1. Let me say at once that this paper has a "covert" theme. I shall talk briefly about it at the beginning only, but it is in fact the leitmotiv of my whole discussion. The theme is: "What today is the place of linguistics in the framework of the cognitive sciences?". I shall mention only what seems to me to be the central issue here.

Artificial Intelligence is not a contemporary invention but derives from a time-honoured intellectual tradition. We might say that the programme of AI was already set forth, in outline, in Hobbes’s semiotics. Specifically Hobbes established the principle that the manifestations of the understanding can all be traced back to an activity that consists in the arbitrary manipulation of symbols. In the interpretation of authoritative theoreticians of AI (see Haugeland, 1985), the idea of the arbitrariness of "mental" signs has had some important consequences: the assumption that semantic rationality and syntactic coherence coincide; a lack of interest in the problem of the "substance" of signs and the (biological, electrical, electronic, etc.) apparatuses that transmit and receive them; the idea that the relationship between the correct utterance of the symbol and its referential link with the external world is essentially irrelevant. What is more, on the assumption that they all share the principle of the arbitrariness of signs, AI has envisaged doing away with the specificity of the various disciplines (including linguistics) and bringing them together in an all-inclusive science of the human mind, namely the science of symbolical manipulation, or "cognitive science".

In what way has contemporary linguistics contributed to this markedly "cognitivist" approach? How has linguistics reacted to this attempt at annexing and incorporating its identity in a more general discipline of mental sciences?

I shall not try to explore the jungle of 20th-century schools of linguistic theory here, partly because I believe that they can, in the end, be reduced to two philosophies of language, one Saussurean, the other Chomskian. Neither seems to add anything new or different to the semiotic presuppositions of AI. Indeed, the two positions can be said to have become the reference points – one philosophical and the other operative – of a formalist-mentalist outlook. With structuralism, in fact, the emphasis on arbitrariness and the privileging of "form" reached their speculative peaks, to the extent of arousing suspicions of idealism. With Chomsky the playing down of semantics and the privileging of syntax have become undisputed scientific standards. Yet neither of the two main sources of contemporary linguistics actually seems capable of suggesting alternative semiotic principles or points of view to AI.

A final remark before coming to the "overt" topic of my paper: the historiography of linguistics accurately mirrors the failure of linguistic semiotics to distinguish itself from the semiotics of AI. For decades now we have been busy back-dating the notion of arbitrariness, tracing the origins of structural semiotics in the most diverse philosophies, surveying and drawing up detailed inventories of works of grammar and syntax great and small. And although many of us might hesitate to admit it, we cannot deny that, even in the field of historiography, we have been totally subjected to the hyper-semiotic hegemony irresistibly exerted by the galaxy of arbitrarism. If proof were needed, it would be enough to mention an absence rather than a presence: the fact that there exists no really authoritative history of phonetics, a science of "matter" as opposed to form.

Let me come now to my overt theme. What I would like to do is pick out one or two fundamental moments in the history of linguistic enquiry when it was still possible to glimpse the outlines of a distinction between "artificial" and "natural" semiotics, an antithesis obliterated at the beginning of this century by the principle of arbitrariness and what followed from it.

First of all I would like to specify what I mean by talking about artificial and natural semiotics, when it would be more radical – and thus clearer perhaps – to talk about an explicit antithesis between semiotics and linguistics.

By "artificial" semiotics I mean all theories of signs based on the principle that it is essentially indifferent what semiotic material is used for constructing a cognitive and/or communicative universe based on form. By "natural" semiotics, on the other hand, I mean all those hypotheses which start from the ineluctable specificity of each single communicative code and of the bio-psychological apparatus which enables it to be implemented.

It is very important to clarify that the artificial/natural antithesis starts from, but is by no means confined to, the issue of the centrality of the material of which signs are made and through which they express themselves. This issue, in fact, is closely bound up with a far broader one, namely how the cognitive apparatus of "communicating" organisms is formed in strict dependence on the biological structure that "constrains" them to use certain expressive modes rather than others. Those who take the opposite position maintain instead that it is not necessary to distinguish between different bio-anthropological structures which think differently, because all, in the last resort, share a single organising principle, a single representing faculty.

In other words, it is not an internal problem of semiotics as a general discipline of signs, but an insuperable antithesis between a semiotic mode and a linguistic mode of conceiving of discursive-communicative activity, with all that it implies for the philosophy and psychology of mind.

In this perspective, an investigation of the literature on and by deaf-mutes and aphasics would seem to provide us with an excellent starting point, since they represent the two facets of language: deaf-mutes illustrate its origin or genesis, aphasics its phenomenology and development.

2. I will begin with the second group by citing an important observation made by J. H. Jackson, the greatest investigator of aphasia. Jackson puts his finger on a central issue when he postulates that every pathological symptom conceals a positive reaction of the organism. Thus attempts made by aphasics to construct sentences in which important elements are missing, para-aphasic behaviour in which the same words are repeated over and again, and the use of syntagmatic stereotypes in place of whole segments of the articulation of sentences, all appear to be forms of reaction and defence of the semiotic faculty.

Jackson, of course, disagrees with all associationist hypotheses and with the more extreme localising theories. According to these schools – of which Broca and Wernicke are the major representatives – aphasia always depends on the lesion of precise centres located in the left hemisphere of the brain. The lesion is supposed to destroy the "areas" (the notorious "auditory", "visual" "articulatory" and "graphic" centres of the Charcot-Bastian quadrilateral), or the conducting channels that carry information from one centre to another. Against this "geometrical" conception, first Baillarger and then Jackson and Alajouanine oppose a series of extremely common clinical cases.

