Sellars, Animals, and Thought
Willem deVries
One of the most common objections against Sellars's philosophy--both in philosophy of mind and in epistemology--is that, because it is so heavily language-centered, it denies both thought and knowledge to all languageless beings, whether animals or small children. This objection was voiced early on by Chisholm in his correspondence with Sellars: "Surely it would be unfounded psychological dogma to say that infants, mutes, and animals cannot have beliefs and desires until they are able to use language" [IM, p. 524].{1} 25 years later it formed a major pillar of Robert G. Meyers's critique of Sellars's attack on foundationalism:
[Sellars] does not think that having a belief or a concept (which we may take as representative of all the intentional abilities) can be analyzed into statements about linguistic ability (SPR, 170). He does think, however, that it is impossible to have beliefs unless you know a language . . . . It may be that all or almost all animal concepts are analogical but it doesn't follow that animals only have concepts in an analogical sense; it is still an open question whether the dog has a concept in the same sense in which the human does. What Sellars needs to prove is not that animals have different concepts from humans or concepts that are only analogous to ours -- that goes without saying. He has to show that attributing any concepts at all to them is just a matter of charity since they do not have concepts in the sense in which we have them. As a result, the sense in which animal concepts are analogical is too weak to show what Sellars needs to show, namely, that they do not really have concepts at all.{2}William Alston also criticizes Sellars's epistemology for placing too rigorous demands on the epistemic sophistication of the knower, demands that animals and infralinguals clearly couldn't meet.{3}
There has been a long history of debate about the abilities of animals and young children in Western Philosophy, but my guess is that most people would be much more willing to agree with Hume that "no truth appears to me more evident, than that beasts are endow'd with thought and reason as well as men" [Treatise, I, 3, xvi, p. 178] than with Descartes that animals are unthinking mechanisms. If Sellars has no response to this objection, his position will be very unattractive.
I
First, we need to ask whether the objection is justified: does Sellars really deny that languageless beings can think? Support for an affirmative answer comes from the psychological nominalism developed in EPM, "according to which all awareness of abstract entities--indeed, all awareness even of particulars--is a linguistic affair" [p. 160]. It is also supported by his remark that, although the theory he develops at the end of EPM can be seen as a variation of the "classical" view of thoughts as internal episodes expressible in language, it could "with equal appropriateness, be represented as a modified form of the view that thoughts are linguistic episodes" [p. 178].{4} And in the correspondence with Chisholm, he writes:
Not only do the subtle adjustments which animals make to their environment tempt us to say that they do what they do because they believe this, desire that, expect such and such, etc.; we are able to explain their behavior by ascribing to them these beliefs, desires, expectations, etc. But, and this is a key point, we invariably find ourselves qualifying these explanations in terms which would amount, in the case of a human subject, to the admission that he wasn't really thinking, believing, desiring, etc. For in the explanation of animal behavior the mentalistic framework is used as a model or analogy which is modified and restricted to fit the phenomena to be explained. It is as though we started out to explain the behavior of macroscopic objects, in particular, gases, by saying that they are made up of minute bouncing billiard balls, and found ourselves forced to add, "but, of course . . ."There is, thus, good evidence for the charge that Sellars does not believe that languageless beings can really think or know.
In his early works Sellars's philosophy of mind focuses almost exclusively on the status of concepts pertaining to persons, developing what has been called a "forensic" concept of persons.{5} Furthermore, Sellars often writes as if (at least in the Manifest Image) the great conceptual cleft is between persons and things.{6} Where would this leave animals or sub-personal humans (if such there be)? Sellars must be able to tell us a convincing story about our concepts of animals and sub-personal humans, or there is a hole in his theory big enough to drive a truck through.
II
I propose to lay out briefly but explicitly how Sellars treats animals and infralinguals. In particular, I want to defend two theses: 1) Sellars's view developed over the years, at least partially in response to his critics; 2) Sellars did eventually develop a view that affords sufficient space for a robust view of animal representational systems, yet retains a special place for persons as thinking beings.
