A creature is said to have a propositional attitude when she stands in some appropriate relation (i.e., hoping, wanting, fearing, believing, etc.) to a proposition.[4] What propositional attitudes are, and who may enjoy them, may well be influenced by what one takes propositions to be. For instance, a skeptic about propositions may deny that anyone has propositional attitudes in the above sense. For our purposes it is not necessary to resolve questions about the nature of propositions, provided that we accept that humans can (and do) have propositional attitudes—meaning that there is something proposition-like to which a thinker can be appropriately related, whether this be a sentence (Fodor 1978), a set of possible worlds (Lewis 1979; Stalnaker 1984), or a state of affairs (Marcus 1990). What remains to be determined is whether appropriate relations to proposition-like entities can be supported in non-linguistic creatures.
Davidson’s arguments are offered in the context of his larger theoretical commitments to the nature of mind and meaning, commitments that stem from his interpretationist philosophy.[5] In general, interpretationist strategies answer the following three questions simultaneously: “In virtue of what does a creature have propositional attitudes?”, “which propositional attitudes do they have?” and “when is one justified in attributing these attitudes to a creature?” According to interpretationism, a creature has propositional attitudes in virtue of being interpretable; the most coherent, charitable interpretation that accurately (or accurately enough) predicts behavior is the justified interpretation; and the contents of that interpretation serve to determine the contents of the creature’s propositional attitudes. As Byrne puts it, in an interpretationist strategy, “there is no gap between our best judgments of a subject’s beliefs and desires and the truth about the subject’s beliefs and desires,” (1998). Thus, if a creature’s behavior can be accurately predicted or explained by an attribution of beliefs and desires in conjunction with the assumption of rationality, we are justified in attributing propositional attitudes to the creature.
Davidson’s strongest arguments for why thought requires language are motivated by his interpretationism. On a strict interpretationist view, meaning does not exist without interpretation; so if a system is uninterpreted, it lacks contentful states. Davidson believes that language is a prerequisite for entering the world of interpretation. If no language, then no interpretation, so no content. But let us consider, from an interpretationist stance, why one might think that language is a prerequisite for interpretation.
One might think that Davidson is moved by the idea that only linguistic behavior can be interpreted. However, this cannot be Davidson’s position. If it were, Davidson’s approach to propositional attitude attribution would be at odds with his own interpretive strategy for attributing content to mental states. The basic idea of Davidson’s interpretationism is that in ascribing content to another person’s mental states, we assume that that person is rational, and we ascribe content to her utterances, behaviors, and mental states in such a way as to maximize the coherence of that person’s beliefs and desires in light of her behavior. Undeniably, there is a class of behaviors that humans have and animals lack, namely linguistic behaviors. However, both humans and animals share a wide range of non-linguistic behaviors that admit of interpretation. On the face of it, those behaviors provide ample evidence upon which to base attributions of mental content, and Davidson himself would not refuse to attribute propositional attitudes to a silent person. However, Davidson pointedly refuses to apply a straightforward interpretationist strategy to non-linguistic animals. To avoid arbitrariness, an independent argument is needed to privilege language over other behaviors.
Perhaps Davidson believes that rationality is impossible without language. If we cannot attribute rationality to a creature, the interpretationist strategy does not apply. More than a few people have argued that animals are not rational, yet there is reason to believe, under some plausible construals of rationality, that they are. To hold that rationality presupposes language commits one to a narrow view of rationality, already colored by a linguistic bias. Such a view implicitly begs the question in which we are interested. Admittedly, what rationality is is a vexed question in philosophy, and determining whether a creature is rational falls prey to the same holistic problems as determining whether it has propositional attitudes. A theory of rationality predicated upon a conception of practical reason instead of upon linguistic manipulation appears to be more neutral. There is abundant evidence for practically rational behavior in the animal world. After all, animals of all stripes are here now because they have been evolutionary successful, and to have succeeded requires in some nontrivial sense that goals are achieved by instrumental behavior.[6] All animals exhibit some degree of rationality, construed in this way. Building on this view of rationality promises to enable us to posit criteria or hallmarks for minimally rational behavior that are independent of language, yet also to concede that some rational behaviors are linguistically dependent, and thus unique to humans. Indeed, one might think that a good way to assess rationality would be to see to what extent an animal’s behavior is predictable or explicable with reference to survival requirements and common sense belief–desire psychology. A wide range of animal behaviors certainly seem apt for explanation with reference to the rational interplay of ecologically-relevant propositional attitudes. If one thinks that aptness for explanation in terms of rationality is sufficient evidence for rationality, and accepts, as Davidson does, that rationality rests on the interplay of propositional attitudes, then we have ample evidence that animals have propositional attitudes, rather than that they do not.
