"It does not matter whether Warsaw was on fire. What does matter is that our rabbi can see all the way from here to Warsaw," said the disciple to the community. He was speaking in the spirit of the anarchist theory of science after the rabbi's claim that Warsaw was burning turned out to be false. Of course, the story is much older than Noam Chomsky's writings, but it could just as well apply to them. "It does not matter whether I am right or not; what does matter is how great my ideas are." I doubt that this sentence occurs literally in Chomsky's books, but it expresses their siprit.

Chomsky's great merit is in developing the Humboldtian idea that all languages have equally complex grammatical structures. He played a key role in making this a self-evident truth, at least to linguists. It is partly due to Chomsky that they have come to agree that the mental capacity which allows us to acquire our native tongue is similar in every healthy individual of our species. It is certainly innate and forms a boundary to the formal, structural characteristics of human languages. That is, the indisputable diversity of human languages has nothing to do with differences of race or culture; rather, it corresponds to variations licensed by universal grammar which are essentially due to historical accidents.

On the other hand, Chomsky also contributed to the transformation of linguistics as a discipline into a branch of science. From the nineteenth century on, linguists have made use of formulae to capture regularities. But the American descriptivist school and Chomsky were the first to overtly label linguistics as a branch of science rather than as part of the humanities. Just like natural sciences, linguistics uses mathematical tools for constructing models of the working of complex linguistics systems which are uncovered by collecting data and experimenting.

Chomsky's Syntactic Structures (1957), which has just appeared in Hungarian, takes us back to the period when Chomsky first formulated these revolutionary ideas about linguistics. (Later he became completely bogged down with the very problems of syntax that the school he established had created.) Syntactic Structures documents the heroic attempt of Chomsky the mathematician formally to encapsulate the views of the American descriptivists. He tries to formalize the descriptive methodology developed by Bloomfield, Hockett, and Harris--i.e. to create a mathematical model that will make their ideas explicit. He failed. Since then, several commentators have pointed out that Chomsky's reconstruction of the method of immediate constituents was not true to the original ideas of the founding fathers. In the book, Chomsky tries sketchily to demonstrate that the model thus developed is inadequate to account for certain phenomena of natural languages. (As a matter of fact, most of these arguments have also been proved wrong since then.) Instead, he sketches a model based on so-called "transformation rules", which he explains in great detail and illustrates with a large number of examples. Throughout the book, he has harsh criticism for the psychological theory underlying his precursors' views. The American descriptivists adopted the aims of behaviorism, which tried to reduce higher-order intellectual capacities to more basic functions, namely, to conditioned reflexes. For Chomsky, the evil of immediate constituents is linked to that of behaviorism, and he wants to get rid of both at one and the same time.

The second part of the volume contains Language and Mind, an unabashedly ideological summary of the history of linguistics and philosophy of language. It attempts to demonstrate that the "rationalist" trend in linguistics has led straight from Descartes to Chomsky himself. It contains the transcript of three lectures, one about The Past, one about The Present and one about The Future.

Its main claim is that the "rationalist" (more closely, Chomskyan) tradition offers the only chance for linguistics to contribute to the philosophical and psychological understanding of the human mind. It takes an almost religious (or rather, dogmatic) faith to believe this of Chomsky. All the more because, as a rule, he is wrong in details, where the proverbial devil dwells. I am not talking about the philological details now, although it would be worth examining those as well. Chomsky's paraphrases of Humboldtian and Cartesian ideas may not all be faithful. But here I will concentrate on his personal ideas about language.

It is hard to tell whether the naive reader will be able to figure out what a mischievous attitude Chomsky has to scientific questions. I will illustrate this in what follows. For example, just read the pages following the Introduction. According to Chomsky's definition, a language is a set of sentences, i.e., of word sequences. It is worth reading Syntactic Structures with this definition in mind and checking just how many times he uses the word language in this technical (mathematical) sense, and how many times in its everyday or philosophical senses. The latter concepts involve considerably more than the above definition. Above all, they do not just refer to sentences as signs that can be written or spoken, but as associations of expressions with meanings, and maybe much more.

