Internal Language and Consciousness
Do the words of inner speech consist of mental images of outer speech? If I say to myself “It’s raining” do I entertain mental images of my saying out loud “It’s raining”? Are inward utterances strings of such images? If so, that would indicate that inner speech is the internalization of outer speech, with the process of internalization consisting in image formation. And that would imply that inner speech is dependent on outer speech for its existence and nature—that it is a derivative phenomenon. On the other hand, if inner speech were not image-like that would imply that it is independent and autonomous, perhaps more basic. It is certainly true that we can form auditory images of the sounds of outer speech, as when you recall how someone pronounces certain words; but are these what inner sentences are made of? I think not: the two things are very different. First, images fade with time, but not so the constituents of inner language—they are not like memory images of words you have heard uttered. Second, they don’t feel the same: using words inwardly is not like hearing yourself speak, or anyone else—it lacks the sensory specificity of such images. Third, it is not to be supposed that a deaf person entertains auditory imagery when engaged in internal dialogue, yet he or she may experience inward speech in much the same way as hearing people. Nor is it that the deaf have mental images of sign language while the hearing have mental images of vocal speech—though these may make an appearance in the minds of the people involved—but rather that their inner dialogue proceeds in a modality-neutral manner. Fourth, such imagery lacks the combinatorial power of words in the formation of phrases and sentences, while inner words combine in natural grammatical ways (how would images distinguish between nouns and verbs?). Finally, people may differ in their powers of mental imagery, some being notably deficient in it, but they don’t thereby differ in their ability to talk to themselves; lacking such imagery altogether is no bar to self-communion. So it would be quite wrong to identify the lexical constituents of inner speech with sensory images of speech events; they are not “faint copies” of heard sounds (or possibly seen gestures). Accordingly, it would be wrong to suppose that internal language use derives from external language use by the process of image formation—by a process of internalization.
What then is the phenomenology of inner speech? What kind of consciousness do we have of it? We are certainly conscious that we have such speech—it is not just a theoretical conjecture—but what is it like to talk inwardly? Here matters grow obscure: the phenomenology of inner speech seems curiously elusive and difficult to pin down. It hardly has a sensory feel at all, except around the edges. To judge from my own case, when I try to introspect its felt quality I come up with nothing definite, except that the harder I try the more I seem to force a sensory interpretation on my experience, as if desperate to find something concrete to cling onto. Most of the time the words slip by noiselessly, neutrally, blandly, a kind of wispy nothingness in the center of my thronging consciousness—a black hole in my sensory awareness. They are speech without the sound of the speech, experience without quality. Why is this? Here is a hypothesis: the internal language is largely unconscious. The lexicon is hidden from us, except for fragmentary connections to heard language, so that we can’t say what word in this language corresponds to the English word “sound” (say). We have no conscious experience of the words of IL (as we may call this internal language). We are conscious that we have such a language, but its actual nature is inaccessible to consciousness—both lexicon and grammar. We are familiar with the idea of unconscious computations conducted in a symbolic medium; well, our inner speech is conducted in a medium whose character is hidden from our consciousness. It is a bit like vision: conscious percepts backed by unconscious symbolic processes. Thus I know what I am saying in my inner code, but I don’t know the nature of the code—I don’t introspect it. The words are nothing like conscious mental images of auditory impressions, but rather inscrutable elements of lived experience, unmistakable yet elusive. The picture, then, is of an unconscious cognitive system that is not derived from external speech and does not share its medium; rather, it is more plausible to suppose that outer speech is the externalization of this system clothed in materials appropriate to the senses. The process of externalization imposes a sensory quality on our awareness of language, or at least one expression of it; hitherto it was largely closed to conscious awareness, possessing zero phenomenology. Language is fundamentally an unconscious phenomenon only recently reaching consciousness in virtue of externalization. It just happens that we are conscious of language in its sensory manifestation—and even here the sensory form is not to be identified with language as such. External perceptible speech is more of a symptom of language than the real thing. Consciousness and language are not natural bedfellows.
