Psycholinguist Steven Pinker: How a great scientist turned into an enemy of himself – and of truth

Steven Pinker’s book „The Language Instinct“ is certainly still one of the best books ever written on the rather elusive subject of language: comprehensive in its wealth of facts, intelligent in its argumentation and fascinating in the refreshing wealth of ideas.

No one wanting to have a say in the matter can possibly do without it although almost 25 years elapsed since it was published in 1995. It was written by a scholar who, as an admirer of Chomsky, was, of course, influenced by what is believed to be this man’s most important contribution to linguistics, that is the universalistic view that all languages are based on man’s generative capacity.

At the same time, however, this book illustrates a fundamental truth invoked by Thomas Kuhn in “The Structure of Scientific Revolutions”. Like all other mental activities, science is exposed to a peculiar danger. Provisional truths may easily be elevated to the rank of irrevocable paradigms, so that, at a certain stage, science may actively prevent its own progress and violate its basic ethos. In fact, Pinker’s book proves two things at the same time: it shows how fruitful a paradigm is whenever an intelligent author explores its most hidden depths and breadths, and how dangerous it becomes when it erects artificial barricades against criticism. Chomsky is known for his poor tolerance of opposition, Pinker seems to be acutely aware of this shortcoming – so he painstakingly avoids any step which could unleash the master’s scorn. That is why “The Language Instinct” so redolent with brilliant ideas turned out to be a stillbirth that up to the present time prevents linguistics from getting rid of a paradigm – or rather the straight-jacket into which Chomsky has enclosed it.

This becomes evident as soon as Pinker expounds the book’s main thesis to which he – following the footstep of others – gives the rather funny name “Mentalese”. Representing for Pinker a basic notion it is mentioned no less than 28 times! According to Pinker, language – every language – is based on a prelinguistic reality defined by him in the following way: „Mentalese: The hypothetical „language of thought,“ or representation of concepts and propositions in the brain in which ideas, including the meanings of words and sentences, are couched“ (p. 509).

A very important insight – and a relatively novel one too, first developed by Jerry Fodor in 1975 and six years later (much closer to natural language) by myself in „Grammatica Nova“! x1x But why does the author pusillanimously turn back from his courage adding the word „hypothetical“? Pinker definitely surprises the reader as his reservation implies that, after all, Mentalese may not even exist. But here, Pinker contradicts himself. At another place, he emphatically argues for the existence of such a prelinguistic reality asking the purely rhetorical question: „If thoughts depended on words, how could a new word ever be coined?” (S.47). Of course, it could not. Nor is there any mention of a fictitious reality when Pinker states: „Are our thoughts couched in some silent medium of the brain – a language of thought, or „Mentalese“ – and merely clothed in words whenever we need to communicate them to a listener?” (p. 45). In the first-mentioned quote, Pinker had already given a positive answer to this question. Spoken language invariably presupposes a prelinguistic reality, he therefore pursues in a perfectly logical way: „No question could be more central to understanding the language instinct“ (p. 45). This is what Pinker really means, and his position becomes even more transparent, when he summarizes it in the brief statement: „We end up with the following picture. People do not think in English or Chinese or Apache; they think in a language of thought“ (p. 72).

Here, there is no more question of a “hypothetical” prelinguistic reality, the latter is most definitely considered an undisputable fact – and rightly so since even the thinking layman never thought otherwise. When Englishmen call a certain tree „birch“, but Japanese call it „shirakaba“, then everybody will agree that no similarity whatsoever obtains between the acoustic vibrations of the air produced in the first case by an Englishman uttering the sound “birch” and in the second case by a Japanese saying “shirakaba” (nor is there any similarity between the symbolic lines on a sheet of paper representing the respective acoustic phenomena). There is but one Tertium Comparationis, and that is the cerebral idea in the minds of an Englishman or Japanese referring to the same kind of reality. Without this idea common to both languages, it would be impossible to mutually replace the two completely different acoustic „word marks“ or pictorial signs (rightly referred to by de Saussure as arbitrary). But, of course, this common idea (belonging to Mentalese) is the exact opposite of hypothetical: it is an ens realissimum, so to speak, the logical prerequisite of any translation.

If it makes any sense at all to distinguish a deep from a surface structure of language then the idea of birches to be found in the mind of both speakers obviously belongs to the first dimension while the two acoustic word marks belong to the second. This fact remains true even if we have to admit that up to the end of the 20th century ideas could only be subjectively perceived, they could not be measured, so their degree of reality seemed to be less pronounced – but even this may change in the near future: neurologists are on the way of proving that mental images such as birch, house, cloud, etc. imprint certain neural impressions on the human brain.

