“The term ‘concept’ itself is a battleground. On the one hand, some, particularly some philosophers deny that non-humans can ever have concepts. For them a concept is essentially bound up with language; only creatures with language, so only humans, can have concepts. This view is now substantially eroded, and a majority of researchers are happy to talk about concepts in non-human animals” (Hurford: The Origins of Language p.61)

Hurford’s cursory mention that a lot of philosophers think that language is necessary for a creature to have concepts is uninformative. He admits that a lot of philosophers hold this view but he doesn’t say who these philosophers are or what arguments they present to support their position. Furthermore, his mention that the view is now eroded; is not very helpful as he does not provide any evidence of why the view is eroded; what arguments and experimental evidence have been presented which make the view seem so untenable. In this blog I will discuss the topic in more detail than Hurford manages to. I will first outline the type of conceptual abilities that Hurford attributes to animals and the evidence he presents for these abilities. Then I will discuss the best arguments against attributing concepts to non-linguistic creatures and evaluate the extent to which these arguments undermine Hurford’s position on concepts for non-linguistic animals. In the final sections I will severely critique the arguments against attributing concepts to non-linguistic creatures. Overall I will argue that Hurford’s position is a defensible one and that criticisms against attributing concepts to non-linguistic creatures largely fail.

Hurford presents a series of important points that he thinks indicate that animals have concepts. A key point for Hurford is the idea of a Detached Representation. A Detached representation is a representation of an object that is no longer present. These Detached Representations involve an Index (a kind of pointer of all the properties associated with a remembered object); the index keeps memory of one object apart from other remembered objects. He argues that an index and its properties are the fundamental basis for the logical distinction between functions and arguments. So the idea of a detached representation is a key ability which helps animals form concepts.

He goes on to argue that some animals not only remember objects but they also remember events. He claims that ‘episodic memory’ is key in this particular skill. He cites experimental data (Raby et al. 2007) which indicates that scrub jays have episodic like memories. They remember the WHAT WHERE and When of events significant to them (ibid. p.67). This episodic memory only lasts a day or so, and is therefore not as powerful as human episodic memory which can last for many years. Non-human animal episodic memory may be limited (in the case of remembering events) but some non-human animals can remember objects years after they encountered them, for example elephants.  This indicates that remembering events is a much more difficult and complex process than remembering objects.

Hurford cites the experimental work of Duane Rumbaugh on ‘reverse learning’ which indicates that some animals can apply an OPPOSITENESS OPPERATION to a rule. Rumbaugh’s comparative experiments were done on apes and monkeys. Now obviously apes are more closely related to humans than monkeys. Rumbaugh’s experiments involved getting the monkey’s to associate a reward with a particular stimulus A and not with a stimulus B. Both Monkeys and Apes can learn this trick pretty easily. Once they have conditioned in this way the researchers reverse the procedure and reward stimulus B and do not reward stimulus A. What is interesting is that when the reversal is taught monkeys need to slowly and painstakingly unlearn what they learned in the first stage before they can learn the procedure in stage two (ibid p.71). Furthermore, the better the monkeys have learned the first procedure the longer it takes them to learn the second procedure. Surprisingly the opposite situation occurs for Apes the better they have learned procedure one the easier they can learn procedure two. The explanation given for this strange fact is that Apes have an abstract ability to reverse rules they have learned by applying what the experimenters call Oppositeness Operation on the rule. This level of abstraction is something apes can perform which most monkeys cannot and Hurford thinks that this ability of Apes may indicate a progression towards the type of conceptual abilities only humans seem to have.

So detached representations made possible by episodic memory and the Oppositeness Operation are used by Hurford as evidence of that non-human animals have concepts. Hurford grudgingly admits that one could call the animal abilities proto-concepts, which provide rich representational information about the world, but basically he argues that the evidence he provides is evidence of conceptual abilities in animals. Because of the limited space in his slim book which was only intended for popular consumption Hurford obviously only tipped the ice-berg in his evidence for conceptual abilities in animals. He touched on the evidence from Alex the Parrett who was very skilled at picking out abstract properties of objects as further evidence of animal conceptual abilities. He could have multiplied examples; the truth is that there is plenty of behavioural evidence which animal ethnologists and cognitive scientists think indicates non-human animals have concepts.

