1In the Dewey Lectures (1994: 488), Putnam said that he had abandoned internal realism (which he defended in his book Reason, Truth, and History) because it was under the sway of the idea that perception presupposes an interface between the mind and the world. In this essay, I will challenge Putnam’s claim that internal realism presupposes such an idea, i.e. the idea that we perceive reality through sense data. On the contrary, I will argue that the thesis of “natural realism of the common man” or “common sense realism” for which he argues in the Dewey Lectures is no different from the thesis of internal realism. At the end of the article, I will conjecture that Putnam pretended to renounce internal realism in order to allow himself to reformulate the key ideas of internal realism in a way that avoided the reproach of idealism that was sticking to it. To put it differently, Putnam has always defended common sense realism, even when he called it “internal realism.” Such a reading sheds a whole new light on Putnam’s thinking in the 1980s.

1. Rejection of the Traditional Picture of Perception to Rehabilitate the Fundamental Intuition of Common Sense Realism

2According to the traditional picture of perception, our senses do not come into direct contact with the objects and properties of reality. Although these objects may be the causes of our perceptions, we do not perceive them directly, but through the interface of mental objects called “sense data.” According to the traditional picture of perception, sense data constitute an interface between the mind and the world. Although our sensory experiences are normally caused by objects and properties of the external world, we sometimes perceive fictitious objects when we dream or when we are under the influence of an illusion or hallucination. It also happens to us, in such situations, to have sensory experiences practically indistinguishable from sensory experiences under normal conditions. It is, among other things, to account for this reality that philosophers adopted the sense data theory; because the latter allowed them to explain the fact that we sometimes have fictitious sensible experiences that are indistinguishable from our true sensible experiences. According to the sense data theory, this is because in both cases we experience the same mental objects, the same sense data. In the case of a true experience, the sense data we experience would be caused by real objects, whereas in the case of a fictitious experience, they would be the result of our imagination. Nevertheless, our fictional experiences and our truthful experiences share a “common factor.” They are both “sense data” experiences.

3The reason why the sense data theory is problematic for realists is that it breeds skepticism about knowledge of the external world because, since the entities we perceive are identical in both illusory and truthful experiences, we cannot distinguish between the former and the latter. Knowing this, how can we know that our sensible experience is a real-world experience when it could all be an illusion?

4In the Dewey Lectures, Putnam claims to have been under the sway of this traditional picture of perception when formulating internal realism, the most accomplished presentation of which is found in his 1981 book Reason, Truth, and History (from now on RTH), and suggests that, for this reason, internal realism failed to account for the fact that “our minds can be in genuine contact with the ‘external’ world” (Putnam 1994: 488). In order to be able to rehabilitate this fundamental intuition of common sense realism, he claims to renounce internal realism and argues for what he calls “natural realism” in philosophy of perception, for which there is simply no need to postulate the existence of an interface between our minds and the objects of reality.

5I would now like to argue that there is nothing in Putnam’s writings that allows one to associate internal realism with the theory of sense data. Putnam does refer to perception several times in RTH. However, at no time does his statement converge towards the adoption of the theory according to which our sensations would refer to mental objects rather than to real objects.

6Putnam says that he renounces internal realism in the Dewey Lectures because, supposedly, it was based on a certain skepticism about our ability to refer to reality through our language. But this is not what he actually argues for in RTH. Instead, RTH’s internal realism aims to account for the fundamental insights of common sense realism without falling into the respective pitfalls of metaphysical realism that claimed to access a “God’s Eye View” and an unbridled relativism that, disheartened by its inability to access such a view (the “God’s Eye View” of metaphysical realism), falls into generalized skepticism. Although Putnam sometimes refers to his position as a non-realist in reference to his sympathy for the criticism of metaphysical realism of Michael Dummett and Nelson Goodman, he has always disagreed with the idealist conclusions of these two philosophers (Putnam 1994: 462; Putnam 1992: 111-5). For Putnam, the failure of metaphysical realism was never the failure of common sense realism. This idea can already be found in “Realism and Reason” (1978: 129-30), the inaugural essay of the period of internal realism.