For example: a young aphasic patient is asked the name of her daughter who is sitting next to her. The patient becomes very agitated and then burst into tears exclaiming: "Oh! my dear little Jacqueline, I don’t know your name any more!". Other aphasic subjects are asked to pronounce the word "five" or "April". Unable to respond to the explicit request, they nevertheless have no difficulty in uttering the series: "one, two, three, four, five, six…" or "January, February, March, April, May…". Aphasics unable to recognise the image of a zebra, answer "zebra" instantly when the association "run like a…" is suggested to them.

Any number of similar cases could be cited. The case history of aphasics abounds in subjects incapable of calling up a word in certain cases but quite capable of rapidly doing so in others. This phenomenon utterly rules out the existence of a "treasury" or "deposit" or "centre" of language in which the patient "fishes for" words and which is "wiped out" by the aphasic syndrome. The words have always been there: aphasics have not "lost" them. What they have lost is a higher-order language activity enabling them to retrieve words voluntarily by exercising what Jackson calls the "power to propositionize" ("loss of speech is, therefore, the loss of power to propositionize. It is not only loss of power to propositionize aloud (to talk), but to propositionize either internally or externally, and it may exist when the patient remains able to utter some few words". Jackson 1866-8: 126).

Thus, above and beyond this rational and voluntary faculty there still remain surviving active language phenomena which can be summed up as follows:

i) emotional language (anger, lamentation, swearing, etc.);

ii) infantile, dialectal, maternal or idiolectal language;

iii) language contextualised in syntactical and vocal-phonic frames or logical or habitual series (numbers, lists, nursery-rhymes, etc.);

iv) a-grammatical language (Pick’s "a-grammatism").

All these forms of non-propositional language elude the will but not meaning. Meaning can be hidden, syncopated, or personalised, but it is always present, so much so that subjects in this state feel a desperate sense of impotence when the listener is unable to understand them.

2.1 Both early and recent studies of other types of language disorder have shown, however, that linguistic survivals exist even in the absence of meaning. I am referring to the work done in the 19th century by Lélut and Séglas on the linguistic behaviour of the insane, but above all to the recent work of Luce Irigaray, which has mercilessly exposed the fragility of Saussurean and Chomskian semiotic frameworks when they are used to interpret the destruction of meaning in the language of schizophrenics.

In these subjects the automatisms already noted in aphasics are laid bare, in the sense that the only things to survive are syntactically-stereotyped rules, transformations and combinations, whereas all reference to meaning or deep structure vanishes. Schizophrenics, as much or more so than aphasics, seem to be "machines programmées une fois pour toutes et qui n’engendront qu’un discours fini, clos, à la fois limité et enveloppant" (Irigaray 1985: 227). Bereft of any link between the inside and the outside of language, the schizophrenic is "le plus rigoreusement syntaxier des linguistes et des sujets parlants" (ibid.: 228): in him syntactical operations take the shape of pure morphological virtuosity. His skill in articulating, assembling, and dismantling all kinds of verbal associations gives rise to a semiotic game based on the signifier alone: "c’est dire qu’il serait marqué par des sons dont les concepts lui resteraient cachés, voilés" (ibid.: 230). As for Jackson’s aphasics, for the schizophrenic "il n’y a pas de langue ni de parole […] pas de dictionnaire imprimé en lui dont on trouverait des exemplaires identiques déposés chez tous les individus d’une même société (ibid.: 224), "plus de signes à double face dans le langage du schizophrène, mais une écriture ou ré-écriture cryptogrammatiques d’inscriptions sonores" (ibid.: 231). In short, in the schizophrenic the signifier is emancipated from the control of meaning; the schizophrenic association no longer even obeys "la loi de l’arbitraire" (ibid.: 234); his signs attempt to set up an economy of signifiers, a "syntaxe idiolectale" (ibid.: 234) aiming simply to tame and "neutraliser la puissance des sons" (ibid.: 235).

2.2 Agnosia victims, too, can be considered "syntactical machines" devoid of semiotic or referential power. In 1833, when studying a syndrome called "verbal blindness", J. M. Charcot, the founder of studies on agnosia and a pioneer of localisation, recorded the case of a cultivated Viennese shopkeeper who had lost the capacity for visual recognition of objects and, although seeing them perfectly, could only grasp their syntactical, definitional and contextually-relevant reality:

A case of this kind is discussed by O. Sacks in his book The Man Who Mistook His Wife for a Hat (1985). His patient suffered from a form of agnosia that prevented him from seeing things directly. For example, if shown a glove, he could define its form and functions, give a minute description of its structure, and imagine its possible uses; however, he was incapable of "recognising" it, of making the statement "this is a glove". Sacks describes the way his patient proceeds by definitions as a cybernetic process that makes use only of the relational structure of semantic features, constructing factorial or componential analyses of objects completely devoid of referental reality.

In Sacks’s interpretation, it is possible to liken man to machine because of the patient’s lack of what J. Lordat, another great pioneer of research on aphasia, had, at the beginning of the nineteenth century, called "the embodying of ideas", namely the syncretic identification of the mass of "ideal" relations which is embodied in the "denomination" and which, as a result, "creates" the object – an operation which, according to Sacks, coincides with the Kantian concept of "judgment".

