and viable position. 215

216

Conclusion
My aim in this thesis has been to examine whether the philosophy of realism can cast light upon the philosophy of metaphor. The issues involved are large ones, and my discussion has often been correspondingly limited, sketchy and provisional. Nevertheless, I hope that some minor progress has been made. I began by endorsing some key aspects of Wrights recommended methodology for prosecuting debates about realism - in particular his suggestion that we should aim to identify and apply a series of diagnostic tests for objectivity - while rejecting his characterization of the relative dialectical positions of the participants in such investigations. I then clarified and defended a broadly Davidsonian position on metaphor, signalling its kinship with error theoretic positions in other areas. I addressed Wrights worry that the availability of minimalism undercut the motivation for such a such a position, and argued that the Davidsonian was best understood as arguing that talk of metaphorical meaning, at least in the semantic sense of that ambiguous term, failed to meet even the most basic hurdle for objectivity, that of coherence in the light of the constitutive standards of its parent discourse. Though there is still much more work to be done, and plenty of room for the opposition to manoeuvre, I provisionally endorsed this aspect of the anti-realist challenge. Key premises of Davidsons master argument against a semantic treatment of metaphorical meaning had proved defensible, and the overall ontological economy of the position had earned it, I claimed, a right to default status in the field. In an effort not to foreclose too quickly on defensible theories, I broadened my avenue of enquiry. Wrights tests provided a useful structure in which to address key issues in the philosophy of metaphor, and I endeavoured to apply them in a way that was maximally neutral between live theories in the field. That methodological decision turned out to be fruitful, when attempting to apply the test to metaphor tests helped bring to the fore several underexplored issues relating, for example, to the realist standing of speaker-intention, to the autonomy of metaphorical interpretation, and the epistemic transparency of context that might otherwise have remained unexamined.

217

In the closing chapter, I set out a more positive vision, identifying a distinctive cognitive and linguistic role that metaphor characteristically discharges. I related Bermudez work on non-conceptual content to Davidsonian worries about realist neglect of the marks of metaphor, and tried to demonstrate how these concerns might be addressed by a pragmatic realist about metaphorical meaning. I briefly treated the role of metaphor in the communication and examination of our emotional states, and suggested some ways in which the position might serve to underwrite the possibility of irreducible religious metaphor, contra Alston. Throughout the thesis, I have tried to address key issues in contemporary philosophy of metaphor. My discussion has been uneven; painfully underdeveloped in some areas, and painfully detailed in others. My focus of attention has often circled round two or three central theories and concerns. My primary stalking horses have been the insightful and original approaches of Davidson, Stern and Denham, with which I have found much to disagree, but much also to applaud. I have often returned to the question of how to pursue profitable discussion about realism, to the proper understanding of the distinctive marks of metaphor, and the connection between figurative meaning and speaker intention. Each of these issues seems barely to have been broached, and yet I hope that I have identified some potentially rewarding routes for further enquiry. Philosophy of metaphor like the philosophy of the emotions, like the philosophy of art often seems beset with a single overriding tension. On the one hand, there is a strong tendency to try to bring metaphor within the compass of some systematic theory, that assigns it a suitably cognitive role, a place at the foundations of our epistemic practice. On the other, there is the tendency to feel that this approach, no matter how sensitively its practitioners try to account for the distinctive nature of the phenomenon, seems to leave everything out that made us interested in the topic in the first place. It is difficult to tell whether this reveals anything important about the nature of such subject matter, paradigmatically close as it is to the expressive, lived life of human beings, or whether it is merely a predictable side-effect of the usual murdering to dissect. In any case, this thesis, like most other work in the area, has made at best limited progress in resolving or dissolving that fundamental tension. Davidson famously described metaphor as the dreamwork of language, but I want to conclude instead with Wittgensteins cautionary remarks about what real dreamwork actually achieves: 218

