Piers Rawling, The Florida State University
What follows is a version of comments presented at the November 10, 2006 meeting of the Florida Philosophical Association. I have modified my original remarks somewhat in light of input at the meeting, and I would like to thank all those who participated – particularly Kirk Ludwig. I owe a far larger debt to Donald Davidson himself, who was my dissertation supervisor, and spent many hours explaining his views to me, and attempting to teach me to think philosophically. He is, of course, not responsible for any errors herein.
I shall endeavour to make my remarks accessible to those unfamiliar with his work. In doing so, I will no doubt sacrifice subtlety and accuracy for simplicity. But if you want the full dose of the former pair, I suggest you consult Donald Davidson: Meaning, Truth, Language, and Reality. This is an excellent book that covers and critiques just about everything Davidson ever wrote. And there is much in it with which I agree. So much, in fact, that I am not going to venture to disagree with anything Ernie Lepore and Kirk Ludwig (hereafter EK) say. Rather, I am simply going to bolster one of their lines of argument with one of my own. In doing so, I’ll be drawing on Rawling 2001 & 2003; the former is followed by a reply from Davidson 1; the latter is in Kirk Ludwig’s edited volume Donald Davidson – so much of this will be familiar to him, but perhaps others may benefit.
The argument concerns whether or not radical interpretation is possible. EK argue that it is not. I agree. But I have a somewhat different argument for its impossibility.
First, what is radical interpretation? Suppose you find yourself stranded on a desert island with another castaway, and rapidly discover that you cannot communicate with her; indeed, it is perhaps not clear initially that she has a language. You must build your interpretation of her, as agent and speaker, from scratch. It is interpretations such as this that are radical.
Davidson, following, to a certain extent, Quine, is famous for pursuing the issue of how such radical interpretation is possible. How is this pursuit undertaken? By exploring the constraints on interpreting the raw data of overt behaviour – constraints that we must respect if our enterprise is to be the interpretation of some other being as a speaker and a thinker. Thus, in understanding these constraints, we can come to learn much about what is required to be a speaker and a thinker.
These constraints involve assumptions that the interpreter must make about the interpretee. Davidson argues, then, that for interaction to be mutual interpretation, the parties must make assumptions about each other that could not turn out to be false lest their enterprise fail to be interpretation at all. In a sense, then, no interpretation is built entirely from scratch, and it is this that makes radical interpretation (hereafter RI) possible.
To avoid confusion, it should be noted that Davidson’s exercise is conceptual. What emerges is not a manual for the field linguist, but a distinctive, and in many respects compelling, picture of language and the mind. Importantly, Davidson sees as impossible the conceptual reduction of our intentional concepts (those we use to describe the meanings and minds of others) to the non-intentional. He seeks, rather, to illuminate the former by examining their relations to one another and to the evidence we use in their deployment.
What are the constraints on RI – the constraints that make it possible? These are summed up under the ‘Principle of Charity’. This is a term borrowed by Quine from N. L. Wilson 2; and it is a purposeful misnomer. There is nothing charitable about its application: it must be applied if we are to be engaged in interpretation.
There is controversy, however, over its best formulation – EK have a very illuminating discussion of it (p. 185 ff) But the basic idea is that the interpretee must be interpreted (1) as being largely rational, and (2) as having largely true beliefs (about her immediate macroscopic environment, at least). For instance, in the absence of evidence that your interpretee is round the bend, an interpretation that has him making inexplicably crazy remarks is likely to be misinterpretation. Much more needs to be said to fill out this principle, of course. Which kind of beliefs, and what proportion (can this really be quantified?) of them, must be true? And what exactly is rationality? Does it go beyond, say, a certain amount of logical acumen and some basic constraints on one’s preferences? I shall not address such issues here: the interested reader can consult EK for more detail.
One of Davidson’s central claims is that the constraints on interpretation are not sufficient to force uniqueness: interpretation is inevitably indeterminate (see EK chapter 15 and the Davidson references therein). There will be many interpretations (perhaps infinitely many) that respect the
principle of charity and account for the behavioural data (non-intentionally described, of course – that is, the behaviour is not to be described using propositional attitude vocabulary since that already is to presume an interpretation). As Quine puts it for translation: “translation [is] indeterminate in principle relative to the totality of speech dispositions.” 3And Davidson sees the same holding true for interpretation (the difference between translation and interpretation is, roughly, that to interpret someone is to understand them, whereas you might know that some sentence is, say, a French translation of some German sentence without understanding either). Radical interpretations, then – including attributions of beliefs – are not uniquely determined by the evidence.
