Tuesday, March 13, 2012

2012.03.13 — Noam Chomsky: A Philosophic Review by Justin Leiber

Justin Leiber.
Noam Chomsky: A Philosophic Overview.
New York: St. Martin's Press, 1975, 0312576102. (Out of Print.)

Began: 2011.10.01; Finished 2012.01.14

This book is not for the faint of heart. Nor is it for those who have become ideologically fixated on empirical / behavioural science in the humanities. Not only is this a book for people interested in how language works, it is far more importantly a powerful book for those who have come to believe that much of what passes for science in the behaviour / humanistic fields has been plagued with a false science that has managed to turn empiricism into a mind numbing ideology. For many readers, NC:APO is likely to be a transformational book in that it provides the solid analysis that supports making the leap from the flaccid so-called truths that behaviourism has provided us with to a resurgence of the scientific attitude of 18th & 19th century rationalism. That rationalism, unlike today's mask of rationalism, does not pretend that their philosophy can explain things beyond what it can.

Chomsky's argument applies to other fields, such as economics and psychology. For example, the behaviourist B.F. Skinner's is mentioned several times in unflattering terms. (In the book of his interviews with Mitsou Ronat, Language and Responsibility, Chomsky goes so far as to say,
Mitsou Ronat
paraphrased, that as far as he knows behaviourism has contributed nothing of meaningful scientific value.)

Empiricism, perhaps especially in fields like linguistics, economics and psychology, act as if all behaviours and characteristics of the human species and the individuals within it, can be explained by stimulus/response theories. The book begins with Leiber succinctly recapping the history of how Chomsky, with the ease of a knife cutting through water, revolutionized linguistics and proved irrevocably that empirical behaviourism is completely inadequate to explain not only the acquisition of language but also its comprehension. Leiber describes Chomsky's argument that, since the sentences of a language that can be created are infinite, that the behavioural linguistic practice of cataloguing them so as to fully describe a language is fruitless. Chomsky extends that argument by pointing out that most sentences that human's comprehend in their lives they will not have ever seen or heard before. He then convincingly argues that the rules of grammar allow for sentences to be constructed that are incomprehensible, whereas sentences are easily created that don't properly follow the rules of grammar but which can be perfectly comprehensible. All of these are extremely strong indictments of some of behaviourism's fundamental tenets of human understanding of language and understanding.

Chomsky's pragmatic rationalism may be most pointedly observed when he describes the real world experience that children learn language before they know the so-called rules of grammar. That repeatedly observed behaviour, from a behavioural model of language acquisition, would ostensibly be unheard of. Chomsky also observes with pragmatic rationalism, that children's language acquisition is largely independent of the oftentimes horrible language usage and training that parents provide. He also suggests with pragmatic rationalism that one might even be able to argue that in extreme cases the acquisition of language skills would appear to be independent of any significant language training because the training skills or environment are so poor that that the child's language acquisition would seem to occur despite their language training behaviour.

The final nail in the behaviourist's coffin, as it pertains to linguistics anyway, is that when the rationale of the behaviourists' practices were questioned vigorously, it was revealed that behavioural linguistic practices were largely preconfigured by the human behaviour and/or psychological bias and preconceptions of those formulating the 'science.'

Rationalistically, as opposed to empirically, Chomsky posits that there is something in the human being that promotes language acquisition independent of race and strict behaviourism. He called it universal grammar.

And this gave me one of the greatest of finds, discoveries, epiphanies, joys I have experienced from reading a book in long time: in exactly the same way, with a nearly identical conceptualization, Chomsky proposes a description of language that is nearly identical to the methods and rational behind Jung's formulation of the Collective Unconscious.
YES! My intuitive prompt, from several years ago, that there was something similar in the philosophy of these two ostensibly disparate thinkers has been beautifully, elegantly, and delightfully affirmed. I wonder, is it just a coincidence that these two thinkers that I highly respect are both ignored or denigrated by our society's political and education leaders?

This commonality is even more strongly affirmed with the idea of a 'deep structure,' which Chomsky posits as providing the pre-language fundamentals of language acquisition. His description reminds me of Jung's descriptions of the common imagery and symbolism of myth, dreams as an expression of the collective unconscious. And when the problem of how to constrain a universal grammar to create only meaningful sentences was discussed, I am again reminded of Jung's theories about the problem of constraining
(not Jung's word) the symbols to being meaningful. A very amusing formulation of that problem is the anecdote attributed to Sigmund Freud: sometimes a cigar is only a cigar.

