In these lectures, I would like to focus attention on the question, What contribution can the study of language make to our understanding of human nature? In one or another manifestation, this question threads its way through modern Western thought. In an age that was less self-conscious and less compartmentalized than ours, the nature of language, the respects in which language mirrors human mental processes or shapes the flow and character of thought – these were topics for study and speculation by scholars and gifted amateurs with a wide variety of interests, points of view, and intellectual backgrounds. And in the nineteenth and twentieth centuries, as linguistics, philosophy, and psychology have uneasily tried to go their separate ways, the classical problems of language and mind have inevitably reappeared and have served to link these diverging fields and to give direction and significance to their efforts. There have been signs in the past decade that the rather artificial separation of disciplines may be coming to an end. It is no longer a point of honor for each to demonstrate its absolute independence of the others, and new interests have emerged that permit the classical problems to be formulated in novel and occasionally suggestive ways – for example, in terms of the new perspectives provided by cybernetics and the communication sciences, and against the background of developments in comparative and physiological psychology that challenge long-standing convictions and free the scientific imagination from certain shackles that had become so familiar a part of our intellectual environment as to be almost beyond awareness. All of this is highly encouraging. I think there is more of a healthy ferment in cognitive psychology – and in the particular branch of cognitive psychology known as linguistics – than there has been for many years. And one of the most encouraging signs is that skepticism with regard to the orthodoxies of the recent past is coupled with an awareness of the temptations and the dangers of premature orthodoxy, an awareness that, if it can persist, may prevent the rise of new and stultifying dogma.
It is easy to be misled in an assessment of the current scene; nevertheless, it seems to me that the decline of dogmatism and the accompanying search for new approaches to old and often still intractable problems are quite unmistakable, not only in linguistics but in all of the disciplines concerned with the study of mind. I remember quite clearly my own feeling of uneasiness as a student at the fact that, so it seemed, the basic problems of the field were solved, and that what remained was to sharpen and improve techniques of linguistic analysis that were reasonably well understood and to apply them to a wider range of linguistic materials. In the postwar years, this was a dominant attitude in most active centers of research. I recall being told by a distinguished anthropological linguist, in 1953, that he had no intention of working through a vast collection of materials that he had assembled because within a few years it would surely be possible to program a computer to construct a grammar from a large corpus of data by the use of techniques that were already fairly well formalized. At the time, this did not seem an unreasonable attitude, though the prospect was saddening for anyone who felt, or at least hoped, that the resources of human intelligence were somewhat deeper than these procedures and techniques might reveal. Correspondingly, there was a striking decline in studies of linguistic method in the early 1950s as the most active theoretical minds turned to the problem of how an essentially closed body of technique could be applied to some new domain – say, to analysis of connected discourse, or to other cultural phenomena beyond language. I arrived at Harvard as a graduate student shortly after B. F. Skinner had delivered his William James Lectures, later to be published in his book Verbal Behavior. Among those active in research in the philosophy or psychology of language, there was then little doubt that although details were missing, and although matters could not really be quite that simple, nevertheless a behavioristic framework of the sort Skinner had outlined would prove quite adequate to accommodate the full range of language use. There was now little reason to question the conviction of Leonard Bloomfield, Bertrand Russell, and positivistic linguists, psychologists, and philosophers in general that the framework of stimulus-response psychology would soon be extended to the point where it would provide a satisfying explanation for the most mysterious of human abilities. The most radical souls felt that perhaps, in order to do full justice to these abilities, one must postulate little s’s and r’s inside the brain alongside the capital S’s and R’s that were open to immediate inspection, but this extension was not inconsistent with the general picture.
Critical voices, even those that commanded considerable prestige, were simply unheard. For example, Karl Lashley gave a brilliant critique of the prevailing framework of ideas in 1948, arguing that underlying language use – and all organized behavior – there must be abstract mechanisms of some sort that are not analyzable in terms of association and that could not have been developed by any such simple means. But his arguments and proposals, though sound and perceptive, had absolutely no effect on the development of the field and went by unnoticed even at his own university (Harvard), then the leading center of psycholinguistic research. Ten years later Lashley’s contribution began to be appreciated, but only after his insights had been independently achieved in another context.
The technological advances of the 1940s simply reinforced the general euphoria. Computers were on the horizon, and their imminent availability reinforced the belief that it would suffice to gain a theoretical understanding of only the simplest and most superficially obvious of phenomena – everything else would merely prove to be “more of the same,” an apparent complexity that would be disentangled by the electronic marvels. The sound spectrograph, developed during the war, offered similar promise for the physical analysis of speech sounds. The interdisciplinary conferences on speech analysis of the early 1950s make interesting reading today. There were few so benighted as to question the possibility, in fact the immediacy, of a final solution to the problem of converting speech into writing by available engineering technique. And just a few years later, it was jubilantly discovered that machine translation and automatic abstracting were also just around the corner. For those who sought a more mathematical formulation of the basic processes, there was the newly developed mathematical theory of communication, which, it was widely believed in the early 1950s, had provided a fundamental concept – the concept of “information” – that would unify the social and behavioral sciences and permit the development of a solid and satisfactory mathematical theory of human behavior on a probabilistic base. At about the same time, the theory of automata developed as an independent study, making use of closely related mathematical notions. And it was linked at once, and quite properly, to earlier explorations of the theory of neural nets. There were those – John von Neumann, for example – who felt that the entire development was dubious and shaky at best, and probably quite misconceived, but such qualms did not go far to dispel the feeling that mathematics, technology, and behavioristic linguistics and psychology were converging on a point of view that was very simple, very clear, and fully adequate to provide a basic understanding of what tradition had left shrouded in mystery.