The basic idea of this sort of indeterminacy is familiar from discussions in the philosophy of science: the data, it is said, underdetermine theory. Many different theories can account for the observations made in any particular realm – and this is true, perhaps, even in the ideal limit when we have made all the observations that are humanly possible. One sort of instrumentalist claims, in light of this, that we should abandon the quest for truth when it comes to scientific theories that go beyond our observational capacities. Rather, such theories should be viewed as convenient instruments that aid us in making accurate observational predictions about, and successful interventions in, our world. Such a view may be tied to an epistemic conception of truth, according to which the truth cannot outrun our ability to know it. If observation is required for scientific knowledge, and theory outruns our powers of observation, then theories cannot be known. If in addition truth-value is tied to knowledge, then theories lack a truth-value – it is not merely that their truth-values cannot be known; rather, they have no truth-value. We might say that their truth-values are not merely epistemically indeterminate, but metaphysically so.
Indeterminacy, then, enters this account of instrumentalism in two places. First there is the claim that observation does not determine theory. And second it is claimed that truth-value is metaphysically indeterminate in the absence of knowledge. Compare this with Davidson’s account of the indeterminacy of interpretation. So far we have a parallel to the first indeterminacy: there are many interpretations that account for the behavioural data. But Davidson goes further to give us, apparently, a parallel to the second indeterminacy – for example “meaning is entirely determined by observable behaviour.” 4 And the same goes for all the propositional attitudes – there’s ‘no more’ to, say, beliefs than what can be gleaned from radical interpretation, and radical interpretations are not unique.
So now it looks as though Davidson must abandon claims to the effect that we really do have propositional attitudes – just as the instrumentalist above abandoned truth-values when it comes to theory, so Davidson must abandon truth-values when it comes to, say, remarks to the effect that so-and-so believes that p. Or, to put it more starkly, there are no beliefs, nor any other
propositional attitudes for that matter. The indeterminacy of interpretation is not only epistemic, but also metaphysical.
But now we have:
(1) Radical interpretation requires charity (this is what makes RI possible by ensuring that radical interpretations are not constructed entirely from non-intentionally described behaviour alone). And charity requires the existence of beliefs: not only must the interpretee have beliefs, but many of them must be true.
(2) There’s ‘no more’ to beliefs than what can be gleaned from radical interpretation.
(3) But radical interpretations – including attributions of beliefs – are not uniquely determined by the evidence.
Together, (2) & (3) entail that there are no beliefs. (1) entails that radical interpretation requires beliefs.
If, then, to say that radical interpretation is possible is in part to say that we have beliefs and they can be determined from the stance of the radical interpreter, then radical interpretation is impossible. Davidson, of course, would disagree. Rather, he claims that the multiplicity of radical interpretations is no more problematic than, say, the multiplicity of scales of length or temperature measurement – radical interpretation is analogous to measurement.
Does the Measurement Analogy Save the Day?
The basic idea of Davidson’s analogical strategy is clear enough. 5We use sentences to track the propositional attitudes of an agent; we use the real numbers to track, say, the lengths of objects. The sentences in the former attribution play the role of the real numbers in the latter. We might maintain that there is an ‘indeterminacy of length’: there are infinitely many serviceable schemes for attributing lengths (feet, inches, metres, etc., and we could come up with infinitely many more). Similarly, there are many serviceable schemes for attributing meanings and the other attitudes to a given interpretee. And the latter indeterminacy, Davidson claims, is as benign as the former. But the details of the analogy need spelling out.
Davidson does not deny that there are many disanalogies between the two cases. There is, for example, an algorithm for moving from one scale of length measurement to another (multiply by the relevant positive constant – for example, multiply by three to move from feet to yards); there is no such algorithm in the case of interpretations. However, we do need at least something invariant
across interpretations if there is to be any appropriate analogy at all. What does invariance amount to in the measurement theoretic case? First, there are invariant relata. In the case of length, for instance, we attach the numbers to physical objects (the relata), and these remain the objects to be measured as we move from one scale to the next. And each scale preserves the relation ‘is at least as long as’: the greater the number on any given scale, the longer the object; and if it’s longer on one (say feet), it’s longer on all the others (metres, inches, etc.).