2.3. From what we have seen so far, it would thus appear that disorders always affect "propositional" activity, the cognitive, voluntary element of semiotic behaviour and expose the empty husk of automatic processes. "So I had lost the memory of the word, but I had retained the memory of the place it occupied", writes Doctor Saloz, an early nineteenth century aphasic who, on recovering his speech, wrote an important autobiographical memoir.

Roman Jakobson has translated this statement into the terms of general linguistics: he sees all pathological phenomena as derangements in the laws of systems: of the phonematic system in the case of sounds, of the lexical system in the case of signifieds, of the morphological and syntactical system in the case of sentences (Jakobson 1944). On this basis, he goes so far as to argue that disorders of signifiers can be predicted from a phonological scheme: where there is a phonetic impairment the subject will re-establish the internal equilibrium of the system by replacing one phonematic "value" with another that has the same function. This point of view is shared by most of those who believe in aphasic holism. Not surprisingly, Jakobson is influenced by K. Goldstein, who claims that there is no categorial difference between sounds, words, thoughts, movements, etc. The elimination of pathological specifities and the re-arrangement of all the components of language on a semiotic-cognitive level is the common denominator what I have called "artificial semiotic" outlook here.

Significantly, however, Jakobson makes no mention of the most important clinical study of the phonetic phenomena of aphasia, written almost five years before his Kindersprache und Aphasie (1944). This study – La syndrome de désintégration phonétique dans l’aphasie by Alajouanine, Ombredane and Durand (1939) – cuts the ground from under the structuralist semiotic approach, in fact. The phonetic disturbances of aphasia have nothing whatsoever to do with abstract phonological schemes, or with wrong mental representations of words. In phonetic aphasia there is no such thing as the one-to-one replacement of the original "values" by substitute "values" that maintain the economy of the system. All we find is defensive behaviours typical of the specific acoustic-prosodic organisation formed during the genesis of the language disorder.

For example: one of the patients analysed started all syllables and ended all words with a ‘stop’ because he needed a fixed point d’appui at the beginning and end of each vocal utterance. As a result, ‘soldat’ became ‘deudat’ (the final ‘t’ being pronounced), ‘jour’ became ‘kouk’, etc.

In another patient what was affected was the duration of all sounds. This rhythm disorder produced defence behaviours that were phonologically inexplicable: the elimination of voiced consonants because the muscular contraction lasted too long; the introduction of glottal stops or initial aspirations to link sounds that were too "detached"; the prolonging of vowels and the merging of phonemes because some of the opening movements of a sound persisted in the sound following, etc. In short, it was a case of "phonetic mutations pertaining only to oral expression" (Alajouanine-Ombredane-Durand 1939: 41): disorders not of the sign system, or of the phonological system in search of its "artificial" equilibrium, but of the primordial phonetic synergisms of speech.

The causal link between this type of natural specificity and human cognitive specificity is precisely what is denied by all the "artificial" hypotheses (in the sense defined above). Actually, the dualism between the cognitive-voluntary component and the mechanical-material component of language smacks of mentalist presumption. That it is indefensible would be immediately obvious if, after enunciating it, we went on to enquire what those linguistic automatisms surviving in pathological subjects represent and how they were formed in the first place.

We can thus reformulate Jackson’s intuition: the disease destroys and simultaneously creates. It destroys the correct mind-body relationship but creates an attempt to re-form the original mechanisms that enabled the setting up of the primordial synergisms between the inside and the outside of signs, between the material of semiosis and the efforts to represent it – mechanisms without which no form of natural intelligence could exist.

I would thus like to put forward a hypothesis based on the other facet of language disorders, the genetic one, as represented by deaf-mutes: any language automatism (whether phonetic, syntactic or semantic) that withstands the disorder is based on the original capacity for sensorial and pre-significational manipulation of the sound material. Biological phoneticity and orality are at the origin not only of the formation of meanings but above all of the procedures for constructing human cognition: not just of contents, then, or ideas, but also of the operations which much later on will make possible that arbitrary symbolical manipulation of which supporters of artificiality speak.

3. We have thus come to the second moment in the history of semiotics that I wish to recall here, namely the debate about deaf-mutes that took place between the beginning of the 17th and the end of the 18th century. Here too we find a deep-rooted disagreement between supporters of artificial semiotics (re-education based on manual signs) and supporters of natural semiotics (pedagogical techniques for rehabilitating subjects to the spoken word).

In its origins, the manual method – backed by L’Epée, Sicard, Bébian and many other 18th-century French educationalists – can be seen as representing a radical, and hence perhaps clearer, version of the Saussurean paradigm of form.

In the Saussurean type of formalist paradigm, in fact, one of the misunderstandings that has prevented us from discerning the full extent of its contradictions is that, in spite of his general semiotic outlook, Saussure always had natural language in mind as his living example.

The advocates of manual methods, on the other hand, no longer relate form to spoken language, which is excluded by definition, but to a conception of the "faculty of language" inspired by an elementary and extremely crude notion of arbitrariness, thanks to which all kinds of signs can be considered interchangeable.

