Perception, Thought, and Language
I will reflect loosely (though not I hope sloppily) on some very general properties of the items listed in the title, with special emphasis on their interrelations. There has been a tendency to assimilate these three things, as if they are all variations on a single theme. Empiricism does this conspicuously: concepts and meanings are regarded as versions of percepts. The senses are taken to be primary and thought and language are modeled on their deliverances. Each type of faculty deals in sensory contents: impressions and copies of impressions, basically. This type of view has been extensively criticized for excessive homogeneity. More recently, language has been taken as primary: concepts are to be linguistically defined, and concepts enter into perception. The mind’s basic form of representation is in language and all other forms incorporate it. Thus people see the world according to the categories present in their language (which usually means speech): they think in language and their thinking shapes their perception of reality. Sometimes theorists bracket one member of the trio, assimilating just a pair of them: thought and language are assimilated but not perception, or perception and thought are assimilated but not language (I don’t know of a view that assimilates perception and language but not thought). It seems to be supposed that deep commonalities must pervade the three faculties; they couldn’t be quite independent of each other. It is allowed that other faculties are separate, such as emotion or desire or will, but it is thought that perception, thought, and language are necessarily bound up with each other; they form a natural unity. The reason for this is perhaps the following: perception leads to belief, which in turn leads to assertion. You see an object, form a belief about it, and then make a statement expressing your belief. The three faculties work together, so percepts, concepts, and meanings must share a basic nature. Hence we have the idea that concepts and meanings are copies of sense impressions, or the idea that concepts and percepts are variations on words. Either we are fundamentally sensing beings or we are fundamentally speaking beings.
It seems to me that this leveling tendency is misguided: there is an irreducible plurality to the three sorts of faculty. First, let’s ask what the purpose of each faculty is. I trust I will not be thought eccentric if I suggest the following: the purpose of perception is to gain information about the world; the purpose of thinking is to solve problems; and the purpose of speech is communication. These are not at all the same purpose. If we think about it from the point of view of evolution, perception evolved so as to get the benefit of up-to-date information about the environment, thought evolved so as to solve problems presented by the environment, and speech evolved so as to exchange information about the environment (as well as possible solutions to problems). Thought uses the information provided by perception in order to solve problems, while language transmits this information to other creatures, along with the conclusions of reasoning. The three faculties intelligibly interact, but it doesn’t follow that each is a variant on the others. They presumably evolved separately and at widely different times (speech evolved only recently), and they serve very different ends. I want to explore the possibility that they are quite unlike in their basic constitution.
The idea that perception and thought are fundamentally different is familiar under the heading of modularity: the senses are not modifiable by the cognitive faculty. I will add to that two other properties possessed by thought (and language) but not possessed by perception: predication and productivity. In thought and speech we predicate properties of objects: we think and say that some identified object is thus-and-so. But we don’t predicate in perception; rather, we are affected by things in a certain way. True, we can be said to represent things in perception, but the same can be said of pictures yet they don’t predicate either. Predication is a certain kind of symbolic act, allowing for negation, truth and falsehood, truth-functional combination, etc., but perception is not capable of this kind of complexity; it is architecturally more primitive than that. Perception presents arrays of properties; it doesn’t single out properties for attribution as propositions do, whether mentally or linguistically. Predication presupposes an evolutionary advance over perception, enhanced cognitive powers. Connectedly, perceptual primitives are not productive in the manner of thought and speech: we don’t have the phenomenon of infinite potential, recursion, grammatical rules, verbs and nouns, transformation, nonsensical strings, etc. If perception were a form of language (“the language of sense”), we would expect that it would display the properties characteristic of human language, productivity chief among them: but it doesn’t. Thought has this kind of productivity (but see below), but perception doesn’t. Thus perception can never be the basis of concepts and meanings, contrary to empiricism. But likewise language and thought can never be the basis of perception, since it lacks productivity. Taste and smell, say, are nothing like thought and language, and the same for the other senses. True, there are sensory primitives, but they are not word-like—they don’t combine into grammatical wholes with meaning and truth-value. They have neither semantics nor syntax. The senses might be said to occupy adjacent mental cells to the cells occupied by thought and language, with communication between cells, but the occupants are different kinds of being. Seeing red, say, is not part of a sentential structure with all that that implies. Predication and productivity are not properties of perception.