This is a brilliant and very important book. That I came across it by accident — I extend a heartfelt Thank you to J&L @ Renaissance Books for catching this one for me without my asking!

Now for some extended citations:
What Chomsky came to show by asking his questions was that structuralist "methods" actually were, in effect, theories about the general nature of human language and about how language could be described adequately. Once the questions were pressed far enough, it became clear that the methods in question could not achieve adequate descriptions and that the tacit theory about human language was false. It is perhaps a measure of Chomsky's success that present editions of Harris's book are titled just Structural Linguistics, the word "Methods" having been dropped; and that linguists now write books and teach courses about language without the —s, about "theoretical linguistics," "the abstract theory of language," "universal grammar," and "universal phonetic theory."

The questions of projection, formulation, and adequacy led to the suggestion that procedure be changed; and, more abstractly, that the presuppositions—behaviourism and empiricism— that underlay structuralist procedure be changed. So let us begin by asking again the first of our questions. Why not simply ask the game players (or speakers of a corpus) what the alphabet and vocabulary are, and what instances of letters and words (or phonemes and morphemes) are the same or different?

The general answer of the structuralist is that he is concerned with what people are doing and not with what people think they are doing. This general answer, of course, emphasizes the structuralist's commitment to empiricism and, more specifically, to behaviourism, that is, he is committed to the view that science should be concerned only with what is observable, and thus that the human sciences should deal with observed human behaviour rather than with "subjective" thoughts and feelings. More specifically, the structuralist would say that people's judgments about what they are doing are in principle essentially irrelevant. The linguist is describing objective noise, a physical phenomenon which does, in the cases that interest the linguist, happen to exhibit certain regularities, that is, certain species, genera, and families of noises that regularly appear in certain patterns. Admittedly, speaker-hearers of the language are aware of some of these regularities and may be able to give the linguist revealing hints about them, but the structuralist insists that since what he is describing has objective existence quite apart from what people may think, quite apart from psychology, he ought to be able to describe it without asking what people think. There is another, less theoretical, reason why the structuralist wants to discover the grammar of a language without asking the native speakers questions, namely, they will very likely give misleading or false answers.

The native speaker is not a linguist and so cannot be expected to understand or answer accurately questions about what are the minimal sound units (phonemes) and minimal grammatical units (morphemes) of his language. Many studies have shown that ordinary language users are likely to 'hear' distinctions in sound that are not there in fact or to mistake variant pronunciations of a phoneme for more significant differences. If we do teach the native speaker structural linguistics, he will then become more accurate but he will be in no better position than the normative linguist. Indeed, he may be in a worse position because his cultural attitudes—about status, etiquette, meaning, and so on—may bias his descriptions. After all, we would be hesitant about accepting conclusions reached by a psychologist examining his own psychology, by a doctor diagnosing his own illness, and so on.

Of course, there are limits to this skepticism. Aside from accepting the corpus itself as a largely regular set of sentences, the linguist may use his ingenuity in "eliciting" new sentences. The corpus may lead the linguist to wonder whether or not a certain sentence is regular, or grammatical. The difficulty then is to see if one can put the native speaker in such a position that he will say the sentence (or indicate that it is regular) only if it is regular. The real problem of "elicitation" is that of ensuring that one does not prejudice the issue: there is a risk that the native speaker will say something irregular to please the linguist, or from confusion, and so on; equally there is a … (38-9).
And, below, Chomsky makes the identical argument about understanding language as Jung makes about understanding human psychology:
As Chomsky puts it in Syntactic Structures, given a corpus of utterances belonging to a natural language, there is neither a discovery nor a decision procedure. That is, given a particular corpus (or set of assumptions) belonging to a natural language, there is no mechanical, step-by-step procedure for discovering the grammar of the language. If there were such a discovery procedure, structuralism would be vindicated, because this would mean that each corpus of noise could be adequately and determinately described. Even if there were a decision procedure this implication would hold, though it would require creativity for the linguist to arrive at the grammar.