A number of specifications follow from this approach:

a) since its signs are not naturally generated, the setting up of a manual code appears explicitly conventional: i.e. it is constructed artificially for the community of sign-users;

b) constructing an explicitly conventional system reinforces the need for an approach that is "calculational" and "abstract" in every possible sense, involving:

c) this set of consequences induces the "formalists" who theorise manual signs to overrate the rational power of manual language: manual signs, they claim, are distinct, precise, self-sufficient, and unambiguous. They are produced by and produce ideas and thoughts as clear as those of "chemical nomenclature" (Bébian 1817: 54). "I [writes Desloges 1779: 17] go directly and necessarily from the perception of the sign to that of the ideas". It is a position endorsed by Condillac, who goes so far as to claim that the manual signs of deaf-mutes even give us "more exact and precise ideas than those usually acquired with the aid of the hearing" (1776: 117).

d) the greater rationality of manual language is contrasted with the subtelty, vagueness, ambiguity, metaphoricalness, and "corporeal heaviness" of oral language:

The language of signs, on the other hand, speaks with "the clarity of an incontestable fact (ibid.: 54).

What the "manualists" fundamentally object to is thus the "bodily", concrete status of the words, sounds and vocality intrinsic to verbal language. It introduces an unnecessary obstacle between the system of symbols and the system of things. From this point of view, then, the language of signs, far from hindering intellectual abstraction, actually favours it:

3.1. The "oralists" – or supporters of the method of rehabilitation to spoken language – respond with three main arguments:

J. K. Amman, the founder of the "oralist" re-educational method at the end of the 17th century, starts, for example, with the idea that the subject learning to speak aims above all at constructing tools before constructing products. The "most intimate relationship" and the "perfect correspondence" existing between the organs of hearing and those of articulation are incapable by themselves of setting in motion the mechanisms of language. Hearing in itself, in fact, plays no active role in imitation: someone who hears does not yet know what he must "do" to produce the sounds. To struggle against the difficult physicality of pronunciation and the easy intuitiveness of meaning is the first step in the oralists’ programme. What takes place, in fact, during the phase known in modern psycholinguistics as the "period of latency", in which the new-born infant begins easily to understand the meaning of many words without being able to pronounce them, is in fact this characteristic mechanical refining of articulatory procedures. Paradoxically, the climax of cognitive evolution is not the big-bang of signifieds – which is an almost banally behavioural event – but the big-bang of the signifiers, which can be verified not by the test of the recognition of an object named and already known for some time, but by the coincidence between sounds originally heard and those finally pronounced:

The strong version of the most radical oralist thesis consists, however, in the argument that the big-bang of signifiers is not a generically semiotic event but a specifically linguistic one. By which is meant that the ensemble of man’s bio-psychological apparatuses can attain to its cognitive specialisation only by means of the unrepeatable embodying of experience in sounds. Their production is no mere historical accident, nor is it a choice conditioned by mere functional convenience. Only through phonic sensoriality is it possible to achieve not circumscribed intellectual units of sense but manifestations of discursive vitality which is instrumental both to the cognitive construction of the world and to the emotional-passional pathos of being. Thus, in addition to the traditional semiotic arguments on behalf of the voice (it needs no tools, it can be heard in the dark, it leaves the hands free, etc.) Amman concludes:

It is to this extraordinary synergism of heart-mind-language that the human bodily apparatus is entirely adapted. Amman constantly talks about "a most intimate, vital correspondence and commerce" of the vocal organs with all their counterparts in the sensory-motor, emotional and cognitive apparatus which produce that "vital character which emanates from the heart and the brain" (ibid.). Other oralists (Deschamp, Ernaud, Pereire, etc.) agree that the production of sound is the anthropological fulcrum of man’s complex cognition. To demonstrate this they "re-condition" deaf-mutes using a completely mechanical method based on training them to recognise the vibrations of their throats by touch and to lip-read, which involves a detailed knowledge of the phonetic apparatus and its functional mechanisms. In this way Amman entirely re-draws the map of theory and application in experimental phonetics.

The impact of oralism does not stop, however, at the clinical-practical antechamber. For oralists, the voice also demolishes the artificialists’ philosophical myth of the combination of elementary features. Just as Locke painstakingly demonstrated in his Essay that no arbitrary combination of signs could enable a blind person to grasp the meaning of the term "scarlet", so no form of "componential" semantics, no combination of features as distinct "as those of chemical nomenclature", will ever succeed in "photographing" the ultimate sense of the outer world:

To teach the term "God", for example, the artificialist has to pick out a limited number of semantic markers: to select the idea that God is being par excellence, that he has created all things, that he is the sum of all perfection. The instructor will thus begin by pointing to the sky where he dwells and the objects he has created, etc. Nevertheless, however much this procedure is refined and complicated, it will never succeed in achieving a sufficiently complex communicative function. No matter how far the meaning is expanded, no matter how many predicates are used to define it, a deaf-mute deprived of the "embodying of the voice" "will not understand his nature or his mode of being any better" (Deschamps 1779: 20).

4. In the 17th- and 18th-century linguistic tradition there are at least two major philosophical referents for a theory of language which is coherent with these principles of applied linguistics. One is Locke, the other Vico. They have in common a philosophy of language based not so much on the indissoluble relationship between senses, imagination and language as on the primacy of this relationship during the ontogenesis and phylogenesis of the understanding. Unfortunately historians of linguistics have interpreted both authors from the standpoint of "artificialist" semiotics and overlooked a number of significant details. It is hardly surprising that in all their reconstructions of the "pre-Saussurean" Locke or of Vico a number of "spurious" topics (which nevertheless appear throughout these authors’ works) are completely ignored, or interpreted as "oddities". The philosophers’ continual recourse to examples involving sense deprivation – Locke’s blind men, Vico’s "mutoli" (mutes) – , the problem of how simple, non-arbitrary ideas are formed on the one hand, and of singing, music, rhythm, and many other Vichian oddities on the other, seem to have been suppressed by the exegetic consciousness of historians of linguistics.

