Can There be Thought Without Words?—Donald Davidson on Language and Animal Minds

Abstract

In a couple of short papers, Donald Davidson holds that a creature cannot think unless it is the interpreter of the speech of another. At first blush, speaking a language is, therefore, a necessary condition for thought. His controversial claims has led many to regard him as a follower of the Cartesian tradition wherein languageless creatures are nothing but mindless machines. Against this widely shared interpretation, in this paper we put forward a more charitable interpretation of Davidson’s claims. According to our reading, Davidson never meant to argue that languageless creatures do not think. Instead, the only thing his arguments purport to show is that one will never be in a position to confirm that they do. This paper consists of a defense of the idea that Davidson is better seen as endorsing radical skepticism as to whether languageless creatures think.

“But the principal argument to my mind, which may convince us that the brutes are devoid of reason is that […] although all of them make us clearly understand their natural movement of anger, of fear, of hunger, and others of like kind, either by the voice or by other bodily motions, it has never yet been observed that any animal has arrived at such a degree of perfection as to make use of a true language; that is to say, as to be able to indicate to us by the voice or by other signs, anything which could be referred to thought alone, rather than to a movement of mere nature; […] now all men […] make use of signs, whereas the brutes never do anything of the kind; which may be taken for the true distinction between man and the brute.” (Descartes 1649, quoted in Fellows 2000, pp. 589–590).

1 Introduction

Noam Chomsky (1976) once said that our ignorance can be divided into problems and mysteries. When facing a problem, although we may not know how to resolve it, we can at least get an inkling of how it might be resolved. In contrast, when faced with a mystery, there is no clue about what is going on and we can only wonder as to the nature of its explanation. The question of how we can come to know the contents of the minds of others, however, seems to be neither a problem nor a mystery. If you feel curious about what people are thinking, their opinion about Scorsese’s most recent movie, or their political views, among other things, you can ask them, read what they have written on the subject, or observe their behavior. Based on all the available evidence, you will likely be in a position to explain and predict a great deal of their future behavior. To a certain extent, it is beyond doubt that all of us somehow have privileged access to the contents of our own mental states. However, when it comes to the minds of others, language is surely what helps us dispel the mystery regarding their contents in the case of those with whom we can communicate.

Further complexities therefore inevitably enter the scene when our attention shifts towards the mental contents of languageless creatures. Putting philosophical concerns aside, most of us would certainly agree that at least some pre- and non-linguistic creatures, such as human babies, apes or dogs, can think. After all, raise your hand all those who are not keen to concede that Normal Malcolm’s (1972–3) dog, which relentlessly barks while staring at the top of the tree in the backyard, thinks that the cat he was chasing may be hidden in its branches. Notwithstanding this belief and however plausible it might be, push really comes to shove when we are asked to justify our affirmation that languageless creatures do (or do not) think.

This paper is devoted to the language–thought relation. What is, after all, the relation between language and thought?¹ Roughly, three different answers can be found in the literature:² lingualism, mentalism, and an intermediate position between the two (for short, we will call it the “intermediate position”). At the root of lingualism lies the belief that thought depends on language, so a creature necessarily lacks thoughts if it lacks language. Davidson, along with other philosophers such as Dummett, is undeniably the best-known contemporary advocate of this view.³ In contrast, the proponents of the opposite view, mentalism, hold that thought does not depend on language, thereby conceding that languageless creatures can potentially think, even though their thoughts might differ in degree from those possessed by human adults (e.g., Peacocke 1992). Lastly, those who embrace the intermediate position accept that languageless creatures can have thoughts and concepts, although these thoughts differ in kind from those possessed by human adults (e.g. Kenny 1989; DeGrazia 1996). In what follows, we focus exclusively on Davidson’s view concerning animal minds. Admittedly, this is not a novel issue since Davidson’s arguments on this topic have been the subject of hot debate over the years; but his texts undeniably constitute one of the most meticulous examinations of the language–thought relation one can find in the literature, which make them still worth discussing.

Davidson is arguably seen as a follower of the Cartesian tradition according to which a creature lacks thoughts if it lacks language. His controversial claims that a creature cannot think unless it is the interpreter of the speech of another (1975, p. 157) and that neither a one-week-old infant nor a snail are thinking creatures (1985, p. 92) have received sharp criticism that calls into question at least one of his arguments (see e.g. Bishop (1980); Carruthers 2008; Tye 1997; Lurz 2009b, pp. 7–8; Glock 2000; Beisecker 2002; Searle 1994, Sect. 3; Jamieson 2009, Sect. 3.1; Horta 2010; Fellows 1998).⁴ Against this backdrop, our purpose here is twofold: first, we aim to show that, although not utterly mistaken, the view that sees Davidson as a proponent of lingualism is misplaced; and second, we put forward an alternative, more charitable reading of Davidson’s arguments which does not commit him to lingualism. If our reasoning is sound, Davidson endorses radical skepticism as to whether languageless creatures literally think.

