The previous post in this series presented Hilary Putnam’s “model-theoretic argument” to the effect that no representational system whatsoever, including natural language and mental states such as thoughts and percepts, can refer to anything definite unless the assignment is made externally by an agent outside the representational system or “Platonically” by means of some non-natural access to the domain of reference. For example, the little airplane icons on an air traffic controller’s screen can be assigned to specific planes because one can see both the icons and the planes—sometimes just by looking out of the control tower window—to map the icons to the planes. But when it comes to thought and perception, we have no such independent access to the intended referents. How in that case is any determinate mapping possible? Putnam’s claim, which the model-theoretic argument is intended to establish, is that, barring some “Platonic” cognitive channel to external reality that cannot be explained by natural science, no determinate mapping is possible. Therefore, our thoughts and percepts have no truth conditions that depend on the mind-independent world being any one way rather than any other. This is what Putnam calls “internal realism.”

In the present post, we critically examine the model-theoretic argument. (The whole paper on which these posts are based is available here. To advance to the next post in the series, click here.)

The Status of the Model-Theoretic Argument

Let us finally assess the merits of Putnam’s model-theoretic argument for “internal realism.” It has been thoroughly critiqued in the extensive philosophical discussion it spawned. The best critique known to me is a pair of papers by Timothy Bays (“On Putnam and His Models,” 2001; “Two Arguments against Realism,” 2008). Their treatment of Putnam’s argument is thorough and devastating. Less detailed but perhaps more philosophically interesting is the criticism by David Lewis (“Putnam’s Paradox,” 1984).

The basic flaw in Putnam’s argument is reasonably clear, I think, and I propose to dispense with minutia and describe it at a fairly high level.

I said in the previous post that Putnam deploys a two-part strategy:

1. first, show that model theory allows any first-order theory to be given radically different interpretations in terms of different sets of objects in the world that nevertheless preserve the truth of all theorems of the theory;
2. second, argue that there is no way consistent with scientific naturalism to limit the profusion of possible interpretations.

What we have seen up to now—which I will call “Putnam’s Proof”—is only the first part. Despite a technical error in Putnam’s use of the Löwenheim–Skolem theorems in the M&R version of Putnam’s Proof pointed out by Bays (2001, section 2), the proof in the permutability version is a slam dunk. Nor is it the first such proof. We will see in the next post that it essentially replicates an earlier result from 1928, and of course the just-mentioned Löwenheim–Skolem theorems are best known for their (seemingly) paradoxical result that the standard axioms of set theory, which entail the existence of uncountably many sets, can be given a countable model (i.e., an interpretation on which the statement “x is uncountable” is true but doesn’t mean that x is uncountable) (Bays, 2014). So, Putnam’s Proof should not even be very surprising.

Putnam’s problems arise in the second part of his argument, where he argues that the profusion of radically different interpretations of a theory cannot be meaningfully constrained. Putnam acknowledges two types of constraints on interpretation, which he calls “operational” and “theoretical” (RT&H, 29–32). Operational constraints are imposed by sentences associated with experience, such as “the thermometer reads 30˚ C” or “sunrise begins at 6:02 a.m.” Our experiences are not completely up to us, and so can impose at least some restriction on the truth-conditions of observation sentences, whatever difficulties and complexities might arise here (such as the difficulty of independent access mentioned above, among many others). Theoretical constraints embody principles like same-cause-same-effect, simplicity, and conservatism. These also, Putnam thinks, are something we can know are satisfied independent of the confirmation of the theory itself. But, he argues, these constraints do little or nothing to address the problem of radical reinterpretation. The reason is that these are constraints on truth-conditions of sentences—on what sentences should be supposed to be true—but the task of interpretation is to assign referents to terms. And as we have seen, what Putnam offers is a proof that multiple, radically different interpretations can always be found for the terms of a given theory that still render all its sentences true. That is, Putnam’s Proof takes a fixed set of sentences to be made true as given, and then shows that, if the theory is formally consistent and there are enough objects in the domain, many radically different assignments of objects to the terms can always be found to make that happen. In Putnam’s words: “no view which only fixes the truth-values of whole sentences can fix reference, even if it specifies truth-values for sentences in every possible world” (RT&H, 33).