2. Sense Data Theory in the Context of Internal Realism

7When commenting on his intellectual journey concerning the question of realism in the Dewey Lectures (Putnam 1994: 457-8), Putnam links his commitment to sense data theory at the time he formulated his internal realism to his commitment at that same time to computational functionalism. The thesis of internal realism was that the understanding of language rests on the mastery of its use and that it is the use of the language that fixes its interpretation. When he first formulated this thesis, in “Realism and Reason” (1978) and “Models and Reality” (1980), and although this aspect is not explicit in these two essays, he represented the mind on the model of the computer and consequently conceived of “use” in terms of a computer program of the brain. According to this model, the mind does not have direct access to the objects and properties of the world, but to “representations,” so that such a model does not allow to account for a true cognitive access (not mediated by representations) to reality.

8Putnam says, however, that he was not aware of such an implication of his commitment to computational functionalism. Even, he did not believe that use could be account for by only referring to the brain’s computer program, because he believed, in accordance with semantic externalism (which he was defending at the time and which I will present in section 3 of this essay), that one should also take into account the interactions of speakers with their environment, and, as we shall see, semantic externalism presupposes that we have the capacity to directly perceive objects in our environment. There was thus, within Putnam’s thinking at the time of formulating his internal realism, incompatibility between his conception of meaning and reference and his (computational) representation of the mind.

9In the Dewey Lectures, Putnam affirms that he does not want to renounce the fundamental thesis of internal realism. It is indeed by learning the use of a language that we understand it, and it is the use of the language that determines its interpretation. However, the fact of representing the mind on the computationalist model of the computer prevents us from recognizing the possibility of a real cognitive access to reality. It is therefore to account for such access to reality that he adopts a new way of conceiving usage inspired by the work of the second Wittgenstein, according to which usage is something that can only be understood from within a linguistic practice, a language game, and which implies that we take for granted our capacity to directly perceive the objects and properties in our environment.

10Putnam thus said in the Dewey Lectures that his commitment to sensory data theory was a consequence of his commitment to computational functionalism of which he was not really aware at the time. On the other hand, a few years later, in his Royce Lectures, presented at Brown University in November 1997 and published as the second part of The Threefold Cord in 1999 (p. 102), he stated that he had rejected sense data theory when he began to study the writings of William James in the 1970s. Consequently, if he was not aware, at the turn of the 1980s, that his functionalism involved a theory that he claimed to have abandoned at that time, it seems difficult to believe that he could have presupposed it in the reflections that led him to internal realism.

11What I will argue for the remainder of this paper is that, even if Putnam was unwillingly committed to sense data theory by virtue of his commitment to functionalism, this does not in fact seem to have had any real impact on his conception of our relationship to reality, which he already at that time represented as “cognitive” and not merely “causal.” In general, RTH’s statement supports the idea that he was already adhering in 1981 to a form of direct realism akin to the one he defended in his 1994 Dewey Lectures, so that there was no reason for him to give up on internal realism on the basis of his alleged commitment to sense data theory.

3. The Use of Direct Realism in the Refutation of the Brain-in-a-Vat Hypothesis1

16What matters here for Putnam’s refutation of the brains-in-a-vat hypothesis is that, if semantic externalism is correct, a community cannot acquire a term of real substance without interacting at least at some point in its history with a sample of the substance it designates. Now, according to the hypothesis considered by Putnam in the first chapter of RTH, the brains in the vat have never been in contact with the real world, therefore, the members of the brains in a vat community cannot have acquired the same concepts of “brain” and “vat” as we have because they have never interacted with real brains and vats. For this reason, their concept of “brain” refers, from our point of view, to virtual brains and their concept of “vat” refers to virtual vats. Therefore, when they think of brains in a vat, they do not think of “brains in a vat” but of “virtual brains in a virtual vat.”

17If we were brains in a vat, we would not be able to say and think that we are brains in a vat because, in order to say and think that we are brains in a vat, we must have acquired the concepts of real “brains” and “vats.” But to acquire these concepts, we must interact with real brains and vats, which brains in a vat are unable to do. Therefore, if we were brains in a vat, we would not have at our disposal the concepts necessary to say and think the assumptions of brains in a vat.

