For Donald Davidson, the concept of truth is central to providing an account of meaning, but remains indefinable yet utterly intuitive.1 For Stephen Stich, on the other hand, truth (or at least the concept of truth as used by Davidson, Alfred Tarski, and others) is idiosyncratic, limited, and best understood as a modestly valuable remnant of folk psychology that awaits either elimination or modification by an improved psychology.2 I wish to evaluate and ultimately challenge precisely this contention by Stich – that our intuitive notion of truth (as expressed in T-sentences3 such as “ ‘the grass is green’ is true if and only if grass is green”) is uninteresting or limited. To do so, however, will require a bit of background to motivate both Stich’s challenge and what I take to be a response that is commensurable with Davidson’s view.
The results of numerous studies conducted in and since the 1970s show that subjects consistently fail a variety of basic reasoning tasks.4 This has led some psychologists and philosophers to declare that, far from being rational animals, human beings are largely irrational and reliant not on some well-founded rational decision making procedure, but on pragmatic and often biased strategies to achieve our ends. Attempts to explain and explore these results have engendered a variety of new lines of research, not the least of which is experimental philosophy. Although the tradition and the programs it generates are diverse, a central theme is that a priori (read loosely) argumentation and thought experiments (i.e., armchair philosophy) are poor contenders to back if we wish to make progress in philosophy. The cognitive diversity that these studies reveal suggests it is a mistake to view human beings as largely rational; subsequently, it allegedly follows that lurking behind many thought experiments, a priori arguments, and so on is a quagmire of faulty reasoning, fallacious inferences, and cultural biases that precludes these arguments from having much credibility.
A central figure in experimental philosophy is Stephen Stich, and a key work to focus on is his The Fragmentation of Reason. In it, he argues against analytic programs in epistemology5 (which are potentially flawed as they may contain hidden biases, etc.) and offers a new, pragmatic epistemology that builds on the results of the experiments in psychology and other social sciences. For the purposes of this chapter, what is of most interest is not his replacement candidate for a reformed epistemology, but the arguments he provides in the hopes of making space for a new epistemology that reflects directly on the issue of truth.
According to Stich, there are two ways that cognitive diversity might threaten Davidson’s project. First, Davidson’s argument against conceptual schemes is either uninteresting or empirically falsified. Second, theories of meaning that are reliant on Alfred Tarski-style T-sentences are idiosyncratic and so only of moderate use. I find Stich’s criticisms of Davidson to be lacking on several fronts and find the account of cognitive diversity, while most certainly true, to not pose a serious problem for Davidson. To substantiate this claim, I will contend that Stich’s arguments against Davidson – both his argument against Davidson’s criticism of such a thing as a conceptual scheme, and his argument that Davidson’s account of the connection between rationality and belief is vacuous – fail.
Before I present Davidson’s argument against conceptual schemes, it is worth mentioning that there is an exegetical mistake in Stich’s reading Davidson’s argument against the possibility of conceptual schemes as geared to support a view that suggests that human beings all reason in similar ways and, for the most part, do it well.6 In many ways, Davidson’s argument has little psychological relevance – he simply is not interested in giving a psychological account of what it is to be a good reasoner.7 Nor does he claim that we all share the same idealized rational conceptual scheme. What he provides is (1) an analysis of the notion of what a conceptual scheme is alleged to be, and (2) an answer to the question of whether radical untranslatability or incommensurability between two languages is coherent. As I will show below, he argues that it is not.
