2. ELEMENTS OF MOSAIC EVOLUTION
3. SELECTIVE VALUE OF LANGUAGE
4. CATEGORICAL PERCEPTION OF SPEECH SOUNDS
5. CROSS-MODAL COMPONENTS
6. INNATE AND ACQUIRED MOTOR PROGRAMS
7. EVOLUTION AND ONTOGENY OF MOTOR/SPEECH NEURAL CONNECTIONS
8. MOTOR CONTROL AND VISION
9. MOTOR CONTROL AND ARTICULATION
10. MOTOR PROGRAMS AND ELEMENTARY MOTOR UNITS
11. MOTOR CONTROL RESEARCH -BIOLOGICAL AND ROBOTIC
This paper amplifies and develops the motor theory of language origin and evolution presented in previous papers for meetings of the Language Origins Society at Cracow and Oxford in 1985 and 1986. This section summarises the main ideas presented in those papers.
Language is taken to be the capacity of one individual to alter, through structured sound emission, the mental organisation of another individual. In considering the origin of language, we should not look for a distinct, datable origin any more than we would look for a distinct, datable origin for the eye. Language is more than speech just as perception is more than the structure and functioning of the eye. In both cases we have also to be concerned with the neural organisation underlying the functions of speech and visual perception.
The fundamental idea is that language was constructed on the basis of a previously existing complex system, the neural motor system. The programs and procedures which evolved for the construction and execution of simple and sequential motor movements formed the basis of the programs and procedures going to form language. At every level of language, from the elementary speech sounds, through the word-forms on to the syntactic rules and structures, language was isomorphic with the neural systems which already existed for the control of movement.
The second principal theme is the mosaic evolution of language, the fitting together of a whole array of elements, anatomical, neural and behavioural. Many of the elements necessary for mosaic evolution of the language capacity can be found in the anatomical and behavioural repertoires of other animals, and particularly of birds.The conclusion drawn is that if birds and other animals have, individually,behavioural elements required for the evolution of human language-capacity, then they must have the neural structures required to produce these behaviours, and in particular the neural motor programs which are required to support them.
That other animals have these elements separately also shows that a mechanism for the development or acquisition of the elements, in evolutionary terms, must exist. Whether or not the individual elements going to form the language capacity each separately had survival-value (the development of language might have been an example of evolution by accumulation of neutral mutations), language as such clearly had a major survival value, not so much for the individual as for the group which possessed language.
Two important examples of mosaic elements required for language are imitation and the categorical perception of speech sound.We take the power of imitation in ourselves, in birds such as the parrot or mynah, and in animals such as the chimpanzee, very much for granted but imitation, of speech or other sound or bodily movement, is in reality a most surprising ability. It involves a remarkable and complex linking of perception and motor organisation. Incidentally, the ability of some birds, mynah birds par excellence, to imitate human speech sounds precisely shows that the production of the sounds of speech is not dependent on any narrowly specified articulatory apparatus, and suggests that one should not expect to trace the development of speech simply in terms of gross anatomical features. The other important mosaic element is the capacity to discriminate categorically between human speech sounds in a way similar to that found in adult speech perception. This ability has been found in a variety of animals, notably in chinchillas, monkeys - and indeed in extremely young human infants.
In these two important elements of the mosaic, and in other behavioural requirements for language, the key observations are the intimate involvement in them of the motor control system and their dependence on cross-modal processes. The development of the language capacity has resulted from the progressive establishment of new cross-modal or trans-functional neural linkages, cerebral reorganisation in the sense that the interconnectedness of different brain regions concerned with what are usually considered distinct functions.has substantially increased.Evolution of language brought together in the human brain homologues or analogues of neural structures spread across a range of animals, and established neural connections between them.
The other related significant feature of the mosaic elements going to form the language capacity is the involvement in each of them of the neural motor control system. This extensive relation between language and the motor system is what one might reasonably expect, given the central role of the motor system in all behaviour and the essentially motor character of speech production, as the outcome of movements of the articulatory apparatus. For speech perception, the existence of a relation with the motor system has long been recognised, in the theory associated with the Haskins Laboratories and Alvin Liberman.
The prominence of the motor system in the mosaic elements which might have gone to form the language-capacity suggested that it would be profitable to undertake a systematic examination of the relation between each aspect of language and corresponding features of motor activity and the motor system. In addition, because of the intimate relation between the use and content of language on the one hand and perception on the other, the examination should extend to the relation between the motor system and perception in all its forms. The motor system is seen as the indispensable mediator between different modalities, and particularly between language and perception.