In Freudian analysis, the dream is as it were dismantled. It loses its original sense completely. You might think of it as performed on the stage, with a plot that is sometimes fairly incomprehensible but also in part quite comprehensible, or at least apparently so, & as though this plot were then torn into little pieces & each part given a completely different meaning. You could also think of it like this: a picture is drawn on a big sheet of paper & the sheet is then folded in such a way that pieces which do not belong together at all in the original picture collide in appearance & a new picture, which may make sense or may not, is formedNow I could imagine that someone might exclaim Yes, that is the solution, that is what I dreamed, without gaps or distortions. It would then be this acknowledgement that made this solution the solutionIn this case it might really be said: only when you have found it, do you know what you were looking for. 273

It seems to me that Wittgensteins construal of dreamwork is a good metaphor for the interpretation of metaphor itself, and also, aptly enough, for the completion of a thesis.

273

Wittgenstein (1998) p. 78. ( MS 136 137a: 22.1.1948.)

219

220

Bibliography
Alston, William P 1989: Irreducible metaphors in theology in his Divine Nature and Human Language: Essays in Philosophical Theology (Ithica, NY: Cornell University Press) Beardsley, Monroe 1958: Aesthetics: Problems in the Philosophy of Criticism (New York: Harcourt, Brace and Co.) 1976: Metaphor and Falsity, The Journal of Aesthetics and Art Criticism, 35, pp. 218-222 Bermudez, Jos Luis Bezuidenhout, Anne Black, Max Blackburn, Simon 1998: The Paradox of Self-Consciousness (Cambridge, Mass.: MIT Press) 2001: Metaphor and What is Said, in French and Wettstein (2001) pp. 156-186 1978: How metaphors work: A Reply to Donald Davidson In Sacks (ed.) (1978) 1984: Spreading the Word (Oxford: Oxford University Press) 1988: Attitudes and Contents, Ethics, 98, 501-17 1993: Circles, Finks, Smells and Biconditionals, Philosophical Perspectives, 7, 259-279 1998: Wittgenstein, Wright, Rorty and Minimalism, Mind, 107, pp. 157-181 Burge, Tyler 1979: Individualism and the mental in P.A. French, T.E. Uehling and H.K.Wettstein (eds.) Midwest Studies in Philosophy (Minneapolis: University of Minnesota Press) 73-121. Churchland, Paul Cohen, L.J Cohen, Ted Cooper, David Crombie, I 1981: Eliminative Materialism and Propositional Attitudes Journal of Philosophy, 78, 1993: The Semantics of Metaphor in Ortony (ed.) (1993) 1979: Metaphor and the cultivation of intimacy in Sacks (1978) 1986: Metaphor (Oxford: Blackwell). 1955: Theology and Falsification, New essays in

221

Philosophical Theology, ed. Antony Flew and Alasdair MacIntyre (London:SCM Press) 1957: The possibility of theological statements in Basil Mitchell (ed.) Faith and Logic (London: Allen and Unwin). Davidson, Donald 1984: Inquiries into Truth and Interpretation (Oxford: Oxford University Press) pp. 245-264 1984a: What metaphors mean in Inquiries into Truth and Interpretation (Oxford: Oxford University Press) pp. 245-264 1984b: Communication and Convention in in Inquiries into Truth and Interpretation (Oxford: Oxford University Press) pp.265-280 1986 A Nice Derangements of Epitaphs, in Ernest Lepore (ed.) Truth and Interpretation, (Oxford: Oxford University Press) pp. 433-446 1993: Locating Literary Language in Reed Way Dasenbrock (ed.) Literary Theory After Davidson (Pennsylvania: Penn State Press) 2001: Subjective, Intersubjective, Objective, (Oxford: Oxford University Press). Davis, Steven (ed.) Denham, Alison 1991: Pragmatics: A Reader, (Oxford: Oxford University Press). 1998: Metaphor and Judgements of Experience in Casati & Tappolet (ed.) (1998) European Review of Philosophy Volume 3: Response Dependence (Stanford: CSLI) pp. 225-253 Denham, Alison Devitt, Michael Divers, John Divers, John and Alexander Miller, 2000: Metaphor and Moral Experience (Oxford: Oxford University Press) 1997: Realism and Truth, revised paperback edition (Princeton: Princeton University Press) 2002: Possible Worlds (London: Routledge) 1994a: Why Expressivists about Value Should Not Love Minimalism about Truth, Analysis, 54, pp. 12-19. 1994b: 'Best Opinion, Intention-Detecting and Analytic Functionalism', Philosophical Quarterly Vol. 44, Issue 175, p. 249-256