Chomsky's claim that there is no discovery or decision procedure for arriving at the grammar of a language, although grammars may be given comparative evaluations, is a way of pointing out that it is impossible to write the grammars of human languages without a very substantial theory of the human mind and its language acquiring, and processing, software. Presumably, Chomsky would accept the claim that adequate and determinate grammars are possible only if such a substantial theory is part of the input for arriving at such grammars. There must be a number of universal features common to the grammars of all natural languages, and these must derive from innate features of man's language-acquisition device, in order for there to be adequately determinate grammars (that is, by rough analogy, for completeness to obtain, regarding the grammar of each language as what may be "proved" from sentences of that language plus the "axioms" of a more substantial theory of language and language acquisition than we now have).

Linguists can only go about determining the universal characteristics of human language in an indirect and conjectural way, of course: there is no direct way of looking into the "black box" of language processing and acquisition. As linguists proceed, in painstaking and piecemeal fashion, to construct and evaluate general, explicit, and formal grammars, or partial grammars, of particular languages, it may be hoped that they will find universal features and thus that they will approach explanatory adequacy. For an analogy, imagine that we wanted to know about the internal structure of some kind of computer found on Mars: the rules of the game are that we cannot look inside at the hardware but can only try to infer the inner structure through observing the inputs and outputs. What we note is that when one of these machines is exposed to a small number of strings (drawn, apparently, from a particular infinite set of strings, a "language") its output is, more or less, the infinite set. We can see that it is impossible, using just logic principles, to arrive at the infinite output by taking a typical input (i.e., no discovery or decision procedure exists). All we can say is that, given a particular input, such and such an output (as generated by a grammar) is more likely than some other sort. As we write partial grammars of outputs A through Z, for inputs A through Z, we would expect various uniformities, and we would reformulate our grammars in terms of these uniformities. The internal software of these Martian computers will be, in particular, the assumptions that must be made in order to get in a regular fashion from these various inputs to their respective outputs (66-7).
And below we see Chomksy's idea of a 'deep structure.' In both imagery and functional description there is very little difference between it and Jung's idea of the collective unconscious:
… (G) English is not a finite-state language.

Note that while this argument is a specific realization of Chomsky's principle that the linguist must show how the infinity of sentences that compose a natural language can be generated through finite means, the argument shows more than that a natural language consists of an infinite number of sentences. … [I]t follows that a natural language cannot be specified except through specifying a generative device, a device that employs recursion. But from the present argument it also follows that the device must be more powerful than a finite-state device with recursive loops, even though such a device does have the capacity of generating an infinite number of sentences.

I have devoted a great deal more space to this argument than Chomsky did in Syntactic Structures. My reason is twofold. First, my experience has been that people who have little or no familiarity with this general sort of argument may fail to grasp the argument in its entirety, to appreciate its force and limits, if the argument is not explained at length and with some repetitions. People have been confused about it even in print. Secondly, the argument already contains, at least in miniature, the most basic elements in Chomsky's approach to language. The basic philosophic and psychological issues, and Chomsky's way of reasoning about theory, are on the table, at least in embryo.

Of course, from the point of view of a practicing linguist particularly interested in some approximation of part of the grammar of English, absolutely nothing of any importance or interest has been said. But as Chomsky's reasoning about the nature of grammar proceeds—in the argument of Syntactic Structures that English is a transformational-generative grammar with particular sorts of rules, in the further developments found in the "standard theory" of Aspects of the Theory of Syntax, and in the "extended standard theory" of his most recent papers —the basic principles of this argument are employed again and again, though the applications are much more complicated, and the results of course much closer to an adequate simulation of our grammatical knowledge than this extremely elementary initial step. The argument that an adequate description of a natural language can only be given through specification of a generative device (whose software the competent speaker of the language must minimally internalize) remains constant, though the specification of the requirements that characterize the device has become much more detailed. Equally constant is the claim that natural language sentences have an abstract or deep structure that cannot be explicated by a physical description of such sentences (as noise) and that is not present in a simple sensory, or observational, characterization of such sequences of noise "bounded by silence" (though deep structure has become ever more abstract and complex as the theory has developed, leading in the past few years to a split between transformational-generative linguists on precisely how abstract deep structure must be) (93-4).
And, also in line with Jung, Chomsky argues that the so-called flakily subjective quality of meaning enters the picture in order for language acquisition, understanding and usage to be possible:
In summary, the grammar sketched in Syntactic Structures consists of three sorts of rules, which operate in sequence in generating the sentences of English and in providing them with phonological realizations in speech.