A. First, it seems to be the case that in the '50's and early '60's (at least), Sellars believed that "Behavioristics" (as he called it) might be able to accomplish much towards the explanation of animal activity without the postulation of a layer of entities in between the neurophysiological and the behavioral. This means that Sellars took seriously the idea that the intentional layer of description that he holds is both irreducible and indispensable for the explanation of human action might not be necessary for the explanation of animal behavior. This is part of the thrust of the remark that follows the passage from the Chisholm-Sellars correspondence quoted above:
The use of the mentalistic framework as the point of departure for the explanation of animal behavior can be characterized as the approach "from the top down." Recent experimental psychology has been attempting an approach "from the bottom up," and while it is still barely under way, it has been made amply clear that discrimination is a far more elementary phenomenon than classification, and that chains of stimulus-response connections can be extremely complex and, in principle, account for such 'sophisticated' forms of adjustment as, for example, the learning of a maze, where these same adjustments, approached "from the top down" would be explained in terms of a qualified use of the framework of beliefs, desires, expectations, etc.{7} [IM, p. 527-28].This opinion, now adjusted to think of neurophysiology as the "bottom," recurs a bit later in PSIM, when Sellars remarks that,
The methodological utility of postulational procedures for the behaviouristics of lower organisms has, perhaps, been exaggerated, primarily because until recently little was known in neurophysiology which was suited to throw much light on correlations at the large-scale level of behaviouristics [PSIM, p. 24].The situation with regard to humans, however, is quite different, Sellars thinks.
In human behaviouristics, however, the situation has been somewhat different from the start, for an important feature of characteristically human behaviour is that any two successive pieces of observable behaviour essentially involve complex, very complex, 'iffy' facts about what the person would have said or done at each intervening moment if he had been asked certain questions . . . . Thus it does prove helpful in human behaviouristics to postulate an inner sequence of events in order to interpret what could in principle be austerely formulated as correlations between behavioural states and properties, including the very important and, indeed, essential 'iffy' ones [PSIM, p. 24-25].According to Sellars, postulating a system of entities is explanatorily useful in understanding the dispositional states of organisms "if the postulated entities are sufficiently specific and can be connected to a sufficient diversity of large-scale behavioural variables to enable the prediction of new correlations" [PSIM, p. 24]. He seems to have thought that this might not be the case with "lower organisms," that is, that the only system of entities sufficiently specific and connected to a sufficient diversity of large-scale behavioral variables to enable the prediction of new correlations among those variables in lower animals would be a system of straightforward neurophysiological entities.
The point here seems to be that animal behavior itself doesn't force the postulation of the kind of inner, unobservable entities Jones comes to attribute to his fellows in Sellars's "myth of Jones" in EPM. Sellars seems ready to accept the inadequacy (à la Chomsky) of a "pure" behavioristic treatment of human activity, while still holding onto the possibility that such a treatment of animals would be adequate. If all they had to worry about was explaining the behavior of animals and other languageless beings, the Ryleans could have moved directly from their behavioristic framework to a neurophysiological framework.
Today, in the light of the development of cognitive science over the past 40 years, very few philosophers would be this sanguine about the prospects of a purely behavioristic treatment of lower animals. There is a widespread belief that any adequate scientific psychology will require the postulation of inner, informationally characterized states (i.e. characterized in terms of their "content") for a wide range of organisms.{8} Sellars certainly would have thought that his belief in the possibility of an adequate behavioristic psychology of lower animals is thoroughly empirical. So if it turns out that the psychological theory applicable to most animals makes mention of inner, semantically characterized states, this result would not contradict any philosophical principle of Sellars's, though it would apparently run counter to his expectations at this point.