The Cracow paper suggested that new light could be thrown on this by using the hypothesis that the motor system, prior to the development of language, was built up from a limited number of primitive elements - units of motor action - which could be formed into more extended motor programs. This would make it possible to look for a direct correspondence between the primitive motor elements and the fundamental elements of spoken language, the phonemic system, and at the same time would allow one to derive the processes of word-formation and syntactic rules for constructing word-sequences from the neural rules governing the union of motor elements into simple and more complex actions. Motor activity and speech activity would thus be shown to have similar and in fact systematically related structures and rules. Language would be one type, though an exceptionally special and valuable type, of skilled action.
The paper for the Oxford meeting dealt with the system of motor control and the nature of motor programming. The effect of linking the system of motor control for bodily movement, to the neural control of the mouth and other anatomical elements which became part of the articulatory system, was that new channels were opened up for the external expression of motor programs The motor system already has externally expressive functions, most notably in facial expression. If language is derived from the motor system, one preliminary but important point is that language cannot be in any way arbitrary. This applies not only to the sound-elements of language, the phonemic system, but also to the words formed from these elements and to the ordering rules which constitute the syntactic structure of language. There is strong experimental evidence that the phonemic system is not arbitrary, suggestive evidence that word-forms are not arbitrary but are expressive or appropriate to their meaning and there is also considerable evidence for a fundamental relation between the syntax of language and physiological syntax, the syntaxes of action and perception. There is no evidence which compels one to accept that phonemes, words or syntax are arbitrary.
What becomes apparent is that the motor theory presented is not only a theory of language origin and development but also a theory of current language function. The proposition that language is completely analogous to skilled motor action opens up a new direction of enquiry, the applicability to language of the extensive and surprisingly successful research into the neural bases of action, of motor control. Recent research strongly supports the concept of motor programs and motor subprograms as real and not merely formal or theoretical bases for the organisation of action. An important task in brain theory is to isolate the substructures of motor behaviour, to identify what might be called the repertoire of detached motor programs and sub-programs, and how these are used by central organising programs.
The elementary motor programs may well be innate, part of standard human (and even vertebrate) neural structure. They may form part of fixed action programs or be formed by a central motor program into novel action-sequences. Motor programs are not necessarily dependent for their functioning upon incoming sensory information. They may run without any afferent information, as research on invertebrates has shown. The similarities in motor programming between a wide range of animals, birds, insects, suggest that common general principles have evolved in neural control of movement. In humans, in the light of evidence bearing directly on the relation between arm and head movements and speech, one may reasonably look for parallel pre-programming of the comparable speech musculature movements.
The relation between motor programming and speech programming can be examined at each level, the phonemic, the lexical and the syntactic. For phonemes, this leads to the idea of an invariant program for each phoneme, or 'auditory targeting', a motor-alphabet underlying speech, related in some way to the elementary motor-patterns underlying other forms of action.The very odd phenomenon (to use Philip Lieberman's phrase) of categorical speech perception has a direct bearing on this. A range of animals and very young infants have displayed, in repeated experiments, the ability to categorise speech-sounds, natural or synthesised, in ways which match the category boundaries in adult speech; very young infants have been shown to discriminate categorically speech-sounds not found in their mother language. On the motor theory presented in this paper, the explanation for this must be that the categorisation of speech-sounds is derived from organisation prior to language, and specifically from the categorisation of motor programs used in constructing and executing all forms of bodily action. What the rhesus monkey, or the chinchilla, share with the young human infant is very similar skeletal and muscular organisation, with very similar processes for the neural control of movement generally.The specificity of the phoneme is the accidental result of the application of the different elementary motor subprograms to the muscles which went to the form the articulatory system.The hierarchical structure of the motor system is built on the basis of a limited set of motor elements, which are combined in an unlimited number of ways (motor-words), just as phonemes can form an unlimited number of spoken words.
Words are a read-out of neural structures in much the same way as actions or facial expressions. Words, as a neural structure, can be formed from the co-activation of the motor subprograms for phonemes which are then melded or shingled together to form a distinct neural program for the whole word. Experimental approaches with the creation of artificial words have suggested that there can be a lawful relation between speech-sounds and auditory or visual percepts. Research into sound-symbolism strongly suggests that there is an isomorphism at the motor level between speech and the contents of perception. The object seen produces a motor-pattern which is readily transferable as a motor-program to the articulatory system and so becomes the associated word for the thing. A similar process is involved to that by which we transfer into our own neural organisation the motor-program underlying the facial expression of others, smiling, yawning or frowning, and so may reproduce in our own expression the expression which we perceive in another.