But let us dwell on the first pages for the moment. On the basis of the examples (Chomsky 1957: 3-5), the author asserts that grammatical correctness may have formal criteria which cannot be denied. For example, yes/no questions in English have to start with an auxiliary verb. Skipping a couple of pages, we find the following statement: "I think that we are forced to conclude that grammar is autonomous and independent of meaning" (Chomsky 1957:17). Do I have to argue in detail about the radical difference between the two statements that I have quoted? On the one hand, he says that grammatical correctness has some formal criteria (independent of meaning). On the other, he says that grammar as a whole is independent of meaning. Do you need to be reminded to what extent this argument depends on whether we adopt the ad hoc definition of language, which defines it as a set of sentences (rather than, say, pairs of sentences and meanings)?

Another favorite idea of Chomsky's is that the data at the child's disposal do not allow him or her to form inductive hypotheses about the grammar of his or her native tongue and to acquire a knowledge of language in this way. We find this statement (sometimes called the "poverty of the stimulus argument") in several places in Chomsky's writings. One of the reasons why this idea is interesting is because it serves as an argument for Chomsky's innatism. According to Chomsky, language acquisition presupposes innate mechanisms that are completely independent of the other intellectual abilities of humans. But there we find no direct or indirect evidence whatsoever for this claim. In Syntactic Structures, as in the other Chomsky books, the only statement used to support this is that the traditional behaviorist theory of learning allegedly cannot account for language acquisition.

For my own usage, I have termed this type of argument the "crop circle argument". It is commonly known that people sometimes find circular patches in wheat fields the origin of which is mysterious. Science has offered no explanation so far for this phenomenon. Every scientific explanation we can think of (special winds, particular behavior of birds, etc.) is highly complex and unlikely to be true. Nobody ever observed the coming into existence of crop circles (except two youngsters who admitted having produced one themselves). However, the UFO people have a straightforward explanation for the phenomenon, in a framework that they think can be motivated independently from a legion of other facts, and which nobody has falsified so far, namely, that UFO's taking off through an opening from below the surface of the Moon regularly land with their circle-shaped spaceships on the planet Earth, and if that happens to take place on a wheat field, it results in a circular trace. With this analogy, I would like to point to the fact that Chomsky's arguments, although he has the merit of classifying linguistics as a branch of science, are often similar to "new age" science in their style.

To show how Chomsky thinks of the role of experience in the use of language, let me quote a longer piece of nonsense test here, from Chomsky (1972:11-12):

[...] The normal use of language is innovative, in the sense that much of what we say in the course of normal language use is entirely new, not a repetition of anything that we have heard before and not even similar in pattern--in any useful sense of the terms "similar" and "pattern"--to sentences or discourse that we have heard in the past. This is a truism, but an important one, often overlooked and not infrequently denied in the behaviorist period of linguistics to which I referred earlier, when it was almost universally claimed that a person's knowledge of language is representable as a stored set of patterns, overlearned through constant repetition and detailed training, with innovation being at most a matter of "analogy".

In reality, the exact opposite of this is true. Obviously, every sentence that we can understand is similar to an astronomical number of sentences heard earlier. This is the case already in early childhood; by the time we understand articulated utterances, we have heard an incredibly large number of (mostly simple) sentences. Moreover, if we look at child language in the first two years, we will see that the child systematically practises the newly-acquired patterns. Finally, we can observe that both in child and adult language, innovations are based on analogies, although, in Chomsky's view, "to refer to the processes involved as "analogy" is simply to give a name to what remains a mystery" (Chomsky 1972:36). Chomsky could go on to argue about various interpretations of the concept of analogy, but he does not, he simply states this as a fact. One of today's main trends in artificial intelligence research is based on the concept of analogy. According to this, intelligent problem-solving behavior consists in discovering the similarities between the new problem and problems solved earlier. The process of solution involves the adaptation of the old solutions to the new problem using the principle of analogy.