18Given the above, Putnam argues that the brain-in-a-vat hypothesis is self-refuting. If it were true that we are brains in a vat, then what we are saying and thinking when we say and think that “we are brains in a vat” is false because, if we are indeed brains in a vat, then what we are saying and thinking is not that we are real brains in a real vat, but that we are virtual brains in a virtual vat. If it is true that “we are real brains in a real vat,” then it is false that “we are virtual brains in a virtual vat.” If we were brains in a vat, the content of our thought when we consider the hypothesis of brains in a vat would be that we are virtual brains in a virtual vat whereas what we are discussing is the possibility that we are real brains in a real vat. There is thus a performative contradiction. If semantic externalism is a reasonable assumption about how we acquire our concepts, then the brains-in-a-vat assumption is not a justified assumption.

19We have seen that, according to the sense data theory, we perceive reality through sense data which are only potentially caused by the objects of the external world. Now, if we accept such a theory of perception, our perceptual inputs would be the same as those of the brains in a vat. In this case, our respective brain concepts would have as their extension the sense datum of a brain (the common factor) and not a real brain for us and a virtual brain for them. But Putnam’s argument presupposes that our concepts diverge precisely because we do not perceive the same objects as the brains in a vat. Putnam’s argument therefore assumes that we can be in direct contact with the objects in our environment. If Putnam had adopted the sense data theory in RTH, he could not have made such an argument without contradicting himself.

20The point here is that if the contents of our perceptions are identical in the sense that we perceive the same thing (the same sense datum), then there is no need to distinguish the content of our respective concepts. What makes it possible to differentiate the content of our concept of “brain” from that of the brains in a vat (which is in fact the concept of “virtual brain”) is the fact that ours originates from the direct perception of a real brain and that of the brains in a vat from the direct perception of a virtual brain. Only the theory of direct perception, which conceives of the content of perception as world-involving, can account for this aspect of semantic externalism. If one rejects the theory of direct perception, then the content of our respective concepts of “brain” would originate in the perception of a “sense datum” (an “mental image” of a brain) and would not be differentiable at all. Let us remark in passing that saying this does not imply that the difference in content is noticeable. We can very well imagine that a brain in a vat can have a sensory experience subjectively indistinguishable from ours when it perceives a virtual brain. The fact remains that it does not perceive a brain, but a virtual brain and, for this reason, Putnam claims, it is not able to refer to real brains.

4. Internal Realism, Semantic Externalism and Direct Realism about Perception

• 7 Note that Putnam continues to defend the idea that perception is theory-laden in the 1990s (Putnam (...)

Internalism does not deny that there are experiential inputs to knowledge; knowledge is not a story with no constraints except internal coherence; but it does deny that there are any inputs which are not themselves to some extent shaped by our concepts, by the vocabulary we use to report and describe them, or any inputs which admit of only one description, independent of all conceptual choices.7

23What internal realism challenges is a certain conception of reference and truth, according to which reference is a physicalist causal relation disembodied from our social practices, our mental words and symbols possess an intrinsic reference, and the world, conceived by the metaphysical realist as a fixed set of predetermined objects and properties, admits of only one and unique description. Against metaphysical realism, internal realism argues that reference cannot be a unique correspondence of this kind because there are several ways of making a discourse true and there are several discourses capable of describing reality. To defend his point of view, he develops, in chapter 2 of the book, his famous model-theoretic argument aiming at showing that the terms of our language do not possess an intrinsic reference, independently of the context of use in which they are inscribed. For this reason, their reference is underdetermined by their truth conditions. Reference is not a unique correspondence because “a slack between truth-conditions and reference […] remains” (Putnam 1981: 45). Our words can say many things in different contexts. Our language admits several interpretations or several models, as the logicians say. Also, as he argues in chapter 3 of RTH (1981: 72-4), we can use different languages to describe the same reality (a phenomenon he will call in other writings “conceptual relativity”). Putnam’s thesis is that, for a language to be able to describe reality, it must be used in a certain way in a certain social and environmental context:

In an internalist view […], signs do not intrinsically correspond to objects, independently of how those signs are employed and by whom. But a sign that is actually employed in a particular way by a particular community of users can correspond to particular objects within the conceptual scheme of those users. “Objects” do not exist independently of conceptual schemes. We cut up the world into objects when we introduce one or another scheme of description. Since the objects and the signs are alike internal to the scheme of description, it is possible to say what matches what. (Putnam 1981: 52)

5. Semantic Externalism Implies the Underdetermination of Reference

26Since our community must interact with a substance in its environment in order to be able to develop its concept, by associating a term with it in its language, it is not impossible that two (or more) communities develop different concepts from the same word, and, as Wittgenstein pointed out, two communities could have different uses of the same word without this being perceived by the speakers of both communities in most usual situations. In other words, the same word can have different meanings (involving different referents) that are indistinguishable from one community to another, i.e. it can admit divergent interpretations.

27One can easily illustrate why semantic externalism involves the underdetermination of reference by truth conditions using Putnam’s famous Twin Earth thought experiment from “The Meaning of ‘Meaning’” (1975b: 223-7), as he does at the beginning of the chapter 2 of RTH (p. 22-9). In this thought experiment, Putnam asks us to imagine two twin Earths, virtually identical, on which each of us would have a replica (a doppelganger) that would lead a life identical to our own, with the difference that, whereas on Earth the substance that fills our lakes and rivers and quenches our thirst is H₂O, the substance that plays the same role on the twin Earth has a different constitution that Putnam calls XYZ. In the vast majority of everyday situations, the inhabitants of Twin Earth share with us the same sensory experiences when they are in contact with this liquid that they call “water” like us. Putnam then asks us to imagine that it is 1750, before the invention of modern chemistry. At that time, a woman, say Eleanore₁, and her replica, Eleanore₂, had exactly the same “perceptual states” as each other when they came into contact with “water.” However, in Eleanore₁’s language, the term “water” referred to the substance referred to as “water” by her community, i.e. H₂O, whereas in Eleanore₂’s language, the same term referred to XYZ. In their respective languages, all common statements about water (“there is a glass of water on the table,” “the water level in the lake is low,” “I am dehydrated, I did not drink enough water,” “salt dissolves in water,” etc.) have the same truth conditions although they do not refer to the same substances. This example illustrates the phenomenon of underdetermination of reference by truth conditions because Eleanore₁’s speech in Earth English would have the same truth conditions on Twin Earth as Eleanore₂’s speech in Twin Earth English, even though the two speeches would refer to different substances.

29Putnam’s technical arguments in “Models and Reality” (1983) and in chapter 2 of RTH (including the appendix) aimed at showing that this phenomenon occurs at all levels in language, so that even if, after the discovery of modern chemistry, a scientist might be able to distinguish the meaning of Eleanore₁’s statements from those of her replica, i.e. by giving details about their respective contexts of use, the vocabulary she would use to do so would still be underdetermined by the truth conditions of her statements. This does not mean that we are incapable of referring, but, as Putnam said, that there is a gap between truth conditions and reference.

30The reason Putnam does not draw the conclusion from the model-theoretic argument that we fail to refer to reality is that he believes that the success of our interactions with reality argues for the approximate truth of our “ordinary world theory,” that is, for the idea that it is an approximately adequate description of reality. What the phenomenon of underdetermination challenges is not our ability to refer approximately to reality, but the idea that truth is based on a unique correspondence between the words of our language and the objects and properties of reality (Putnam 1981: 38-41).

31The above argues for the idea that the Putnamian critique of metaphysical realism is based on semantic externalism. If I am right in arguing that the latter presupposes direct realism about perception (as I argued in section 3 of this essay), then there is every reason to believe that internal realism presupposed a form of direct realism about perception rather than the sense data theory.