The neuromuscular sequences which are the immediate motor programs underlying words are derived from the integration of the neural structures underlying perception in all its forms (visual, auditory, tactile etc.) and motor organisation. The assumption that the last stage of the perceptual process and the first stage of the motor process are one and the same is attractive because it solves the problem of imitation. The subparts of the perception and action systems are thought of as pieces of a jigsaw puzzle that are made to fit each other.
As regards syntactic systems, if phonemes and word-forms are derived ultimately from the motor system (if necessary modulated by the perceptual system) it seems inevitable that there must be a close relation between the organisation of motor activity, motor syntax, and the organisation of language, speech syntax. Putting the bits together, beside speech-elements (phonemes), speech-element compounds (words) and speech sequences (syntax word-strings) on this theory one can now set a motor-alphabet (of elementary motor programs for bodily action), an array of motor-words (actions formed from motor-elements) and motor-sentences (formed from sequences of motor-words).
Because the motor theory of the origin and development of language is at the same time a theory of the current functioning of language, it is also potentially a theory of the ontogenetic as well as phylogenetic development of language. It can be a useful instrument for illuminating and investigating both speech and motor organisation, at the neural level and in extended speech and action sequences.
The first section of this paper summarises the contents of the two earlier papers which contain the main discussion of the essential features of the motor theory. The remainder of this paper provides supplementary material on different aspects of the theory and, in some instances, extends the scope of the theory.
The Cracow paper presented evidence for the existence of elements required for language found in the behavioural repertoire of animals other than man.Emphasis was placed on the abilities of birds in particular. In this I was following in the footsteps of many other writers. In the Descent of Man Darwin anticipated much of the material now thought to be relevant e.g. birdsong and its analogy to language, birdsong dialects, birdsong learning, parrots able to relate words and objects, birds' powers of imitation, and he gave a number of curious illustrations of birds' abilities:
Wonderful as is the possession of the voice by Man, we should remember that even birds can imitate the sounds surprisingly well... The sounds uttered by birds offer in several respects the nearest analogy to language..the actual song, and even the call-notes, are learnt from their parents or foster-parents. These sounds.. are no more innate than language is in man... Nestlings which have learnt the song of a distinct species, as with the canary birds educated in the Tyrol, teach and transmit their new song to their offspring. The slight natural differences of song in the same species inhabiting different districts may be appositely compared...to 'provincial dialects' and the songs of allied, though distinct species, may be compared with the languages of distinct races of man. I have given the foregoing details to show that an instinctive tendency to acquire an art is not a peculiarity confined to man.... Even the unmelodious sparrow has learnt to sing like a linnet....There can be no doubt that birds closely attend to each other's song...the case of a bullfinch which had been taught to pipe a German waltz.. when this bird was first introduced into a room where other birds were kept and he began to sing, all the others, consisting of about twenty linnets and canaries, ranged themselves on the nearest side of their cages and listened with the greatest interest to the new performer. (1871 Vol. 1: 55-56 Vol. 2: 52,55)
The ethologists have noted many similar patterns of behaviour. Eibl-Eibesfeldt in discussing the imprinting of motor patterns gave some thought-provoking examples of bird abilities:
Heinroth.. once recorded the song of blackcaps.. In the same room he kept 12-day-old nightingales.. which only uttered the begging call at that time. The total time these birds were exposed to the blackcaps was about one week. When they began to sing the following spring, they surprised Heinroth with the complete blackcap song, which was identical with the recorded song... In motor 'imprinting' apparently the relevant information is acquired at a time when the behavior that is learned, the song, does not even exist in rudimentary form... E. Gwinner.. had a raven that was called to the wire of its cage with the German word "Komm". Later he called his female in the same manner. B. Grzimek's.. raven called all children Gregor after the first child he had come to know. / Birds display remarkable cross-modal abilities/ One gray parrot learned to open as many food dishes out of seven and take out as many kernels as was indicated by the number of times a bell rang. The bird also learned to take food kernels in response to two or three light signals, and then transpose, without further training, to one acoustic signal (Eibl-Eibesfeldt 1975: 126, 298).
The remarkable abilities of some birds in imitating human speech,despite the radically different anatomical organisation in the bird of the apparatus for breathing and producing song and other sound, were referred to in the previous papers.
These often curious examples of the behavioural repertoire of birds (the number of them could be very much extended) are included to re-emphasise the point that many individual components of those which together go to make up human language-capacity can be found at the present day in the behavioural repertoire of animals other than man.