To this end, the rest of this paper is structured as follows. In Sect. 2, we introduce the pressing question at issue when it comes to justifying whether languageless creatures think. Then, in Sect. 3, we summarize Davidson’s views on language, and in subsections 3.1, 3.2 and 3.3. we briefly recap his three related arguments supposedly in favor of lingualism.In Sect. 4, we put forward a more charitable reading of those arguments that does not commit him to lingualism. Lastly, in Sect. 5, we consider a possible objection steaming from our suggested reading that seriously threatens Davidson’s whole theory of mind, and in Sect. 6 we show that Davidson’s philosophy is not in fact undermined by it. The paper ends with the conclusion that we are better off seeing Davidson as endorsing skepticism as to whether languageless creatures think than as endorsing lingualism.

2 Setting the stage

Do languageless creatures think? ⁵ Here is how Jamieson 2009 addresses this problem, which will prove to be interesting for our current purposes. There are two powerful assumptions that may be at work when it comes to determining whether languageless creatures think or not:

• (a) Many languageless creatures think.

• (b) What languageless creatures think cannot reliably be characterized.

Assumption (a) is typically grounded in the fact that many creatures manifest goal-directed behavior, so belief ascriptions successfully account for their behavior. Think, for instance, of Fido who starts frenetically wagging his tail when his owner gets home. A plausible explanation for Fido’s behavior may include that he believes that his owner is home. In such a scenario, we ascribe to Fido the belief that his owner is home, and such belief ascription partly explains his behavior. Following what appears in the literature, we will call this the predictive utility of thought ascriptions:

• (c) Predictive utility of thought ascriptions: Belief ascriptions are pragmatically justified as long as they successfully account for a creature’s behavior.

It is important to distinguish assumption (a) from the explanatory role of (c). This distinction is clearly made by Fellows in the very first paragraph of his 2000 paper:

Can non-language-using animals have thoughts? Can a cat, for instance, literally believe that there is a mouse in the house? The question has been a persistent one in the Western philosophical tradition. But this question needs to be distinguished from another question, which is: ‘Is there predictive utility in ascribing thoughts to creatures which lack a language?’ An affirmative answer to the latter question is sometimes taken to provide evidential support for an affirmative answer to the firmer, but I shall argue below that this is not correct. (Fellows 2000, p. 587)

In other words, the difference can be stated as: (a) suggests that languageless creatures do literally think insofar as our belief ascriptions successfully account for their behavior, whereas (c) simply states that the ascription of beliefs to languageless creatures is merely pragmatically justified insofar as they successfully account for a good deal of the behavior of those creatures. Unlike (a), (c) on its own says nothing about whether or not languageless creatures literally think.

Before moving on to the next section, let us make it perfectly clear that our only concern here is how to justify the claim that languageless creatures have (or do not have) thoughts. Thus, the empirical question of whether or not languageless creatures literally think lies out of the scope of this discussion. It is crucial to bear this in mind in order to understand much of what we are going to argue in the rest of this paper. Davidson’s arguments are typically seen as entailing an outright denial of (a); yet we believe that there is a more charitable construal of those arguments that lead us to a slightly different conclusion. Before diving into that issue, next, I briefly recap Davidson’s arguments.

3 Davidson on language and animal minds

The notion of interpretation plays a pivotal role in Davidson’s philosophy. Interpretation, as he sees it, is intimately linked to the notion of rationality and, by extension, to the notion of mind. Minded creatures are rational creatures—or, speaking in his own words, rational animals (Davidson 1982)—and rational animals are those we can interpret. Davidson was highly selective as to what is a rational creature. As we have said above, thought and language are interdependent, which suggests that there is no thought—or mind—unless there is language. As far as the argument goes, only linguistic creatures are rational—or minded—creatures.

Minds are essentially public and, therefore, publicly accessible from a third-person viewpoint. Likewise for language and meaning: the meaning of words is not inside our heads. Rather, it must be understood in relation to external objects and events. For Davidson, there are fundamental connections between mind, a creature’s linguistic and non-linguistic behavior, and the external world, which means that the mental contents of a creature, as well as the meanings of its words, are to be found outside, in a single external and shared world.

Davidson’s views on language and mind reflect his skeptical stance towards the scientific study of the brain. Unlike other philosophers such as Chomsky for whom the analysis of the syntactic structure of language must be carried out by cognitive and biological sciences, Davidson believes that the study of the brain is irrelevant to understanding language and mind. Both Davidson and Chomsky hold that the task of explaining linguistic behavior of creatures is an empirical one⁶ (see Knowles 2015). However, when it comes to linguistic understanding, Davidson shifts his attention to communication via radical interpretation. In this respect, Davidson coincides with Dennett for whom “the rational for natural language was communication”, and contrasts with Fodor’s language of thought according to which language is an internal articulation of thought (see Miguens 2021, this issue). Davidson’s emphasis on radical interpretation steams entirely from his skepticism about the traditional and commonsensical understanding of language as something shared by a community of speakers. Davidson’s “A Nice Derangement of Epitaphs” ends with the provocative statement:

there is no such thing as a language, not if a language is anything like what philosophers and linguists have supposed. There is therefore no such thing to be learned, mastered, or born with. We must give up the idea of a clearly defined shared structure with language-user acquire and then apply to cases. (Davidson 1986, p. 107).