The obvious rejoinder to Putnam is to say that we need constraints that apply to the interpretation, not just to the set of sentences that our theory regards as true. The most straightforward suggestion, I think, is to appeal to causation. For example, we might require that terms for perceptible objects and properties refer only to those objects and sets of objects that cause the relevant perceptual states. Terms for nonperceptible properties like mass and electric charge might be required to refer to properties, if such there be, that belong to the objects the theory says possess them and that cause those objects to behave as the theory predicts (e.g., be subject to gravitational acceleration or swerve in magnetic fields). Terms for unperceived or unperceivable objects of a theory could likewise refer to objects, if such there be, specified in terms of perceptible properties like spatial relations and that have the causal effects attributed to them by the theory. The base of this proposal is the existence of a determinate, identifiable causal relation between percepts and external things. I have already said something in defense of this claim, and I’ll say more below.

There are other constraints on interpretation that could be proposed, instead of or in addition to causation. David Lewis (1984) suggests that not all objects in a domain should be regarded as equally eligible candidate referents. Some, such as the fundamental properties (mass, charge, quark color) and objects (electrons, quarks) of physics, are both highly eligible in themselves and also are the basis of causal chains whose length determines the eligibility of other candidates (horse is more eligible than horse–or–paramecium–or–dandelion, which is more eligible than cubic–centimeter–at–the–center–of–the–sun–or–what’s–under–my–right–thumbnail–on–8/7/2023, etc.). For any given theory, interpretations that employ more eligible referents are to be preferred to interpretations that employ less eligible referents. Still other constraints might impose modal requirements such as necessity and lawlikeness on the satisfaction conditions for sentences of a theory that are supposed to have that status. Any of these modifications of the interpretation procedure for first-order formal theories would break Putnam’s Proof.

Of course, Putnam is alive to the prospect that people will suggest added constraints as a way around his argument. And he has an answer waiting, which is that the added constraint amounts to just more theory and so is subject to a further iteration of his argument. “The problem is that adding to our hypothetical formalized language of science a body of theory entitled ‘Causal theory of reference’ is just adding more theory” (M&R, 18). Then the theory thus supplemented is subjected to Putnam’s Proof to show that it too can be interpreted in innumerable, radically different ways, because terms like “causes” and “refers” of the supplemented theory are as subject to radically different interpretations as any other vocabulary.

To this Putnam’s opponent is right to reply that the suggestion is not to add more sentences to the existing theory, but to change the method of interpretation of that theory. Surely this is an important and clear distinction. Sometimes we make statements in our theory, and sometimes we make statements about how a formal theory is to be interpreted, and these are not the same. The former are object statements; the latter are meta-statements. Putnam himself makes meta-statements when he lays down his preferred principles of interpretation. If he can do so, why can’t we? Why does he get to make meta-statements in specifying his own preferred version of model theory, but when we do so, ours get reclassified as object statements?

Surely it would be unfair to assume that Putnam is pigheadedly disregarding the distinction between object statements and meta-statements (when it suits him to do so). And surely he does not unthinkingly assume that since we need a theoretical understanding of a constraint such as causation—in sense-perception, say—to employ it in interpreting the terms of our theories, therefore the constraint must be incorporated into the theory as opposed to the method of interpretation. But if neither of these is what’s going on, then what makes Putnam think he can get away with this sleight of hand?

Let’s look more closely at what Putnam actually says. In RT&H (45–47), he considers a causal approach to reference put forth by Hartry Field (1972).

Suppose there is a possible naturalistic or physicalistic definition of reference, as Field contends. Suppose

(1) x refers to y if and only if x bears R to y

is true, where R is a relation definable in natural science vocabulary without using any semantical notions (i.e., without using “refers” or any other words which would make the definition immediately circular). If (1) is true and empirically verifiable, then (1) is a sentence which is itself true even on the theory that reference is fixed as far as (and only as far as) is determined by operational plus theoretical constraints. (1) is a sentence which would be part of our “reflective equilibrium” or “ideal limit” theory of the world. (45)

This is the “just more theory” reply. He says that statement (1), “if true and empirically verifiable,” qualifies for inclusion in our full theory of the world. As such, it suffers from the referential indeterminacy of that theory. Putnam pounces accordingly:

If reference is only determined by operational and theoretical constraints, however, then the reference of “x bears R to y” is itself indeterminate, and so knowing that (1) is true will not help. (45–46)

But then Putnam acknowledges that this misconstrues Field’s intentions.