In the '50's and '60's Sellars was, perhaps, too ready to abandon the commonplace treatment of animal mentality in favor of a less intentionalistic theory. But if cognitive science-style, representationalistic theories of animal mentality do turn out to be more adequate than the purer behaviorisms Sellars contemplated, Sellars still has an account of how they fit into the picture he is painting. For, as he makes clear in the passage from IM, p. 527 quoted above, he thinks of the mentalistic theory of animal intelligence we commonly apply as itself modeled on the mentalistic theory of human intelligence that Jones originated.{9} Thus, we get a picture in which our theory of animal mental states is at a double remove from explicit and overt linguistic events in a three-tier structure:
(Theory of)
Prelinguistic Animal Thoughts
Modeled on
||
\/
(Theory of)
Human Thoughts
Modeled on
||
\/
(Metalinguistic descriptions of)
Overt Verbal EpisodesThus, Sellars has available to him a strategy for explaining the nature of our attributions of thoughts to animals that is perfectly consistent with his strategy for explaining the nature of our attributions of thoughts to persons. The basic picture is that when animals do the kinds of things we do, it is because they are in internal states very similar to ours. Furthermore, given Sellars's realistic treatment of theories, he should hold that one has the right to treat these inner states as real until a better theory comes along in which they are not mentioned. Indeed, given Sellars's realism, it is important to explain why he is as dismissive of prelinguistic thoughts as he is, if they are part of the best theory of animal behavior available before the rise of modern experimental psychology.
The problematic feature is specifying just how similar the internal states of the animals are to the internal states of persons. For with any model, as we have seen, some aspects of the model are relevant to the explanatory power of the theory while others are not. Jones's theory of thoughts already abstracted away from most of the salient properties of overt utterances, their sound, the actual natural language they are couched in, etc. What was preserved was that thoughts possess semantic properties that are essentially identical to those of the utterances that would normally express them. The determinateness of the "inner speech" attributed to someone relies in part on the fact that it is the kind of inner state that would be expressed overtly by a certain sentence of a public language. One of the constraints on the kinds of unobservable states Jones' theory attributes to people and the determinate nature of those states is that episodes of Mentalese are tied in certain ways to (possible) linguistic episodes of some public, natural language. Thus, it makes no sense to attribute thoughts about the relations between bosons and leptons to Anaxagoras, because the vocabulary of the Standard Model of contemporary quantum physics did not exist in Ancient Greece and such thoughts would be inexpressible without major and radical revision of the available public language.
But this constraint has to be abandoned in using Jones's theory as itself the model for a theory of animal behavior.{10} For animals and infralinguals have no language in which their thoughts are naturally or normally expressed. How, then, can we adequately constrain what contents, concepts, or thoughts we can attribute to infralinguals so that the theory isn't riotously permissive about the inner states of animals?
Without some constraint on the kinds of internal states attributable to animals, it is, of course, open to say that they are just the same as the internal states attributable to humans, so whenever an animal does "the same thing" as a human, we can assume that the animal is in just the same kind of mental state as the human with the corresponding behavior. This would essentially eliminate all significant differences between humans and animals other than the fact that animals are mute. Now, Sellars clearly believes that there are significant, even qualitative differences between persons and animals. But in EPM he does not have available to him any resources that would enable him to grant that there is good reason and plenty of evidence for attributing propositional attitudes to infralinguals while drawing a significant distinction between them and full-fledged, language-using persons. For in EPM, Sellars pretty much equates the linguistic with the propositional. If animals have propositional attitudes, then, they have as good as a complete "language of thought" that differs in no significant way from ours except for, perhaps, a smaller vocabulary. Thus, in EPM Sellars is caught between the Scylla of denying that animals think and the Charybdis of attributing animals a fully human range of thought. Saying that animals "think," but they don't really think, or that "think" is equivocal when applied to persons and animals is unsatisfactory unless backed up with far more argument than Sellars gives.
I believe that Sellars was conscious of this problem and developed a set of distinctions in his later work, principally "Mental Events," designed to enable him to steer a path between this Scylla and Charybdis. That is, these distinctions enable him sensibly to attribute to animals and other sublinguistic beings propositional attitudes, or internal representational states such as those regularly posited by cognitive scientists, without granting infralinguals the full range of human conception. Animals and infralinguals remain importantly distinct from full-fledged persons.