For an up-to-date theory of the difference between thought and language, Locke’s and Vico’s formulations, however, are nothing more than distant philosophical referents. In times of mentalist hegemony it is not at all easy to understand the damage that can be done by confusing or denying that which is typical of thought and that which is typical of language. Nevertheless, in some branches of contemporary psychology this distinction is beginning to emerge.

The psychologist Gaetano Kanizsa, for instance, has clarified the difference between seeing and thinking, which can be said to mirror the distinction between speaking and thinking. The main line of Kanizsa’s reasoning runs as follows: in cognitive psychology a broader sense of the term "visual perception" is beginning to prevail, according to which there are no essential differences between the process whereby visual clues are formed and the process whereby they are interpreted. Each are seen as "ratio-morphic procedures" typical of discursive and scientific thought (categorisation, formation of hypotheses, production of inferences). It would thus appear that the rules of reasoning condition perception in all its phases: "what we see not only appears to be used by inferential processes in the phases of interpretation but also to be the product of unconscious inferences during the primary process" (Kanizsa 1991: 22).

Kanizsa contests this dominant thesis of cognitive psychology by citing a series of visual experiences of perceptual completion that unequivocally reveal the existence of cases in which visual results openly contradict both contextual rules and rules of experience interpreted on the basis of purely cognitive models. This demonstrates, he argues, that "seeing follows a different logic", or more precisely, that it does not follow any logic at all, but simply "functions according to autonomous principles that are not the same as those regulating thinking" (ibid.: 38).

Of course, in the ordinary phenomenology of adult sight, "seeing" and "thinking" continually interact, giving rise to an impression of the "rationality" of sight that leads cognitivists, as Kanizsa puts it, to imagine "the mechanism of perception processes as if it were itself a computer programmed to carry out inferential operations of a logical kind" (ibid.: 47). Here too, this is possible because "the physical representation of stimuli is replaced by a symbolical representation; and the sense experience is conceived as the set of conclusions reached by a sequence of procedures for the manipulation of symbols" (ibid.: 47). On the other hand, the genesis of sense experience is rooted in that sort of pre-symbolic magma which in perception has been called the "primary" or "pre-attentive" process.

Sight and language are not the same thing, of course. In the first place, because sight is a sense and language is not. In the second, because the use of the terms "symbol" or "sign" for sight is almost a metaphor of biological processes, whereas for language it is the appropriate term for representative projection. Lastly, because it seems legitimate to suppose that there are bio-chemical and neuro-physiological correlates for sight that set up the genetic code of the "primary processes", whereas for language we do not know whether these exist or not.

In spite of this, the need to identify a linguistic prius that cannot simply be equated with a thought, an idea, a concept, a meaning, a sense, or with the cognitive operations that use this set of notions, should, I feel, be one of the main goals today of a theory of language that is not prevalently semiotic.

The oralist instructors of 18th-century deaf-mutes seem to have located this prius in the voice. However, present-day linguistic theory – which has been drawn irresistibly into the orbit of semiotics – has not paid much attention to this claim on behalf of the power of the signifier. Only experimental phonetics, when it has broken away from the suffocating embrace of phonology (as for instance in the school of Ph. Lieberman), has opportunely revealed the shortcomings of semiotic approaches to the processes of coding and decoding speech. Nevertheless, these studies seem so far to lack a genetic outlook. And yet it is precisely when it first comes into being that the tumultuous syncretism of vocality produces the cognitive-emotional "embodiment" mentioned by Amman and Lordat. It is thus likely that we will have to look even further back in order to explain the ontogenesis of language. The only linguists who has so far attempted to follow this path is I. Fónagy (1983), who, in his psycho-phonetics, has begun to explore the impervious terrain of the pulsional bases of phonation, prosody, proxemics and vocal kinesics. Beyond this threshold we find only a number of major studies by non-linguists. It is to audiologists, neonatologists, embryologists, neurophysiologists, child neuropsychiatrists and even to psychoanalysis that we must turn today in order to try to set up a genetic theory of the oral basis of human cognition and reconstruct a non-arbitrarist scientific paradigm in the sphere of semiotics and philosophy of language.

4.1. The idea that to be able to understand the mechanisms giving rise to the primary physio-psychological synergisms of language we need to probe into the neo-natal or pre-natal realm began to make headway around the nineteen-twenties. The first studies in the embryology of communication confined themselves (i) to observing that the new-born infant and the foetus (from as early as its fourth month of life in the womb) react to sound stimuli from both the inside and the outside, and (ii) to setting up a method of experimental observation devised for measuring this type of physiological reflex.

Very soon, however, the accumulation of a vast body of data from experiments under way made it possible to set up an extremely rich general theoretical framework, and to approach the understanding of this phenomenon from the various standpoints of biological sciences. Let me now try to give a synopsis of the most important results achieved so far.

4.1.1. In the first place, it has now been clearly shown that at the anatomical-physiological level hearing occupies a special, primary position among the various sense systems.

For example, it has been ascertained that the vestibule and its connection with the tectum and crown are formed extraordinarily early. Between the fourth and fifth month of gestation the inner ear, the vestibule, the tympanum and the chain of ossicles have already reached adult size. The osseous structures of the ear have, in fact, a unique characteristic: unlike the rest of the Haversian osseous system, they are not formed from a nucleus of medullar substance around which, through the system of nutrition, a laminar shell forms, growing stronger and ossifying definitively only over a long span of time (the tibia, for instance, may take twenty years or more to reach full length). On the contrary, the malleus, the incus, the stapes and the rigid part of the acoustic labyrinth already show osseous consistency at birth. From a histological viewpoint the ear ossicles form a kind of whole with the skull and in particular with the osseous tissue of the temporal bone. From a morphogenetic point of view they seem to be remodelled elements of the set of branchial arcs forming a single piece with the primitive facial walls, the styloid apophysis, the stylohyoid ligament and the thyroid cartilage.