In “On the Very Idea of a Conceptual Scheme,” Davidson argues that the notion of a conceptual scheme is fatally confused. A conceptual scheme, supposedly, provides the means by which different peoples are able to understand the world. It suggests there is a neutral and independent realm of content or experience, which language or concepts filter or organize. Strictly speaking, Davidson does not claim that a conceptual scheme is impossible, but rather that the notion of there being such a thing is unintelligible. He does this by first examining the metaphors that are used to support the claim that conceptual schemes do, in fact, exist (and the corollary claim, of course, that these schemes would remain untranslatable). There are two chief metaphors deployed: either conceptual schemes organize experience and the world, or they fit the world and accord with experience.8
The first metaphor collapses quickly under its own internal inconsistency. Davidson notes that only pluralities can be organized; the world and experience, taken as monoliths, are not intelligibly open to organization. Rather, objects in the world, experiences of something, can be organized. Yet once we have that ontology in common with others, what could prevent translation? Even if we have different experiences and ways of organizing the world – and, of course, we do – why should that present a barrier to translation? This recognition of difference explicitly acknowledges the fact that different people will have varying beliefs about what they experience and what is in the world, and even, perhaps, will hold different standards with respect to assessing those claims. Far from denying cognitive pluralism, Davidson’s argument against conceptual schemes can be seen as an acknowledgement and acceptance of it. Thus, he should not be read here as arguing a priori against weak forms of cognitive pluralism, as he never denied the uncontroversial idea that cultures and peoples differ in substantial ways. Instead, he can be read as arguing against what I will call strong cognitive pluralism, in which we could grant that other beings could have beliefs, be rational, and so forth, but could still remain unintelligible to us.
At this stage, irrespective of the argument’s soundness, it should be evident that Davidson has not claimed that cognitive diversity is impossible or that we must all adhere to one true form of rationality to be genuine reasoners. That Stich comes close to endorsing this interpretation is puzzling,9 but I think he needlessly conflates two separate issues. The first is the normative question of how we ought to reason; it seems plausible to me that he is correct in suggesting that better strategies than those that currently exist are intelligible and worth developing. The second issue, however, centres on the notion of what we need to be properly understood as a reasoner and as a believer – the necessary conditions for being described legitimately as a rational being.10 Davidson focuses on this concern, not on the normative question of how we ought to reason.
Hence the failure of numerous subjects to follow deductively valid paths to reach conclusions in the selection task and conjunction problem is largely irrelevant with respect to the latter issue; even if they fall short of the deductive ideal, they are rational enough in their beliefs and actions that we are able to ascribe content to their utterances. Furthermore, it is worth asking how genuinely irrational the subjects of these studies actually are, and whether we can substantiate this claim of cognitive pluralism in such a way that it might be a threat to Davidson’s project. It seems that far from being irrational, the subjects of the studies are largely rational – after all, they were rational enough to communicate with their interviewers, read the instructions of the test, make their way to the interview room, etc. Failing a handful of logical questions hardly seems evidence enough to justify the claim that human beings are largely irrational. It does show, uncontroversially, that people do not follow deductively valid inferences most of the time, but it is difficult to find in any of Davidson’s writings the very strong claim that in order to think and speak we must do so unvaryingly when reasoning about daily affairs.
To genuinely challenge Davidson’s program, Stich would need to argue that a being could have contentful cognitive states but remain uninterpretable or unrecognizable as having those states. Davidson denies that this is actually a coherent claim, as it would be tantamount to endorsing a distinction between scheme and content. To see why, Davidson’s analysis of the second metaphor is most helpful. To say that schemes fit the world or face experience amounts to the claim that various sentences are true (or false). The intermediary appeal to “fit” or to “facts” that this metaphor relies upon does not, Davidson states, add anything “intelligible to the simple concept of being true.”11 Moreover, if we grant that the sentences of others can be true – and this is what the advocates of conceptual schemes must do (otherwise the idea of a conceptual scheme does no work) – then the notion of conceptual schemes differing radically amounts to the claim that the sentences of two different languages could be largely true and yet untranslatable. This way of putting it suggests that truth and translation are separable – a claim that Davidson, following Tarski, finds incoherent.
In Fragmentation, Stich wants to both “soften up the conviction that true beliefs are obviously valuable”12 and move past intuition-based accounts of truth by revealing the limitations and weaknesses of what he considers the best of such theories. As before, his goal is to create space for the possibility of a radically improved epistemology that does not rely on folk intuitions in the same way that the best scientific theories are not held responsible to folk intuitions about physics, chemistry, and so on. Ideally, he will have the resources to show that Tarskian T-sentences (what he calls the interpretation function) and causal theories of reference à la Putnam and Kripke are inextricably tied to our folk intuitions and, because of this, are limited in their usefulness. In contrast, what I hope to show is that while Tarski (and Davidson) rely on a “folk” notion of truth (although, as I will discuss, the applicability of this claim to Davidson is contentious), that notion is best understood as a fundamental norm of rationality that we simply cannot do without. Moreover, as I will show, even Stich can be read as relying on such a norm, and his blindness to this can be diagnosed by an overly epistemological reading of truth that, I will argue, neither Davidson nor Tarski intended it to have. In short, I will argue that Stich has embraced the second metaphor detailed above – the idea that something is true when it fits the facts. If I am correct in this diagnosis, then Stich’s criticism of Tarski, and by extension Davidson’s truth-theoretic semantics, fails and Davidson’s argument against the intelligibility of a conceptual scheme is left untouched.