In "Mental Events" Sellars explicitly confronts the problem:
17. Now, of course, animals, e.g. our useful friend the white rat, also have representational systems. They do not, however, have languages. Do they think? Have they mental events? Are they perhaps fraught with intentionality? Do they refer and characterize? [ME, p. 328]He introduces a distinction unfamiliar from his earlier work between propositional form and logical form. "To have propositional form, a basic representational state must represent an object and represent it as of a certain character" [ME, p. 336]. The functions of referring and characterizing are by no means solely the province of linguistic entities, however, Sellars emphasizes. Both referring and characterizing are functional roles that can be played by many different kinds of things, but always within a more or less complex representational system in interaction with a world around it. A mark on a map, for instance, can refer to a city and characterize it as located in a certain place, as of a certain size, etc.{11} Thus, Sellars has no problem attributing internal states with propositional form to animals and other sub-lingual beings. In this sense, the later Sellars clearly has room for and believes that animals can have propositional attitudes, intentional states, or thoughts. Our concept of such states, he insists though, is still derived from our metalinguistic semantic concepts.
But, as we have seen, Sellars also wants to be able to draw a significant distinction between infralinguals and fully rational humans. One might think that the distinction is to be drawn by separating those RSs [representational systems] that utilize the linguistic subject-predicate structure to perform or embody the more general referring and characterizing functions. But Sellars does not adopt this strategy. For the subject-predicate structure, the concatenation of separate symbols playing the referring and characterizing roles respectively, is just one way--the linguistic way--of performing those functions. Nonlinguistic representational systems can be in .this is f. states, even though the functional state they are in does not have the subject-predicate structure of a linguistic episode.{12}
Sellars draws the distinction he is looking for at a different place: between those representational systems that use logic, and those that are merely logical in the sense that their description mentions logical terms, but themselves possess and use no such devices.
Thus we carve nature at the joint by distinguishing between those RSs which contain representational items which function as do logical connectives and quantifiers, i.e. which have logical expressions 'in their vocabularies', and those which do not [ME, p. 341].Representational systems that lack items functioning like logical operators can nonetheless "ape reason" by possessing complex sets of dispositions to change their representational states in ways that parallel many of the inferential connections a full-fledged reasoner acting on general principles exhibits.{13} Without the logical connectives and quantifiers, a representational system cannot formulate or explicitly obey conditional principles (If C, then D) or generalizations (All C are D; Whenever C do D), though the system could have dispositions to represent D if (or whenever) it represents C as the case.
Logical operators are, Sellars seems to believe, essentially linguistic, and their presence in a representational system marks a significant increase in the expressive power of the system. For it is the logical operators that enable the formulation of principles and generalizations, and this is the first step towards dealing with the general or the abstract as such. Only RSs that possess the logical operators can reason in the classical sense of acting on a general principle of inference. Thus, Sellars can point to a plausible language-based capacity that marks a significant distinction between persons and animals.
Insofar as animals, like persons, can have propositional attitudes, the concept of thought seems to apply equally to both. But there is a certain kind of thinking activity, namely full-fledged reasoning from general principles, that is not available to animals. To the extent that that activity is the paradigmatic form of thought (as it was for Plato, Aristotle, and Kant, for instance), it is misleading to say that the animals are thinking. Animal concepts differ significantly from human concepts, Sellars thinks, because he believes that human concepts are always bound up with general truths that employ those concepts.{14} Most humans can articulate at least some of those truths explicitly for most of their concepts. Thus Sellars can point to a distinction between human and animal concepts in general, namely playing the characterizing function in a logic-using RS as opposed to playing the characterizing function in a nonlogic-using RS. Is this a sufficient distinction to rebut Meyers' criticism?