Language is essentially and unavoidably a group possession, a group behavioural attribute. As such, it serves to increase a group's competitiveness and to promote evolution by way of group-selection. Any group of animals with an effective means of communication within the group e.g. ants or bees as well as humans, are in a position to react to events external to the group with a group-reaction, and so also to be subject to evolution by way of group-selection rather than individual selection. Hence the evolution of diverse castes of ants, bees and termites. 'The local population occupies a really pivotal position for the genetic nature of the species' (Carson 1957: 32). In the case of humans, "it seems necessary to invoke a selective process acting between group and group, with the groups persisting as semi-permanent units, giving time for the better-integrated ones to prosper and supplant those that are less vigorous" applying Wynne Edwards' comment (1986: 1) to the usefulness of language to competing human groups. Eibl-Eibesfeldt refers in this connection to Bigelow's view that the origin of man's rapid evolution lies in inter-group warfare... "man's tendency to cluster into small groups (pseudospeciation) and to compete aggressively with others certainly provided a motive force for this evolutionary development. In a tragic way we are indebted to aggression for the rapid development of our intellect" (Eibl-Eibesfeldt 1975: 509)- and I would add that one of the most powerful forces in pseudospeciation and in aggressive intergroup struggle must have been the development of language.
One of the most striking experimental findings, which has been repeated on many occasions over the last 10 years is the existence of categorical perception of speech sounds in a variety of animals and also in extremely young infants. As Kuhl (1979) commented: "The findings are relevant to the origins and evolution of speech and language in general, to a theory of speech perception in particular, and to the notion of an innate predisposition for the perception of auditory signals that are part of an organism's communicative repertoire". As she has remarked in a very recent extended survey of this work (1987) "the phonetic structure of language is diverse but restricted in interesting ways across languages. Certain articulatory maneuvers never appear as phonetic units in the world's languages. Others are very popular. Moreover, the acoustic forms that have been used in the phonetic inventory of language form a restricted set..What was the nature of the pressure that guided the selection of the phonetic inventory?".."If language is viewed as a biologic process that evolved gradually, rather than one which apparently emerged in a discontinuous evolutionary jump, then the cognitive, physiologic, and perceptual substrates underlying language skills might exist as 'bits and pieces' of the whole in more 'primitive' mammals..There is increasing evidence that while language occurs in only one species, man, some of its substrates may not.." (Kuhl 1979: 375, 397; 1987: 287) This is remarkably in line with the thesis advanced in this and previous papers.
The early work on speech sound discrimination in chinchillas (chosen as subjects because their basic auditory abilities are similar to man's) has been replicated in the case of other animals and in other conditions: see for instance Howell and Rosen "Natural auditory sensitivities as determiners of phonemic contrasts" (1984) and Morse, Molfese and Laughlin who say that their findings with Rhesus monkeys suggest that "human and nonhuman primates have in common not only the behavioral but also the underlying neocortical control processes involved in the perception of voicing contrasts. Categorical perception of speech sounds may not be wholly dependent on language processes." (1987: 76-77).
This section describes more recent experiments on infants' abilities and the relationships that they show between speech-sound perception and production, vision and action. Except where otherwise noted, the findings are from papers by Kuhl. A recent account of infants' perception of speech (1986) reports new data on very young infants' abilities revealing an even more advanced response to speech. The data show that human infants' perceptual organisation of speech is quite sophisticated - infants categorize discriminably different auditory events as equivalent, they reflect the adult's rather complex definition of a 'good' stimulus, and they relate speech information presented by eye to that presented by ear. What infants can do is not only categorise speech sounds (including sounds phonemic in languages other than the mother tongue) according to adult categories but also generalise this categorisation over widely differing presentations of the speech-sounds and imitate, with some consistency, the production of speech sounds by adults. The experiments reported show that 18- to 20-week-old infants can detect the correspondence between auditorily and visually perceived speech; "in other words, they too manifest some of the components related to lip-reading phenomena in adults."(Kuhl 1986 (a): 241; 1986 (b): 258)
Experiments indicate a cross-modal representation of speech in infants: infants demonstrated categorisation abilities that go beyond those involving auditory perception; they demonstrated the cross-modal recognition of speech categories.. /infants preferred to look at face drawings which matched the speech-sounds heard/... four month-old infants perceive auditory-visual concomitants for speech. They appear to know that /a/ sounds go with faces displaying wide-open mouths and /i/ sounds with faces displaying retracted lips. In relation to experiments on the imitation of speech sounds by infants: "Our characterisation of the infant's initial state has to account for the infant's apparent ability, at four months of age, to match the sound produced by another with a vocal tract it cannot see, articulators (like the larynx responsible for the matched pitch contours) it cannot feel, and without inordinate numbers of attempts to do so. The infant appears capable of directing his articulators to achieve an auditory target. Any theory of infants' initial state will have to account for this."(Kuhl 1986 (a): 238-240)