In the United States at least, there is little trace today of the illusions of the early postwar years. If we consider the current status of structural linguistic methodology, stimulus-response psycholinguistics (whether or not extended to “mediation theory”), or probabilistic or automata-theoretic models for language use, we find that in each case a parallel development has taken place: a careful analysis has shown that insofar as the system of concepts and principles that was advanced can be made precise, it can be demonstrated to be inadequate in a fundamental way. The kinds of structures that are realizable in terms of these theories are simply not those that must be postulated to underlie the use of language, if empirical conditions of adequacy are to be satisfied. What is more, the character of the failure and inadequacy is such as to give little reason to believe that these approaches are on the right track. That is, in each case it has been argued – quite persuasively, in my opinion – that the approach is not only inadequate but misguided in basic and important ways. It has, I believe, become quite clear that if we are ever to understand how language is used or acquired, then we must abstract for separate and independent study a cognitive system, a system of knowledge and belief, that develops in early childhood and that interacts with many other factors to determine the kinds of behavior that we observe; to introduce a technical term, we must isolate and study the system of linguistic competence that underlies behavior but that is not realized in any direct or simple way in behavior. And this system of linguistic competence is qualitatively different from anything that can be described in terms of the taxonomic methods of structural linguistics, the concepts of S-R psychology, or the notions developed within the mathematical theory of communication or the theory of simple automata. The theories and models that were developed to describe simple and immediately given phenomena cannot incorporate the real system of linguistic competence; “extrapolation” for simple descriptions cannot approach the reality of linguistic competence; mental structures are not simply “more of the same” but are qualitatively different from the complex networks and structures that can be developed by elaboration of the concepts that seemed so promising to many scientists just a few years ago. What is involved is not a matter of degree of complexity but rather a quality of complexity. Correspondingly, there is no reason to expect that the available technology can provide significant insight or understanding or useful achievements; it has noticeably failed to do so, and, in fact, an appreciable investment of time, energy, and money in the use of computers for linguistic research – appreciable by the standards of a small field like linguistics – has not provided any significant advance in our understanding of the use or nature of language. These judgments are harsh, but I think they are defensible. They are, furthermore, hardly debated by active linguistic or psycholinguistic researchers.
At the same time there have been significant advances, I believe, in our understanding of the nature of linguistic competence and some of the ways in which it is put to use, but these advances, such as they are, have proceeded from assumptions very different from those that were so enthusiastically put forth in the period I have been discussing. What is more, these advances have not narrowed the gap between what is known and what can be seen to lie beyond the scope of present understanding and technique; rather, each advance has made it clear that these intellectual horizons are far more remote than was heretofore imagined. Finally, it has become fairly clear, it seems to me, that the assumptions and approaches that appear to be productive today have a distinctly traditional flavor to them; in general, a much despised tradition has been largely revitalized in recent years and its contributions given some serious and, I believe, well-deserved attention. From the recognition of these facts flows the general and quite healthy attitude of skepticism that I spoke of earlier.
In short, it seems to me quite appropriate, at this moment in the development of linguistics and psychology in general, to turn again to classical questions and to ask what new insights have been achieved that bear on them, and how the classical issues may provide direction for contemporary research and study.
When we turn to the history of study and speculation concerning the nature of mind and, more specifically, the nature of human language, our attention quite naturally comes to focus on the seventeenth century, “the century of genius,” in which the foundations of modern science were firmly established and the problems that still confound us were formulated with remarkable clarity and perspicuity. There are many far from superficial respects in which the intellectual climate of today resembles that of seventeenth-century Western Europe. One, particularly crucial in the present context, is the very great interest in the potentialities and capacities of automata, a problem that intrigued the seventeenth-century mind as fully as it does our own. I mentioned above that there is a slowly dawning realization that a significant gap – more accurately, a yawning chasm – separates the system of concepts of which we have a fairly clear grasp, on the one hand, and the nature of human intelligence, on the other. A similar realization lies at the base of Cartesian philosophy. Descartes also arrived, quite early in his investigations, at the conclusion that the study of mind faces us with a problem of quality of complexity, not merely degree of complexity. He felt that he had demonstrated that understanding and will, the two fundamental properties of the human mind, involved capacities and principles that are not realizable by even the most complex of automata.
It is particularly interesting to trace the development of this argument in the works of the minor and now quite forgotten Cartesian philosophers, like Cordemoy, who wrote a fascinating treatise extending Descartes’ few remarks about language, or La Forge, who produced a long and detailed Traité de l’esprit de l’homme expressing, so he claimed with some reason, what Descartes would likely have said about this subject had he lived to extend his theory of man beyond physiology. One may question the details of this argument, and one can show how it was impeded and distorted by certain remnants of scholastic doctrine – the framework of substance and mode, for example. But the general structure of the argument is not unreasonable; it is, in fact, rather analogous to the argument against the framework of ideas of the early postwar years, which I mentioned at the outset of this lecture. The Cartesians tried to show that when the theory of corporeal body is sharpened and clarified and extended to its limits, it is still incapable of accounting for facts that are obvious to introspection and that are also confirmed by our observation of the actions of other humans. In particular, it cannot account for the normal use of human language, just as it cannot explain the basic properties of thought. Consequently, it becomes necessary to invoke an entirely new principle – in Cartesian terms, to postulate a second substance whose essence is thought, alongside of body, with its essential properties of extension and motion. This new principle has a “creative aspect,” which is evidenced most clearly in what we may refer to as “the creative aspect of language use,” the distinctively human ability to express new thoughts and to understand entirely new expressions of thought, within the framework of an “instituted language,” a language that is a cultural product subject to laws and principles partially unique to it and partially reflections of general properties of mind. These laws and principles, it is maintained, are not formulable in terms of even the most elaborate extension of the concepts proper to the analysis of behavior and interaction of physical bodies, and they are not realizable by even the most complex automaton. In fact, Descartes argued that the only sure indication that another body possesses a human mind, that it is not a mere automaton, is its ability to use language in the normal way; and he argued that this ability cannot be detected in an animal or an automaton which, in other respects, shows signs of apparent intelligence exceeding those of a human, even though such an organism or machine might be as fully endowed as a human with the physiological organs necessary to produce speech.