       This is, of course, not what Field intends. What Field is claiming is that (a) there is a determinate unique relation between words and things or sets of things; and (b) this relation is the one to be used as the reference relation in assigning a truth value to (1) itself. …

       …let us consider the view that (1), understood as Field wants us to understand it (as describing the determinate, unique relation between words and their referents), is true. If (1) is true, so understood, what makes it true? Given that there are many “correspondences” between words and things, even many that satisfy our constraints, what singles out one particular correspondence R? Not the empirical correctness of (1); for that is a matter of our operational and theoretical constraints. Not, as we have seen, our intentions (rather R enters into determining what our intentions signify). It seems as if the fact that R is reference must be a metaphysically unexplainable fact, a kind of primitive, surd, metaphysical truth.

       […]

       To me, believing that some correspondence intrinsically just is reference (not as a result of our operational and theoretical constraints, or our intentions, but as an ultimate metaphysical fact) amounts to a magical theory of reference. (46–47)

Here Putnam accurately identifies the view of a realist proponent of a causal theory of reference and others who object to his model-theoretic argument along the lines I am suggesting; viz., they challenge his view of the proper method of interpretation of a theory. And then he counters it with (yet another iteration of) the impotence of natural mechanisms argument identified at the end of the previous post: there are many physical or natural correspondences that could realize R, so a correspondence—as specified by a causal theory of reference, say—is insufficient to fix reference. And we have seen that a theory of R is unable to specify the nature of the correspondence (because operational and theoretical constraints are just more sentences, and fixing the truth-values of whole sentences will not fix the reference of the terms in those sentences). Thus, we are left with no choice but to regard the determination of reference by R as a metaphysical primitive. And this is what Putnam regards as akin to magic (“Neo-Platonism,” “medieval essentialism” (1983, xii), etc.).

In light of the foregoing, we can perhaps see why Putnam would regard the distinction—between adding statements about reference to one’s theory and modifying the method of interpretation of that theory—as neither here nor there. He might say, “when you propose that your theory must be interpreted by relation R, you assume that ‘relation R’ has a determinate referent, but what I have shown is that it doesn’t!” Fundamentally, his view is that we are stuck inside language and can’t get out. We have no independent access to the objects, properties, and relations in the world that would enable us to know which of them our terms refer to. So, since Putnam’s Proof shows that the terms of our language and thought do not interpret themselves, they simply have no determinate interpretation—at least, not if that is to consist of an assignment of mind-independent objects and properties. Putnam expresses this clearly in a key passage from the “Introduction” to Realism and Reason (1983). Discussing the realist philosopher’s claim “that reference is fixed by causation itself,” he writes:

       Here the philosopher is ignoring his own epistemological position. He is philosophizing as if naïve realism were true of him (or, equivalently, as if he and he alone were in an absolute relation to the world). What he calls “causation” really is causation, and of course there is a fixed, somehow singled-out, correspondence between the word and one definite relation in his case. Or so he assumes. But how this can be so was just the question at issue. (xi)

To sort out everything that has gone wrong here is a large job. Let us begin by observing that the reference of our thoughts and percepts does not depend on our “epistemological position” if that is supposed to mean, for example, that to have a percept of a determinate object, we must be able to identify the causal correspondence between percept and object, much less justify knowledge of its existence in a particular case. If this were so, then it would be impossible for a toddler playing with its blocks to have percepts of those determinate blocks! Rather, if perception depends on the causation of percepts by their objects, then this is a process that operates to determine the reference of percepts regardless of the perceiver’s knowledge or ignorance of this process. It is sufficient that the process produces percepts of determinate objects. Understanding of the causal processes involved in perception can come later, in species capable of the abstract thought necessary to achieve such understanding—exclusively the human species, as far as we know. Other species have percepts of determinate objects without theoretical understanding of causal relations or anything else. Nothing in Putnam’s Proof provides any reason to think that percepts do not have determinate referents in this way.