To motivate his criticism, Stich intends to cast doubt upon the credibility of the link between truth and translation by suggesting that our intuitive understanding of the truth of T-sentences is simply one out of a multitude of different ways of understanding truth. The core of Stich’s critique of Tarski-style truth theories is that they are (a) limited and (b) idiosyncratic. Their limit is revealed by their non-obvious application to non-declarative sentences. While Stich grants that sentences like “grass is green” and “snow is white” can be given truth conditions via appeal to T-sentences, the application of T-sentences to counterfactual sentences, interrogatives, etc., is not so easily done – which is not to say that it cannot be done, but rather that a straightforward appeal to intuitiveness is not readily available for such sentences. Stich’s second critique – that T-sentences are idiosyncratic – is far more damaging (if it succeeds) and is where he focuses his attention.
Davidson aims to arrive at the truth conditions of novel sentences by appealing to the base axioms of a truth theory as those base axioms are also used to generate the theorems employed to express the truth conditions of sentences already known in the language. Similarly, Tarski generates his extensional definition of truth by constructing T-sentences from a large, but finite, set of axioms. For example:
1 (∀ x)(x satisfies “is a philosopher” iff × is a philosopher).
2 “Plato” denotes Plato.
Unfortunately, says Stich, “Tarski tells us too little about what it is to get these axioms right. He does not tell us what relationship must obtain between a name and a person for the former to denote the latter.”13 So too with satisfaction. Without a theoretically defensible account of reference to ground T-sentences, their truth (or falsity) remains subject to the whims of our intuitions and what we find intuitively plausible as an account of reference. Thus, Stich’s claim here is that in generating a T-sentence, such as:
3 “Plato is a philosopher” is true iff Plato is a philosopher.
the use of the predicate “is true” is warranted, or the concept articulated, by appeal to the theory of reference that makes the assertion of the simple axioms of (1) and (2) epistemically justified. Since Tarski is silent about what justifies such an axiom, we are left in limbo.
Stich suggests that a possible remedy to this problem might be found in a causal/historical account of reference that would make it evident how and why, for example, “Plato” denotes Plato.14 He is quick to note that such a move will quickly flounder on, again, the fact that such theories are held accountable to commonsense intuitions (which certainly have been shown to vary across cultures)15 and, more damningly, are but a tiny subset of possible, although unintuitive, accounts of reference that might be able to generate meaning via alternative T-like sentences.16
To be explicit, there are numerous causal connections (the vast majority of which will not be intuitive) that will link “Plato” to the ancient Greek philosopher, to my former neighbour who ran Plato’s Pizza, or to no one at all. Moreover, Tarski did not tell us which theory of reference was right, and so we have no way to exclude the unintuitive T-sentences such axioms generate. For Stich, this suggests that the meaning we associate with a sentence (if it is given by the satisfaction conditions generated by T-sentences) is but one of a potentially infinite array of other interpretations that different functions (say T*- and T**-sentences)17 might provide. Since the authority that T-sentences are claimed to have rests on their intuitiveness (which may very well be a product of socialization or other factors), we must allow that these other semantic interpretations – which “vastly [outrun] those that our intuitive semantics is prepared to admit”18 – are genuine alternatives.
To adapt an example from Stich, within the statement “there is water on the moon” the referent for “water” could be, on one theory of reference, H2O; XYZ on another; and H2O and XYZ on a third. What Stich wishes to argue is that truth conditions for the statement “there is water on the moon” are dependent on the theory of reference to which we subscribe. Since there is no general consensus as to which theory of reference is correct – only appeals to intuitiveness – we must multiply truth to allow for these non-intuitive, but meaning-generating, truth conditions to be captured (as we have no non-idiosyncratic way of precluding them). To illustrate:
4 “There is water on the moon” is true (on one theory of reference) iff there is H2O on the moon.
5 “There is water on the moon” is true* iff there is XYZ on the moon.