Evidently, Davidson does not deny that there are natural languages such as French, Galician or English, but he believes that natural languages in this sense are of no help in providing us a philosophical understanding of language: “people don’t need to speak the same language in order to understand each other” (Davidson 1998, p. 323; Davidson 1991, p. 210). In order to understand each other, it is not enough to mean the same by the same expressions. Rather, speakers must ascribe the same meaning to the speaker’s words.

The theory of (radical) interpretation is thus at the core of Davidson’s philosophy (see Davidson 1973). It is concerned with the question of what a speaker must know that will enable her to interpret the utterances of another. Tarski made important contributions to set this theory by showing us how an extensionally correct definition of a truth predicate can be given to formal languages. Drawing on Tarski, Davidson held that a semantic theory for natural languages could take the form of a Tarski-style truth theory (“T-theory”, for short), which would specify the conditions under which the speaker’s utterances are true. Therefore, in order to interpret the speaker’s utterances of a given language L, the interpreter is supposed to construct a T-theory for L. The radical interpreter is thus the hero of Davidson’s philosophy (Glüer 2011, p. 4).

The central tenet behind radical interpretation is that language and belief are interdependent. Broadly speaking, this means that in order to assign meanings to a speaker’s utterances, the interpreter must ascribe beliefs to the speaker while, at the same time, in order to ascribe beliefs to the speaker, the interpreter must be able to assign meanings to a great deal of the speaker’s utterances. For instance, if a speaker of an alien language utters “Gavagai” while pointing to a rabbit in the grass, the interpreter would ascribe him the belief that there is a rabbit in the grass. In doing so, the speaker would simultaneously ascribe meaning to the speaker’s word (for instance, “Gavagai” means “Rabbit”). Belief and meaning go hand in hand, and a creature is a thinking creature if it can triangulate (Davidson 1974, p. 154; 1997a).

Davidson’s triangulation is central to his views on thought and language. Roughly, triangulation is the thesis that meaning and mental contents can be fixed and determined only on the basis of a “threefold interaction, an interaction which is twofold from the point of view of each of the two agents: each is interacting simultaneously with the world and with the other agent” (Davidson 1997, p. 128; see also Davidson 1997; 1991, p. 213; 1988, p. 44–45 for more on this). It is this three-fold interaction what makes interpretation and language learning possible. It may be worthwhile quoting Davidson at length here:

[T]he learner is rewarded, whether deliberately or not, when the learner makes sounds or otherwise responds in ways the teacher finds appropriate in situations the teacher classes together. The learner is subsequently caused to make similar sounds by situations the learner instinctively classes together. Corrections are possible, of course. Success at the first level is achieved to the extent that the learner responds with sounds the teacher finds similar to situations the teacher finds similar. The teacher is responding to two things: the external situation and the responses of the learner. The learner is responding to two things: the external situation and the responses of the teacher. All these relations are causal. Thus the essential triangle is formed which makes communication about shared objects and events possible. But it is also this triangle that determines the content of the learner’s words and thoughts when these become complex enough to deserve the term. The role of the teacher in determining the content of the learner’s attitude is not just the ‘determine’ of causality. For in addition to being a cause of those thoughts, what makes the particular aspect of the cause of the learner’s responses the aspect that gives them the content they have is the fact that this aspect of the cause is shared by the teacher and the learner. Without such sharing, there would be no grounds for selecting one cause rather than another as the content-fixing cause. A noncommunicating creature may be seen by us as responding to an objective world; but we are not justified in attributing thoughts about our world (or any other) to it (Davidson 1990, p. 203).

Nevertheless, triangulation is not only a story about how children learn native language. Most importantly, it is a theory which purports to explain how words refer and acquire meaning (see Davidson 1988, pp. 44–45). If meanings of our words and the contents of our thoughts were nothing but objects inside our heads or mere neural stimulations in our brains, then no one would ever be in a position to know what we mean by what we say or what we think⁷. Meanings and thoughts would not be publicly accessible and, hence, interpretation—and communication—would not be possible: “Until a base line has been established by communication with someone else, there is no point in saying one’s own thoughts or words have a propositional content” (Davidson 1991, p. 213). Yet, since it is a fact that people communicate and understand each other, meanings and thoughts must be accessible to the interpreter: “what makes the distal [or external] stimulus relevant determiner of content is […] its social character” (Davidson 2001, p. 10).

Ultimately, it is the rationale behind the triangulation argument what underlies Davidson’s famous—and infamous!—claim that “a creature cannot have thoughts unless it is an interpreter of the speech of another” (1975, p. 157). If taken literally, Davidson’s words entail that a creature that does not speak cannot think. For if there is no other way of obtaining the contents of a creature’s beliefs than interpreting its speech, then there are no grounds for ascribing beliefs to languageless creatures.

The following three subsections are devoted to a brief description of Davidson’s three related arguments that allegedly support lingualism. We will not, however, spend much time expanding on these arguments, and less till assessing the soundness of their conclusions. A more precise and thorough analysis and discussion of Davidson’s arguments can be found in Glock (2000).