1. Phrase-structure rules which rewrite single, non-terminal symbols into, eventually, terminal symbols or words, in this manner creating a tree diagram or phrase-structure bracketing.

2. Transformational rules, which operate upon the phrase structures produced by (1), deleting, reshuffling, and joining portions of such structures. Singulary [sic] transformations, whose input is single phrase-structure (kernel) strings, are either obligatory or optional, optional transformations including passive, negative, and question transformations. Generalized transformations, which are always optional, join two or more kernel strings. As opposed to the phrase-structure rules, transformations are ordered, in that some must be applied after others—the passive transformation, for example, must apply before the transformation that makes the verb plural or singular, and so on, because it is the noun that is put into subject position by the passive transformation that determines the form of the verb.

3. Morphophonemic rules, which convert the output of (2), the sentences of the language from a syntactical viewpoint, into the actual sounds of speech. (These rules, of which no account has been given here, are similar to the phrase-structure rules. But they allow the rewriting of more than one symbol, and they are "context-sensitive" in that they may indicate that a symbol is to be rewritten in a particular way only if certain symbols precede or follow that symbol. For example, the purely syntactical rules (1) and (2) will generate strings such as take + past-tense; the rule that will convert that segment into the sounds that we write as "took" is a morphophonemic rule.)

If one has grasped the nature of these rules, and the general structure of the grammar that is summarized here, one will not find it difficult to follow the changes that are brought in with the "standard theory" of Aspects of the Theory of Syntax (1965), and in still more recent work. I mention this because the reader may feel that he has been burdened with enough of a technical apparatus, and so he has. Perhaps the major change that takes place is that meaning becomes a respectable and central part of linguistics: to the syntactical and phonological components that may be found in the first version of the theory. Aspects adds a third semantic component. The formulation of this component, and its relationship to the others, will be seen as a major theatre of controversy. The phonological component undergoes radical changes in The Sound Pattern of English (1968), which Chomsky coauthored with his MIT colleague Morris Halle. In that book considerable evidence is marshalled for the view that the input… (107-8).
A further statement that parallels Jung. In this case the problem of everything having the possibility of symbolizing or alluding to everything and having no meaning at all. Meaningfulness is what makes the distinction. Leiber does not cite Chomsky connecting the constraint, here, to meaningfulness, but it seems obvious that meaningfulness is in the end the singular constraint. In an odd way, I see Chomsky struggling with a similar problem to Jung's examination of the meaning of dreams as an expression of the unconscious: for Chomsky, it is the struggle to see how a 'deep structure' can express language meaningfully.
… the constraint would be universal. Universal constraints have seemed particularly important to transformational grammarians because, as Chomsky has often emphasized, transformational rules are so powerful that unless such constraints are established there will be by far too many ways of writing the grammars of particular languages. It has been established that, given a few powerful and unnatural transformational rules, one can write a perhaps highly unnatural, but nonetheless observationally adequate, grammar for any possible human language. The existence for universal constraints would be one of the most powerful ways of eliminating such excesses and reducing the range of solutions in establishing grammars: to put it another way, if one wants to determine the internal structure of a generative device of considerable power, it is likely to be more helpful to find what the device cannot do than what it can.

But the drive toward universality (and reactions toward its excesses) has extended beyond the standard theory of Aspects, and it has led to conflict and reformulation, particularly in controversy between generative semantics and interpretive semantics.