I will return to this argument and the ways in which it was developed. But I think it is important to stress that, with all its gaps and deficiencies, it is an argument that must be taken seriously. There is nothing at all absurd in the conclusion. It seems to me quite possible that at that particular moment in the development of Western thought there was the possibility for the birth of a science of psychology of a sort that still does not exist, a psychology that begins with the problem of characterizing various systems of human knowledge and belief, the concepts in terms of which they are organized and the principles that underlie them, and that only then turns to the study of how these systems might have developed through some combination of innate structure and organism – environment interaction. Such a psychology would contrast rather sharply with the approach to human intelligence that begins by postulating, on a priori grounds, certain specific mechanisms that, it is claimed, must be those underlying the acquisition of all knowledge and belief. The distinction is one to which I will return in a subsequent lecture. For the moment, I want merely to stress the reasonableness of the rejected alternative and, what is more, its consistency with the approach that proved so successful in the seventeenth-century revolution in physics.
There are methodological parallels that have perhaps been inadequately appreciated between the Cartesian postulation of a substance whose essence was thought and the post-Newtonian acceptance of a principle of attraction as an innate property of the ultimate corpuscles of matter, an active principle that governs the motions of bodies. Perhaps the most far-reaching contribution of Cartesian philosophy to modern thought was its rejection of the scholastic notion of substantial forms and real qualities, of all those “little images fluttering through the air” to which Descartes referred with derision. With the exorcism of these occult qualities, the stage was set for the rise of a physics of matter in motion and a psychology that explored the properties of mind. But Newton argued that Descartes’ mechanical physics wouldn’t work – the second book of the Principia is largely devoted to this demonstration – and that it is necessary to postulate a new force to account for the motion of bodies. The postulate of an attractive force acting at a distance was inconsistent with the clear and distinct ideas of common sense and could not be tolerated by an orthodox Cartesian – such a force was merely another occult quality. Newton quite agreed, and he attempted repeatedly to find a mechanical explanation of the cause of gravity. He rejected the view that gravity is “essential and inherent to matter” and maintained that “to tell us that every species of things is endowed with an occult specific property (such as gravity) by which it acts and produces manifest effects, is to tell us nothing.” Some historians of science have suggested that Newton hoped, like Descartes, to write a Principles of Philosophy but that his failure to explain the cause of gravity on mechanical grounds restricted him to a Mathematical Principles of Natural Philosophy. Thus, to the common sense of Newton as well as the Cartesians, physics was still not adequately grounded, because it postulated a mystical force capable of action at a distance. Similarly, Descartes’ postulation of mind as an explanatory principle was unacceptable to the empiricist temper. But the astonishing success of mathematical physics carried the day against these common-sense objections, and the prestige of the new physics was so high that the speculative psychology of the Enlightenment took for granted the necessity of working within the Newtonian framework, rather than on the Newtonian analogy – a very different matter. The occult force of gravity was accepted as an obvious element of the physical world, requiring no explanation, and it became inconceivable that one might have to postulate entirely new principles of functioning and organization outside the framework of what soon became the new “common sense.” Partly for this reason, the search for an analogous scientific psychology that would explore the principles of mind, whatever they might be, was not undertaken with the thoroughness that was then, as now, quite possible.
I do not want to overlook a fundamental distinction between the postulation of gravity and the postulation of a res cogitans, namely the enormous disparity in the power of the explanatory theories that were developed. Nevertheless, I think it is instructive to note that the reasons for the dissatisfaction of Newton, Leibnitz, and the orthodox Cartesians with the new physics are strikingly similar to the grounds on which a dualistic rationalist psychology was soon to be rejected. I think it is correct to say that the study of properties and organization of mind was prematurely abandoned, in part on quite spurious grounds, and also to point out that there is a certain irony in the common view that its abandonment was caused by the gradual spread of a more general “scientific” attitude.
I have tried to call attention to some similarities between the intellectual climate of the seventeenth century and that of today. It is illuminating, I think, to trace in somewhat greater detail the specific course of development of linguistic theory during the modern period, in the context of the study of mind and of behavior in general.1
A good place to begin is with the writings of the Spanish physician Juan Huarte, who in the late sixteenth century published a widely translated study on the nature of human intelligence. In the course of his investigations, Huarte came to wonder at the fact that the word for “intelligence,” ingenio, seems to have the same Latin root as various words meaning “engender” or “generate.” This, he argued, gives a clue to the nature of mind. Thus, “One may discern two generative powers in man, one common with the beasts and the plants, and the other participating of spiritual substance. Wit (Ingenio) is a generative power. The understanding is a generative faculty.” Huarte’s etymology is actually not very good; the insight, however, is quite substantial.