Chomsky's ideas on the acquisition and use of language are relevant because they also underpin his philosophical views. Indeed, if his views on analogy were correct, then philosophy and psychology should be more innatist than they are today. But Chomsky is demonstrating pure arrogance in using the properties of his own model of grammar (which, as I have mentioned, he proposes as an improvement on the theory of immediate constituents) to justify the idea that human language cannot be acquired using our innate intellectual capacities.

It may not be immediately clear how the characteristics of Chomsky's "transformational generative grammar" are related to the acquisition and use of language. Let me briefly explain this. Chomsky's transformational grammar is based on the assumption that sentences have several structures. Some are closer to the structure of their meaning, while others resemble more their surface structure. Transformations are regular relationships that hold between these structures (also called "levels of representation"). Without going into mathematical details, let me emphasize that the set of grammars which produce sentences using such transformations are significantly more complex than the types of grammars Chomsky attributes to his precursors. These grammars fall into simpler mathematical classes. Mathematically complex rule systems are more difficult to acquire and handle (if they can be acquired and handled at all) by finite devices like a human brain.

Ironically, Chomsky is very famous for his work in precisely this area of mathematics. He invented the so-called "Chomsky hierarchy", a formal classification associating types of languages and the devices to handle them. The hierarchy runs from the simplest type of languages, called regular languages (which can be handled by finite state machines) to what are termed recursive languages, which can be analysed using Turing machines, the most powerful machines imaginable. As we move from simpler to more complex devices or abstract machines, analysis becomes more problematic. Only finite state automata are well-behaved: they can analyse sentences in linear time (the time/space required for analysing a sentence is proportional to its length). As we go further up in the hierarchy, the complexity of the analysis problem increases. Most unrestricted languages are so complex that it is impossible even to determine whether a Turing machine will be able to analyse them or not. Now, this is the class containing languages produced by transformational grammars. And Chomsky, not unlike other apostles of the science of the "new age", uses his authority as a professor of mathematics to mislead the reader by being silent about the detailed nature of the type of grammar he is proposing.

This raises two problems, one theoretical and one practical (not to mention: moral). The theoretical question is that the human brain is a huge yet finite device, so it can only carry out operations that require a finite number of states. Human languages must thus be regular. Chomsky rejects this problem by arguing that the non-regularity of human languages is an assumption that simplifies description (and consequently, acquisition and use of) human languages (i.e., regular subsets of their theoretical model). But it is not clear how a finite device can arrive at the assumption that a language is very general (say, type zero), and thus make it easier to generate a regular subset of that hypothetical language. Chomsky does not go into details about this. The practical/moral problem, on the other hand, is that Chomsky, who rejects an entire theory in one sentence on the basis of its alleged "complexity" and "arbitrariness" (cf. Chomsky 1957:23), completely ignores the mathematical, well-defined complexity of the solution he himself proposes.

Moreover, from the Chomskyans' perspective it is disappointing that those of today's computer systems which show any sign of "intelligence" (e.g., which can recognize a hand-written address on an envelope) all work on a stochastic, inductive basis instead of applying rule systems. The same is true for computer programs that are supposed to "learn". This proves nothing about the operation of the human brain; what it does show, however, is how mistaken Chomsky was, in the sixties, when he tried to foresee the future of science.

Chomsky's thinking is an example of the fact that the style of science characteristic of the "new age" need not be accompanied by a wish that reality contains more entities than we can justify using "traditional" science. The "new age" science can also go hand in hand with an anarchist view of scientific development. That is, it can proceed with the view that earlier ideas can be rejected without rigorously1 proving them false. I am not sure whether this can be related to Chomsky's anarcho-syndicalist political views. Maybe it would have been a good idea for this new translation to include a political pamphlet of Chomsky's, to give an opportunity to reflect on possible connections.