Indeed, Putnam’s Proof—part 1 of his strategy, with its appeal to imposing theorems from metalogic with names like “Löwenheim–Skolem” and “Shoenfield Absoluteness”—in fact plays only a minor part in his argument for internalism (a point emphasized by Bays 2008). It is perfectly obvious that a merely syntactical, abstract formal language like first-order logic does not interpret itself, and only slightly less obvious that, if the satisfaction relation is as standardly defined in first-order logic (described in the previous post), then any sufficiently large domain of objects whatsoever can provide a model and indeed countless models of any system that is logically consistent through permutation. Putnam’s picture of the semantics of our thoughts and percepts is that there is a set of “mental signs” that need to be interpreted by being assigned to objects in the external world as referents. In the case of a formal system, such as a theory stated in a first-order formal language, the assignments are made by the theorist, who has independent access to both the signs of the language and objects in the domain. But in the case of our thoughts and percepts, we have no independent access to the objects, and so there is no simple way—and perhaps no way at all—to make the assignments. This is what Putnam means in referring to the epistemological position of the philosopher. Putnam thinks that the “metaphysical realist” takes a “God’s eye view” in assuming he can somehow step outside his own skin and gain direct access to the objects in order to make the assignments. Or rather, to select a preferred relation R (causation or some other natural correspondence) as the “natural” determiner of the assignments. But, Putnam thinks, this is plainly illegitimate: causation makes no unique assignments, nor similarity nor any other natural mechanism, and non-natural means such as grasping Platonic Forms are no longer accepted in polite society. Therefore, internalism stands as the only viable option. Case closed.

We see that (a) the analysis into the three options (Platonism, causation or other natural correspondence, internalism) and especially (b) the argument against the viability of natural correspondence do all the real work in the argument. The model-theoretic proof is a sideshow. The question is not whether the proof is valid or what it means, but why we should think that the semantics of our thoughts and percepts is like the semantics of first-order logic. I think it’s clear that they are very dissimilar and that, accordingly, Putnam’s model-theoretic proof is far from showing that internalism is true. Nevertheless, an examination of some of the key dissimilarities is philosophically illuminating. In what follows, I discuss what seem to me to be the most important disanalogies.

We have already encountered the objection that the standard satisfaction relation of first-order model theory is unrealistically weak and must be supplemented with additional constraints if it is to serve as a facsimile of reference fixation for our thoughts and percepts. However, we have also seen that Putnam has a reply from the impotence of natural mechanisms to determine reference. This is the most important argument he deploys in defense of his internalist philosophy, and in fact the argument does raise challenges. For one thing, it is difficult to specify what causal relations are necessary to fix the referents of our percepts. For another, it is hard to see how any causal relations could ever confer intentionality on anything. That is, just because certain external causes give rise to our percepts, why does that mean our percepts are about those causes? Let’s consider these two questions in order.

First, then, Putnam wants to know what sort of causal relation—or any other natural correspondence—could bring it about that a given percept represents a given kind of thing. For example, suppose we have a mental sign, say ♣, which is to be the visual percept of being a body or having a certain shape. We need a causal correlation that brings it about that when ♣ occurs in perception, it represents being a body or being a certain shape. Thus, this correlation needs to distinguish the occurrence of bodies or certain shapes in the environment from distinct yet co-occurring elements of the total causal situations that produce occurrences of ♣, such as certain patterns of retinal stimulation or cortical activity, as well as all the many artificially gerrymandered “properties” that could be invented that co-occur with one’s sensory receptors being stimulated by a body or a certain shape. What is there in causal relations alone that can pinpoint a certain environmental feature as the referent of a percept?