Yet they are properties of both language and thought. Does that imply a fundamental unity between the two? There are two possible ways in which such unity might be thought to obtain: speech is essentially the expression of thought, or thought is the internalization of speech. It seems to me that these are at best exaggerations. No doubt speech evolved for the same kinds of reasons that animal communication systems evolved (which is not to say that language evolved for these reasons), viz. to share information about the environment and the animal’s inner states, as well as to intimidate, impress, cajole, command, vent, etc. There is no necessity for thought to exist in order for these functions to be performed by speech, i.e. the kind of problem-solving thought that mature humans possess. So speech is not necessarily the expression of belief. Also, speech need not replicate the structure of thought in order for it to perform its function—it could be a lot more primitive (as much of our speech actually is). It is true that human language has the same kind of structure that thought appears to have (though we know little about the nature of thought), but there is no necessity about the two sharing the same defining traits: so language is not required to mirror thought. Speech is a sensorimotor system that evolved separately from thinking and serves a different purpose, so there is no a priori reason why the two should share their structure. As to the idea that thought is internalized speech, that theory would certainly entail that thought and speech share their inner architecture, but it is highly implausible: we don’t begin by speaking and end up thinking by some process of internalization (still less do other animals). What is true is that human thought and speech seem to be made for each other, natural partners, but that does not entail that concepts are words or words concepts. People can think in the absence of speech and speaking (vocalizing) doesn’t require thinking. Problem solving is not the same as communicating. Speaking is a sensorimotor skill designed for communication; thinking is an internal capacity designed for reasoning. The two are joined, but they are not variants of each other. So it would be wrong to assimilate thinking and speaking. Thinking is silent while speaking is noisy, and there is a deep reason why this is so, issuing from their very different purposes. It could easily be that the first forms of human speech were disconnected from our problem-solving capacities, as is the case for other species with problem-solving abilities and communication skills; only later did communicative speech and thinking become connected. Thinking and speaking are not indissoluble talents possessed of the same basic structure.
It therefore seems to me that perception, thought, and language (speech) are separate faculties of mind, as ordinary language suggests. It is wrong to assimilate them. They are connected, to be sure, but they are not versions of each other. Each is composed of elements that do not compose the others, and their combinations don’t follow the same rules (thoughts cannot be nonsensical or ungrammatical in the way sentences can be, for example). It is actually surprising that thoughts and speech acts can share such properties as truth and reference, given how different they are (perception does not have these properties, though it may have analogues of them). Maybe their application to speech is entirely derivative upon thought and is not literally correct (“derived intentionality”). Sentences can be grammatical or ungrammatical, but can they really be (intrinsically) true or false? In so describing them are we thinking of them as embodied thoughts in some way? In any case, it is an interesting fact that we so readily talk in this way, given how different thoughts and utterances are—as different as both are from percepts, one would have thought. This is not something to be taken for granted but rather to be explored and possibly questioned. The default assumption should be that all three faculties are made of different stuff, organized differently.
 Notice that I say “speech” not “language”: the reason is that I am not concerned with language in the sense that includes a putative language of thought. Such a language would clearly be constitutive of thought by definition. I am concerned with language as most philosophers use the term, i.e. public spoken languages like English. Even if there is a language of thought, its syntax and semantics might be quite different from external spoken languages, so there is no guarantee that spoken language would provide any insight into this internal (presumably innate) language.
 Language as an abstract structure could have evolved as an aid to thought not as a means of communication, possibly coexisting with a pre-existing system of communication (as Chomsky has long urged). A language of thought could evolve without having any input into communication. Speech as a sensorimotor system, on the other hand, almost certainly arose as a communicative vehicle, which is why it is audible (or visible in some cases).
 There is a well-known tradition that regards language as merely a more or less inadequate means of expression of thought, which is in its nature more nuanced and sublime than language can ever be. According to this tradition, language can never be the stuff of thought, only its inept vehicle; it is what we must perforce resort to in trying to get our thoughts across to other people. Such a view certainly seems plausible for animals and their means of expression: the inner lives of animals are clearly more complex than their relatively primitive signal systems can convey. Assimilating animal cognition to animal communication would be particularly unappealing. It is only in humans that language (speech) even approximates to capturing what is going on at the level of thought.