What remains invariant across radical interpretations? We needn’t insist on pushing the analogy all the way, but at least we should be able to find a class of relata whose membership is fixed across different interpretations. It might initially appear that these relata are the propositional attitudes of the interpretee. But do these remain invariant across different interpretations? No. It is precisely they that vary. On Davidson’s picture, indeterminacy is simply the fact that we can use different locutions to locate the same node in some pattern. But what are the invariant nodes? They cannot be propositional attitudes: the belief that p, say, under one scheme, will be the belief that q under another – two different propositional attitudes (both attributions are couched in the one idiolect of the interpreter). The relata cannot comprise the propositional attitudes themselves since we identify propositional attitudes in part by their content. And their content varies between schemes of interpretation.
One might respond here by wondering whether there is not some ‘neutral’ way of identifying propositional attitudes – so that, despite appearances to the contrary, the belief that p and the belief that q are in fact the same belief. But the problem is worse than so far suggested. Propositional attitudes in one scheme need not even map one to one onto propositional attitudes in another. Consider the following possibility: on one interpretation, a piece of behaviour, B, is interpreted as a signal; on another interpretation (that also saves all the relevant phenomena),B is interpreted as a simple scratch. 6Thus, on the first interpretation, B is explained by a complex of propositional attitude states that is simply larger than that invoked in the second. Or, indeed, the explanation of B on some scheme might invoke no propositional attitudes at all.
What is Invariant?
What, then, are the invariants across interpretations on Davidson’s account? Quine has us tracking “men’s dispositions to respond overtly to socially observable stimulations.” 7And according to Davidson, “… a satisfactory theory is one that yields an acceptable explanation of verbal behaviour and dispositions.” 8 Behavioural dispositions are key; and they are apparently the invariant nodes across different schemes of interpretation – they are what propositional attitude attributions ‘measure’. Just as we attribute numbers in the form of lengths to track the relation ‘is at least as long as’, so we attribute propositional attitudes to track behavioural dispositions.
The invariants, then, have the following form: a disposition to B under circumstance C, where Bis a bodily movement non-intentionally described, and C is specified without reference to propositional attitudes of the agent. A full interpretation of an agent – something unattainable by mere mortals – would account for the totality of the agent’s behavioural dispositions. And all adequate interpretations must agree on these dispositions; but (modulo charity, of course) that is all they must agree on. Propositional attitude states seem to be mere posits; they vary from interpretation to interpretation. That we cannot re-identify a belief across interpretative schemes is a consequence of the fact that there is no such state to re-identify.
To return to instrumentalism, an instrumentalist in the philosophy of science holds something akin to the following. The constraints on a theory are that it should save the phenomena and meet some criteria of coherence and simplicity (and elegance, perhaps). Theories will (usually) posit entities that underlie the appearances; but, according to the instrumentalist, these are only posits – constructions designed to save the phenomena. If two theories were to save the phenomena, score equally well in other respects, but disagree as to posits, the instrumentalist would have no difficulty in countenancing both: the posits, being merely that, would not compete.
Davidson’s view of the mind, then, might appear to be instrumentalist, the relevant criteria of coherence being provided by the principle of charity; the posits comprising, amongst other things, propositional attitudes; and (non-intentionally described) behaviours and behavioural dispositions constituting the phenomena to be saved. About some posits Davidson is quite explicit:
words, meanings of words, reference, and satisfaction are posits we need to implement a theory of truth. They serve this purpose without needing independent confirmation or empirical basis …. [Satisfaction and reference are] notions we must treat as theoretical constructs whose function is exhausted in stating the truth conditions for sentences. 9
degree of belief is a construction based on more elementary attitudes. 10
When it comes to the propositional attitudes he seems, however, more wary. He does say:
In thinking and talking of the weights of physical objects we do not need to suppose there are such things as weights for objects to have. Similarly in thinking and talking about the beliefs of people we needn’t suppose there are such entities as beliefs. 11
But in the same paper, he speaks of beliefs as mental states with identity conditions. However if Davidson is correct in his views that
(i) interpretation is epistemically indeterminate: many different interpretations will save the phenomena; and
(ii) all of these will be correct, since there is no more to correctness here than saving the phenomena
then there are no propositional attitudes. (i) implies that the propositional attitudes are epistemically underdetermined in the sense of out-running the evidence. Yet (ii) implies that there is no more to the propositional attitudes than the evidence provides, so that the evidence cannot be out-run. But nothing can fit both these bills.
Can Davidson, however, really adopt instrumentalism concerning the propositional attitudes? Is the position even coherent? The first difficulty is that the principle of charity would have to be reformulated – it could no longer commit us to the existence of beliefs, let alone true ones. And then there is a more overarching concern that there would be no perspective from which the view itself could be stated since there would be no statements. Furthermore, we couldn’t know any of this since (as Eve Garrard pointed out to me) there would be no states of knowledge.