According to László Kálmán, Chomsky's "ars poetica" is: "It does not matter whether I am right or not; what really matters is how great my ideas are". He goes on to say that Chomsky is the world champion of scientific mischief. And so on, and so on. This is a sad and rather provincial affair. For Chomsky is, without doubt, one of the greatest scientists of our times. To speak of such a major figure as if he were a half-witted liar at the very least places a question mark over the rationality of the enterprise. It is possible, of course, to disagree with Chomsky; for the most part, I have done exactly that, and have even written a book explaining why. But the sort of arguments that Kálmán resurrects have been repeatedly refuted in detail in respected international publications. The one-sided and gratuitously aggressive popularization of these arguments in the lay market does not serve the academic cause, nor indeed any cause at all.

Kálmán is not alone in failing to understand what Chomsky is getting at; this is to some extent an international phenomenon. However what is special about Kálmán is that he writes as if he does not read. I mean, as if he has not read what he is writing about.

Kálmán could easily have written this review in the sixties: it is obvious to the expert that he has not followed the relevant literature published since then. He argues, for example, that language is not a set of sentences and lists of words, as Chomsky says, but perhaps a set of sentence/sentence meaning pairs. He does not even mention that although Chomsky occasionally used such a definition in his earlier works, he later systematically rejected it, and now uses the term "language" as a synonym for "grammar"--an element of the mind or brain. (If I "know Hungarian", then it is not that I "know" an infinite list of words, but I have the mental capacity to form Hungarian sentences.) Or perhaps he illustrates the concept of transformation (omitted in the English version) with examples no one in the field has used for about twenty years, or would consider as a possible transformation, because he is reviewing a very old book? Or he writes that Chomsky "perhaps does not speak of a deep structure any more" (omitted in the English version). Does he or doesn't he? Perhaps it would be worth finding out.

So Kálmán did not pursue the literature closely, presumably because in his opinion it is not worth it. He thinks that there are fundamental problems with the whole approach; he says that Chomsky "later became completely bogged down with the very problems of syntax that the school he established had created." The huge and divergent school founded by Chomsky is that of generative grammar. Most of Kálmán's work also belongs here. Within this, Chomsky played a decisive role in, amongst many other things, the development of the principles and parameters theory and the more recent "minimalist" model. Kál mán blandly asserts that he does not understand these theories and does not like them.

Chomsky is still very much alive and active. But I doubt that he often reads the Budapest Review of Books. He does, however, read international journals, and systematically and conscientiously responds, again and again, to the sort of arguments that Kálmán digs up, often far more seriously than the arguments deserve. Off the cuff, at least three of his books spring to mind, in addition to the larger-scale specialist works, in which he has dealt in detail since 1975 with a variety of general reflections and objections including those raised by Kálmán.

Should I go into more detail about the problems with Kálmán's reasoning? Is this of interest to the fond reader of the Budapest Review of Books? In case it is, I shall say a few words about one or two of the questions he discusses. Kálmán accepts the popular dogma which says that the innate capacity to learn languages (he does recognize the existence of this) is part of some general intelligence. He thinks that this capacity has no separable, language-specific components--contrary to what Chomsky claims. Language acquisition occurs with the help of general intelligence. As to how this happens, Kálmán, not surprisingly, has little to say, apart from attributing an important role to similarity and analogy. Readers not involved in linguistics may nonetheless be surprised by the fact that should have long since ceased to surprise linguists: namely, that the concept of analogy, which is so attractive to many, is completely useless in explaining language acquisition.

If I say "He saw his mother", this can mean that the man saw his own mother; more formally, that "x saw x's mother", or that he saw someone else's mother, as in, for example, "Peter saw Mary's mother", that is, "x saw y's mother", where x and y are different. If I ask "Who saw his mother?", the sentence can similarly refer to the person's own mother or to a different mother: "For which x is it true that x saw x's mother?", or "For which x is it true that x saw y's mother?". But if I ask "He saw whose mother?" or "Whose mother did he see?", that can no longer mean "For which x is it true that x saw x's mother?", only "For which x is it true that y saw x's mother." (Please read this paragraph once again. I guarantee that it is actually dead easy.)