I think the present state of vision science allows us to give a general answer to this question. Let us begin by distinguishing perception from mere sensation. In our sensory interactions with the environment, sometimes we merely register information about it, such as coldness or hotness, quietness or loudness, which we do by means of sensations. Sensations can result from proximal stimulation alone, such as stimulation of temperature-sensitive receptors in our skin. Thus, sensations of cold and hot can register information about temperature and do so by means of a relatively simple causal relation between coldness and hotness in the environment and our sensations of cold and hot, not unlike the simple causal relation by which a thermostat registers temperature or a gasoline gauge registers the level in a gas tank. And some philosophers, such as Dretske (1995), have argued that this is in fact the ground level of mental representation, at least when the receptors in question have evolved to have the function of registering temperature; i.e., they were selected for in evolutionary history because their registering temperature promoted the survival and reproduction of our ancestors, so registering temperature is arguably their biological function. But whether sensations of hot and cold really represent temperature can be questioned on grounds such as Putnam raises: because they correspond merely to proximal stimulation, they may as well be taken to “represent” that proximal stimulation as the objective temperature of anything in the external environment. By the same token, if we have hot or cold sensations when there is nothing hot or cold or fail to have them when there is, in virtue of what would it be right to say that the sensations are inaccurate? True, they have not performed in the way they were selected for performing in the evolutionary past, which could disrupt the other biological functions of the organism. But that is a pragmatic failure, not a representational one. Rather than regard sensations of hot and cold as representing the ambient temperature, one might as well say—and a skeptic would say—that they function merely as a proxy for it. Thus, on occasions when the causal link between sensations and temperature fails, we might well say that the organism’s trick of using the sensations as a guide to temperature didn’t work out this time, but it is less clearly correct to say that the sensations themselves were inaccurate.

By contrast, the complex causal-computational processes that generate percepts render them independent of even large differences in proximal stimulation, so that it becomes much more plausible to say that they track and therefore represent features of the distal environment. I have discussed this point already at some length in the first post in this series on brains in a vat. Percepts must discriminate and track objects, properties, and relations, despite wide variations of distance, perspective, orientation, lighting, and other environmental conditions. Think of the varied conditions in which organisms forage for food or mates, fight with antagonists, chase prey, flee predators, build nests, seek feeding grounds, etc. In all these different conditions, the proximal stimulation of receptors from the same distal objects and properties will vary widely. To isolate, identify, and represent objective, distal features of the environment on the basis of fluctuating, noisy, partial, and perspectival sensory stimulation requires complex causal-computational processing by the perceptual system. These capacities have been extensively studied, especially for vision. They are called perceptual constancies. Our visual system has constancies for distance, depth, shape, size, orientation, position, lightness, and chromatic color, among others. (Basic discussions of many visual constancies will be found in visual perception textbooks such as Palmer 1999 and Bruce, Green, and Georgeson 2003.) The exercise of perceptual constancies depends on isolating and processing aspects of proximal stimulation that are relevant to determining distal aspects of the environment. That is, the perceptual system itself works to produce a state—the percept—that is causally related to and tracks a specific feature of the distal environment in the face of variation in other features. Burge supplies a criterion in terms of causal explanation:

A sensory state is perceptual and representational only if it is associated with a capacity to connect with a specific distal environmental attribute or particular in a way that is distinctive to that attribute or particular and that figures in causally explaining occurrences of that sensory state better than a naturally specified type of proximal stimulation alone. (Perception 2022, 60)

The key difference that Burge is pointing to is between one class of sensory states—sensations—that are best explained by their proximal stimuli without regard to distal stimuli and another class of sensory states—percepts—that are best explained by their distal stimuli without regard to their proximal stimuli. It is thus the nature and identity of percepts to track distal features of the environment. This is presumably a key factor in making percepts, as opposed to sensations, truly representational states, and it certainly is key in determining their content and accuracy conditions. Burge again:

I think that causal-computational explanation in science specifies sensory states in terms of their representational content and accuracy conditions when and only when sensory states are constituents in perceptual constancies. (Perception 2022, 67)

Here then is a direct reply to Putnam’s challenge to explain how a causal relation fixes reference. Yes, there are many causal relations between our sensory states and other things, but they are not all created equal. In the case of a percept, the sensory state in question is the product of causal-computational processing that functions to maintain a specific causal link between the state and a distal environmental feature. That is what “singles out” one particular correspondence (RT&H, 46). No “magic” or dreaded “Platonism” is required. This should not be surprising. The world has structure that matters to living organisms. To perceive that structure at a distance is an obviously beneficial ability in Darwinian terms. It is no wonder that such an ability has evolved.