According to some studies carried out mainly with radiographic techniques, this "fantastic advance of the ear in time" (Tomatis 1981: 228), can be explained by the need for an instrument for frequency analysis in the sensorially "activated" foetal organism. The morphological linkage between the temporal bone and the osseous part of the inner ear in fact forms a perfect mechanism for setting off a ringing vibration in the rigid structures of the organism on which the adult faculty of the osseous ear depends (and which is often used to enable deaf-mutes suffering from profound deafness "hear" music). It is to this structure that man owes his ability to monitor – through the cheekbones and mandibles – the modulations in intensity of his own voice, i. e. his articulatory feedback.

The most recent studies have gone even further towards explaining the precocious formation of the most archaic and internal part of the skull-ear, postulating that the vestibule has an extremely particular function.

As is well known from studies in anatomic audiology, the primary hearing process is characterised by a dual level of perception. The first involves the "quantitative" analysis of the variations in pressure due to vibration; the second involves the "qualitative" analysis of frequencies. Quantitative analysis, which measures the periodicity or a-periodicity of sound stimuli, in effect makes possible the perception of regularities within our acoustic environment: rhythms, time sequences, relative intensities, and intonation curves. In other words, quantitative analysis processes the musical and prosodic component of language, abstaining from, or preparing for, the recognition of the "content" of sound, its articulation. This type of processing is based above all on low tones and receives inputs above all through rigid channels of access, laminar and osseous tissues that relay the vibrations "internally". The final, physiologically-internalised product of this processing is a kind of "calculated" regularity of "sonorous" recurrences, which has its morphological counterpart in the astonishing similarity of the acoustic labyrinth of foetuses. The vestibular apparatus thus prepares and exercises the cochlea by means of quantitative-rhythmical training: at a given moment what can be properly considered a relay comes into play and transfers the processing of the musical, continuous substance to the organ which will convert it into articulated and discrete phonic substance.

The results obtained by the embryology of hearing in its functional reconstruction of the roles of the various components of the inner ear thus reveal an important feature of the primary activity of the vestibular labyrinth, namely its capacity to store, during the foetal stage, rhythms and cadences that will form the basis of the perceptual faculties of the infant. To put it another way, the first nucleus of the ear – which is also genetically the first sense instrument to form itself – by means of an exhaustive engrammatic reconnaissance of a "musical (or rhythmic or prosodic) topic", produces a sort of diffuse (i.e. not yet specialised) sensorial expectation which prearranges the modes of access to the tabula rasa of the nervous system, to its initialisation, or, in more specific terms, to the future processes of myelination of the neural system.

4.1.2. Very recently a number of important psycho-physiological investigations of the behavioural reactions of infants only a few days old to acoustic hearing have clearly confirmed the rhythmical-physiological origins first of the nervous processes and then of the fully cognitive processes . I am referring in particular to the studies carried out in France at the "Laboratoire de Sciences Cognitives et Psycholinguistique" in Paris under the guidance of J. Mehler, and to the identification of the characteristics of "motherese language" which many researchers have been exploring since the nineteen-eighties.

Mehler and his team have shown that we must turn to neonatology or cognitive embryology if we want to understand how adults are able to use their grammatical knowledge in the actual practice of natural speech. The most striking theoretical outcome is that traditional linguistic conceptions based on the idea of a "pyramidal" hierarchy of minimal units starting from the simple (phonemes) and ending with the complex (sentences) are not only ineffective for the practical analysis of communicative interaction, but also inadequate for constructing a theoretical model that might be of use in artificialist simulations of human language.

The novelty of this work lies above all in the use of items deriving from natural pluri-syllabic chains in experiments. In fact, as long as experiments attempted to probe the reactions of infants using phonemes or synthetic syllables (i.e. syllables detached from real sentence chains) as stimuli, the only significant result was the demonstration that the neonate’s capacity for phonemic discrimination is always linked to the presence of "vocalic kernels" in the inputs. In other words, that neonates distinguish sounds, but not as phonemes, given that the presence of a vowel is essential for minimal recognition to take place. These results have reinforced the hypothesis that the syllable is the unit of discrimination, since the definition of the concept of syllable is precisely the presence of a vowel around which the most elementary phonematic variation can be organised. Two-day-old infants, in fact, can only recognise global, syllabic-type units, whereas at four months they begin to show signs of a more analytical and detailed differentiation of consonantal sounds.

As in the transition from vestibular ear processing to cochlear ear processing revealed by embryological audiology, here too the evolutionary phases indicate that the infant’s initial hearing is preceded by a global "musical" or rhythmical training, which paves the way for an increasingly discrete articulatory functionalisation or specialisation.