The general thrust of the generative-semanticist proposal for improving transformational grammar is very simple; if some semantic features of sentences can be specified in their syntactical deep structure why can not all such features be specified? Why split the syntactic and semantic components at all? Why not equate ultimate syntactic deep structure with semantic representation? Or, more speculatively, one can ask, why not take the system of semantic representation to be something like the familiar predicate logic (with perhaps a few additions), and the base to be such a system supplemented with a relatively small number of "atomic predicates," or semantic primitives, universal to human thought?
I am going to interrupt this paragraph to propose a Taoist / Jungian answer to these questions: because the foundations of language, like dreams, is pre- and/or non-verbal. Philosophically the problem of words struggling to represent non-verbal meaning has been wrestled for a very long time, most especially by artists and poets and philosophers. And there is a very interesting East/West split when it comes to the approach to wrestling with that 'problem'. In the east, they write koans such as "Does a dog have Buddha nature or not?"; but in the west they write tomes like Being and Time — (or even Jung, with his 20 volumes!). Now back to Leiber:
The words of particular languages, Just as their surface syntactical structures, would decompose into extremely abstract and complex syntactic-semantical deep structures; the features constituting the lexical-syntactical peculiarities of a language would be given as a series of transformations relating the syntactic-semantical deep structures (or "natural logic formulas") of the "universal base" to their particular realizations in the language in question, similarly for the peculiarities of other human languages (122).
And now for Chomsky's unequivocally argument that linguistics belongs to a branch of psychology and not behavioural science.
Chomsky has said, both in the paper quoted above and in other recent papers, that the crucial problem of present transformational-generative grammatical work is that transformational grammars are too powerful. There is not a sufficient number of restrictions on the construction of such grammars, and thus such grammars are not testable against each other (though, of course, they are selected by the evidence, over other sorts of grammars—for example, the transformational grammars of Aspects and Syntactic Structures are inadequate on the evidence, and, of course, phrase-structure and finite-state grammars are certainly inadequate). Many critics, who are quite convinced of the general validity of the transformational-generative approach to language, have made this point; it is a point on which much work must be done. But it is wrong to think of this as a criticism of any stage in Chomsky's development of transformational-generative grammar. Chomsky has always maintained that this was the crucial area for work in linguistic theory. The fundamental problem of linguistic theory is to describe the essential, as against purely happenstance, features of natural (human) language in as circumscribed a way as possible. In other words, to approach explanatory adequacy by accounting for the choice of a particular grammar on the part of a human language-learner given the data about the particular human language that he is exposed to is to specify the software of man's language acquisition device.

The problem is not how a universal thinking machine, programmed with nothing more specific or contentious than the "universal notation" of, say, predicate logic, would determine the grammar of a particular human language, on exposure to a limited number of its sentences. Such a device, which simulates, in that it makes use of no species-specific, "innate" principles for limiting its choice of grammars, the seventeenth-century empiricist view of man as learning essentially everything from experience, would produce countless numbers of "wrong" solutions, "wrong" meaning, "a 'solution' which a normal human language-learner would not, essentially could not, think up." For that reason, of course, Chomsky found structural linguistics wrong, and for that reason Chomsky has maintained that linguistics is a branch of psychology and part of the study of the human mind. It is also for that reason that Chomsky has found behaviourism and radical empiricism wrong, at least in their stronger forms—those forms that propose to make serious theoretical and empirical claims as opposed to essentially terminological stipulations (such as, for example, the arguments of those who define their ]'argon in such a way that no discovery about human beings, or anything else, could ever constitute the slightest evidence against behaviourism and empiricism). What is the problem is to discover the principles, particularly specific to man in being much narrower and circumscribed than those of a generalized calculating device, that operate in language learning and are the basis for the universal features of human language.

It is particularly clear that Chomsky thinks that this is quite a problem. He has been much more reluctant than some transformational-generative linguists in making confident claims about which features of natural languages are universal and still more as to what universal features of language acquisition give rise to this universality. Chomsky has maintained that all natural languages are transformational and have rules that operate on abstract, as opposed to explicit surface, structures;

that all human languages make use of cycles of transformations working on successively more inclusive structures within sentences; and he has suggested that nouns, verbs, and adjectives are likely to prove universal deep-syntactic categories of human languages, and that at least one constraint on grammatical rules, which prohibits transformations that move material in or out of various conjunctive structures, is probably universal. But he has emphasized, particularly in his most recent papers, that the confident identification of particular universals is not possible at this point (132-4).
Now for one of Chomsky's pet peeves: the failure of university professors to honour the integrity of learning and become courtiers to the halls of wealth and power.
CHAPTER 3 Psychology, Philosophy, Politics

OF course, we have been talking about these three topics all along; for Chomskyan linguistics is above all an attempt to characterize a significant portion of human psychology (as, substantially, the study of the software of human nature), and this sort of characterization can be seen as establishing some of the claims of traditional rationalism. The view of man that results might be thought to have political significance.