Although the phenomenon of perceptual constancy may explain how a percept has the representational content it does and corresponding accuracy conditions, I think it is less clear how percepts have intentionality; that is, how they get to be representations in the first place. This is the second of our two questions. It has two aspects. First, how does a percept get to be about whatever it is about? That is, how does it have the status of being representational at all? Intentionality—aboutness—is a mysterious property for anything to have. Nothing in nature except mental states appears to have original intentionality in the sense of being constitutively about something. I don’t think a successful reductive analysis of intentionality has ever been given, and I don’t know if one ever will be. Perhaps as psychology advances, we will come to understand something about how intentionality comes into existence, but for now I think it just has to be accepted as a fact. A fact it certainly is. You can think about your mother, about 2+2 being 4, about what you will do after you finish reading this, about how hungry you are right now, and many other things. These are intentional states. In addition, you are typically aware of their contents. This is the second aspect of this question. Our percepts as well as our thoughts typically have content that is manifest to consciousness. I have percepts as of my coffee cup, a book I hold in my hand, a tree in my yard, and I am immediately conscious of these contents. It could hardly be otherwise, given the central role of consciousness in guiding our behavior. This is not to say that the entire content of all of our percepts necessarily reaches consciousness, nor that the way it is presented is necessarily veridical. Rather, the point is that our percepts immediately present the world as having certain contents and being a certain way. Percepts are not “mental signs” (♣) in need of interpretation. The force of saying that percepts and thoughts have original intentionality is that their semantics is done; it is constitutive of their identity.

Obviously, on this point the contrast with Putnam’s picture could not be more stark. Putnam is working with a traditional empiricist picture of the relation between mind and world that descends from Locke, according to which the immediate objects of consciousness are always our own “ideas.” So, in sense-perception the immediate objects are our percepts, which in the case of vision are usually conceived as a two-dimensional field of color patches. Naïvely, we take this field for the world itself, but “the slightest philosophy” soon teaches us “that nothing can ever be present to the mind but an image or perception” (Hume, Enquiry, sec. XII). These are Putnam’s mental signs. They are not presentations of the world, not intentional states. Rather, they are the objects of presentation. And so it is a further task—the philosopher’s “epistemological position”—to determine their meaning, if any. Beyond them lies a vast black box, the contents of which can be discerned, if at all, only through inference. Maybe, if we’re lucky, there will turn out to be a unique correspondence, R, that links our mental signs to fixed referents in the black box. But how will we know of it? Only through the best theory we can construct, though this can be guided only by “operational and theoretical constraints,” which turn out to be pretty loose. Anyway, what could possibly “single out one particular correspondence R?” (RT&H 46).There are many correspondences, even many that satisfy all our operational and theoretical constraints. Therefore, unless we are willing to believe in magical “metaphysical glue” that sticks our percepts onto fixed referents, the only viable alternative is internalism, according to which our mental signs actually have no fixed referents in the black box. Accordingly, the mental signs of sense-perception have no specific contents. Fortunately, they don’t need to, even as a practical matter of survival (RT&H 38–41). We can construct a conceptualization of our mental signs that satisfies all of our operational and theoretical constraints and that enables us to live successfully in a conceptual world of our own creation. Thus, Putnam’s view devolves into the phenomenalism I described in the preceding post.

At the root of Putnam’s problems lies the old empiricist tenet that the ultimate empirical base of all theory and all thought are mental signs (or, for behaviorists like Quine, proximal stimuli). I would not go so far as to say that from this starting point, Putnam’s end conclusions are inevitable. For one thing, modifying the method of interpretation in any of the ways suggested by Putnam’s critics would block his proof of indeterminacy. For another, the impotence of natural mechanisms argument is weak. In spite of our attempt to give it its due, it turns out that, as a matter of empirical fact, we can give naturalistic causal explanations of the content of our percepts, the leading example being the identification of perceptual constancies in vision science. This should not be too surprising. After all, perception is one of Chalmers’s “easy problems” of consciousness. There is little reason to doubt that, as the science progresses, so will the causal explanations it is able to generate. It is odd that Putnam—and others—are so ready to believe an argument like impotence of natural mechanisms in the case of perception, when presumably they would never apply it to other natural phenomena. Would Putnam say we can’t explain the rain cycle or plants growing toward a light source because of the impotence of natural mechanisms? Perhaps this argument might seem to have more force if the mind is regarded as non-natural, although I’m not sure why this should make a difference. In any event, Putnam doesn’t seem to regard the mind as non-natural.