So, why does Pinker weaken his own point of view by speaking of a merely “hypothetical” reality? He does so, because otherwise he would be in danger of coming into conflict with Chomsky. If Englishmen, Chinese or Apaches – even before uttering a single word – „think in the language of thought“ (p. 72), then, of course, Chomsky should have based his so-called generative grammar on Mentalese because this would have been the true deep structure and generative foundation every historical language is based upon.

Pinker shrinks back from this obvious conclusion, although he knows better, and for the same reason he continues to use the almost comical term „Mentalese“, though a ready-made concept used from the very beginning of linguistics is at his disposal and would, indeed, be much more appropriate. Whenever an Englishman or a Japanese express the same mental reality through different word marks such as birch and shirakaba, linguists (and laymen too) used to say that these word marks have the same “meaning”. The generative background or Mentalese to which Englishmen, Chinese or Apaches all refer when they communicate with acoustic marks is therefore nothing more than “mental meaning” (beg your pardon for the pleonasm).

Thus, meaning constitutes the mental (and even neurologically impressed) reality on which all language is based. In an indirect way, Pinker expresses this truth clearly enough, but he recoils from replacing the curious term „Mentalese“ with the common word “meaning”, because he would again collide head-on with Chomsky’s paradigm. The generative mental source, his „language of thought“, would consist of prelinguistic purely semantic contents such as birch, house, car, etc. (bundled together in a comprehensive semantic class such as „substance“). „Meaning“ would comprise the entire „language of thought“ (Mentalese) as the true deep structure, while „form“ would refer to its material representation or surface structure consisting of acoustic vibrations or the signs on a piece of paper. 

This conclusion Pinker definitely embraced when contrasting meaning and form: „Knowing a language, then, is knowing how to translate Mentalese into strings of words and vice versa“ (73, 108). In my words: “Knowing a language, then, is knowing how to translate a deep structure of meaning into a surface structure of acoustic or other formally realized signs and vice versa.”

Moreover, the author is well aware that meaning on the one hand and form on the other are indeed „structures“ – each of them representing quite complex phenomena. Language does not merely consist of individual contents of meaning such as birch or house on the one hand and, on the other, individual acoustic word marks representing them on the level of linguistic form. Both meaning and form are highly structured entities. Pinker himself stresses the point with regard to Mentalese: „even a wordless thinker does well to chop continuously flowing experience into things, kinds of things, and actions“ (pp. 149, 150). In other words, Mentalese, the deep lying language of thought, is already highly structured before it pops up in form appearing as the surface structure of some definite language like Chinese, English, Japanese etc. (It should, of course, be added that only the overall structure of meaning is the same for all languages – they all distinguish substances from actions, qualities etc. and differentiate questions from statements and so on – but the mental analysis of reality giving rise to the structure of meaning is susceptible of almost infinite variety on a more concrete level!)

The difference between Mentalese or the general structure of meaning and its possible formal realization may be illustrated by means of basic examples. In a sentence like „Men eating rice (on Wednesday generally) eat rice (on Tuesdays too), the same core of logical meaning „men, eat, rice“ is to be found twice in the formal surface structure, namely, first, as an independent sentence and, second, as a dependent clause. „Men eat rice“ represents the independent occurrence while “men eating rice … ” may only occur as part of a sentence. To this difference on the formal level corresponds a difference on the deeper semantic level, which is not logical but “informational” (see Principles of Language revised). Obviously, the formal contrast between a “nominal phrase“ (men eating rice … ) and an independent sentence (men eat rice) does not explain anything – rather it is itself in need of explanation. Such explanation is provided by the “informational” difference on the level of Mentalese, that is the semantic deep structure. In order to be understood on the level of formal realization, this difference in meaning must be unambiguously expressed by the surface structure as well. We may add that the informational semantic dichotomy in question needs to be formally realized in all developed languages; in some of them, such as English, even two formal alternatives are available (a) men eating rice … , and b) men who eat rice … ).

Similar considerations come to the mind with regard to formal differences like, for instance, the English noun „withdrawal“ as opposed to the English verb „to withdraw“. The core of logical meaning is identical in both cases: it is an action (like to run, to go, to give etc.), so why can it be represented on the level of form either as a noun or a verb? Otto Jespersen was the first to try an explanation (namely “rank lifting”). But Chomsky was by his very method prevented from finding it as he would have to dive deeply into Mentalese (i.e. the structure of meaning), more precisely into its informational requirements (see Principles revised). As long as we cling to formal surface structure (verb/noun), nothing at all is explained – not even the fact that not all languages allow such an alternative realization as both verb or noun.