"Who saw his mother?" and "Whose mother did he see?" both generate a question in the same "analogous" fashion from "He saw his mother". If general intelligence and analogy were the bases for language learning, then we would expect the two sentences to behave in the same way. But they don't. Both questions can relate to the formula "x saw y's mother"; the former asks about x, the latter about y. But strictly only the former can ask for which x is it true that x saw x's mother. Why? It is possible, of course, that "Whose mother did he see?" is analogous to something else than that which "Who saw his mother?" is analogous to. But then we would like to know what it is analogous to, and why, and how this follows from the nature of general intelligence, or from the situation of the language learner. This is not a special case: any serious practising linguist's head is full to the brim with such problems, and he or she would be happy to list them for tenpence a time. While it remains the case that one cannot even attempt to provide an answer to the myriad of similar problems (any issue of any mainstrem international journal on syntax offers an ample taste of such problems), the idea of learning by analogy remains meaningless. Until then we have every reason to assume that if we were to make the concept of (innate) analogy precise enough for the task in hand, we would no longer be able to distinguish it from an (innate) language-specific system of rules.

Kálmán cites but does not understand Chomsky's highly relevant remark: "to refer to the processes involved as 'analogy' is simply to give a name to what remains a mystery". His reply is that analogy, the link created between similar problems, could be the essence of intelligent behavior. But what is similar to what? As long as we say nothing about this, to appeal to analogy, whether in relation to language or in general, is just empty chatter coated with a translucent academic gloss.

Contrary to this, Kálmán believes that the argument that we have no other acceptable explanation for the properties of language and language learning than the hypothesis of innate language structures is similar to the reasoning which attributes circular patches of unknown origin found in wheat fields to the existence of UFOs. But we have better explanations than UFOs: for example the effect of local cyclones or, on occasion, deliberate fraud; Kálmán himself refers to the latter possibility. Let us suppose, however, that it is a question neither of cyclones nor of fraud. Even then we do not accept the UFO explanation if it contradicts other scientifically justified assumptions about how the world works. We would expect, for example, that radar is capable of detecting UFOs. But the assumption of language-specific innate structures is not at odds with other scientifically-justified assumptions, only with dogmas. Nowadays no one tries to deny the fact that the morphology of the heart or our ability to see the world in three dimensions are questions not of learning, but of genetic coding. If someone believes that the situation is different in the case of the fundamental properties of language, he should at least try to produce some sort of argument to explain why he posits such a difference. Kálmán does not do this; others have tried, but with little success.

If the long-suffering lay reader still has the patience for such abstract discussion, let us look at the distinction between language, or knowledge of language, and the use of language. This distinction is a rather basic abstraction in modern linguistics. In the case of the repeatedly-embedded relative clauses Kálmán speaks of (this discussion is omitted in the English version) this distinction is immediately useful. A relative clause like "the dog that barks" can be embedded in a main clause: "The dog that barks doesn't bite". But it can also be embedded in a second relative clause: "The cat that was barked at by the dog that doesn't bite is whining a lot". This is already pretty dreadful, but it can be taken further, embedding again and again another relative clause into the previous one: "The pig, which was scratched by the cat, which was barked at by the dog that does not bite, grunts a lot", and so on.

The more the embedding, the worse the sentences. But where should we draw the line between a good sentence and a bad sentence? After one embedding? No, because in English, even that can be awkward: "The dog the cat left left". The Hungarian translation of this would be: "A kutya, amelyiket a macska elhagyott, elment". The transitive sense of leave is expressed by elhagyott, the intransitive by elment. This sentence is more difficult in English than in Hungarian, because in Hungarian we cannot omit the relative pronouns which make explicit the relations between the parts of the sentence, and the verbs help break up the sentence, with different lexical forms for the transitive and intransitive. Whether we regard such sentences as good or bad depends on intonation, short-term memory, stylistic sense and concentration level of the speaker and listener, and probably many other things. But why should these things be directly related to the mechanism of embedding? The natural hypothesis is that language (grammar) can allow any number of embeddings, since any finite limit would be arbitrary. At the same time the use of language cannot fully exploit this possibility, and the number of embeddings is restricted in the course of using language (grammar) by the limitations of systems other than grammar.