In this final chapter I want to talk about what Chomsky's work may mean for psychology, philosophy, and politics; about his views of these subjects, particularly as ramifications and interpretations of his work in linguistics; about the work in these fields that is convergent with his approach; and about various criticisms and misunderstandings that have been made respecting his general rationalist view of man. Psychology, philosophy, and politics are not strange bedfellows, in Chomsky's view, and this attitude makes many people find Chomsky's work exciting and important (or infuriatingly pretentious and misguided). If anything has been characteristic of analytic philosophy, or the rather more traditional (and perhaps less subtle) empiricism, and behaviourism, that has been very common among social and psychological scientists of all sorts, it is the hardly questioned conviction that no psychological discovery, no psychological fact, can establish, or refute, any philosophic claim (and the reverse), and that neither can properly determine the answer to any political question (at least in Aristotle's sense of politics as the practical job of deciding, and achieving, what is good for man, within and between nation states).

In part, of course, this subconscious positivism has been a result of the increasing professionalization of knowledge; our century has seen the concentration of all theoretical scientific and intellectual activity in the university, which is divided into professional departments (or unions), each zealously insisting on its independence and importance. The professional academic is institutionally pressed to believe both that the claims and suppositions of each discipline are independent of the others and that generalizations which are not professional, in that they are interdisciplinary, are not really "scientific" or "objective." But, equally, this century has been dominated by a rather skeptical empiricism (often expressed in analytic terms) that insists that the truths of reason (the logical software) are quite independent of those of (or about) human nature, and that neither sort, properly considered, can tell us what we ought to decide to do politically or morally.

Skepticism, empiricism and behaviourism, the professionalization and departmentalization of knowledge—these can seem a worthy commitment to slow, "solid," painstaking progress in detail, shorn of grandiose mysticism, "moralistic ranting," and "ideology." A familiar metaphor can be inverted aptly by those sympathetic to analytic empiricism or to straightforward behaviourism: it is that of the careful, humble "hedgehog" against the wild and dubious "fox." The understood moral is that the attempt to make general rational claims about man's condition, with psychological, philosophical, and political implications—"ideology"—is inevitably a "foxy" affair: charlatanism is the only refuge of the generalist. An obvious implication of the metaphor that is not often stressed by the radical empiricist-behaviourist is that if one hires a university faculty of hedgehogs, they will defend the status quo. At least they will defend the status quo in the sense that they will firmly maintain that no attack on it can be rational and objective: rationality and objectivity come into politics only in that the hedgehogs can tell the government (any government with any goals) the specific techniques of "behavioural reinforcement" that will achieve the goals which are themselves beyond the judgment of the poor, short-sighted hedgehog.

For that is the other, darker side of the radical empiricist's skepticism: as an organized social phenomenon it is little more than a rationalization of the economic subservience of the university to the political and economic power of the status quo; for though, from the hedgehogian point of view, he can do research in behavioural reinforcement, and so on, for the capitalist-imperialists or commissar-imperialists, he in fact will do the research that boards of trustees and governmental granting institutions find proper. Thus, though the faculty and staff of Michigan State University who trained Diem's secret police could equally have undertaken research and instruction helpful to Diem's opponents, or to the American peace movement, they would not be paid for it (or even allowed to do it at all as American citizens). Seen in this light, the short-sighted hedgehog would seem to be—as Martin Luther once wrote of reason— a whore: he works for whoever pays and he does not question his employer's tastes, except as to practicality. And—strangely enough—he is likely to call any attempt to make the university an independent general critic of society and government, "politicization" or "ideologizing." How convenient a philosophy that allows the hedgehog to believe that any research that might lead to a quarrel with the basic character of the established political and economic order is necessarily unscientific and bogus foxiness! The hedgehog is very likely to be a self-satisfied hedgehog; his philosophy teaches him that the fox—since there is no real scientific and objective knowledge beyond what hedgehogs can get while being dutiful and doughty hedgehogs—is really an irresponsible rogue hedgehog, stumbling about and presumably suffering from psychiatric problems and an inflated ego, while pretending to see truths beyond the resolution of vision proper to hedgehogs (132-7).
A brilliant book in many, many ways. I have directed my local used bookstore proprietors to keep their open for Chomsky's books on linguistics.

Finally, it was a complete delight that in this summary of Chomsky my intuitive thought that there was something Jungian in Chomsky has been confirmed so tangibly.