Huarte goes on to distinguish three levels of intelligence. The lowest of these is the “docile wit,” which satisfies the maxim that he, along with Leibnitz and many others, wrongly attributes to Aristotle, namely that there is nothing in the mind that is not simply transmitted to it by the senses. The next higher level, normal human intelligence, goes well beyond the empiricist limitation: it is able to “engender within itself, by its own power, the principles on which knowledge rests.” Normal human minds are such that “assisted by the subject alone, without the help of anybody, they will produce a thousand conceits they never heard spoke of . . . inventing and saying such things as they never heard from their masters, nor any mouth.” Thus, normal human intelligence is capable of acquiring knowledge through its own internal resources, perhaps making use of the data of sense but going on to construct a cognitive system in terms of concepts and principles that are developed on independent grounds; and it is capable of generating new thoughts and of finding appropriate and novel ways of expressing them, in ways that entirely transcend any training or experience.
Huarte postulates a third kind of wit, “by means of which some, without art or study, speak such subtle and surprising things, yet true, that were never before seen, heard, or writ, no, nor ever so much as thought of.” The reference here is to true creativity, an exercise of the creative imagination in ways that go beyond normal intelligence and may, he felt, involve “a mixture of madness.”
Huarte maintains that the distinction between docile wit, which meets the empiricist maxim, and normal intelligence, with its full generative capacities, is the distinction between beast and man. As a physician, Huarte was much interested in pathology. In particular, he notes that the most severe disability of wit that can afflict a human is a restriction to the lowest of the three levels, to the docile wit that conforms to empiricist principles. This disability, says Huarte, “resembles that of Eunuchs, incapable of generation.” Under these sad circumstances, in which the intelligence can only receive stimuli transmitted by sense and associate them with one another, true education is of course impossible, since the ideas and principles that permit the growth of knowledge and understanding are lacking. In this case, then, “neither the lash of the rod, nor cries, nor method, nor examples, nor time, nor experience, nor anything in nature can sufficiently excite him to bring forth anything.”
Huarte’s framework is useful for discussing “psychological theory” in the ensuing period. Typical of later thought is his reference to use of language as an index of human intelligence, of what distinguishes man from animals, and, specifically, his emphasis on the creative capacity of normal intelligence. These concerns dominated rationalist psychology and linguistics. With the rise of romanticism, attention shifted to the third type of wit, to true creativity, although the rationalist assumption that normal human intelligence is uniquely free and creative and beyond the bounds of mechanical explanation was not abandoned and played an important role in the psychology of romanticism, and even in its social philosophy.
As I have already mentioned, the rationalist theory of language, which was to prove extremely rich in insight and achievement, developed in part out of a concern with the problem of other minds. A fair amount of effort was devoted to a consideration of the ability of animals to follow spoken commands, to express their emotional states, to communicate with one another, and even apparently to cooperate for a common goal; all of this, it was argued, could be accounted for on “mechanical grounds,” as this notion was then understood – that is, through the functioning of physiological mechanisms in terms of which one could formulate the properties of reflexes, conditioning and reinforcement, association, and so on. Animals do not lack appropriate organs of communication, nor are they simply lower along some scale of “general intelligence.”
In fact, as Descartes himself quite correctly observed, language is a species-specific human possession, and even at low levels of intelligence, at pathological levels, we find a command of language that is totally unattainable by an ape that may, in other respects, surpass a human imbecile in problem-solving ability and other adaptive behavior. I will return later to the status of this observation, in the light of what is now known about animal communication. There is a basic element lacking in animals, Descartes argued, as it is lacking in even the most complex automaton that develops its “intellectual structures” completely in terms of conditioning and association – namely Huarte’s second type of wit, the generative ability that is revealed in the normal human use of language as a free instrument of thought. If by experiment we convince ourselves that another organism gives evidence of the normal, creative use of language, we must suppose that it, like us, has a mind and that what it does lies beyond the bounds of mechanical explanation, outside the framework of the stimulus-response psychology of the time, which in relevant essentials is not significantly different from that of today, though it falls short in sharpness of technique and scope and reliability of information.
It should not be thought, incidentally, that the only Cartesian arguments for the beast-machine hypothesis were those derived from the apparent inability of animals to manifest the creative aspect of language use. There were also many others – for example, the natural fear of population explosion in the domains of the spirit if every gnat had a soul. Or the argument of Cardinal Melchior de Polignac, who argued that the beast-machine hypothesis followed from the assumption of the goodness of God, since, as he pointed out, one can see “how much more humane is the doctrine that animals suffer no pain.”2 Or there is the argument of Louis Racine, son of the dramatist, who was struck by the following insight: “If beasts had souls and were capable of feelings, would they show themselves insensible to the affront and injustice done them by Descartes? Would they not rather have risen up in wrath against the leader and the sect which so degraded them?” One should add, I suppose, that Louis Racine was regarded by his contemporaries as the living proof that a brilliant father could not have a brilliant son. But the fact is that the discussion of the existence of other minds, and, in contrast, the mechanical nature of animals, continually returned to the creative aspect of language use, to the claim that – as formulated by another minor seventeenth-century figure – “if beasts reasoned, they would be capable of true speech with its infinite variety.”
It is important to understand just what properties of language were most striking to Descartes and his followers. The discussion of what I have been calling “the creative aspect of language use” turns on three important observations. The first is that 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.” The fact surely is, however, that the number of sentences in one’s native language that one will immediately understand with no feeling of difficulty or strangeness is astronomical; and that the number of patterns underlying our normal use of language and corresponding to meaningful and easily comprehensible sentences in our language is orders of magnitude greater than the number of seconds in a lifetime. It is in this sense that the normal use of language is innovative.