Yet despite the fact that it is pretty much a consensus in science that a wide variety of non-human animals have at least some conceptual abilities, as Hurford correctly noted a lot of philosophers disagree with this view. A variety of different philosophers have expressed scepticism about whether non-human animals have concepts: Quine, Rorty, Brandom, and Donald Davidson amongst others. Davidson has probably gone deeper into the matter than any philosopher. So it is probably best to evaluate his arguments against non-human animals having conceptual abilities attributed to them.

Davidson makes a number of arguments which he argues make it implausible to think (he admits that his arguments don’t amount to a proof on the matter) that animals are concept mongers.

• Attributing concepts to animals because they can reliably distinguish aspects of their environment leads to a situation where one is forced to attribute concepts to flowers who can distinguish sunlight from non-sunlight, or thermostats which can distinguish degrees of heat. In none of these cases is reliable discrimination of aspects of an environment an indication of having a concept of x or y. To have a concept one needs to be able to make claims about x or y and justify them to others.
• In order for a creature to have one concept it needs to have many other concepts connected to it. Having a concept of a dog, for example, entails having the concept of an animal, and the concept of an alive thing, and a concept of entities who do not count as dogs etc. To have a connected series of concepts is to have a series of beliefs, desires and intentions towards these concepts. These abilities cannot emerge without an intersubjective language that can be used to make claims about the world that can be judged true of false.
• The Referential Opacity Objection: Davidson mentions the case of a dog running after a cat; while chasing the cat he momentarily loses sight of the cat and sees the rustling in a large tree near him. He starts barking at the rustling tree. Meanwhile the cat he was chasing has ran up a different tree. Davidson notes that someone viewing this situation may make the claim that this shows that the dog is entertaining the proposition ‘The dog is up the tree’, and holds the attitude of belief towards the proposition. Davidson notes that there is a difficulty with adopting this approach. In natural language quantifying into belief sentences results in the sentences becoming referentially opaque and this effects the truth value of the statement. Consider the sentence (A) Bruce Wayne is the crime fighter Batman who caught the Joker. Now consider sentence (B) John believes that Bruce Wayne is the crime fighter Batman who caught the Joker. Sentence A is true because Batman and Bruce Wayne are the same person; the statement is simply one of identity. While sentence (B)’s truth value changes depending on what the beliefs of John are. Attaching ‘X believes that’ changes the sentences truth value from referentially transparent to referentially opaque. However in the case of the dog chasing the Cat up the tree things are different. Suppose in our imagined case that the cat has gone up a tree and the dog is barking at that tree. In this case the sentence (A) ‘The cat is up that tree’ would be true. Now suppose the tree the Dog was barking at was an oak tree. The sentence (B) ‘The cat is up that Oak tree’ will be true as well. If the tree the cat is up is the tallest tree in the forest then the sentence (C) ‘The cat is up the tallest tree in the forest’ will be true. As we know sentences change from a transparent to an opaque reading when attached to belief contexts. However when we attach belief sentences to non linguistic creatures this logical asymmetry does not occur. So take (D) ‘The dog believes that the cat is up the oak tree’, (E) ‘The dog believes that the cat is up the tree’ (F) ‘The dog believes that the cat is up the tallest tree in the forest’. In these sentences we have a series of attributions to the dog and we have no evidence to say that D-F are true or false, this contrasts with sentence 2 above where we can discover what John believes by simply asking him and getting him to justify his claim. There is little evidence that the dog has any of the above concepts of ‘tree, cat, forest, tallest, oak tree, or any of the other possibilities attributable to his behaviour. Thus we have little reason to attribute any concepts to the dog in this case. As Davidson puts it in his (1982) ‘Rational Animals’:

“In a popular if misleading idiom, the dog must believe, under some description of the tree that the cat went up the tree. But what kind of description would suit the dog? For example, can the dog believe of an object that it is a tree?  This would seem impossible unless we suppose the dog has many general beliefs about trees:  that they are growing things, that they need soil and water that they have leaves or needles that they burn. There is no fixed list of things someone with the concept of a tree must believe, but without many general beliefs, there would be no reason to identify a belief as a belief about a tree, much less an oak tree. Similar considerations apply to the dog’s supposed thinking about the cat.” (Rational Animals p. 98)