3.1 Intensionality of thought and de dicto belief ascriptions

A curious feature of belief contents is that they yield referentially opaque or intensional contexts where co-referential terms cannot be substituted salva veritate. According to this fine-grained conception of belief, the belief that Superman caught Lois Lane is not the same as the belief that Clark Kent caught Lois Lane, even though both beliefs are about the same object. Belief ascriptions in this sense aim to describe the way a creature thinks or conceives of some object. In terms of belief ascriptions, this semantic feature may lead us from a true belief ascription to a false one (e.g., since someone can rationally believe that Superman caught Lois Lane while failing to believe that Clark Kent caught Lois Lane, the belief-sentence “Michael believes that Superman caught Lois Lane.” may be true while the belief-sentence “Michael believes that Clark Kent caught Lois Lane.” may be false.)

The problem with ascribing de dicto beliefs to languageless creatures arises due to the fact that “unless there is behaviour that can be interpreted as speech, the evidence will not be adequate to justify the fine distinctions we are used to making in the attribution of thoughts” (Davidson 1974, p. 164). To illustrate the case, recall the previous example of Fido, who starts wagging his tail frenetically when his owner gets home. Part of a plausible explanation for Fido’s behavior may be that he believes that his owner is home. However, it turns out that Fido’s owner is also the President of the USA. Hence, if we are to ascribe to Fido the belief that his owner is home, then, in a manner of speaking, we could also ascribe to him the belief that the President of the USA is home, or that the President of the USA is in the White House, or that Joe Biden is in the White House, or even that the tallest man in the neighborhood is in the official residence of the President of the USA, and so on so forth. But are we really to ascribe to Fido the belief that Joe Biden is in the White House, or that the President of the USA is home? The problem with ascribing beliefs to languageless creatures is that there are countless systems of beliefs that could successfully explain their behavior and no way of deciding which is the correct one, viz., which is the one that would rightly describe the way the creature thinks or conceives of a certain object. The upshot is that our de dicto ascriptions of belief to languageless creatures are unwarranted, or so Davidson argues.

3.2 Holism and de re belief ascriptions

Davidson’s second argument relies on the holistic nature of thought, viz., the view that the content of a belief can only be determined in connection to a wide web of other related beliefs. It rests on the assumption that in order for a creature to have a belief, say a belief about trees, it must have a large stock of tree-related beliefs, such as trees are not flying creatures, they have leaves, grow large roots that allow them to take up water and nutrients from the soil, and so on. There is no fixed list of beliefs a creature must possess for it to have a belief about trees, but unless it has a general web of other beliefs, there is no reason to identify a given belief as a belief about trees (Davidson 1982, p. 98). Applying the same rationale to our example, if we are to ascribe to Fido the belief that his owner is home, then we must credit him with a wide range of other related beliefs, such as owners are individuals who own pets and take good care of them, home is the place where people usually live, and so on. However, unless the creature is able to speak, we will not be in a position to tell whether it has the necessary beliefs that would make sense of the initial belief (Davidson 1985, p. 98). As a consequence, our de re belief ascriptions to languageless creatures are unjustified, Davidson concludes.

There is a principled, and not merely a practical, obstacle to verifying the existence of detailed, general and abstract non-linguistic beliefs and intentions, while being unable to tell what a speaker’s words mean. We sense well enough the absurdity in trying to learn without asking him whether someone believes there is a largest prime, or whether he intends, by making certain noises, to get someone to stop smoking by that person’s recognition that the noises were made with that intention. The absurdity lies not in the fact that it would be very hard to find out these things without language, but in the fact that we have no good idea how to set about authenticating the existence of such attitudes when communication is not possible. (Davidson 1974, pp. 143−144)

3.3 The concept of belief

Davidson’s last argument focuses on the conditions for belief. It rests on the following two assumptions that, as he (1985, p. 102) concedes, can be called into question:

• (i) The possession of beliefs requires the possession of the concept of belief.

• (ii) The possession of the concept of belief requires the capacity to speak a language.

From (i) and (ii) Davidson infers that languageless creatures cannot have the concept of belief and, therefore, cannot have beliefs.

The concept of belief plays a central role in Davidson’s overall argument. By concept of belief, he refers to the capacity of grasping the concept of objective truth, viz., the capacity of grasping the contrast between what objectively is the case and what one actually thinks the case to be (1982, p. 104). To put it briefly, the main idea behind the argument is that since beliefs have truth-values, a creature cannot have beliefs unless it is able to recognize that it can be the case that its beliefs be false.⁸ But this capacity requires the possession of the concept of objectivity which, according to Davidson, can only be reached via triangulation—that is, through the creature’s capacity to contrast its beliefs with the beliefs of others. Triangulation of this sort, however, demands that the creature possess the ability to ascribe beliefs to another, which—as we have seen at the beginning of Sect. 3—requires language (see e.g. Davidson 1982, p. 105). The upshot of the argument is that since languageless creatures do not have the concept of belief, they do not have beliefs.