Davidson, however, obviously thinks that the reasons to deny animals propositional attitudes supersede reasons to attribute rationality to them; he applies modus tollens to my modus ponens. Since he denies that animals have propositional attitudes, and he thinks rationality requires propositional attitudes, he denies that animals are rational. We are led to very different conclusions about the nature of animals’ mental lives depending upon whether we take ourselves to be more justified in attributing rational behavior to them or in refusing to attribute to them propositional attitudes. Because the questions of propositional attitudes and of rationality are both equally troubling and closely linked, arguments against animal thought based on assumptions about rationality are not compelling.
Thus, we have as yet failed to find ample reason to refuse to apply the basic interpretationist strategy to non-linguistic animal behavior. Perhaps Davidson thinks that, in the absence of language, we have insufficient evidence for attributing propositional attitudes to animals. Perhaps it is because Davidson thinks that “having the gift of tongues” is both necessary and sufficient for having propositional attitudes (1984, p. 156, 2001a, p. 104), he views language possession as the evidential criterion for propositional attitude attribution. He consequently denies that we can be justified in attributing propositional attitudes to creatures on the basis of non-linguistic behavior. Language gives the radical interpreter the green light: evidence of linguistic behavior licenses application of the radical interpretive strategy.
Even if we grant that language is the best evidence for propositional attitudes, we should be immediately suspicious of the presumption that the only evidence relevant for deciding whether something has propositional attitudes is the presence of a necessary and sufficient condition for having them. In normal empirical inquiry, criteria that are necessary and sufficient are rarely the only ones that qualify as evidence for assessing empirical claims. For instance, a rash may be relevant evidence for determining whether a person has Lyme disease, despite the fact that not all people with rashes have Lyme disease, and not all people with Lyme have rashes. Might there not be evidence highly indicative of whether a creature has propositional attitudes, despite the fact that the evidence is not decisive? Reasonable, predictable behavior is surely a clear source of evidence for the existence of propositional attitudes, despite the fact that it only provides defeasible reasons for thinking they exist.
Furthermore, unlike instrumentalists like Dennett, Davidson seems to favor the idea that beliefs are real; his anomolous monism posits a physical-causal substrate for mental states, albeit one that exempts psychology from being reduced to physical laws.[7] One might think, nonetheless, that it would be reasonable for a realist to accept the possibility that beliefs involve some internal representational structures, and that there could therefore be other types of reliable evidence besides linguistic evidence for the presence of propositional attitudes. Thus, Davidson’s exclusive focus on language is in tension with his realist leanings. Furthermore, if one is a realist about thought, it is not the evidential question, but rather the question of the grounds of possibility for having propositional attitudes that should be of primary interest. The Davidsonian mix of interpretationism and realism creates an uneasy tension, for while he tends toward realism about belief, he often seems to think the metaphysical and epistemological construals of the question amount to the same thing: a creature has propositional attitudes if we ought to interpret him as having them. I suspect that this collapsing of the issues accounts for Davidson’s view that the question of whether a creature has propositional attitudes is closely tied to the evidential question of what evidence is relevant for deciding whether something has propositional attitudes.
There is, as fard as I can tell, a lack of a substantive argument for requiring that a creature has language to be a candidate for interpretation, as well as for holding that only the presence of language provides sufficient evidence for attributing propositional attitudes. Thus, neither the interpretationist strategy itself, nor Davidson’s concerns about evidential warrant justify the position that only language-speaking creatures can be candidates for propositional attitudes. Now let us turn to the specific arguments Davidson offers for denying animals propositional attitudes: the reasons he offers for holding that language is necessary for thought.