222

1995: Platitudes and Attitudes, Analysis, 55, pp. 37-44 Dummett, Michael 1973: Frege: Philosophy of Language (London: Duckworth). 1978: Truth and Other Enigmas (London: Duckworth) 1991: The Logical Basis of Metaphysics (Cambridge, Mass.: Harvard University Press) Edwards, Jim 1992: Best Opinion and Intentional States Philosophical Quarterly, Vol. 42, Issue 166, p. 21. 1994: Debates about Realism Transposed to a New Key, Mind, 103, 59-72 Evans, Gareth 1975: 'Identity and Predication', Journal of Philosophy 72 pp.343-63. Fodor, J and Lepore, E 2002: The Compositionality Papers (Oxford: Oxford University Press) Fogelin, Robert J. French, Peter A. and Wettstein, Howard Gallois, Andre 1988: Figuratively Speaking, (New Haven, CT: Yale) 2001: Midwest Studies in Philosophy Vol XXV: Figurative Language, (Oxford: Blackwell) 1998: Does Ontology Rest on a Mistake?, Proceedings of the Aristotelian Society Supplementary Vol. LXXII pp. 263-283. Gazdar, G. Gibbard, Allan Goodman, Nelson Grice, Paul 1979: Pragmatics: Implicature, Presupposition and Logical Form (New York: Academic Press) 1990: Wise Choices, Apt Feelings (Cambridge, Mass.: Harvard University Press) 1981: Metaphor as Moonlighting, in Johnson (1981) pp. 175-180 1975: Logic and Conversation, in D.Davidson and G. Harman (eds.) The Logic of Grammar (Encino, Cal: Dickinson) Griffiths, Paul Greenough, Patrick Haack, Susan 1991: What Emotions Really Are (Chicago UP, 1997) 2003: 'Vagueness: A Minimal Theory', Mind, Volume 112, Issue 446 1988: Surprising Noises: Rorty and Hesse on Metaphor, Proceedings of the Aristotelian Society LXXXIII, pp 291-301 Hagberg, Garry L. Hale, Bob 2001: Metaphor in The Routledge Companion to Aesthetics (London: Routledge) 1986: The compleat projectivist, Philosophical Quarterly,

223

36, 65-84 1992: Can there be a logic of attitudes? in Haldane and Wright (eds.) (1992): Reality, Representation and Projection (Oxford: Oxford University Press) 1997: Realism and its Oppositions in Hale and Wright (1997) Hale, Bob and Wright, Crispin (eds.) Harman, Gilbert Hopkins, R. 1997: A Companion to the Philosophy of Language (Oxford: Blackwell). 1977: The Nature of Morality (New York: Oxford University Press) 2000: Beauty and Testimony, in A. OHear, (ed.) Philosophy; the Good, the True and the Beautiful, Cambridge: Cambridge University Press . 2001: Kant, Quasi-Realism and the Autonomy of Aesthetic Judgement, European Journal of Philosophy 9:2, pp. 166-189 Horwich Paul Jackson, Frank Jackson, Frank, Oppy Graham, and Smith Michael Johnson, Mark Kemp, Gary Kittay, Eva Feder Kripke, Saul Lamarque, Peter Levinson, Stephen C. Lewis, David 1981: Philosophical Perspectives on Metaphor (Minneapolis, MN: University of Minnesota Press) 1991: Metaphor and Aspect Perception, Analysis 51:2 1987: Metaphor; Its Cognitive Force and Linguistic Structure (New York: Oxford University Press) 1979: Speakers Reference and Semantic Reference, reprinted in Davis (1991) pp. 77-96 1994: Truth, Fiction and Literature (Oxford: Clarendon). Page references to (1996) paper back edition. 1983: Pragmatics (Cambridge: Cambridge University Press) 1979: Scorekeeping in a Language Game, reprinted in Davis (1991) pp.416-427 1986: On the Plurality of Worlds (Oxford: Blackwell) Lycan, William 2000: Philosophy of Language (London: Routledge) 1990: Truth (Oxford: Oxford University Press). 1994: The Essence of Expressivism, Analysis, 54, pp. 19-20. 1994: Realism, Truth and Truth Aptness, Philosophical Books 35, pp. 162-169. 1994: Minimalism and Truth Aptness, Mind, 103, pp. 287302.

