Davidson argues that language is required for thought. His Master Argument posits that having a concept of belief is a necessary intermediary for having propositional attitudes, and that language is necessary for having a concept of belief. Of the various conceptions of “concept of belief” that might play a role in Davidson’s argument, the robust conception is too strong, and empirically falsified. While the robust conception may require language, we attribute propositional attitudes before children are clearly in possession of such concepts. Davidson’s examples and arguments support only deflationary interpretations of the concept of belief and the associated concept of objective truth: those that involve distinguishing between appearance and reality, or those that involve attributing mental content. However, as numerous studies in developmental and comparative psychology have shown, the deflationary conception is one that many creatures without language enjoy. Even an intermediate conception does not seem to play the role Davidson’s argument requires, for the ability to attribute mental content does not require language, and neither does the ability to attribute to others representational mental states, though here the evidence is less clear. Davidson’s arguments seem compelling because their plausibility relies upon a slide between less and more demanding conceptions of the concept of belief. For instance, a weak conception of “concept of belief” in M2 and a robust one in M3 yields an argument with apparently true premises, but because the argument equivocates on “concept of belief”, the argument is invalid. This analysis, as well as an appreciation of the methodological considerations for using non-linguistic behavior as evidence of propositional attitudes, supports the view that some mental states of non-linguistic animals can aptly be classified as propositional attitudes.
In empirical circles it seems to be taken for granted that at least some non-linguistic animals have mental states best described as propositional attitudes. But this acceptance is merely the first step in a larger project. For example, even if there is good reason to think that non-linguistic creatures have propositional attitudes, how they could have these remains to be elucidated. That is, what is the nature of the representational resources available to them? And given these representational resources, what sorts of contents are they capable of representing? What kinds of reasoning and inference could such representations support? What are the cognitive limitations necessitated by their representational architectures? One can begin addressing these fascinating questions empirically either at the functional psychological level or at the level of representation, and from either level one can work toward answering questions about the other.
Instead of thinking that language itself is what makes complex, structured, or propositional thought possible, we should consider: 1) how non-linguistic capacities could underlie complex representational abilities 2) the unique elements of linguistic competence and what they may or may not make possible vis-à-vis thought. In an example of the first, Proust (1999); see also this collection) provides an illuminating philosophical discussion of structured non-linguistic representational abilities (or “structured competences”) and how they could make possible objective representations. Structured representations as such could form the building blocks of propositional attitudes. Bermudez argues for abilities and for certain logical limitations on both the inferential and representational abilities of non-linguistic representers (Bermúdez 2003). Whether such limitations necessarily obtain is a matter of dispute (Lurz 2007).
When considering how linguistic abilities could augment thought, it is useful to identify elements of language that could contribute to representational complexity even if present without all the components of language. For example, Clark suggests that the human language-like ability to use symbols to represent abstract objects allows us to objectify our own thoughts and operate upon them (2000). Depending on what things can be symbolized, this could make possible metacognition or higher-order thought that might not otherwise be possible. Thus, the ability to represent symbolically can influence the kinds and complexity of reasoning available to a creature, even if that creature is not linguistic in Davidson’s sense. Symbolic capacities are necessary but not sufficient for linguistic competence, and could be present even when language is not. And if mere use of symbols is taken to be sufficient for language, then some nonhuman primates are capable of language and thus again can have propositional attitudes. Indeed, it is clear that some nonhuman primates can be trained to use abstract symbols, even if they do not do so naturally. Boysen and colleagues, for example, relate how naïve chimps fail to learn to make second-order generalizations about object classification, but those trained to associate objects with symbols (for relations of “same” and “different”) are able to succeed on a second-order classification task (Thompson et al. 1997). These interesting results give causal punch to the notion that symbolic objectification is a prerequisite for higher-level or abstract thought, and help to explain the competences that appear to come along with linguistic abilities.
Focusing less on the vehicles and more on the ways in which they can be exploited, Fitch and colleagues argue that recursion, which is a core element of natural language processing, can only operate on symbolic structures subject to rules, and that neither rules nor the objects on which they operate can exist without language-like representations (Hauser et al. 2002). If so, one might expect that forms of reasoning that rely on recursion may only be possible for creatures that also possess linguistic capacities. Thus, use of symbols and recursive rules are two candidates that could help explain the different representational capacities of linguistic and non-linguistic creatures.