4 An alternative reading of Davidson’s arguments

These three arguments of Davidson’s are typically judged as a defense of lingualism. To recall, the argument from intensionality shows that our de dicto ascriptions of belief to languageless creatures are not compelling, since fine distinctions between thoughts cannot be made in the absence of language. In the same vein, the argument from holism leads to the conclusion that our de re ascriptions of belief to languageless creatures are dramatically indeterminate, because in the absence of language a creature can be assigned as many systems of beliefs as we can imagine, thereby leaving us with no grounds on which to tell whether it is manifesting intentional behavior or merely a causal disposition. In his last argument, the argument from the concept of belief, Davidson lays out the conditions for belief, which rest on the assumptions that (i) the possession of belief requires the concept of belief, and (ii) the possession of the concept of belief requires language. From (i)-(ii) we end up with the conclusion that languageless creatures cannot have beliefs. In summary, belief ascriptions to languageless creatures, despite successfully accounting for the behavior of the creatures, fall short of demonstrating that they literally have beliefs.

In Sect. 2, we said that determining whether or not languageless creatures think is not as simple as it may seem. To recall, the root of the predicament lies in the following assumptions:

• (a) Many languageless creatures think.

• (b) What languageless creatures think cannot reliably be characterized.

Assumption (b) is quite often taken as a strong reason against (a). Ultimately, this is the reasoning those who regard Davidson as an advocate of lingualism endorse. Admittedly, there does seem to be a good point here. Davidson clearly states that due to their lack of language, exactly what languageless creatures think—if they do—will remain unknown and, consequently, we will never be in the position to distinguish between any of the countless systems of beliefs that successfully fit their pattern of behavior.

Nevertheless, we believe that this construal of Davidson’s arguments may be too hasty. In the rest of this paper, we pursue a different line of interpretation that suggests that in spite of endorsing (b), Davidson never meant to undermine (a). Rather, we believe the only thing his arguments purport to show is that our belief ascriptions to languageless creatures are pragmatically justified at best, if they successfully account for the behavior of those creatures. Yet, that a creature’s behavior is pragmatically justified does not count as a reason in favor of (a): pragmatic justification does not count as evidence that a languageless creature literally thinks, because in the absence of verbal responses we can never abandon (b). Therefore, due to the dramatic indeterminacy in the explanation of the creature’s behavior, we will never be in a position to distinguish between any such belief ascriptions. This reasoning does not amount to an outright denial of (a). As we see it, it commits Davidson to skepticism regarding animal minds more than to lingualism.

A terminological clarification must be given here, concerning our use of the expressions “pragmatic justification”, “justification”, and “explanation”. To explain something is to make it comprehensible. Therefore, when we explain a creature’s behavior by ascribing mental states to it, we are making its behavior intelligible in light of those mental states. In turn, to justify something is to provide a suitable explanation for it. For instance, when ascribing the belief that his owner is home to Fido, we are partly explaining Fido’s behavior in light of that belief content. In addition, this belief ascription is also justified in a practical sense provided that it successfully explains Fido’s behavior. However, such ascription does not make it plain that Fido literally believes that his owner is home, and that is why it is merely pragmatically justified. Belief ascriptions to languageless creatures are pragmatically justified when, even though they successfully account for the creature’s behavior, the same behavior could be explained by employing non-intentional terms. We tend to account for their behavior using intentional vocabulary because we might not have a better way to do so. Thus, we ascribe beliefs to languageless creatures to explain their behavior just as we could ascribe beliefs to heat-seeking missiles or thermostats. In other words, the explanations used to account for the behavior of languageless creatures need not appeal to normative concepts, such as the concept of reason. Conversely, psychological explanations of the behavior of linguistic creatures are justified in a way that outstrips a practical sense because the mental terms we ascribe to them to make their behavior intelligible cannot be reduced to or replaced by non-intentional terms. This is mainly why ascriptions of belief to linguistic creatures are justified in a way that demonstrates that these creatures literally have those beliefs, at least from a Davidsonian point of view. The explanation of the behavior of linguistic creatures is normative and necessarily invokes the concept of reason. (Of course, all this may seem to entail verificationism, as many have contended (see e.g. Williamson 2004; Jamieson 2009): we will not, however, tackle this objection here.)

Many passages can be found in Davidson’s writings in support of our reading; and we believe they have been overlooked by some of his interpreters. Before considering them let us first take a quick look at a possible counterexample to one the arguments previously introduced. It will help us understand better the significance of the argument from the concept of belief.⁹

It could be said that if languageless creatures are capable of discriminating between objects in their surroundings, then this suffices to justify ascriptions of beliefs to them. For such creatures to do this, they need not recognize objects under a given description, since the discrimination could be achieved in virtue of perceptual features of the object. For instance, Fido can believe that his owner home and be able to discern his owner from other individuals or objects in his surroundings, without believing or knowing that his owner is the President of the USA or, so to speak, without having the concept owner or other related concepts.