5. Some Textual Evidence in Favour of the Idea that Internal Realism Would Have Presupposed Direct Realism in Philosophy of Perception

32In addition to what I have just said, there are several passages from the period of internal realism that also support the idea that internal realism would have presupposed direct realism about perception. One is Putnam’s description of how he conceived of an ideal epistemic situation in RTH in the preface to Realism With a Human Face (1990: viii):

By an ideal epistemic situation, I mean something like this: If I say “There is a chair in my study” an ideal epistemic situation would be to be in my study with the light on or with daylight streaming through the window, with nothing wrong with my eyesight, with an unconfused mind, without having taken drugs or been subjected to hypnosis, and so forth, and to look and see if there is a chair there. Or, to drop the notion of “ideal” together, since that is only metaphor, I think there are better and worse epistemic situations with respect to a particular statement. What I just described is a very good epistemic situation with respect to the statement “There is a chair in my study.”

33We can also establish a correspondence between the view of common sense realism defended by Putnam in his 1994 Dewey Lectures and the internal realism in the first of his 1985 Carus Lectures (published as The Many Faces of Realism in 1987). In this lecture, Putnam appears to be arguing the same thesis that he defends in the Dewey Lectures. To do so, he traces the history of sense data epistemology and holds it responsible for the impasse in which Western philosophy has found itself since the 17th century, an impasse that, as he will say in the Dewey Lectures, has led it to “oscillate […] between equally unworkable realisms and idealisms” (Putnam 1994: 488). The difference between the two lectures is that while the latter classifies internal realism as a sterile position, the former identifies internal realism as the ultimate way out of this impasse:

The key to working out the program of preserving commonsense realism while avoiding the absurdities and antinomies of metaphysical realism in all its familiar varieties […] is something I have called internal realism. […] it is a view that takes our familiar commonsense scheme, as well as our scientific and artistic and other schemes, at face value […]. (Putnam 1987: 17)

34If Putnam believed in 1985 that internal realism was the way out of the impasse generated by the epistemology of sense data, how could he have coherently argued in 1994 that it presupposed it?

35My hypothesis on this subject is that, faced with his inability to get rid of the prejudices that stuck to the skin of internal realism, Putnam decided in the early 1990s to pretend to renounce internal realism in order to allow himself to reformulate the key ideas of internal realism in such a way as to avoid the reproaches that it was an idealism that reduced truth to justification.

6. What is Changing in Putnam’s Thinking in the 1990s

36As we saw in section 2, what really changed in Putnam’s thinking in the 1990s was that he abandoned the cognitive science interpretation of use in favour of an interpretation of use inspired by the work of the second Wittgenstein that makes it a notion irreducible to the discourse of cognitive science and closer to the spirit of commonsense realism according to which, as he already argued in his Carus Lectures in 1985, we must be content to “take our familiar schemas at face value” without philosophical reinterpretation.

7. The Thesis of Common Sense Realism is not Different from the Thesis of Internal Realism

38I said in the introduction that I would argue that the Dewey Lectures thesis is not fundamentally different from the RTH thesis. In order to do this, I need to mention two things. The first is that Putnam retains most of the flagship theses of internal realism in the Dewey Lectures, as evidenced by the note 41 of the lectures:

Am I then giving up “internal realism”? Well, while in Reason, Truth and History I identified “internal realism” with what I am here calling “moderate verificationism,” in The Many Faces of Realism I identified it with the rejection of the traditional realist assumptions of (1) a fixed totality of all objects; (2) a fixed totality of all properties; (3) a sharp line between properties we “discover” in the world and properties we “project” onto the world; (4) a fixed relation of “correspondence” in terms of which truth is supposed to be defined. I rejected those assumptions not as false assumptions, but as, ultimately, unintelligible assumptions. As will become clear in the sequel, I still regard each and every one of those assumptions as unintelligible, although I would argue for that conclusion in a different way. So whether I am still, to some extent an “internal realist” is, I guess, as unclear as how much I was including under that unhappy label. (Putnam 1994: 463)