6 “There is water on the moon” is true** iff there are H2O and XYZ on the moon.
Although I am not interested in discussing the merits of Stich’s entire project, it is worth noting that he intends to develop, once this pluralistic account of truth is well-supported, a pragmatic account of cognitive evaluation that does not require that rational agents (qua rational agents) always (or even for the most part) prefer that their beliefs be true. Perhaps, instead, they prefer that their beliefs are true* or true**. All that really matters is that in being judged successful, a cognitive system is able to achieve its goals – whatever they might be – and true beliefs need not (and ideally would not) play a central role in that achievement.
That being said, Stich is correct in remarking that T-sentences say nothing whatsoever about a theory of reference. Still, it remains to be seen whether the “intuitive” concept of truth is idiosyncratic in the way he suggests. If it is not, we are not obliged to allow for multiple conceptions of truth that, consequently, might reveal our “folk semantics” to be somehow impugned.19 Fortunately, in “Reality without Reference,” Davidson anticipates this sort of critique and dissolves the objection. He does so, as I will show, by exposing the desire for a theory of reference that justifies T-sentences as a throwback to the building block theory of meaning.
Interestingly, Davidson grants from the outset that T-sentences and the truth theory they rely on presuppose “a general and pre-analytic notion of truth.”20 It is this pre-theoretical notion of truth that allows us to grasp that T-sentences are in fact true. So there is at least one point of agreement between Stich and Davidson, insofar as they both grant that the truth of T-sentences is in some sense intuitive. Furthermore, Davidson grants, also like Stich, that T-sentences simply do not disclose much about reference.21 He remains uncertain, however, that this is in fact a problem. To see why, Davidson explores what he terms the paradox of reference. In trying to give an account of language and meaning, we have (crudely put) two options. The first, the building block theory of meaning, insists we start with an account of reference to generate the meanings of words and then of sentences. The second and more holistic approach begins with truth to give the meaning of sentences, which can then be used to determine the meanings of individual words (although it does not provide a theory of reference).
The first approach requires that we be able to give some non-linguistic characterization of reference (perhaps the causal-historical one), and it seems likely that something like this view of meaning is harmonious with Stich’s goal. He does grant that T-sentences do generate truth conditions – what he wants is an explanation of what it is to get them right, and an account of reference would provide just that.22 Insofar as sentence-based truth theories are silent on that issue, he thinks they are exposed to criticism of idiosyncrasy. Unfortunately, the reductive account of providing the meaning of sentences via their words has failed, which leaves the holistic or sentential account the only credible alternative. This is because if we want to give an account of the meaning of words, we must recognize that words appear only in (or sometimes are) sentences. Since we begin to interpret another’s unknown language from the standpoint of the radical interpreter (or at least we do in Davidson’s view), we simply do not know what counts as a word for the language under consideration. We must begin with sentences, from which we can move to words once we have constructed T-sentences and have an account of the truth conditions of the base axioms. Thus, we must turn to sentences first and observe the “extra-linguistic interests and activities that language serves.”23 Nonetheless, such sentences tell us nothing of what it is for individual words to refer – but to what extent is this a problem? For Davidson, it is not one at all; the key to recognizing this is to pay attention to the theoretical role that reference plays within a given theory of meaning, and not independently of, nor prior to, a theory of meaning.24
Simplistically, a theory of meaning, if constructed properly, tells us what sentences (and then words) mean. Within that theory, however, we can ask different questions about how and why the internal features of that theory work, how its parts function together, and more.25 But, “it makes no sense . . . to complain that a theory comes up with the right truth conditions time after time, but has the logical form (or deep structure) wrong. We should take the same view of reference.”26 While for Davidson reference is theorized about only within a theory of meaning,27 Stich wants to begin with it prior to such a theory’s construction – he has confused an internal question with an external one.