So, the empiricist underpinnings of Putnam’s view don’t render his argument and conclusions inevitable—but it’s easy to see how they help. If our empirical base was really a collection of meaningless mental signs beyond which was a giant inscrutable black box, our epistemological situation would indeed be dire. No doubt we have achieved our extensive empirical knowledge of the world only because our epistemological situation is in fact nothing like that ideology depicts. This shows the difference intentionalism makes vis-à-vis traditional empiricism. According to intentionalism, the objects of perception are not percepts, but particulars, properties, and relations in the external world. Thus, we start epistemologically with perceptions of many objects and features of the world that we wish to know and to know more about. Clearly, our perceptions need not be all veridical for this to be a good start on the road to knowledge. The world is not a black box to us. Also, intentionality begins relatively early in our mental life, both ontogenetically and phylogenetically, since percepts have intentionality. This makes sense evolutionarily. Animals without higher cognition presumably still have perception. This would seem to be common sense, and it is backed by scientific findings. According to Burge (2022, 8), the earliest creatures evolutionarily that are known to have visual perception (not just sensation) are arthropods (e.g., bees and spiders). Moreover, as noted previously, the ability to perceive distal features of the environment is highly advantageous from an adaptive point of view. Thus, intentionality begins with percepts, not concepts, and indeed presumably the intentionality of concepts is in part inherited from that of percepts. As I noted earlier, for Putnam and others it must be the other way around. But how concepts are supposed to derive their intentionality is a question these philosophers rarely if ever ask.

A last disanalogy between the semantics of first-order logic and that of our thoughts and percepts I will mention only briefly. It is that first-order logic treats sentences as fundamental and treats individuals and predicates as constituents of sentences. This may well make sense for formal languages and natural languages both. But it is not reflective of percepts and concepts/thoughts, where percepts evidently are the more fundamental form of awareness. As I mentioned a moment ago, percepts are developmentally earlier, both ontogenetically and phylogenetically. They do not have their being only as “parts” of concepts or thoughts, and their representational capacities do not depend on those of concepts or thoughts. Indeed, if anything it is the other way around. I only want to mark this point, not elaborate it. It seems worth mentioning in view of the dominance of the opposite view in so much of philosophy from Kant onward.

References

• Bays, Timothy. 2001. “On Putnam and His Models.” Journal of Philosophy, 98: 331–350.
• ———. 2008. “Two Arguments against Realism.” Philosophical Quarterly, 58: 193–213.
• ———. 2014. “Skolem’s Paradox.” The Stanford Encyclopedia of Philosophy.
• Bruce, Vicki, Patrick R. Green, and Mark A. Georgeson. 2003. Visual Perception: Physiology, Psychology, and Ecology. Psychology Press.
• Burge, Tyler. 2022. Perception: First Form of Mind. Oxford U.P.
• Dretske, Fred. 1995. Naturalizing the Mind. MIT Press.
• Field, Hartry. 1972. “Tarski’s Theory of Truth.” Journal of Philosophy, 69: 347–375.
• Lewis, David. 1984. “Putnam’s Paradox.” In Papers in Metaphysical and Epistemology, Cambridge U.P., 1999: 56–77.
• Palmer, Stephen E. 1999. Vision Science: Photons to Phenomenology. MIT Press.
• Putnam, Hilary. “Realism and Reason.” 1976. In Meaning and the Moral Sciences, Routledge & Kegan Paul, 1978: 123–138. [R&R]
• ———. 1977. “Models and Reality.” In Realism and Reason: Philosophical Papers, Vol. 3, Cambridge U.P., 1983: 1–25. [M&R]
• ———. 1981. Reason, Truth, and History, Cambridge U.P. [RT&H]
• ———. 1983. “Introduction.” In Realism and Reason: Philosophical Papers, Vol. 3, Cambridge U.P., 1983: vii–xviii.