However, in the Cartesian view even animal behavior is potentially infinite in its variety, in the special sense in which the readings of a speedometer can be said, with an obvious idealization, to be potentially infinite in variety. That is, if animal behavior is controlled by external stimuli or internal states (the latter including those established by conditioning), then as the stimuli vary over an indefinite range, so may the behavior of the animal. But the normal use of language is not only innovative and potentially infinite in scope, but also free from the control of detectable stimuli, either external or internal. It is because of this freedom from stimulus control that language can serve as an instrument of thought and self-expression, as it does not only for the exceptionally gifted and talented, but also, in fact, for every normal human.
Still, the properties of being unbounded and free from stimulus control do not, in themselves, exceed the bounds of mechanical explanation. And Cartesian discussion of the limits of mechanical explanation therefore took note of a third property of the normal use of language, namely its coherence and its “appropriateness to the situation” – which of course is an entirely different matter from control by external stimuli. Just what “appropriateness” and “coherence” may consist in we cannot say in any clear or definite way, but there is no doubt that these are meaningful concepts. We can distinguish normal use of language from the ravings of a maniac or the output of a computer with a random element.
Honesty forces us to admit that we are as far today as Descartes was three centuries ago from understanding just what enables a human to speak in a way that is innovative, free from stimulus control, and also appropriate and coherent. This is a serious problem that the psychologist and biologist must ultimately face and that cannot be talked out of existence by invoking “habit” or “conditioning” or “natural selection.”
The Cartesian analysis of the problem of other minds, in terms of the creative aspect of language use and similar indications of the limits of mechanical explanation, was not entirely satisfying to contemporary opinion – Bayle’s Dictionary, for example, cites the inability to give a satisfactory proof of the existence of other minds as the weakest element in the Cartesian philosophy – and there was a long and intriguing series of discussions and polemics regarding the problems that Descartes raised. From the vantage point of several centuries, we can see that the debate was inconclusive. The properties of human thought and human language emphasized by the Cartesians are real enough; they were then, as they are now, beyond the bounds of any well-understood kind of physical explanation. Neither physics nor biology nor psychology gives us any clue as to how to deal with these matters.
As in the case of other intractable problems, it is tempting to try another approach, one that might show the problem to be misconceived, the result of some conceptual confusion. This is a line of argument that has been followed in contemporary philosophy, but, it seems to me, without success. It is clear that the Cartesians understood, as well as Gilbert Ryle and other contemporary critics understand, the difference between providing criteria for intelligent behavior, on the one hand, and providing an explanation for the possibility of such behavior, on the other; but, as distinct from Ryle, they were interested in the latter problem as well as the former. As scientists, they were not satisfied with the formulation of experimental tests that would show the behavior of another organism to be creative, in the special sense just outlined; they were also troubled, and quite rightly so, by the fact that the abilities indicated by such tests and observational criteria transcended the capacities of corporeal bodies as they understood them, just as they are beyond the scope of physical explanation as we understand it today. There is surely nothing illegitimate in an attempt to go beyond elaboration of observational tests and collection of evidence to the construction of some theoretical explanation for what is observed, and this is just what was at stake in the Cartesian approach to the problem of mind. As La Forge and others insisted, it is necessary to go beyond what one can perceive or “imagine” (in the technical, classical sense of this term) if one hopes to understand the nature of “l’esprit de l’homme,” just as Newton did – successfully – in trying to understand the nature of planetary motion. On the other hand, the proposals of the Cartesians were themselves of no real substance; the phenomena in question are not explained satisfactorily by attributing them to an “active principle” called “mind,” the properties of which are not developed in any coherent or comprehensive way.
It seems to me that the most hopeful approach today is to describe the phenomena of language and of mental activity as accurately as possible, to try to develop an abstract theoretical apparatus that will as far as possible account for these phenomena and reveal the principles of their organization and functioning, without attempting, for the present, to relate the postulated mental structures and processes to any physiological mechanisms or to interpret mental function in terms of “physical causes.” We can only leave open for the future the question of how these abstract structures and processes are realized or accounted for in some concrete terms, conceivably in terms that are not within the range of physical processes as presently understood – a conclusion that, if correct, should surprise no one.
This rationalist philosophy of language merged with various other independent developments in the seventeenth century, leading to the first really significant general theory of linguistic structure, namely the general point of view that came to be known as “philosophical” or “universal” grammar. Unfortunately, philosophical grammar is very poorly known today. There are few technical or scholarly studies, and these few are apologetic and disparaging. References to philosophical grammar in modern treatises on language are so distorted as to be quite worthless. Even a scholar with such high standards as Leonard Bloomfield gives an account of philosophical grammar in his major work, Language, that bears almost no resemblance to the original and attributes to this tradition views diametrically opposed to those that were most typical of it. For example, Bloomfield and many others describe philosophical grammar as based on a Latin model, as prescriptive, as showing no interest in the sounds of speech, as given to a confusion of speech with writing. All these charges are false, and it is important to dispel these myths to make possible an objective evaluation of what was actually accomplished.
It is particularly ironic that philosophical grammar should be accused of a Latin bias. In fact, it is significant that the original works – the Port-Royal Grammar and Logic, in particular – were written in French, the point being that they formed part of the movement to replace Latin by the vernacular. The fact is that Latin was regarded as an artificial and distorted language, one positively injurious to the exercise of the plain thinking and common-sense discourse by which the Cartesians set such store. The practitioners of philosophical grammar used such linguistic materials as were available to them; it is noteworthy that some of the topics that were studied with the greatest care and persistence for well over a century involved points of grammar that do not even have an analogue in Latin. A striking example is the so-called rule of Vaugelas, which involves the relation between indefinite articles and relative clauses in French. For 150 years the rule of Vaugelas was the central issue debated in the controversy over the possibility of developing a “rational grammar,” one which would go beyond description to achieve a rational explanation for phenomena.