Davidson does not deny that there is some relationship between words and objects that might be best characterized by the term reference. The issue at heart is whether we can determine what that relationship is prior to understanding the role of the sentence that contains those words as uttered by a particular person. And that is precisely what we cannot do. When coming to interpret a speaker, we cannot know in advance what language she will speak or the idiosyncratic ways she might use her words. Therefore, to base T-sentences on an account of the reference of the terms they contain is to ask them to do the impossible: to provide the meaning of a word – which can be put to any use in any language – before knowing the truth conditions or meaning of its sentence. Unless Stich has an argument to suggest that this is in fact possible, we must concur with Davidson when he concludes, “we must give up the concept of reference as basic to an empirical theory of language.”28
At root, I think, the origin of Stich’s misstep is that he has demanded too much of truth (or at least of the conception of truth that is involved in T-sentences). He has incorrectly read into (semantic) truth the need for a deeper theoretical explication of what makes something true true – and he turns to an account of reference to do it.29 But for Davidson, the notion of something making something true is a secondary concern and does not precede a theory of truth. As such, in the sentence “ ‘grass is green’ is true iff grass is green,” the intuitive notion of truth that the biconditional captures is primitive, and it is a mistake to ask for an account of reference for the terms contained therein. I do not know how or why grass is green; nor is it obvious that I need to know to grasp the conception of truth the T-sentence expresses. T-sentences are intuitive because they do not rely on or require familiarity with a theory of reference to understand that they do, in fact, generate satisfaction conditions. If they did, it would demand too much. If, in order to know what others meant, we had to have a robust understanding of reference and truth, then it seems difficult to imagine how we could ever come to learn a language in the first place.
Indeed, both Tarski and Davidson share this non-epistemic interpretation of truth. Davidson states, “To know what it would be for a proposition to be true (or false), it is not necessary to be able to tell when it is true or false (much less know whether it is true or false). If the world will come to an instantaneous and unforeseen end, no one will or could know that it came to an end at that instant. This does not prevent our understanding the proposition that the world will come to an end at that instant.”30 Similarly, Tarski states:
In fact, the semantic definition of truth implies nothing regarding the conditions under which a sentence like [“snow is white”] can be asserted. It implies only that, whenever we assert or reject this sentence, we must be ready to assert or reject the correlated sentence [“the sentence ‘snow is white’ is true”].
Thus, we may accept the semantic conception of truth without giving up any epistemological attitude we may have had; we may remain naive realists, critical realists or idealists . . . The semantic conception is completely neutral toward all these issues.31
This is why, as alluded to earlier, I think it best to view the notion of truth as Tarski and Davidson discuss it as something more akin to a fundamental norm of rationality than a concept that requires epistemic justification in order to be correctly deployed. Davidson makes such an interpretation more plausible in light of his comment in “Truth and Meaning”: “I hope that what I am saying may be described in part as defending the philosophical importance of Tarski’s semantical concept of truth. But my defence is only distantly related, if at all, to the question of whether the concept Tarski has shown how to define is the (or a) philosophically interesting conception of truth, or the question whether Tarski has cast any light on the ordinary use of such words as ‘true’ and ‘truth.’ ”32 While truth remains, for Davidson, primitive, intuitive, and basic,33 its intuitiveness arises not from the pragmatic facility we have with our ordinary concepts – although this likely plays a part – but rather from its essential normative character that grants us a glimpse into how we might be able to make sense of our words meaning what they do – i.e., through the construction of T-sentences. It provides us with a means of distinguishing beliefs and so of specifying the meanings of utterances, and it does so without requiring any theoretical intermediaries (i.e., reference).
The above passage shows that Davidson does not engage in ordinary language philosophy; nor does he want to provide an explication of truth as it is used by the folk or even by epistemologists. Nonetheless, his notion of semantic truth remains primitive and intuitive. In some ways, this puts him beyond the reach of Stich’s criticism, as Stich primarily argues against “accounts that assign truth conditions largely compatible with commonsense intuitions,”34 and that does not seem to be quite what Davidson is doing (although it is not miles away, either). Yet, insofar as Davidson deems the semantic concept of truth “obvious,” “intuitive,” or any of other such cognates, he does seem to be within Stich’s crosshairs – it is just that “intuitive” for Davidson is not synonymous with “commonsensical,” if the latter is meant to capture what people commonly think and to derive justification from it.35 I do not know if the notion of semantic truth as Davidson uses it, in fact, identical to the common sense concept of truth; presumably this is an empirical question.36 I raise these issues solely to point out the threat of a possible equivocation that lurks in these discussions. Not all appeals to intuitions are appeals to folk intuitions, just as not all uses of “truth” are epistemic uses.