Sellars has surely conceded a great deal to the friends of the animals. If the difference between fullfledged conceptuality and a "merely analogous" animal approximation is only that between representational states of logic-using vs. non-logic-using representational systems, the obvious objection would be that it is more appropriate to say that human conceptualization is just a subspecies, a special case of animal conceptualization, and not that human conceptualization is the real thing and animal conceptualization a mere analogue.
Sellars does have a response to this line of attack. Despite Sellars's talk of "carving nature at the joints," how we sort these things out is ultimately a matter of norm-based decision, not of empirical truth, I believe. Just as we could have decided that only jadeite is real jade, and nephrite is "fool's jade" or some base imitator of the real thing, rather than accepting both of these specifically different minerals as forms of jade, the empirical facts about animal representation systems do not dictate one particular reading of the relation between animal, non-logic-using and human, logic-using representational systems. When Sellars emphasizes the chasm between human and animal thought, rather than their equally deep similarities, it is because he has his eyes on the importance of rational agency to the being of humans. If it is our ability to guide our conduct by general principles, which requires using logic, and not just aping it, that is the distinctively human trait and essential to any fullfledged conception of thought and cognition, then animals have only an analog of what we possess. If our focus strays from humans as centers of practical reason, however, it is easy, and for some purposes proper, to regard ourselves as a species of animal representational systems. The "right" or "wrong" way to think of the human-animal relation will depend on the purposes for which we engage the distinction.
In his later work Sellars has developed what seems to be a coherent conception of the relation between the cognitive capacities of humans and animals. It still leans towards rationalism in putting a decided limit on the capacities of animals, but provides a greater scope for animal thought than EPM. Whether it correctly adheres to current usage of such terms as "think" and "know" is not, of course, his concern. The real test of his conception of such matters is whether it is compatible with our increasing knowledge of animal and human capacities, our best explanatory theories of their activity, and the distinctions we are compelled to make to sustain a viable conception of human agency.
[Return to Main Page][Return to Sellars Forum]
Notes {1} I shall use the following set of abbreviations for citing Sellars's essays in the body of this paper:
- EPM:
- "Empiricism and the Philosophy of Mind," originally published in The Foundations of Science and the Concepts of Psychology and Psychoanalysis, H. Feigl and M. Scriven, Eds., Minnesota Studies in the Philosophy of Science, Vol. I. Minneapolis: University of Minnesota Press, 1956: 253-329. Reprinted in W. Sellars, Science, Perception and Reality, London: Routledge and Kegan Paul, 1963: 127-96.
- IM:
- "Intentionality and the Mental" A Correspondence with Roderick Chisholm. Minnesota Studies in the Philosophy of Science, Vol. II, Edited by H. Feigl, M. Scriven, and G. Maxwell. Minneapolis: University of Minnesota Press, 1957: 507-39.
- PSIM:
- "Philosophy and Scientific Image of Man" Frontiers of Science and Philosophy, edited by R. Colodny, Pittsburgh: University of Pittsburgh Press, 1962. Reprinted in W. Sellars, Science, Perception and Reality, London: Routledge and Kegan Paul, 1963: 1-40.
- SM:
- Science and Metaphysics: Variations on Kantian Themes, London: Routledge and Kegan Paul, 1967.