What is structurally peculiar about the syllable, in fact, is its musical or prosodic nature: the syllable carries stress, it is the smallest unit in which intonation variation can already be detected. Indeed, rather than a unit in the (phonological, morphosyntactical, semantic) sense that we normally ascribe to this concept, the syllable is simply the mark of auditory attention based on broad segments of speech. For this reason, in experiments, syllabic perception is most convincingly demonstrated when syllables are presented as items within "natural" configurations: i.e. pluri-syllabic sequences incorporated in whole sentences or segments of sentences. In such cases the human voice is represented by its most significant acoustic properties: co-articulation between syllables, full stress variation, and the appearance of a melodic contour. These features form the basis not only of phonetic perception and recognition, but also of syntactical recognition as such, in the sense that natural prosody makes it possible to identify the boundaries of syntactic constituents and the order of words. This obviously does not mean that syntactical units do not exist as purely combinatory entities with their own, autonomous grammatical definition. What it means is that in first stage of ontogenesis, they are no more than rhythmical sequences which are perceived by registering the pattern of their pauses; this prepares the way for the subsequent perception of content and form (of a grammatical kind) as such.

As a result of this, our theories of language require modifying in two respects.

The first modification is of a practical-descriptive nature: the analytical order in which contemporary linguistic theories tend to process communicative events needs to be inverted. Instead of taking the phoneme as our starting point and seeing sentences as the result of combining it with other phonemes, we need to start from the continuity of the sentence (or at least from acoustic continuums) and proceed first to distinguish syllables, and then to distinguish phonetic elements.

The second modification is a theoretical one and consists in acknowledging the primacy of the substance with respect to the form of language, and, within the substance, in recognising the function of non-articulated vocality as opposed to articulated vocality.

At the outset, in fact, items of language are not only not perceived semantically or phonologically: they are not even perceived as phonetically articulated. Four-day-olds have been exposed to their mothers’ natural sentences and show reactive behaviours to them (recognition); when the same voices are filtered at frequencies (400Hz) high enough to destroy not only all kinds of semantic signs (indices) but all phonetic articulation too, leaving only prosodic signs intact, the neonates are still able to recognise their mothers’ voices.

Studies of "motherese language" seem to point in the same direction as Mehler’s work. These consist of a series of researches that have registered a remarkable number of differences in the way a mother speaks when she turns from an adult listener to her new-born child. Above all these changes consist of a higher basic frequency (FO), a slower rhythm, an exaggerated intonation – in other words, what has been termed "an expanded intonation contour" (Fernald & Simon 1984). This type of signal carries a high amount of perceptual saliency for infants, even those only a few days old (ibid.: 105). The interaction between maternal musical language and the neonate’s perceptual preference for an expansion of intonation contours seems to constitute a true "language universal" both of production and of reception. (Grieser-Kuhl 1988: 19). In fact, it has been recorded in both European and Asiatic languages, in tonal and non-tonal types of language, and no differences have been recorded between primiparous and multiparous mothers (Fernald & Simon 1984: 110). From a functional standpoint, these studies have demonstrated a number of important features which it is worth listing one by one:

1) prosody increases the redundancy of the message: whereas two syntactically identical sentences pronounced with a neutral intonation arouse no reaction, two syntactically different sentences pronounced with an expanded intonation contour trigger the neonate’s response;

2) the exaggerations of wave-crests in "motherese" maximise the attention of the neonate and introduce a kind of "turn-taking" into the interaction;

3) the expansion of intonation contours enhances affective motivation: in adults, too, high intonation peaks and an increase in the range of prosodic variation marks behaviour of greater affective intimacy;

4) the prosodic patterns of the mother amplify the perceptual abilities of the neonate thereby paving the way for phonological, semantic and syntactic organisation; in particular, the continuity of intonation segments is what most likely foreshadows the idea of the sentence;

5) the exaggerated rhythmical patterns of maternal language play a basic role in the formation of the infant’s metrical competence, and hence in its manipulation of the speeds most suitable to the perception and production of language: "this temporal and tonal redundancy may aid the infant in auditory pattern recognition, an essential skill in the development of speech perception" (Fernald & Simon 1984: 112).

Taken as a whole, these observations seem to suggest that what has been called the "prosodic bootstrapping hypothesis" fits in perfectly with the view that an auditory-musical apprenticeship takes place during the foetal stage. The point of contact between post-natal research into the characteristics of "motherese language" and experiments on the embryo’s perception of language is represented by the work of A. J. De Casper who has probed even deeper into the nature of voice recognition by infants immediately after birth. In particular, in his study entitled Histoire de foetus par un nouveau-né (1990), which sums up many years of experience in psycholinguistic and embryological research, De Casper reveals that the perceiving organism is sensitive to the mother’s voice not only in the immediately post-natal phase of life but also in the pre-natal phase; and, within the mother’s voice, it is sensitive to the most recurrent expressions, the "maternal melodies" it has learned to recognise in the womb.

What is more, De Casper puts forward and experimentally tests a hypothesis that may provide an essential key for understanding the totally unconscious nature of the original mechanism by which communication is first implemented and for explaining the rhythmical, prosodic, or, in the most archetypal sense of the word, musical origin of language.

As we have already seen, Amman – the oralist healer of deaf-mutes – had already sketched out a sort of metaphysics of the voice in the 17th century: the voice, which expresses the bodily synergism of heart and brain, passion and reason, is the life of language.