Finally, and in the spirit of the argument Davidson generates in “On the Very Idea of a Conceptual Scheme,” there is no reason to think that the semantics generated by non-intuitive T-sentences should remain untranslatable into our own language. We can capture them quite nicely by, instead of pluralizing truth, relativizing it to a person, language, and time:
7 “There is water on the Moon” is true as spoken by S1 at t1 in L1 iff there is H2O on the Moon.
8 “There is water on the Moon” is true as spoken by S2 at t1 in L2 iff there is XYZ on the Moon.
9 “There is water on the Moon” is true as spoken by S3 at t1 in L3 iff there are H2O and XYZ on the Moon.
Thus the intuitive notion of truth expressed by T-sentences is not so idiosyncratic. Properly relativized, it can provide interpretations for whatever theory of reference we might come up with. And so the need for an account of reference to ground truth dissolves, once we recognize that Stich’s desire for such an account of what it is for the axioms to be “right” runs in parallel with the desire for sentences to “fit” experience or the facts in giving an account of truth. In short, Stich endorses the second metaphor, which Davidson, in his argument against conceptual schemes, shows to be irrelevant in providing an account of truth. Moreover, Davidson’s view of truth has the theoretical advantage of being substantially simpler than embracing an infinitude of T-like sentences to capture multiple theories of reference. We need to choose among these possibilities in order to give a determinate meaning to the sentence, but we do not need to turn to reference to do so. Rather, we can appeal to the “human ends and activities”37 that the sentence is put to – we ought to turn to the speaker’s other utterances and actions.
After all, T-sentences alone do not generate meaning. They help set boundaries that a theory of meaning must meet. But we require evidence as well that a person’s vocalizations are genuine utterances – we need empirical data to flesh out our theory – and that, says Davidson, can only be arrived at through the application of the principle of charity. To assign determinate content to another’s utterances, we must locate those utterances in a network of actions and patterns of behaviour that are accessible to us in a shared public world. This gestures to the holistic nature of language and suggests that the meanings of utterances and the contents of beliefs are dependent on their relationships to other sentences and beliefs. We cannot isolate one sentence from the mix and give it determinate content without assuming an atomistic conception of language. Yet it seems that it is this conception of language Stich endorses when he suggests that the meaning of the word “water” could be determined by its referent alone and removed from the actual use of its speakers.
Of course Stich is right that we can ascribe to our speaker the beliefs that there is H2O, XYZ, and whatever else we might like, on the moon. But we should have the necessary evidence to do so, and that cannot come from a single sentence in isolation, let alone from a solitary word. Nor is it clear that a universally agreed upon theory of reference would provide it. After all, our speaker may idiosyncratically use the word “water” to mean something other than what we would use it to mean in our language. There is no way to determine what language our speaker speaks in advance, and so the fact that Stich bases his criticism on the vagaries of reference in the way that he does seems largely beside the point.
In conclusion, in Stich’s attempt to generate an objection to what he (rightly or wrongly) considers to be a largely a priori or intuitive understanding of truth, he makes a misstep. If Davidson is correct in subsuming a theory of reference under a larger and more central theory of meaning and truth, Stich’s move to pluralize reference in the hopes of pluralizing truth is conceptually backward, as we can theorize about reference only after we have a theory of meaning. As such, his plan to reveal the idiosyncrasy of truth and thus question its value can be seen as mistaken and unsuccessful. This leaves the intuitive, primitive, and, admittedly, modest account of (semantic) truth Tarski and Davidson champion untouched and safe from harm.
1 See, for example, Donald Davidson, “The Folly of Trying to Define Truth,” in Truth, Language, and History (Oxford: Oxford University Press, 2005), 19–38.
2 Stephen Stich, The Fragmentation of Reason (Cambridge, MA: MIT Press, 1990). Stich, of course, changed his mind about eliminativism in Deconstructing the Mind (Oxford: Oxford University Press, 1996) but remains committed to the belief that truth is limited and idiosyncratic – see his “Replies,” in Stich and His Critics, ed. Dominic Murphy and Michael Bishop (Oxford: Wiley-Blackwell, 2009), 190–252.