Granted, this is a strong argument, and it is not our intention to debunk it here. Likewise, there might be some truth in the claim that thought does not require language; we do not want to make a case for this claim here either. What we do think, however, is that those who adopt such a line of argument against Davidson are missing the point. For one thing, the issue is not whether the creature is capable of discriminating between objects in virtue of their perceptual features or under certain descriptions, but the nature of that discrimination. Belief ascriptions need to be normative or intentional in order to be justified. That is, the creature must have a reason to act as it does or be able to do otherwise. The possession of the concept of belief is called for because it introduces the normative force that allows the interpreter to determine when the i creature is acting for a reason or simply discriminating, and hence merely acting as if it were for a reason: “to have a belief it is not enough to discriminate among aspects of the world, to behave in different ways in different circumstances; a snail or a periwinkle does this” (Davidson 1991, p. 209; see also, e.g. Davidson 1997b, p. 139). Discrimination of this sort is performed even by plants and heat-seeking missiles, and can be explained (at least in part) as a certain causal relation to a stimulus without the need to invoke propositional content. Normativity emerges on the linguistic level, or so Davidson goes on to argue (see Davidson 1997a).

Some may still resist Davidson’s approach. After all, some empirical data suggest that social learning has a significant impact on vervet monkeys, for example. Young vervet monkeys often make classification mistakes when producing an alarm call and are subsequently corrected by adult vervet monkeys. On the face of it, this apparently corrective behavior seems to allow young members of the group to improve their performance¹⁰ (see e.g. Cheney & Seyfarth 1980). Moreover, the behavior of plants or heat-seeking missiles is clearly far outstripped both in complexity and unpredictability by the behavior of vervet monkeys and many other creatures, and therefore they should not be judged on the same basis.

At first blush, the vervet monkeys’ behavior looks like a primitive or simple awareness of error that may be seen as a pre-linguistic distinction between truth and falsehood which does not require the possession of concepts (for more on this, see e.g. Beisecker 2002, p. 118; Macintyre 1999, p. 16). Surprisingly, Davidson does not explicitly deny this (see Davidson 1982, p. 105). The issue, however, is not that simple because, even though the creature’s behavior looks like self-corrective behavior from our point of view, we will never be in a position to ensure that it is in fact an instance of such behavior from the monkey’s point of view. Unless a creature is capable of recognizing its error itself, there are no grounds on which to determine whether its behavior is causal or intentional. In other words, normativity is not in place when language is not present and, consequently, there is no reason to hold that it is in fact self-corrective behavior rather than a causal reaction to a stimulus.

Corrections, whether administered by teacher, parent, playmate, or nature, can in themselves do no more than improve the dispositions we were born with, and dispositions, as Wittgenstein emphasized, have no normative force. […] To correct behavior is not, in itself, to teach that the behavior is incorrect. Toilet training a child or a dog is like fixing a bathtub so it will not overflow; neither apparatus nor organism masters a concept in the process. (Davidson 1997b, pp. 138–139)

Returning to the case of the vervet monkeys, we feel tempted to account for their behavior by appealing to the presence of eagles in the surroundings, because this seems to be the most plausible explanation from our point of view. Yet we cannot be sure that believing that there is an eagle in the surroundings is exactly what is going on inside the heads of the monkeys, so to speak. Even though the youngest monkeys seem to be corrected by the oldest ones in a way that allows the former to improve their performance in the presence of a given threat, the question of whether the older moneys are in fact spotting and correcting the errors of the younger ones or simply adapting their behavior remains endlessly open. The possession of beliefs requires that a creature have the capacity not only to behave according to a rule, but to be able to follow a rule. And to put it crudely, in order to be able to follow a rule, the creature must be capable of understanding the rule and its applications, viz., the creature must know that it is acting according to a rule and be able to do otherwise. This is nothing but the famous Wittgensteinian motto: discriminating between P and ~P must be caused by a reason. Based on all overt behavior alone, it remains unclear whether the monkeys simply acquire new cognitive capacities, or if they actually acquire a particular kind of new capacity: the capacity to make judgements. It might be that languageless creatures are capable of acting for a reason, recognizing their errors or even having done otherwise. What empirical findings show us is at best that the observation of a creature’s behavior alone is in principle not enough to determine, from a conceptual point of view, whether it does in fact have thoughts (see Dreckman 1999). Ultimately, this is precisely the reason why our belief ascriptions to languageless creatures are merely pragmatically justified, and do not show that these creatures literally have beliefs. As we see it, none of this amounts to an outright denial that languageless creatures may have or can have thoughts. This reasoning may help us understand Davidson’s provocative claim that: “Neither an infant 1 week old nor a snail is a rational creature” (1982, p. 95).

In his early paper “Thought and Talk”, we can already find some passages that cast doubt on the common view of Davidson as a proponent of lingualism:

The usual assumption is that one or the other, speech or thought, is by comparison easy to understand, and therefore the more obscure one (whichever that is) may be illuminated by analysing or explaining it in terms of the other.