Kálmán's question is why, if we consider as arbitrary a finite restriction to the number of embedded relative clauses in grammar, it is not just as arbitrary if it happens in the use of language based on some system other than grammar. But there is no arbitrary, determinate restriction in the use of language. Sometimes even one embedding is a great deal (as in the previous example), sometimes two are acceptable. Professionals in the field of linguistics might, amongst themselves, occasionally come to grips with three embeddings, or on a good day, even four. In fact, then, Kálmán's question does not even arise. Instead of the logical contradiction which Kálmán believes he has discovered, there arises the empirical question of why, in a given situation, a given sentence, in which there are, say, x embedded relative clauses, might appear good or bad, and which characteristics of which systems play a role in the native speaker's evaluation. It would be good to know more about this than we do at the moment, but the fact itself makes no difference to the logic of the situation.

Apart from the imputed but non-existent logical problem, Kálmán also unearths "practical", and even "moral" (sic) questions. He begins: "Chomsky [...] rejects an entire theory in one sentence on the basis of its alleged 'complexity' and "lack of plausibility" (as in Chomsky 1957:23) [...]". To start with, if this "and" were an "or", then the proposition would not contradict the fact that on the page cited the expression "lack of plausibility" is nowhere to be found (corrected to "arbitrariness" in the English version). I make this otherwise unimportant observation only because Kálmán, as part of a crusade against Chomsky, even insinuates philological imprecision on the part of his target, again without even the most minimal attempt at substantiation: "I am not talking about the philological details now, although it would be worth examining those as well. Chomsky's paraphrases of the Humboldian and Cartesian ideas may not all be faithful." There is no question that these paraphrases are light years from matching the precision of Kálmán's interpretation of Chomsky.

The single sentence of dismissal postulated to occur on the twenty-third page of Chomsky's work is also rather at odds with what Kálmán writes somewhere else in his review (and omits in the English version), namely that Chomsky "argues that [certain types of] models are unsatisfactory because they arbitrarily narrow the range of acceptable sentences". So according to Kálmán, Chomsky makes the case for his "simplistic" dismissal in a single sentence that goes on for several pages (in fact, for a whole chapter). Let us proceed further. Chomsky "allows himself to completely disregard the mathematical, well-defined complexity of the solutions he himself proposes". At the risk of wallowing with Chomsky in the mire of impracticality and lack of ethics, I think the primary question is which theory is true. We do not need to get upset if our theory has a component that proves to be mathematically more "complex" in certain ways than the imagined alternative which can only externally be squared with the basic facts, if at all. No theory exists for which an alternative like this cannot be found. (When Chomsky objects to the "complexity" of the model he rejects, he uses the word in a rather different sense, as a synonym for lack of explanatory power.)

Kálmán's further critical comments also arise from the misunderstanding and the forcefully dogmatic dismissal of Chomsky's ideas. His desire to connect Chomsky's academic work with his anarcho-syndicalist views is nothing short of ludicrous. In his generosity--but I hope not without blushing--Kálmán concedes that "Chomsky's theory" does "not necessarily" rest on thinking which is "spurred by the desire to see genuine phenomena as richer than the scientifically verifiable facts (that is, to postulate paranormal phenomena)". He thinks that "the view that earlier ideas can be rejected without exhaustively falsifying them", which he attributes to Chomsky, is the "anarchist view of scientific development". There is nothing anarchist in this at all. Science simply works by replacing less successful theoretical frameworks and paradigms with more successful ones. The idea that only theories which have been exhaustively disproved can be dismissed--what, in any case, do we mean here by "exhaustively"?--is surprisingly naive and idiosyncratic.

As one of his many untrue and unfair remarks, Kálmán jokes that the story of the rabbi who thought he saw that Warsaw was on fire, and whose pupils were proud that even if mistaken, at least he could see that far, could as well be about Chomsky. This is not the case. The truth is that there really is a very big fire, but Kálmán still cannot see it.