No doubt it is a complete misunderstanding of the issue of rational explanation that leads to the charge of “prescriptivism” that is leveled, quite erroneously, against philosophical grammar. In fact, there is no issue of prescriptivism. It was well understood and frequently reiterated that the facts of usage are what they are, and that it is not the place of the grammarian to legislate. At stake was an entirely different matter, namely the problem of accounting for the facts of usage on the basis of explanatory hypotheses concerning the nature of language and, ultimately, the nature of human thought. Philosophical grammarians had little interest in the accumulation of data, except insofar as such data could be used as evidence bearing on deeper processes of great generality. The contrast, then, is not between descriptive and prescriptive grammar, but between description and explanation, between grammar as “natural history” and grammar as a kind of “natural philosophy” or, in modern terms, “natural science.” A largely irrational objection to explanatory theories as such has made it difficult for modern linguistics to appreciate what was actually at stake in these developments and has led to a confusion of philosophical grammar with the effort to teach better manners to a rising middle class.
The whole matter is not without interest. I mentioned earlier that there are striking similarities between the seventeenth-century climate of opinion and that of contemporary cognitive psychology and linguistics. One point of similarity has to do with precisely this matter of explanatory theory. Philosophical grammar, very much like current generative grammar, developed in self-conscious opposition to a descriptive tradition that interpreted the task of the grammarian to be merely that of recording and organizing the data of usage – a kind of natural history. It maintained – quite correctly, I believe – that such a restriction was debilitating and unnecessary and that, whatever justification it may have, it has nothing to do with the method of science – which is typically concerned with data not for itself but as evidence for deeper, hidden organizing principles, principles that cannot be detected “in the phenomena” nor derived from them by taxonomic data-processing operations, any more than the principles of celestial mechanics could have been developed in conformity with such strictures.
Contemporary scholarship is not in a position to give a definitive assessment of the achievements of philosophical grammar. The ground-work has not been laid for such an assessment, the original work is all but unknown in itself, and much of it is almost unobtainable. For example, I have been unable to locate a single copy, in the United States, of the only critical edition of the Port-Royal Grammar, produced over a century ago; and although the French original is now once again available,3 the one English translation of this important work is apparently to be found only in the British Museum. It is a pity that this work should have been so totally disregarded, since what little is known about it is intriguing and quite illuminating.
This is not the place to attempt a preliminary assessment of this work or even to sketch its major outlines as they now appear, on the basis of present, quite inadequate knowledge. However, I do want to mention at least a few of the persistent themes. It seems that one of the innovations of the Port-Royal Grammar of 1660 – the work that initiated the tradition of philosophical grammar – was its recognition of the importance of the notion of the phrase as a grammatical unit. Earlier grammar had been largely a grammar of word classes and inflections. In the Cartesian theory of Port-Royal, a phrase corresponds to a complex idea and a sentence is subdivided into consecutive phrases, which are further subdivided into phrases, and so on, until the level of the word is reached. In this way we derive what might be called the “surface structure” of the sentence in question. To use what became a standard example, the sentence “Invisible God created the visible world” contains the subject “invisible God” and the predicate “created the visible world,” the latter contains the complex idea “the visible world” and the verb “created,” and so on. But it is interesting that although the Port-Royal Grammar is apparently the first to rely in a fairly systematic way on analysis into surface structure, it also recognized the inadequacy of such analysis. According to the Port-Royal theory, surface structure corresponds only to sound – to the corporeal aspect of language; but when the signal is produced, with its surface structure, there takes place a corresponding mental analysis into what we may call the deep structure, a formal structure that relates directly not to the sound but to the meaning. In the example just given, “Invisible God created the visible world,” the deep structure consists of a system of three propositions, “that God is invisible,” “that he created the world,” “that the world is visible.” The propositions that interrelate to form the deep structure are not, of course, asserted when the sentence is used to make a statement; if I say that a wise man is honest, I am not asserting that men are wise or honest, even though in the Port-Royal theory the propositions “a man is wise” and “a man is honest” enter into the deep structure. Rather, these propositions enter into the complex ideas that are present to the mind, though rarely articulated in the signal, when the sentence is uttered.
The deep structure is related to the surface structure by certain mental operations – in modern terminology, by grammatical transformations. Each language can be regarded as a particular relation between sound and meaning. Following the Port-Royal theory to its logical conclusions, then, the grammar of a language must contain a system of rules that characterizes deep and surface structures and the transformational relation between them, and – if it is to accommodate the creative aspect of language use – that does so over an infinite domain of paired deep and surface structures. To use the terminology Wilhelm von Humboldt used in the 1830s, the speaker makes infinite use of finite means. His grammar must, then, contain a finite system of rules that generates infinitely many deep and surface structures, appropriately related. It must also contain rules that relate these abstract structures to certain representations of sound and meaning – representations that, presumably, are constituted of elements that belong to universal phonetics and universal semantics, respectively. In essence, this is the concept of grammatical structure as it is being developed and elaborated today. Its roots are clearly to be found in the classical tradition that I am now discussing, and the basic concepts were explored with some success in this period.
The theory of deep and surface structure seems straightforward enough, at least in rough outline. Nevertheless, it was rather different from anything that preceded it, and, somewhat more surprising, it disappeared almost without a trace as modern linguistics developed in the late nineteenth century. I want to say just a word about the relationship of the theory of deep and surface structure to earlier and later thinking about language.