The assumption is, I think, false: neither language nor thinking can be fully explained in terms of the other, and neither has conceptual priority. The two are, indeed, linked, in the sense that each requires the other in order to be understood; but the linkage is not so complete that either suffices, even when reasonably reinforced, to explicate the other. To make good this claim what is chiefly needed is to show how thought depends on speech, and this is the thesis I want to refine, and then to argue for” (Davidson 1975, pp. 155–156).

What remains to be explained in this passage is how, or in what sense, thought depends on language in a way that does not lead to lingualism. This has been our purpose in this section.

We understand that Davidson’s insights into animal minds may be taken as a defense of lingualism at first glance. However, we believe that if we look deeper into his words, we will conclude that placing him within the group of lingualists is a hasty interpretation that does not comply with what he aimed to convey. Davidson did not mean to say that languageless creatures do not literally have thoughts; all he said as to the requirement for literal thought was that his “considerations point in the direction of language, but they do not amount to a demonstration that language is necessary to thought. Indeed, what these considerations suggest is only that there probably can’t be much thought without language” (Davidson 1982, p. 101).

In an earlier paper, he had already pointed out that his “considerations will probably be less persuasive to dog lovers than to others, but in any case they do not constitute an argument. At best what we have shown, or claimed, is that unless there is behaviour that can be interpreted as speech, the evidence will not be adequate to justify the fine distinctions we are used to making in the attribution of thoughts” (1975, p. 164). That is, as we see it, utterly different from endorsing lingualism. When discussing the scenario of a tribe of monkeys in which a certain member responds to the threat of danger by emitting a certain cry, Davidson clearly states that “to explain the behavior of the monkeys we do not need to attribute intentions or beliefs to them (I am not arguing that they don’t have intentions or beliefs)” (emphasis is ours). It may be that the monkeys have intentions or beliefs irrespective of being unable to speak a language. Who knows? All Davidson’s argument purports to say is that “whether or not intention [or belief] is present, [in the absence of language] not enough is in place to insure that” (Davidson 1987, p. 116).

To complete the ‘argument’, however, I need to show that the only way one could come to have the belief-truth contrast is through having the concept of intersubjective truth. I confess I do not know how to show this. But neither do I have any idea how else one could arrive at the concept of an objective truth. (Davidson 1982, p. 105)

In a nutshell, we believe that Davidson’s claims concerning animal minds are better seen as a defense of skepticism than of lingualism. As he writes, “we set out to find an argument to show that only creatures with speech have thoughts”, but “what has just been outlined is not an argument, but a proposal, and a proposal we need not accept” (1975, p. 167). It must be stressed that the difficulty in justifying belief ascriptions in a way that goes beyond (c) when language is not present “lies not in the fact that it would be very hard to find out these things without language, but in the fact that we have no good idea how to set about authenticating the existence of such attitudes when communication is not possible” (Davidson 1973, p. 144). If one accepts that there is a divide between intentional actions and adaptive behavior, then a criterion for distinguishing between the two is called for, and room must be made for it. Davidson appeals to language. Evidently, one can disagree with him. Why on earth can Fido be said to have the belief that the cat ran up the tree only if he is capable of verbally reporting that belief? It may very well be the case that Fido has that and many other beliefs. However, the only way Davidson finds of ensuring that Fido actually has those beliefs is by appealing to language. Ultimately, this is why Davidson sides with Stich when he said: “Do animals have beliefs? To paraphrase my young son: ‘A little bit they do. And a little bit they don’t’” (Stich 1979, p. 28).

5 Triangulation, action and indeterminacy: a threat to Davidson

So far we have been making a case for an alternative and more charitable construal of Davidson’s arguments concerning animal minds. According to our reading, Davidson tacitly endorses a skeptical stance as to whether languageless creatures think, rather than lingualism. It has also been suggested that his skepticism stems from a dramatic indeterminacy in the explanation of a creature’s behavior when speech is not present. A clear advantage of our reading over lingualism is that it does not rule out the presence of thought in the absence of language. Yet, not everything is good news for Davidson, as our suggestion can be seen as seriously undermining his entire theory of mind!

To see why, recall that Davidson’s whole philosophy is built upon his theory of radical interpretation, at the core of which lies the idea that meanings and beliefs are interdependent. This means that one must know a great deal about a speaker’s beliefs in order to assign meanings to her words, and vice versa. The interpretation of speech is as much a domestic matter as it is an exotic one, as speakers do not have to speak the same language in order to understand each other (Davidson 1998, p. 323; 1991, p. 210; 1973 p. 125). The indeterminacy is common to all forms of interpretation: “If there is indeterminacy, it is because when all the evidence is in, alternative ways of stating the facts remain open” (Davidson 1973, p. 154).

Indeterminacy is therefore a byproduct of interpretation, and it is independent of the creature being linguistically competent or not. Not surprisingly, the degree of indeterminacy may be higher when we try to account for a creature’s behavior on the basis of its non-linguistic behavior only, regardless of whether it is a language user. Imagine a person who pulls on both ends of a string, giving us no clue to why she is acting as she does (see Davidson 1975, pp. 159–160). If we are to interpret her behavior, among the various plausible explanations, we could say that she may want to move the string in opposite directions, or that she is struggling against herself and wants to find out which side will win. Either of these two explanations fits in well with the behavior we have witnessed. Things become clearer when we are told that all the person wants is to break the string. As Davidson puts it: “Even a generous sample of actions threatens to leave open an unacceptably large number of alternative explanations” (Davidson 1975, p. 160).