If that is true for inner speech, it is true a fortiori for what is called the language of thought. We should distinguish this idea from that of inner speech: LOT is not IL. The LOT theory, as propounded by Fodor, is motivated by considerations about the logical structure of propositional attitudes, particularly related to productivity: and any creature capable of propositional attitudes is credited with a LOT. It is not supposed that in thinking and belief formation we are saying the words of LOT to ourselves as we say the words of our inner language to ourselves. LOT is a theoretical construct not a datum of human conscious life (it is meant to apply to animal thinkers too, whether they engage in internal dialogue or not). LOT belongs at the level of computations in the visual system, i.e. not connected with consciousness at all. Our thoughts are indeed conscious but the symbolic medium in which they are encoded is supposed far removed from conscious awareness, so that we are not even aware that LOT exists. So the language in which we are alleged to think is not the same as the language in which we talk to ourselves: we might even call the latter language English (or Swahili), while that description is presumably false for LOT. The idea of LOT is logically independent of the idea of IL—neither entails the other. The existence of IL is a matter of lived experience, while LOT is a daring hypothesis unconfirmed by direct awareness. Accordingly, we can say that a typical human speaks three languages (individuating languages intuitively): LOT, IL, and EL. We have a language of thought, possibly universal among thinkers, human and other; a language for inner dialogue (or monologue), which is not possessed by all thinkers but which may be possessed by all normal humans; and a particular so-called natural language that is not universal to all human speakers (English or Swahili). These occupy different compartments of the mind, with two of the compartments hidden to consciousness and the other compartment open to conscious perception (at least as so manifested). The study of language therefore has three separate parts, with the third part distinctly marginal to the general phenomenon of language. Of course, it may turn out that these three compartments are at a deep level unified by a common set of linguistic principles, so that there is really only one language at work in all three areas; but prima facie we have three distinct linguistic faculties, operating in different realms and serving different purposes. Putting it chronologically, LOT is as old as thought itself, perhaps going back to the dinosaurs and beyond, while IL is of more recent vintage, date of origin unknown (let’s say two millions years ago), and EL is the new kid on the block, a mere 200,000 years old (and highly dependent on the larynx). Each resulted from suitable mutations driven by natural selection at vastly different times, but they each now sit in the human head guiding behavior: one language used as a vehicle of thought, another used to talk inwardly, and the third used as a means of communication. It would be quite wrong to think that the last adaptation is somehow more fundamental than the other two; they are simply different. It is as if evolution discovered language three times for different reasons and installed it in our nervous systems to serve three sorts of end. LOT is designed to enable thought to exist, IL is designed to improve thinking and solve problems, and EL is designed to get information across from one organism to another. Different functions, different languages—unless by some miracle they all converge on a set of linguistic universals. Consciousness plays a minor role in this grand narrative, being mainly clueless except in peripheral ways. Nor need these languages obtrude themselves on conscious awareness given their job description: basically they are combinatorial systems designed to generate unlimited output, thus enabling a potential infinity of thoughts, internal conversations, and meaningful utterances. These are different sorts of output, and the means by which they are generated is of no immediate concern to the organism benefitting from them (compare vision). We happen to have three linguistic organs or organ-systems installed in our brain, a necessary duplication in the circumstances–as we have two eyes, two hands, etc. It is widely accepted that we have a number of memory systems, also designed with different goals in mind, possibly with different evolutionary origins, and so language is hardly anomalous in its variety (we also have five senses). We pay more attention to our communicative system for various reasons, but that is no indication of theoretical centrality or evolutionary priority. Language extends far beyond this domain and penetrates into every corner of the human mind (as well as the minds of many animals). Its relative unconsciousness is no impediment to this ubiquity. In particular, inner speech is not to be identified with some kind of echo of outer speech, as the image theory would suggest; it is a robust mental faculty in its own right, plausibly underlying the very possibility of outer speech.
 We can also say, more generally, that the inner speech faculty is not the same as the imaginative faculty: imagination extends beyond language, obviously, but also internal language use is not a special case of imagination—which is not to say that the two don’t interface.
 Chomsky writes: “There’s good reason to believe that the inner language that we use all the time is inaccessible to consciousness, and that the fragments that reach consciousness reflect externalization—which, strictly speaking, is not part of language but of an amalgam of internal language and some sensorimotor system; it’s like the relation between a program in a computer and the particular printer that’s used to externalize it (so philosophers of language aren’t really discussing language).” (Personal communication, June 2020)
 I mean recent in evolutionary time: prior to the onset of vocal speech about 200,000 years ago language was silent and unperceived, not present to consciousness except obliquely; when it connected to the vocal apparatus it became an object of conscious apprehension (though not in its full nature).
 We speak this way because of the obvious connections to the language we grow up to speak not because IL is literally identical to spoken English (or Swahili), but LOT is not connected to our spoken language in that way, operating as it does entirely behind the scenes.
 LOT is regarded as constitutive of thought, part of its inner essence, while IL is best seen as an aid to thought, and strictly external to it. Thus one language operates to aid in the deployment of another language—IL helps LOT form the right beliefs. In other words, by speaking to ourselves we regulate what LOT will enter in the belief box. If we call the first language L1 and the second L2, then we can say that L1 helps organize L2 so as to serve the purposes of the belief system. That is, we get better beliefs if we use our inner speech faculty to think.
 A possible hypothesis is that IL is the origin of both EL and LOT instead of being quite separate from LOT. That would imply that LOT did not exist before IL came into existence, which would imply either that dinosaurs didn’t think or that they spoke to themselves. I suppose these are possible scenarios, but the more plausible theory is that thought came into existence before inner speech evolved, thus ruling out the hypothesis that IL is the sine qua nonof LOT. Still, we should keep an open mind on the question. Alternatively, it may be supposed that LOT is the basis of both IL and EL: I think this would be wrong for various reasons, but I won’t go into the question now.
 This paper is offered as a summation of several controversial lines of thinking not as a complete defense of the positions presupposed, for which the interested reader would have to look elsewhere.