There are also more specific problems for others of Davidson’s own views. For example, he relies upon the existence of propositional attitude states in his (justly famous) causal theory of action – roughly speaking, they’re what does the causing. I’ll conclude with a brief discussion of another aspect of his views, however.
There is an anti-Cartesian thread that runs through Davidson’s work on interpretation. 12 As EK put it:
If Davidson is right, then the central mistake of the philosophical tradition is the assumption of the Cartesian standpoint, and, in particular, the central place in epistemology accorded to the epistemic priority of knowledge of our mental states to knowledge of the world and other minds (EK, p. 418).
The Cartesian epistemic perspective is first-person: you start from knowledge of your own mind and attempt to move outward. Davidson, by contrast, proposes (roughly speaking) to begin in the external world. When it comes to the problem of other minds, for example, one way to look at matters is to see Davidson as putting the shoe on the other foot: the problem is not so much one of other minds as a problem about your own. The minds of others are accessed through interpretation of their behaviour, but how do you access your own mind? It’s not as though you observe your own behaviour and compile a self-interpretation (or, at least, such goings on are very much the exception). 13 This anti-Cartesianism is heady stuff, and potentially revolutionary. But, as EK go on to point out (pp. 418-419):
The central difficulty, we believe, lies in the apparent underdetermination of the content of our thought about meaning and thought content by behavioural evidence, including interactions with objects in our environment. What must be done is to show that the [external] standpoint can yield acceptably determinate assignments of contents to beliefs and interpretations to sentences.
Yet again, I agree. One could hold that there really are propositional attitudes and there’s nothing more to them than is accessible to the radical interpreter if it could be shown that radical interpretation is not epistemically indeterminate (or, at least, is far less so than Davidson argues in some places 14).
Davidson, Donald. Inquiries into Truth and Interpretation. Oxford: Clarendon Pr, 1984.
Davidson, Donald. “The Structure and Content of Truth.” Journal of Philosophy 87 (1990): 279-328.
Davidson, Donald. Subjective, Intersubjective, Objective: Philosophical Essays Volume 3. Oxford: Clarendon Pr, 2001.
Davidson, Donald. “Reply to Karlovy Vary Papers.” In Interpreting Davidson, edited by Petr Kotatko, Peter Pagin & Gabriel Segal, 285-307. Center for the Study of Language and Information, 2001.
Quine, W.V.O. Word and Object. Cambridge, Mass.: MIT Pr, 1960.
Rawling, Piers. “Davidson’s Measurement-Theoretic Reduction of the Mind.” In Interpreting Davidson, edited by Petr Kotatko, Peter Pagin & Gabriel Segal, 237-255. Center for the Study of Language and Information, 2001.
Rawling, Piers. “Radical Interpretation.” In Donald Davidson, edited by Kirk Ludwig, 85-112. Cambridge, U.K.: Cambridge UP, 2003.
- Donald Davidson, “Reply to Karlovy Vary Papers,” in Interpreting Davidson, eds. Petr Kotatko, Peter Pagin & Gabriel Segal (Center for the Study of Language and Information, 2001), 298-300.
- See W.V.O. Quine, Word and Object (Cambridge, Mass.: MIT Pr, 1960), 59.
- Quine 221.
- Donald Davidson, “The Structure and Content of Truth,” Journal of Philosophy 87 (1990): 314. This is discussed by EK on p. 166 ff.
- See, e.g., Donald Davidson “What is Present to the Mind?” in Subjective, Intersubjective, Objective: Philosophical Essays Volume 3 (Oxford: Clarendon Pr, 2001).
- This example is similar to one suggested to me by Kirk.
- Quine ix.
- Donald Davidson, “The Inscrutability of Reference,” in Inquiries into Truth and Interpretation (Oxford: Clarendon Pr, 1984), 237.
- Donald Davidson, “Reality without Reference,” in Inquiries into Truth and Interpretation (Oxford: Clarendon Pr, 1984), 222-223.
- Davidson, “The Structure and Content of Truth,” 322. See EK, p. 225 ff for further discussion.
- Davidson, “What is Present to the Mind?”
- See his “What is Present to the Mind?” for example.
- See, e.g., Donald Davidson, “First Person Authority,” in Subjective, Intersubjective, Objective: Philosophical Essays Volume 3 (Oxford: Clarendon Pr, 2001) and EK, chapter 20, and the references therein.
- See, e.g., Davidson, “The Inscrutability of Reference.”