39What he claims to reject from internal realism at the turn of the 1990s is, on the one hand, the fact that it would have presupposed the theory of sense data and, on the other, the idea that truth “could never be totally recognition-transcendent” (Clark & Hale 1994: 243). I have already given my reasons for believing that internal realism does not presuppose sense data theory. It is also far from clear that Putnam argued within internal realism that truth “could never be totally recognition-transcendent.” Since I do not have the space in this article to establish this latter thesis, I will simply raise the following point. The thesis that Putnam defended in the late 1970s and 1980s was not that truth cannot transcend verification, but that it cannot overtake our cognitive capacities to the point where metaphysical statements can be truly intelligible and, therefore, possess truthfulness. This is how he presented this idea in the preface of Realism With a Human Face:

What bothered me about statements of the sort I rejected, for example, “There really are (or ‘really aren’t’) numbers,” or “There really are (or ‘really aren’t’) space-time points,” is that they outrun the possibility of verification in a way which is utterly different from the way in which the statement that, say, there was a dinosaur in North America less than a million years ago might outrun the possibility of actual verification. These former statements are such that we cannot imagine how any creature with, in Kant’s phrase, “a rational and sensible nature” could ascertain their truth or falsity under any conditions. (Putnam 1990: ix)

40Now the discourse he holds about truth in the 1990s is not different. What he admits is that an empirical statement of the kind he quotes in this excerpt about the presence of a dinosaur in North America at least a million years ago may have a truth value that transcends verification, but he still rejects the possibility that metaphysical statements such as the brain-in-a-vat hypothesis could be intelligible, because we cannot imagine what it would be like for beings with our cognitive capacity to verify them (Putnam 2002: 107-8).

41On the basis of the above, it seems to me difficult to argue, as Putnam did from the 1990s onwards, that he had renounced internal realism. Rather, in light of what I have argued in this article, it seems fair to say that the thesis of the “natural realism of the common man” or “common sense realism” of the Dewey Lectures is in fact the same as that of the internal realism of RTH, though Putnam has refined his way of defending it over the years. Knowing this, and knowing that Putnam had great difficulty in getting his ideas understood in the 1980s, it seems to me that the most plausible explanation for his pretending to renounce internal realism is that, faced with his inability to get rid of the misinterpretations of internal realism, he decided to pretend to renounce it in order to be able to reformulate his original ideas in such a way as to avoid the misinterpretations that he had faced in the previous decade.

8. Conclusion

42In this article I have criticized Putnam for leading his readers to believe that he had renounced internal realism. I have mainly attacked the idea that the latter (internal realism) presupposed the theory of sense data. I have argued that the discourse of RTH, which is the seminal work of internal realism, rather supports Putnam’s adherence, at this time, to a form of direct realism of the type he puts forward in his Dewey Lectures. To this end, I have shown how the refutation of the brains-in-a-vat hypothesis, which is based on semantic externalism, rests essentially on the recognition that perception gives us cognitive access to the objects of reality. Secondly, I wanted to show that the model-theoretic argument on which the Putnamian critique of metaphysical realism was mainly based at the end of the 1970s and during the 1980s should not be interpreted as a skeptical argument. The point of this argument is not to deny that we have cognitive access to the objects of reality, but only to deny that this access is not influenced by our concepts. To illustrate this point, I have shown how semantic externalism involves the phenomenon of underdetermination of reference on which Putnam bases his model-theoretic argument to show the unintelligibility of a conception of truth as a unique correspondence between words and the objects and properties of a ready-made world. Since semantic externalism presupposes direct realism in the philosophy of perception, there is no reason to see the model-theoretic argument as a skeptical argument based on the theory of sense data. After saying this, I pointed out that Putnam described his position in much the same way when he claimed to be an internal realist and when he later claimed to be a common sense realist, so much so that I have argued that there is no real difference in his thinking on the question of realism other than the fact that his thinking evolved on the interpretation of the notion of linguistic usage, which led him to abandon computational functionalism for good in the 1980s. For all these reasons, I think that the common sense realism of the 1990s is in fact a reformulation of the internal realism of the 1980s, and that Putnam’s so-called rejection of internal realism is in fact a feint, the purpose of which was to allow him to reformulate his thinking in such a way as to avoid the misinterpretations that were sticking to him. I think this clarification will allow for a greater appreciation of some of the subtleties of Putnam’s thought that have too often been crowded out by a widespread interpretation that Putnam renounced realism in the 1980s and returned to it in 1990.