3 The canonical form of a T-sentence is “S is true if and only if p.” T sentences are important because the truth-functional connective (i.e., if and only if, or iff as some philosophers prefer it) that links S with p guarantees that the entire T-sentence is true just in case both S and p have identical truth values. Thus, we know exactly under what conditions S would be true (or false). When confronted with a foreign sentence, we may replace “S” with a description or name of S, because we are discussing the sentence and not uttering it ourselves, while “p” is replaced by S or a translation of S. So, for example, “ ‘la neige est blanche’ is true iff snow is white.” This allows us to see when the sentence “la neige est blanche” is true, and so we begin to get an insight into how knowing the truth conditions of a sentence can provide an insight into the meaning of that sentence.
4 For example, Peter C. Wason, “Natural and Contrived Experience in a Reasoning Problem,” Quarterly Journal of Experimental Psychology 23 (1971): 63–71 developed the selection task and Tversky and Kahneman developed the conjunction problem in Amos Tversky and Daniel Kahneman, “Extensional versus Intuitive Reasoning: The Conjunction Fallacy in Probability Judgment,” Psychological Review 90 (4, 1983): 293–315. Richard Samuels and Stich provide an excellent summary of this research in “Rationality and Psychology,” in The Oxford Handbook of Rationality, ed. Alfred Mele and Piers Rawlings (Oxford: Oxford University Press, 2004), 279–300.
5 As Stich defines it in The Fragmentation of Reason, 91, analytic epistemology “[denotes] any epistemological project that takes the choice between competing justificational rules or competing criteria of rightness to turn on conceptual or linguistic analysis.” As a chief example, he points to Alvin Goldman’s 1986 Epistemology and Cognition (Cambridge, MA: Harvard University Press, 1986).
6 In The Fragmentation of Reason, 15, Stich states: “So if the [conceptual] argument works, I suppose we would have to conclude that, contrary to appearances, descriptive monism and pluralism are not genuine empirical theses. This is, near enough, the view urged by Davidson in his argument against ‘the very idea of (alternative) conceptual schemes.’ ”
7 Davidson stresses this in “Could There Be a Science of Rationality?,” in Problems of Rationality (Oxford: Oxford University Press, 2004), 117–34, especially 128–34. In giving an account of rationality, by way of interpretation, we do not ask how people interpret others, or how people reason (although those are interesting questions), but rather what could make interpretation possible.
8 Donald Davidson, “On the Very Idea of a Conceptual Scheme,” in Inquiries into Truth and Interpretation, 2nd ed. (Oxford: Oxford University Press, 2001), 191–2.
9 I say “comes close” because while Stich states that Davidson is a defender of the minimal rationality hypothesis, he also suggests, “there are a number of passages in which Davidson sounds instead like an advocate of the perfect rationality view” (The Fragmentation of Reason, 41). In a way, Stich is right on both accounts as even Davidson’s minimal account of rationality demands a great deal.
10 In his final chapter in Fragmentation, Stich sketches an alternative account of cognitive evaluation that does not engage in conceptual analysis of such terms as justification, truth, and so on, but rather hopes to centre our evaluation of cognitive strategies on their success in meeting our goals. That he sees a priori arguments linking belief and rationality together as an obstacle to this end is evident from Fragmentation’s second chapter.
11 Davidson, “On the Very Idea of a Conceptual Scheme,” 194.
12 Stich, The Fragmentation of Reason, 102, emphasis original.
13 Ibid., 108.
14 Stich is not suggesting that he subscribes to the causal/historical account of reference (although he allows that it may correctly explicate the folk concept of reference). Indeed, he is entirely silent, at least in Fragmentation, on what theory of reference he does endorse (if any). Rather, his goal is to show that the best possible accounts of reference and truth are limited and idiosyncratic as, more than anything else, they are the products of intuition-based theorizing.
15 See for comparison Edouard Machery et al., “Semantics, Cross-Cultural Style,” Cognition 92 (2004): 1–12.
16 “There are indefinitely many possible patterns of formally specifiable causal interactions among mental sentences and thus indefinitely many possible mental sentence constructions, which admit of no intuitively plausible semantic interpretation at all. Most purely formal, syntactically characterizable patterns of interaction among sentences or well-formed formulas have no intuitively plausible semantics.” Stich, The Fragmentation of Reason, 113; emphasis original.
17 The asterisks here suggest that there might be other conceptions of truth that are unintuitive to us but nevertheless legitimate. We could use these “deviant” notions of truth to arrive at different interpretations for the sentence contained within the T-sentence in question.