There is a similarity, which I think can be highly misleading, between the theory of deep and surface structure and a much older tradition. The practitioners of philosophical grammar were very careful to stress this similarity in their detailed development of the theory and had no hesitation in expressing their debt to classical grammar as well as to such major figures of renaissance grammar as the Spanish scholar Sanctius. Sanctius, in particular, had developed a theory of ellipsis that had great influence on philosophical grammar. As I have already remarked, philosophical grammar is poorly understood today. But such antecedents as Sanctius have fallen into total oblivion. Furthermore, as in the case of all such work, there is a problem of determining not only what he said but also, more importantly, what he meant.
There is no doubt that in developing his concept of ellipsis as a fundamental property of language, Sanctius gave many linguistic examples that superficially are closely parallel to those that were used to develop the theory of deep and surface structure, both in classical philosophical grammar and in its far more explicit modern variants. It means, however, that the concept of ellipsis is intended by Sanctius merely as a device for the interpretation of texts. Thus, to determine the true meaning of an actual literary passage one must very often, according to Sanctius, regard it as an elliptical variant of a more elaborate paraphrase. But the Port-Royal theory and its later development, particularly at the hands of the encyclopedist Du Marsais, gave a rather different interpretation to ellipsis. The clear intent of philosophical grammar was to develop a psychological theory, not a technique of textual interpretation. The theory holds that the underlying deep structure, with its abstract organization of linguistic forms, is “present to the mind,” as the signal, with its surface structure, is produced or perceived by the bodily organs. And the transformational operations relating deep and surface structure are actual mental operations, performed by the mind when a sentence is produced or understood. The distinction is fundamental. Under the latter interpretation, it follows that there must be, represented in the mind, a fixed system of generative principles that characterize and associate deep and surface structures in some definite way – a grammar, in other words, that is used in some fashion as discourse is produced or interpreted. This grammar represents the underlying linguistic competence to which I referred earlier. The problem of determining the character of such grammars and the principles that govern them is a typical problem of science, perhaps very difficult, but in principle admitting of definite answers that are right or wrong as they do or do not correspond to the mental reality. But the theory of ellipsis as a technique of textual interpretation need not consist of a set of principles represented somehow in the mind as an aspect of normal human competence and intelligence. Rather, it can be in part ad hoc and can involve many cultural and personal factors relevant to the literary work under analysis.
The Port-Royal theory of deep and surface structure belongs to psychology as an attempt to elaborate Huarte’s second type of wit, as an exploration of the properties of normal human intelligence. The concept of ellipsis in Sanctius, if I understand it correctly, is one of many techniques, to be applied as conditions warrant and having no necessary mental representation as an aspect of a normal intelligence. Although the linguistic examples used are often similar, the context in which they are introduced and the framework in which they fit are fundamentally different; in particular, they are separated by the Cartesian revolution. I propose this with some diffidence, because of the obscurity of the relevant texts and their intellectual backgrounds, but this interpretation seems to me correct.
The relation of the Port-Royal theory to modern structural and descriptive linguistics is somewhat clearer. The latter restricts itself to the analysis of what I have called surface structure, to formal properties that are explicit in the signal and to phrases and units that can be determined from the signal by techniques of segmentation and classification. This restriction is a perfectly self-conscious one, and it was regarded – I believe quite erroneously – as a great advance. The great Swiss linguist Ferdinand de Saussure, who at the turn of the century laid the groundwork for modern structural linguistics, put forth the view that the only proper methods of linguistic analysis are segmentation and classification. Applying these methods, the linguist determines the patterns into which the units so analyzed fall, where these patterns are either syntagmatic – that is, patterns of literal succession in the stream of speech – or paradigmatic – that is, relations among units that occupy the same position in the stream of speech. He held that when all such analysis is complete, the structure of the language is, of necessity, completely revealed, and the science of linguistics will have realized its task completely. Evidently, such taxonomic analysis leaves no place for deep structure in the sense of philosophical grammar. For example, the system of three propositions underlying the sentence “Invisible God created the visible world” cannot be derived from this sentence by segmentation and classification of segmented units, nor can the transformational operations relating the deep and surface structure, in this case, be expressed in terms of paradigmatic and syntagmatic structures. Modern structural linguistics has been faithful to these limitations, which were held to be necessary limitations.
In fact, Saussure in some respects even went beyond this in departing from the tradition of philosophical grammar. He occasionally expressed the view that processes of sentence formation do not belong to the system of language at all – that the system of language is restricted to such linguistic units as sounds and words and perhaps a few fixed phrases and a small number of very general patterns; the mechanisms of sentence formation are otherwise free from any constraint imposed by linguistic structure as such. Thus, in his terms, sentence formation is not strictly a matter of langue, but is rather assigned to what he called parole, and thus placed outside the scope of linguistics proper; it is a process of free creation, unconstrained by linguistic rule except insofar as such rules govern the forms of words and the patterns of sounds. Syntax, in this view, is a rather trivial matter. And, in fact, there is very little work in syntax throughout the period of structural linguistics.
In taking this position, Saussure echoed an important critique of Humboldtian linguistic theory by the distinguished American linguist William Dwight Whitney, who evidently greatly influenced Saussure. According to Whitney, Humboldtian linguistic theory, which in many ways extended the Cartesian views that I have been discussing, was fundamentally in error. Rather, a language is simply “made up of a vast number of items, each of which has its own time, occasion, and effect.” He maintained that “language in the concrete sense . . . is . . . the sum of words and phrases by which any man expresses his thought”; the task of the linguist, then, is to list these linguistic forms and to study their individual histories. In contrast to philosophical grammar, Whitney argued that there is nothing universal about the form of language and that one can learn nothing about the general properties of human intelligence from the study of the arbitrary agglomeration of forms that constitutes a human language. As he put it, “The infinite diversity of human speech ought alone to be a sufficient bar to the assertion that an understanding of the powers of the soul involves the explanation of speech.” Similarly, Delbrück, in the standard work on Indo-European comparative syntax, denounced traditional grammar for having set up ideal sentence types underlying the observed signals, referring to Sanctius as the “major dogmatist in this domain.”