However, this poses a serious predicament for Davidson’s whole theory of mind.¹¹ For one thing, if indeterminacy springs from all forms of interpretation—interpretation of speech included—then to ground skepticism regarding the possession of belief by languageless creatures in the indeterminacy of interpretation or explanation of a creature’s behavior forces us necessarily to accept skepticism regarding the possession of belief by linguistic creatures as well. Therefore, we would never be in a position to guarantee that any creature, linguistic or not, is a rational creature, as far as this rationale goes. Things get even worse when we think of Davidson’s “Three Varieties of Knowledge” (1991), where he maintains that knowledge of our own minds, of other minds, and of the world are mutually dependent. That being so, if we were never in a position of knowing whether another creature were a thinking creature due to the indeterminacy of interpretation, then we would end up having no reason to deem ourselves rational creatures either. Here we must state clearly that Davidson never wanted to hold any such view.

6 A possible way out

There is a possible rebuttal to the foregoing objection that a defender of Davidson could take. It might explain why the indeterminacy of interpretation of verbal and nonverbal behavior does not threaten the attribution of thought. Part of the explanation lies on Davidson’s ontology of events (see e.g. Davidson 1970; 1985; 1987, p. 114). Davidson endorses a conceptual rather than a metaphysical ontology of events. For Davidson, the mental and the physical are only conceptual descriptions; there are no such things as mental or physical objects or properties inside our heads. Events per se are ontologically neutral: they are neither physical nor of mental. An event is physical if it can be couched in physical terms, and mental if it can be couched in mental terms. Nothing more and nothing less.

In contrast to what many have contended, Davidson’s arguments never purport to undermine the idea that languageless creatures literally think. Perhaps languageless creatures think; perhaps they do not. Strangely, perhaps, the fact is that Davidson just could never engage in a discussion as to whether languageless creatures literally have beliefs, as if beliefs were objects floating inside the creatures’ heads. Davidson’s approach to mind is hermeneutical or interpretative (see Evnine (1991, ch. 19); Malpas (1992 at the Introduction)). The only thing his theory of interpretation is concerned with is what must we know in order to be able to interpret other creatures’ behavior, and not with exactly what the contents of the beliefs that this creature is entertaining right now are. Evidently, quite often languageless creatures exhibit behavior that could perfectly well fall under mental terms. But as Davidson acknowledged, it may be that all events could be described using mental vocabulary (1970, pp. 211–212). When it comes to interpretation, something more is needed in order to take a creature as a thinking creature, otherwise thermostats, heat-seeking missiles or sunflowers could be taken as rational creatures as well. Davidson appeals to language, as language is what introduces the normative force that allows us to judge whether or not a creature is rational. This is mainly why, in the absence of verbal responses, the creature’s behavior is at best pragmatically justified. This last claim, nevertheless, does not purport to say that languageless creatures do not literally think. We will never overcome the indeterminacy, and this reply may not save Davidson from the objection raised in the previous section; but this explanation may help us see why it is harmless for belief ascriptions.

Some might find this answer utterly unconvincing, and fair enough. However, Davidson clearly says in some passages that it might be possible to have a theory of belief without a theory of interpretation, although belief itself is intelligible only by reference to language (Davidson 1974). Such passages have been largely overlooked by most of Davidson’s critics. In opposition to the widespread interpretation that relates him to lingualism, what Davidson actually asserts is only that “men and woman are alone in having language, or anything enough like a language to justify attributing propositional thoughts to them” (1982, p. 96, fn. 1 (emphasis is ours)). And this seems to be quite different from rejecting the notion that languageless creatures can have thoughts altogether.

7 Conclusion

This paper is concerned with the question of whether languageless creatures think. According to the widespread interpretation, Donald Davidson answers negatively to this question. His controversial claims that neither a one-week-old infant nor a snail are thinking creatures has led many to see him as a proponent of lingualism, viz., the view according to which thought depends on language, so a creature necessarily lacks thoughts if it lacks language. However, having Davidson’s arguments concerning animal minds as a backdrop, we have shown that he never intended to provide a straightforward yes-or-no answer to this question. In a way, languageless creatures think; in another, they do not. More specifically, we have closely examined Davidson’s three related arguments that a creature can have thoughts only if it is the interpreter of the speech of another—namely, the argument from the intentional nature of thought, the argument from holism, and the argument from the concept of belief—, and we have put forward an alternative and more charitable construal of them. As we see it, the only thing his arguments purport to show is that we will never be in a position to confirm that languageless creatures literally think. Or, to put it in other words, we have no grounds on which to tell whether a languageless creatures is manifesting intentional behavior or merely a causal disposition. In a nutshell, we have argued that Davidson is better seen as endorsing skepticism as to whether languageless creatures think than lingualism.