224

Margalit, Avishai and Goldblum, N, Martin, M. Martinich, A.P. McNaughton, N. Miller, Alexander

1994: Metaphors in an Open-Class Test, in Jakko Hintikka (ed.) Aspects of Metaphor, (Dordrecht: Kluwer) pp. 219-241 1992: Perception, concepts, and memory Philosophical Review 101, 745-763. 1984: A Theory for Metaphor in Davis (1991) pp. 507-518 1989: Biology and Emotion (Cambridge: Cambridge University Press) 2003: An Introduction to Contemporary Metaethics (Cambridge: Polity Press) 2002: What is the Manifestation Argument?, in A.Barber (ed.) Epistemology of Language (Oxford: Oxford University Press) forthcoming, (a): The Significance of Semantic Realism, Synthese forthcoming, (b): Realism and Antirealism, in Oxford Handbook of the Philosophy of Language, E. Lepore and B.Smith (eds) (forthcoming) (Oxford: Oxford University Press)

Moran, Richard

1989: Seeing and Believing: Metaphor, Image, and Force, Critical Inquiry, Vol. 16 pp. 87-112. 1997: Metaphor in Hale and Wright (1997).

Nogales, Patti D. Ortony Andrew (ed.) Peacocke, Christopher

1999: Metaphorically Speaking, (Stanford, California: CSLI Publications) 1993: Metaphor and Thought 2nd Edition (Cambridge: Cambridge University Press). 1993: A Study of Concepts (Cambridge, Mass.: MIT Press) 1994: Non-conceptual content: Kinds, rationales and relations Mind and Language 9, pp. 419--429. 1999: Being Known (Oxford: Oxford University Press)

Pettit, Philip, Putnam, Hilary

1990: "The Reality of Rule-Following", Mind, Vol. 99, 393, Jan. pp. 1-21. 1975: The meaning of 'meaning' in Philosophical Papers, ii: Mind Language and Reality (Cambridge: Cambridge University Press);

Quine, W.V

1951:

Two dogmas of empiricism, Philosophical Review, Vol. 60, pp. 20-43.

225

1960: Word and Object (Cambridge, Mass.: M.I.T Press) Recanati, Franois Reimer, Marga Richards I. A Rorty, Richard 1993: Direct Reference (Oxford Blackwell). Page references are to the 1997 paperback edition. 1996: The Problem of Dead Metaphors, Philosophical Studies. 1965: The Philosophy of Rhetoric, (Oxford: Oxford University Press) 1991: Unfamiliar noises: Hesse and Davidson on metaphor, in his Objectivity, Relativism and Truth: Philosophical Papers Vol.1 (Cambridge: Cambridge University Press) pp. 162-172 Sacks, Sheldon (ed.) Sag, I 1978: On Metaphor (Chicago: University of Chicago Press). 1981: Formal Semantics and Extra-Linguistic Content, in P. Cole (ed.) Radical Pragmatics, (New York: Academic Press) Sainsbury, R.M. and Williamson, Timothy Searle, John 1978: Literal Meaning, Erkenntis, 13: 207-24 1993: Metaphor, revised edition, in Ortony (1993), pp. 83111 Shapiro, Stuart and Taschek, William Sider, Ted, and Braun, David Smith, Michael 1996: Intuitionism, Pluralism and Cognitive Command, Journal of Philosophy, 93, pp. 74-88 2004: Vague, so Untrue, draft of 27th August 2004, available at http://fasphilosophy.rutgers.edu/~sider/cv.html#in_progress 1994a: Why Expressivists about Value should Love Minimalism about Truth, Analysis, 54, pp. 1-11. 1994b: Minimalism, Truth-aptitude and Belief, Analysis, 54, pp. 21-26. Soskice, Janet Martin Sperber, Dan, and Wilson, Deirdre Stalnaker, Robert Stern, Josef 1985: Metaphor and Religious Language (page references to pb. edition), (Oxford: Clarendon) 1986a: Relevance: Communication and Cognition, (Oxford: Blackwell) 1986b: Loose Talk, reprinted in Davis (1991) pp. 540-549 1976: Possible Worlds, Nos, 10, 65-75 1991: What Metaphors Do Not Mean, Midwest Studies in Philosophy, XVI, pp. 13-52 1997: Sorites in Hale and Wright (1997)