Today, De Casper (and others) are postulating that non-articulated vocality, prosodic melody, pre-frequential rhythm, the musical alternation between regularity and irregularity – phenomena which, as we have seen, necessarily precede the analysis or assembling of combinations of "discrete" atoms in the natural evolution of language – all, in fact, derive from the monistic relationship between mother and foetus, united by the silent transmission and reception of the sole acoustic source that can be "discriminated" from the medley of intra-uterine noises: the heart-beat. This is no mere archetypal image whose explanatory power lies in its undefinable suggestivity and fascination. Far from it: as De Casper has shown in a series of repeatable and rigorously circumscribed experiments, non-nutritional sucking (the symptom of a "perceptual" reaction of the foetus) measured by probes linked to an electronic data-recording system, is rigorously synchronised with the mother’s heart-beat. Not only does the foetus react (by increasing the number of suctions) to the pulsing of the mother’s heart, but it stops to listen to it, making a clearly discernible pause between systole and diastole. This ecstatic, unconscious starting and stopping in conjunction with the sound of the heart-beat is, moreover, absolutely specific. Within the uterine acoustic environment, which is brimming with noises of all kinds (breathing, gastric activity, movement of fluids, etc.), the mother’s heart-beat is "the only unspoken stimulus that it has been possible to recognise as having reinforcement value for the new-born infant" (De Casper 1990: 170).

How far this primordial experience will affect the cognitive future of that essentially physiological mass which is the foetal organism is today the most stimulating object of enquiry of contemporary embryology. There can be no doubt, however, that the path taken by the embryology of communication is calling in question some of the theoretical cornerstones of contemporary semiotic frameworks: the assumption that participants in communication are separate; that we share a code understood as a system of rules for combining minimal entities (of whatever kind); that the nature of the signifying material is essentially irrelevant (the minimal version of arbitrariness); that the articulated substance (or articulation itself) has priority; that consciousness or will or communicative intention is fundamental.

In positive terms, embryological research has sketched out the lines of a development which has barely begun, but which obliges linguistics to break free of the formalist fixation that has characterised it ever since Saussure. It has brought to light the pre-articulatory nature of language, its irreplaceable function in the formation of human cognition, thereby confirming Amman’s early intuition of the centrality of the voice: the voice as the life of language, the body of language, the concreteness of language.

5. Having a mind without life, without body, without concreteness is precisely what pathological subjects deprived of their voice complain of. From their autobiographical writings – which offer the most extraordinary testimony of unconscious semiotic theorising – three major points emerge with which I would like to conclude this paper:

As Abbé Copineau noted, summing up a long passage by Desloges, an 18th-century deaf-mute trained with the method of manual signs, the language of signs is "une définition perpétuelle des idées qu’on y exprime" (Desloges 1770: 56) and the reality of the senses is seen "comme à travers une glace transparente" (ibid.: 57).

In the Mémoires of doctor Saloz, an 18th-century aphasic I have already mentioned, the loss of the spoken word and the persistence of the "thought" word gave rise to a feeling of confusion and indefiniteness, a weakening of ideas so that they became ungraspable, "comme flou" (Saloz 1919: 9).

Abbé Carton, an educator of blind deaf-mutes, defines their language of touch as a vicious circle, a "synonymie toujours incomplète" (Carton 1840-41: 138).

Similar views are expressed by Saboureaux de Fontenai, Laura Bridgman, and above all by the famous blind deaf-mute Helen Keller, who complained of the absolute lack of concreteness of manual sign language.

In Helen Keller the loss of oral concreteness and the overcrowding of abstract signs accentuates the impression that manual language has become a semiotic prison on which the coercive structure of a purely mental reasoning depends.

The loss of access to the spoken word arouses the same feeling in all of them: they see their minds as an unorganic, random entity, a jumble of fragmented, unconnected signs:

c) the spoken word, a sense of concreteness and cognitive organicity – these are the main privations felt by subjects who have gone through all the levels of artificial re-training. This training has certainly produced arbitrary symbols and intelligences capable of manipulating them, but it has left a cognitive gap that can be filled only by the recovery of the voice. It is extraordinary that all the "deprived" persons I have mentioned choose spontaneously at some point to break out of the cage of artificial signs (gestures, writing, etc.) in order to learn spoken language in spite of the enormous effort required. Only this can make them feel that they have fully regained that sense of life which Amman attributed to the voice, and which Helen Keller identified as free expression in rapid and winged words, for which manual signs can never be a substitute.

6. This paper began with the admission that its real purpose was to assign to linguistics the place that it deserves among the cognitive sciences. I hope the line of reasoning I have followed has made it clear that this place ought to be a pre-eminent one, provided that linguistics ceases to be confounded with semiotics, which is a theory not essentially different from that of the cybernetics of information, although richer of course.

To achieve this goal we need to give renewed dignity to the study of the specificity of the material of signs. To consider the material they are made of and above all the kind of bio-psychological procedures they impose on the living organism, and hence the kind of anthropological cognition they give rise to, as irrelevant, is as detrimental to cognitive epistemology as it is to linguistic epistemology.

Something is stirring in the stronghold of AI as well. The ups and downs of its first forty years has led to a re-examination of the sources of artificialism. It is no accident that connectionism has revived biologically-orientated aspects of Wiener’s cybernetics in which the formalist element tends to be rooted in ethology. Although it still seems extremely reductive, the shift from the simulation of the mind (i.e. essentially the operations of logic) to the simulation of the emotional, intuitive, imaginative, non-logical and not-strictly-semiotic features of the brain, is an important development that linguistics cannot ignore.

In any event, in the new epistemological architecture of the human sciences there is very little space left for a pure linguistics of the sign. Events in the distant world of the embryology and pathology of language are an invitation to us to bring the priceless historical accumulation of reflections on language to bear on psycholinguistics, phonetics, and the anthropology and biology of language – to embrace a new philosophy capable of reviving and remoulding the exhausted energies of formal and artificial thought too.