- SK:
- "The Structure of Knowledge: (1) Perception; (2) Minds; (3) Epistemic Principles,"Action, Knowledge and Reality: Studies in Honor of Wilfrid Sellars, edited by Hector-Neri Castañeda Indianapolis: Bobbs-Merrill, 1975: 295-347
- ME:
- "Mental Events," Philosophical Studies 39 (1981): 325-45. [Back]
{2} Robert G. Meyers, "Sellars' Rejections of Foundations," Philosophical Studies 39 (1981): 73-74. [Back]
{3} W. Alston, "What's Wrong with Immediate Knowledge" in Epistemic Justification. Ithaca: Cornell University Press, 1989: 57-78. [Back]
{4} This sentence needs to be compared with his later assertion that,
I find that I am often construed as holding that mental events in the sense of thoughts, as contrasted with aches and pains, are linguistic events. This is a misunderstanding. What I have held is that the members of a certain class of linguistic events are thoughts. The misunderstanding is simply a case of illicit conversion, the move from 'All A is B' to 'All B is A' [ME, p. 325].Small wonder that Sellars has found himself interpreted as holding that thoughts are linguistic events, since he expressly says in EPM that his theory can be seen as one form of that position! Moreover, the analogy with the illicit conversion does not seem to work, for Sellars's revised statement of his position is that some linguistic events are thoughts, whereas his interpreters standardly take him to hold that all thoughts are linguistic episodes, which is not the same kind of (fallacious) conversion, since the quantifier changes. If all Sellars ever maintained was that
1) Some linguistic events are thoughts,it certainly doesn't follow that he believes that2) All thoughts are linguistic events,for (1) leaves open the possibility that some thoughts are not linguistic events. But it seems impossible to square with the text of EPM that (1) is all he intended to hold. [Back]{5} See Joseph Margolis, Persons and Minds: The Prospects of Nonreductive Materialism. Dordrecht, Holland: D. Reidel, 1978. p. 16.[Back]
{6} See PSIM, pp. 6-14. It is no accident that Science and Metaphysics has two chapters entitled "Appearances and Things in Themselves: 1. Material Things" and "Appearances and Things in Themselves: 2. Persons". [Back]
{7} Note that this "bottom-up" effort is not at the micro- or neurophysiological level, but still at the molar, behavioral level. Thus Sellars's distinction here is quite different from the top down/bottom up distinction to be found in more recent philosophy of mid, e.g., the work of Dennett, Fodor, Churchland, et al. [Back]
{8} A materialism that denies this can be properly called eliminative. [Back]
{9} Sellars reiterates this approach at SK, Lecture I, #33, p. 304:
Thus, our common-sense understanding of what subconceptual thinking--e.g., that of babies and animals--consists in, involves viewing them as engaged in 'rudimentary' forms of conceptual thinking. We interpret their behavior using conceptual thinking as a model but qualify this model in ad hoc and unsystematic ways which really amount to the introduction of a new notion which is nevertheless labeled 'thinking'. [Back]{10} There is no problem in principle with using one theory as a model for another. Bohr took the Copernican/Newtonian theory of the solar system as a model for his theory of atomic structure. [Back]
{11} "But, of course, the map is a parasitical RS [representational system]. It depends for its mappishness on its use by human RSs" [ME, p. 337]. [Back]
{12} The dot-quotes here are used in accordance with Sellars's long standing usage to create a functional sortal term, "and while the sortal in terms of which we classify it is built from an expression in our background language, it must be remembered that such functional sortals apply not only to expressions in any language which play in that language a relevantly similar role to that played in our language by the dot-quoted expression, but, as we can now put it, to representational states in any RS which play in that RS the relevantly similar role in question" [ME, p. 340]. [Back]
{13} Sellars distinguishes inferences that proceed from basic representation to basic representation by some sort of direct association from inferences that connect basic representations via standing representations of general or conditional connections. "Let me therefore call the sequence 'Smoke here', 'Fire nearby' a 'Humean' inference, and contrast it with the 'Aristotelian' inference which involves the quantified premise 'If smoke anywhere, then fire nearby there'" [ME, p. 342]. Sellars's distinctions seem clearly intended to evoke a Kantian conception of rationality. "Everything in nature works in accordance with laws. Only a rational being has the power to act in accordance with his idea of laws--that is, in accordance with principles . . . ." [Groundwork of the Metaphysics of Morals, trans. H. J. Paton. p. 412 in the Akademie edition]. [Back]
{14} It is no accident that an early article of Sellars's is entitled "Concepts as Involving Laws and Inconceivable without Them," Philosophy of Science 15 (1948): 287-315. [Back]
[Return to Main Page][Return to Sellars Forum]