It would have been easy for Steven Pinker to adopt these insights – they are obvious and he has already hinted at them in the above-mentioned quotes. But then he would have been forced to add that verb, noun, nominal phrase, etc. cannot be concepts of Mentalese, i.e. the true depth-structure of meaning, presupposed by the speakers of all languages in their mental analysis of reality. In the statement quoted above that „even a wordless thinker does well to chop continuously flowing experience into things, kinds of things, and actions“, Pinker rightly omits mentioning verbs, nouns, nominal phrases etc. (pp. 149, 150).  In other words, he would have been forced to admit that nouns, verbs, adjectives, nominal phrases etc. have no place in Mentalese or the structure of meaning, because they belong to the level of formal realization. By this admission Pinker would, of course, fundamentally contradict Chomsky.

It was a red line the author of the „Language Instinkt“ did not dare or did not wish to cross. For this reason, his book represents more than just a missed chance – it is nothing less than a flawed paradigm that presents a formidable obstacle to the advancement of linguistics. Pinker writes against his own better knowledge as he constantly cowers before Chomsky’s overpowering shadow.

Why is Pinker so afraid of defending his own convictions? Why couldn’t he put his mind at ease?  After all, Chomsky has achieved so much more than most of his colleagues. His undeniable merit is to have prepared and paved the way for modern linguistics’ greatest achievement: automated translation. A machine need not know that birch and shirakaba “mean” the same thing. Machines only deal with material, measurable elements – so they are barred from knowing anything about meaning. They simply receive the order to replace the formal element „birch“ with the formal element „shirakaba“ when translating it into Japanese. And they proceed in basically the same way when they come across a nominal phrase such as „Men eating rice” (are usually healthy) in English replacing it with Japanese „Rice eat men (are usually healthy)“. The formal realization used in English is merely replaced with the formal realization prescribed for Japanese. Thus, Chomsky’s formal surface structures have become the indispensable tools for translating any language A into any other language B.

In this respect, Chomsky is nothing less than the founder of modern linguistics as a practical instrument. For a single man’s honor and fame this should be praise enough. But not content with this great achievement, Chomsky wants more – and this is where he failed. Chomsky definitely does not live up to his own claim of having created a truly universal and generative grammar. He never came to terms with the structure of meaning or, in Pinker’s words, the “language of thought”. Nor did Chomsky explain how Mentalese arises, i.e. how the human brain proceeds in the analysis of reality in order to arrive at a general structure of meaning. And he does not deal with the formal constraints the structure of meaning has to obey at the moment of its realization as sequences of sounds (for only single word marks are “arbitrary” in the sense of de Saussure). Understanding language, beyond its automated transfer from one surface form to another, implies looking into the relationship of meaning and form (Mentalese versus strings of sound). This dimension has altogether escaped Chomsky’s attention.

Steven Pinker is far too intelligent not to have gained these rather basic insights, in fact he is well aware of them. That is why he says on page 95 that he is appalled when coming across all those “countless popular books on mind, language, and human nature that refer to ‘Chomsky’s deep structure of meaning common to all human languages’.” Right, that is what Chomsky should have said but what he definitely did not say. And the reader will now understand the two main reasons why Chomsky was unable to do so. If he had chosen as the true deep structure the “general structure of meaning”, he would have had to abandon all familiar concepts of traditional grammar like nouns, verbs, adjectives etc. as these belong to the surface. His primary task would have consisted in creating entirely new ones. But these new – purely semantic – concepts would, of course, be useless in machine translation precisely because they refer exclusively to the sphere of meaning. Machines only describe transformations on the level of surface structure, they have no place for meaning. So, renouncing traditional surface concepts, Chomsky would not have become what he actually is: the father of machine translation.

Now, you may understand the precarious situation Pinker had to confront when he was writing “The Language Instinct”. He understood clearly enough that the real language of thought could not be described by categories belonging to the surface of any given natural language (“People do not think in English or Chinese or Apache; they think in a language of thought“ and he was perfectly aware that “even a wordless thinker does well to chop continuously flowing experience into things, kinds of things, and actions“). But he recoiled from the inevitable conclusion that Chomsky’s method to evolve a universal grammar was simply invalid and wrong as it relied on concepts belonging not to the semantic deep but to the formal surface structure of language!