226

2000: Metaphor in Context (Cambridge, Mass.: MIT Press) Szab, Zoltn 2004: Compositionality, The Stanford Encyclopedia of Philosophy, http://www.seop.leeds.ac.uk/entries/compositionality/# 3.2 Tirrell, Lynne Travis, Charles Walton, Kendall 1991: Reductive and Nonreductive Theories of Metaphor, Journal of Philosophy Vol. 88, Issue 7, pp. 337-358 1997: Pragmatics in Hale and Wright (1997) 1990: Mimesis and Make Believe (Cambridge, Mass.: Harvard University Press). 1993: Metaphor and Prop Orientated Make-Believe, European Journal of Philosophy, Vol. 1, No. 1. pp. 3957. Weiss, Bernard White, Roger M. Williamson, Timothy 2002: Michael Dummett (London: Acumen) 1996: The Structure of Metaphor (Oxford: Blackwell) 2004: Manuscript. 1994: Vagueness (London: Routledge) 1996a: Knowing and Asserting, The Philosophical Review, 105, 4, pp. 489-523 1996b: Unreflective Realism, Philosophy and Phenomenological Research, Vol. 56, No.4, 905-909 1997: Prcis of Vagueness, Philosophy and Phenomenological Research, Vol. 57, No. 4, December 1997 pp. 921-928. 2000: Knowledge and Its Limits (Oxford: Oxford University Press) Wittgenstein, Ludwig Wright, Crispin 1998: Culture and Value, revised second edition with English Translation (Oxford: Blackwell) 1987: 'On Making Up One's Mind: Wittgenstein on Intention', in Logic, Philosophy of Science and Epistemology, P. Weingartner and G. Schurz (eds.), (Vienna: HolderPichler-Tempsky), pp. 391-404 1989a: Wittgenstein's Rule-Following Considerations and the Central Project of Linguistics', in Alexander George (ed.), Reflections on Chomsky (Oxford: Blackwell), pp. 233-64

227

1989b: Critical Notice, Wittgenstein on Meaning, by Colin McGinn, Mind 98, pp. 289-305 1989c: 'Wittgenstein's Later Philosophy of Mind: Sensation, Privacy and Intention', Journal of Philosophy 86, pp. 622-35 1992: Truth and Objectivity (page references to pb. edition), (Cambridge, Mass.: Harvard University Press). 1993: Realism, Meaning and Truth 2nd Edition, (Oxford: Blackwell) 1994: Reply to Jackson, Philosophical Books 35, pp. 169 175. 1996: Response to Commentators, Philosophy and Phenomenological Research, Vol. 56, No.4, 911-941 1998: Comrades Against Quietism: reply to Simon Blackburn on Truth and Objectivity, Mind, Vol. 107, pp. 103-203 1999: Truth: A Traditional Debate Reviewed in Truth, Simon Blackburn and Keith Simmons (eds.) (Oxford: Oxford University Press) pp. 203-38 2001: On Being in a Quandary, Mind, 110, pp. 45-98 Yablo, Stephen 1998: Does Ontology Rest on a Mistake?, Proceedings of the Aristotelian Society Supplementary Vol. LXXII pp.229-2

228