With the expression of such sentiments as these, we enter the modern age of the study of language. The death-knell of philosophical grammar was sounded with the remarkable successes of comparative Indo-European studies, which surely rank among the outstanding achievements of nineteenth-century science. The impoverished and thoroughly inadequate conception of language expressed by Whitney and Saussure and numerous others proved to be entirely appropriate to the current stage of linguistic research. As a result, this conception was held to be vindicated, a not unnatural but thoroughly mistaken conviction. Modern structural-descriptive linguistics developed within the same intellectual framework and also made substantial progress, to which I will return directly. In contrast, philosophical grammar did not provide appropriate concepts for the new comparative grammar or for the study of exotic languages unknown to the investigator, and it was, in a sense, exhausted. It had reached the limits of what could be achieved within the framework of the ideas and techniques that were available. There was no clear understanding a century ago as to how one might proceed to construct generative grammars that “make infinite use of finite means” and that express the “organic form” of human language, “that marvellous invention” (in the words of the Port-Royal Grammar) “by which we construct from twenty-five or thirty sounds an infinity of expressions, which, having no resemblance in themselves to what takes place in our minds, still enable us to let others know the secret of what we conceive and of all the various mental activities that we carry out.”
Thus, the study of language had arrived at a situation in which there was, on the one hand, a set of simple concepts that provided the basis for some startling successes and, on the other, some deep but rather vague ideas that did not seem to lead to any further productive research. The outcome was inevitable and not at all to be deplored. There developed a professionalization of the field, a shift of interest away from the classical problems of general interest to intellectuals like Arnauld and Humboldt, for example, toward a new domain largely defined by the techniques that the profession itself has forged in the solution of certain problems. Such a development is natural and quite proper, but not without its dangers. Without wishing to exalt the cult of gentlemanly amateurism, one must nevertheless recognize that the classical issues have a liveliness and significance that may be lacking in an area of investigation that is determined by the applicability of certain tools and methods, rather than by problems that are of intrinsic interest in themselves.
The moral is not to abandon useful tools; rather, it is, first, that one should maintain enough perspective to be able to detect the arrival of that inevitable day when the research that can be conducted with these tools is no longer important; and, second, that one should value ideas and insights that are to the point, though perhaps premature and vague and not productive of research at a particular stage of technique and understanding. With the benefits of hindsight, I think we can now see clearly that the disparagement and neglect of a rich tradition proved in the long run to be quite harmful to the study of language. Furthermore, this disparagement and neglect were surely unnecessary. Perhaps it would have been psychologically difficult, but there is no reason in principle why the successful exploitation of the structuralist approach in historical and descriptive study could not have been coupled with a clear recognition of its essential limitations and its ultimate inadequacy, in comparison with the tradition it temporarily, and quite justifiably, displaced. Here, I think, lies a lesson that may be valuable for the future study of language and mind.
To conclude, I think there have been two really productive traditions of research that have unquestionable relevance to anyone concerned with the study of language today. One is the tradition of philosophical grammar that flourished from the seventeenth century through romanticism; the second is the tradition that I have rather misleadingly been referring to as “structuralist,” which has dominated research for the past century, at least until the early 1950s. I have dwelt on the achievements of the former because of their unfamiliarity as well as their contemporary relevance. Structural linguistics has enormously broadened the scope of information available to us and has extended immeasurably the reliability of such data. It has shown that there are structural relations in language that can be studied abstractly. It has raised the precision of discourse about language to entirely new levels. But I think that its major contribution may prove to be one for which, paradoxically, it has been very severely criticized. I refer to the careful and serious attempt to construct “discovery procedures,” those techniques of segmentation and classification to which Saussure referred. This attempt was a failure – I think that is now generally understood. It was a failure because such techniques are at best limited to the phenomena of surface structure and cannot, therefore, reveal the mechanisms that underlie the creative aspect of language use and the expression of semantic content. But what remains of fundamental importance is that this attempt was directed to the basic question in the study of language, which was for the first time formulated in a clear and intelligible way. The problem raised is that of specifying the mechanisms that operate on the data of sense and produce knowledge of language – linguistic competence. It is obvious that such mechanisms exist. Children do learn a first language; the language that they learn is, in the traditional sense, an “instituted language,” not an innately specified system. The answer that was proposed in structural linguistic methodology has been shown to be incorrect, but this is of small importance when compared with the fact that the problem itself has now received a clear formulation.
Whitehead once described the mentality of modern science as having been forged through “the union of passionate interest in the detailed facts with equal devotion to abstract generalization.” It is roughly accurate to describe modern linguistics as passionately interested in detailed fact, and philosophical grammar as equally devoted to abstract generalization. It seems to me that the time has arrived to unite these two major currents and to develop a synthesis that will draw from their respective achievements. In the next two lectures, I will try to illustrate how the tradition of philosophical grammar can be reconstituted and turned to new and challenging problems and how one can, finally, return in a productive way to the basic questions and concerns that gave rise to this tradition.
1 For additional details and discussion, see my Cartesian Linguistics (New York: Harper & Row, 1966) and the references cited there.
2 These examples are taken from the excellent study by Leonora Cohen Rosenfield, From Beast-Machine to Man-Machine (New York: Oxford University Press, 1941). The quotes are her paraphrases of the original.
3 Menston, England: Scolar Press Limited, 1967.