• In his (1995) ‘What Thought Requires’ Davidson offers further arguments against the view that non-linguistic animals have concepts. He argues that even if a creature has a proto language consisting of a finite number of names, and predicates, and the ability to use demonstrative devices to pick out an in principle infinite amount of objects. And even if this creature could use truth-functional connectives to give this proto-language creativity (the ability to use the finite vocabulary in a way to make an in principle infinite amount of statements). We would still not be justified in attributing to such a creature an ontology according to Davidson. This is because for Davidson such a creature does not have a concept of an object. To have a concept of a creature needs to have grasped the syntax of quantification. It is according to Davidson these devices which help us construct complex predicates and map them on to objects in the world. If a creature cannot do this then Davidson believes that we are not justified in thinking that the creature is a concept mongering thinker until they have mastered all these apparatus and only creatures who have a language have this apparatus.
• Davidson offers one final piece of evidence which he believes shows that only language using creatures have concepts. He argues that in order for a creature to have a concept of x the creature needs to have a concept of truth. The creature needs to know when he is correct or incorrect in applying the concept. If a creature has no concept of when he is using a concept correctly or incorrectly then we have no reason to think that the creature has said concept. On Davidson’s picture a creature has the concept of true and false applications of a concept only when he can triangulate on a shared object of experience, with another self. So a necessary condition of being a rational creature with concepts is having a self, another and a shared object of experience. The other and the shared world are what give the creature the ability to recognise that he has applied a concept incorrectly. This is provided by and only by having another subject (or community of such subjects), against whom one can check his reactions to the stimuli for correctness. Being able to triangulate on shared objects of experiences and make claims that are true or false is only possible according to Davidson for creatures who have a language. Therefore he concludes only of creatures who have a language are we justified in saying they have concepts.

I have obviously condensed Davidson’s arguments above but this is necessary in a short blog. Overall I think his arguments do not really touch the claims made by Hurford on animals having concepts. I will now briefly say what I think is wrong with each of Davidson’s arguments and show how they do not seriously undermine any of Hurford’s claims.

Davidson’s first complaint, that if we attribute concepts to non-linguistic creatures because they can reliably distinguish between aspects of their environment then we will be forced to attribute concepts to thermostats and flowers, is a serious point. It shows that we cannot simply use reliably distinguishing between aspects of the environment as a standalone justification for attributing a concept to something or someone. However in Hurford’s case he is not merely pointing to a creature reliably distinguishing between x and y. Hurford provides evidence that creatures can remember objects which are no longer present. That some creatures (Scrub-Jays) not only reliably distinguish between x and y but can use this distinction to judge that since x is perishable there is no point in digging it up, while since y is not perishable I will dig that up. This requires more than just the ability to distinguish between x and y but the ability to use this distinction to make inferences. If Davidson denies that this is the case he needs to find a way of predicting the behaviour of scrub-jays in a more accurate manner without using the language of the scrub-jay forming inferences. Incidentally Davidson mentioned that a creature being surprised by something may be good evidence that they believed that x is the case, and are surprised that x is false. The scrub jay experiment provides a way of testing whether they are capable of being surprised by deceiving them into thinking you are burying non-perishable food and actually burying perishable food. Monitoring the behaviour of the scrub-jay when it finds nothing may be instructive (though it would be far from conclusive).

Davidson’s second argument that order for a creature to have one concept they must have many concepts, which hang together in interconnected web of beliefs and intentions; is far from conclusive. I agree with Davidson’s claim that concepts are connected together in an interconnected web. But I don’t see how this fact has any consequences for the attribution of concepts to non-linguistic animals. If we start with the assumption that animals are genetically programmed with certain key concepts; causality, agency, object, dominance, submission (depending on the particular animal), and use these concepts to develop basic theories about the world. We have a perfectly reasonable schema with which to try and empirically test our assumptions. It could be argued that by arguing for this innate set of concepts I am begging the question against Davidson. I agree that I am making a question begging assumption; but one that can be tested by experimental evidence. If Davidson wants to start with the opposite assumption he is welcome to; as long as he is willing to test his assumptions empirically as well. However merely pointing to the fact that concepts are holistic is not a sufficient basis to provide any evidence pro or con for non-linguistic animals having concepts.