I tried to outline such a deep structure of meaning and the processes and restraints governing its representation („formal realization“) on the level of surface structures. Unfortunately, Pinker either did not know or did not bother to mention my work. The “Language Instinct” was published in 1995. My thoughts on this topic first appeared in 1981, more than ten years earlier – albeit in German (“Grammatica Nova”). “Principles of Language” was, however, published in English in 1993, two years earlier than the “Language Instinct”. Everything Pinker had to say about Mentalese was explained in the Principles (above all in the new version Principles revised– in detail and due emphasis on broad detail.x2x Last year, I tried to draw Mr. Pinker’s attention to this fact writing a letter, which, I believe, was as polite as it should be between scholars with similar interests. Was it because of arrogance or of helplessness that to this very day I never got an answer?


x1x You may have noticed with disapproval that I mentioned Jerry Fodor only in passing. Indeed, my own thoughts on the matter originated quite independently. No specific theory of mind is implied in my work on the “General Structure of Meaning”. The latter is simply – I don’t mind if you call it “naively” – accepted as a given fact and described as such in its twofold aspect of what I call its logical and informational parts (the latter certainly leading to the more original suggestions). In other words, the “General Structure of Meaning” serves the sole purpose of providing the indispensable fundament for truly explaining the generativeness of natural languages. Nevertheless, you will want to know what distinguishes the ”General Structure of Meaning” from Fodor’s LOT (1 or 2) or his and Pinker’s “Mentalese”? My answer is based on a purely linguistic and quite unequivocal criterion: “formal relevancy”. Only those parts of the semantic deep structure belong to the “General Structure of Meaning”, which will obligatory be expressed by specific formal means in the surface structure of every developed natural language. The “action synthesis with or without spatial and temporal determinants” thus belongs to the logical part of the semantic “General Structure”, so does the informational dichotomy giving rise in the surface structure to English “men eat rice” versus “men eating rice (or “who eat rice“)….“ Such formal distinction is obligatory in every natural language. But this does not apply to the semantic distinction found in Japanese that expresses the rank of the speaker with regard to that of the person addressed: it therefore does not belong to the “General Structure of Meaning“. 

x2x Pinker only superficially touches upon the actual difficulties raised by pre-linguistic Mentalese or the “Language of Thought”. Merely in passing does he say that ”even a wordless thinker … chop/s/ continuously flowing experience into things, kinds of things, and actions”. In fact, language presupposes a twofold pre-linguistic activity, consisting of both analysis (chopping continuously flowing experience) and synthesis built upon it. „The bear now runs towards the abyss“ is a unique event on the level of sensual experience.  Mentalese, however, has analytically broken down the sensual totality into constant mental images (bear = substance, running = action, now = temporal determinant, towards the abyss = spatial determinant), which it afterwards recomposes into a synthesis made of these images.

Difficulties do not end at this point. Neither the unique sensual experience nor its mental reconstruction in Mentalese or the „Language of Thought“ represents a „sentence“ – as sentences, nominal phrases, etc. denote structures belonging to the level of formal surface. For this reason, we need a new term. I speak of a synthesis, more precisely of an “action synthesis extended in time and space“ that reverses the preceding analytical segmentation of “continuously flowing experience”.

An action-synthesis like “The bear now runs towards the abyss” certainly belongs to the core of Mentalese, that is to the general structure of meaning (Mentalese) as it may be expressed in all developed natural languages. Besides the general core many more semantic contents may, of course, be added. In some languages the speaker is required to specify whether he means a male or female bear, whether the latter runs towards him or away from him, whether he addresses a person of higher, minor or equal rank, whether he believes that the statement represents a fact or is known only by hear-say and so on. Pre-linguistic Mentalese may thus be said to be at the origin of both the general core and a true infinity of semantic specifications characterizing specific natural languages.

To be sure, at this level we do not yet speak of interlinguistic differences arising at the level of formal realization, that is when Mentalese is expressed in orderly strings of acoustic signals. Differences such as between isolating, non-isolating, agglutinating, polysynthetic languages etc. belong to the level of formal realization above Mentalese. The ordering of formal elements as such is quite independent of semantic differences; it obeys its own specific laws (as I have tried to show in my work: The same structure of meaning may be expressed in various formal ways – a fact made quite evident in computerized artificial languages where different operating systems are freely used to express identical contents. Japanese, Russian and Chinese are, so to speak, different operating systems but they may convey the same meaning). This is the main point: meaning and form are subject to laws originating independently from each other in their own spheres, and both together create the tremendous variety of natural languages.

These are basic facts of meaning and its expression in linguistic form which Pinker does not speak about – and he certainly need not do so. After all, everybody is free to choose his or her particular field of interest but it was never deemed to be good and responsible science to just overlook or belittle insights that contradict one’s own paradigm, especially if the latter turns out to be a hindrance to a deeper understanding of language.