18 Ibid.
19 I am not certain that designating Davidson’s program as a continuant of “folk semantics” is entirely correct. Although he does use concepts that derive from folk psychology (such as meaning, truth, and belief), to characterize his theory as a folk psychological theory seems unapt.
20 Davidson, “Reality without Reference,” in Inquiries into Truth and Interpretation, 223.
21 In “Reality without Reference, 218, Davidson writes: “There remains the claim, which I also grant, that the theory does not explain or analyse reference.”
22 In The Fragmentation of Reason, 112, Stich writes: “What Tarski did not provide was any general account of what it was for a name to denote an object or for a predicate to have a certain extension. Thus Tarski did not explain what it was to get the base clause lists right, and without such an account we could not even begin to apply Tarski’s strategy to a newly discovered language” (emphasis original). This suggests to me that what Stich is after is an account of reference that would precede an account of meaning or truth – we would analyze the concept of truth via an account of the concept of reference.
23 Davidson, “Radical Interpretation,” in Inquiries into Truth and Interpretation, 2nd ed., 127.
24 Davidson, “Reality without Reference,” 221.
25 For example, an internal feature of a theory of meaning would be a definition of the theoretical role that reference or satisfaction plays within that theory. But that is all there is to reference – it is a theoretical posit that gets its life from the larger theory of which it is a part (Davidson, “Reality without Reference,” 222). Davidson makes a comparison to physics and the role theoretical physical phenomena play in the larger physical theory – they are important, but not prior. Perhaps as an analogy, this might serve to make things clearer: we search for confirmation of the existence of quarks, neutrinos, and so on in the pursuit of explaining the macroscopic features of reality that we are confronted with. We do not begin with theoretical posits about sub-atomic particles, as such posits are unintelligible without the larger theory to provide them with content. Similarly, words, satisfaction, and reference only make sense against the background of a theory of truth (or of meaning).
26 Davidson, “Reality without Reference,” 223.
27 Or, perhaps, not at all. He does say, for example, “there is no unique reference of a singular term, nor unique extension of a predicate . . . Referring to a unique entity is not an intelligible task.” Donald Davidson, “Pursuit of the Concept of Truth,” in Truth, Language, and History (Oxford: Oxford University Press, 2005), 72.
28 Davidson, “Reality without Reference,” 221.
29 As noted above, Stich wants to know what it is for one to “get the base clause lists right.” The Fragmentation of Reason, 112, emphasis original.
30 Davidson, “The Problem of Objectivity,” in Problems of Rationality (Oxford: Oxford University Press, 2004), 9–10, emphasis original. This idea is expressed also in Davidson, “In Defense of Convention T,” in Inquiries into Truth and Interpretation, 70. The question is not, “What makes a sentence true?,” but rather, “What is it for a sentence to be true?” T-sentences are meant to help us explicate the latter question, while avoiding the former and the problems it gives rise to.
31 Alfred Tarski, “The Semantic Conception of Truth and the Foundations of Semantics,” Philosophy and Phenomenological Research 4 (3, 1944): 361–2.
32 Davidson, “Truth and Meaning,” in Inquiries into Truth and Interpretation, 24.
33 Cf. Davidson’s “The Folly of Trying to Define Truth,” in Truth, Language, and History, 19–38.
34 Stich, The Fragmentation of Reason, 106.
35 To take a cue from Richard Rorty, “Davidson between Wittgenstein and Tarski,” Crítica: Revista Hispanoamericana de Filosofía 30 (88): 49–71, it might be useful to stress the Wittgensteinian flavour of Davidson’s view here instead of the Tarskian influence. The shared judgments that we have about truth are where we reach bedrock and no further justification can be provided. That bedrock allows the generation of normativity. Whether people are cognizant of such a bedrock is irrelevant to its necessity.
36 Interestingly, in “The Semantic Conception of Truth,” 360, Tarski states that he thinks any given T-sentence “does conform to a very considerable extent with the common-sense usage.” He also discusses early research – although he, unfortunately, does not provide a citation – that finds 90 per cent of the folk “agreed that a sentence such as ‘it is snowing’ is true if, and only if, it is snowing.”
37 Davidson, “Reality without Reference,” 222.