Davidson’s third argument amounts to nothing more than an appeal to intuition. He says that he can think of no way of telling whether a dog has a concept of ‘tree’, ‘an Oak Tree’, ‘the tallest tree in the forest’ etc. He says that since we have no way of deciding which if any of the concepts to attribute to the dog, then there is no fact of the matter and we have no justification for attributing any concepts to the animal. This argument simply doesn’t work. Davidson doesn’t even consider whether we can use a variety of different behavioural tests to whittle down what concepts the dog has of tree. He just assumes without trying that it cannot be done and draws his negative conclusion. This appeal to incredulity is a dreadful argument. We are only justified in drawing such negative conclusions if years of testing show that we cannot draw the relevant distinctions. But there is no reason to draw the conclusion, as Davidson does, by simply stating that we know in advance that there is no empirical way of deciding the issue.

Davidson’s forth argument that we have no reason for attributing an ontology to a creature until he has mastered quantification is derived directly from Quine.  Quine, like Davidson, argued that mastering the syntax of quantification was essential to a creature if we were to attribute an ontology to him. And that mastering the syntax of quantification was made possible through learning a language. This claim has been tested empirically.  In their 1990 paper ‘‘Ontological categories guide young children’s inductions of word meaning’’[1] Soja, Carey, and Spelke tested whether the ontological distinction between objects and non-solid substances conditions the projection of word meanings prior to the child’s mastery of count/mass syntax.  If before the child has grasped the syntax of quantification the child differentiates in the above manner, this shows that the child is not generalizing the word-based perceptual similarity, but is doing so based on the type of object he is presented with. So, for example, if he was generalizing according to an innate perceptual similarity quality space which focuses on shape then why does this not work for substances? The answer is because the child recognises that objects and substances are distinct ontological categories. Soja et al. summed up their results as follows:

“In sum, the children chose according to object type when the stimulus was an object and according to substance type when the stimulus was a non-solid substance. There was no effect of the syntactic context: performance was neither facilitated nor hindered by the additional syntactic information.

The data from Experiment 1 show that different inferences about the meaning of a newly heard word are drawn according to the ontological status of its referent. If the word refers to an object, the child’s projection respects shape and number, and ignores texture, color, and substance. If the word refers to a non-solid substance, the child’s projection ignores shape and number, respecting texture, color and substance. (ibid., 192)”

From this experiment, Soja, et al. claim to have shown that Quine’s view of how children learn language is incorrect because the experiment shows that, contrary to what Quine claims; children do indeed have a distinction between different ontological categories prior to grasping count/mass syntax (which for Quine and Davidson is the same thing as mastering the syntax of quantification). It also shows that these innate ontological categories are what help a child learn a language and not the apparatus of quantification. Soja et al.’s experiment purports to have shown that children learn words according to ontological distinctions which they exhibit knowledge of prior to learning a language.  This experiment indicates that, despite what Davidson and Quine, think a child has an ontology prior to mastering the syntax of quantification, and this ontology helps them learn their language. The experiment also shows that Davidson’s claim that a creature cannot have an ontology without having a language and the rules provided by the syntax of quantification; is false.

Davidson’s fifth point is about triangulation is simply too vague to evaluate. Non-linguistic creatures do seem to prefer dyadic interactions when communicating. But if Davidson wants to make any large scale claims on the topic he will need examine detailed cases like ‘Vervet Monkey’s’ seeming to triangulate with conspecifics,  and discuss things like the bee waggle dance etc. It is hard to even evaluate Davidson’s[2] claims here until he fills in much needed detail.

Overall Davidson’s arguments do not really present a serious challenge to Hurford’s views on non-linguistic animals having conceptual abilities. In my next blog I will evaluate Hurford’s discussion of how language developed from the conceptual abilities of our ancient ancestors.

[1] I have discussed this experiment in detail in my blog-post ‘The Indeterminacy of Translation and Innate Concepts’. Here I will just point to the conclusion of the experiment those looking for more detail should read my earlier blog-post.

[2] Obviously Davidson is dead so won’t be filling in any details but I refer to people who are convinced by Davidson’s arguments and the work they need to do on this topic.