What conclusions are to be drawn from these results? Kuhl and Meltzoff suggested that these abilities have a common origin - the infant's intermodal representation of speech: Future studies should test the extent to which auditory-visual and auditory-motor equivalence classes are related and the extent to which experience plays a role in their development.. Studies on infants' intermodal organisation of the auditory, visual and motor concomitants of speech may bring us closer to understanding the development of the human capacity to speak and comprehend language. (Kuhl and Meltzoff 1982: 1138-40) Most recently (1987), Kuhl has considered how one might account for the detection of these complex equivalences /visual/auditory/motor/: Categorisation, cross-modal perception, and imitation are competencies that have been demonstrated to exist in infants outside of the world of language and speech. They are not specific to speech, and thus, even complex behaviors such as these may not be due to a domain-specific 'speech-module.' It appears, then, that even if speech is inter-modally represented in infants, speech may not require 'special mechanisms' to be organised in that way. Rather, speech may draw upon a natural cognitive proclivity to represent information intermodally... What matters is that very young infants do indeed respond to speech as though the categories are already complexly mapped - representing them as a set of auditory, visual, and motor equivalents.(Kuhl 1987: 373)
The experiments suggest that in speech perception, there seems to be operating a process of perceptual constancy analogous to the constancy found in vision, that is the perception of an invariant characteristic, some relation that remains constant over many transformations; in the case of speech perception, the perception of an elementary sound, a phone, remains constant over transformations produced by particular syllable contexts and talkers. "The underlying assumption in vision is, of course, that there is a 'real' object out there; correspondingly, we must argue that there is some reality to the notion that a set of abstract prototypical acoustic cues can be specified for each phonetic entity, even though we have yet to do so." (Kuhl 1979: 384-5) It is interesting that in a quite different experimental context Gallistel refers to the parallel between action constancy - the same action performed using different means- and perceptual constancy - the same percept identified under different aspects.(Gallistel 1980: 390)
In the light of findings such as those described above, Liberman and Mattingly have recently constructed a revised motor theory of speech perception: The first claim of the motor theory, as revised, is that the objects of speech perception are the intended phonetic gestures of the speaker, represented in the brain as invariant motor commands that call for movements of the articulators through certain linguistically significant configurations...To perceive an utterance, then, is to perceive a specific pattern of intended gestures...The second claim of the theory is a corollary of the first: if speech perception and speech production share the same set of invariants, they must be intimately linked. This link, we argue, is not a learned association, a result of the fact that what people hear when they listen to speech is what they do when they speak. Rather, the link is innately specified, requiring only epigenetic development to bring it into play...Perhaps, then, sensitivity of infants to the acoustic consequences of linguistic gestures include all those gestures that could be phonetically significant in any language, acquisition of one's native language being a process of losing sensitivity to gestures it does not use.((Liberman and Mattingly 1985: 23-24)
The motor theory is predicated on the prior existence in the neural system of elementary motor units. the subprograms which go to form more complex action-sequences and which also serve to structure the development of language. There is extensive evidence that many aspects of motor behaviour, and particularly expressive motor behaviour, are found in new-born infants and the neural connections to support the behaviour must have been established before birth. A necessary consequence of the existence of pre-wired expressive motor programs is that the motor organisation controlling the production of the expressive movements must also be pre-wired, and in turn this requires that the elementary motor subprograms, motor units, from which the more complex movements are constructed, must also be pre-wired. The following material is largely drawn from Eibl-Eibesfeldt. Some of the best evidence that expressive behaviours are innate and not learned comes from observation of unfortunate children born deaf and blind. Expressive motor behaviour in such children is the same as in normal children. Smiling, laughing and crying, also the expressions of anger, pouting, fear, and sadness, looked the same in blind-born children, although they could not have imitated anyone...These unfortunate children grow up in eternal night and silence. They have no means for imitation and education is very difficult. In spite of the lack of formal training these children nevertheless show a number of well-coordinated motor patterns...a whole array of even quite complex behavior patterns, which are typical for human beings, have also developed in the deaf and blind and are therefore present as phylogenetic adaptations..The hypothesis that the expressive movements of these children are learned thus lacks all foundation.(Eibl-Eibesfeldt 1975: 474-5)
These observations accord with the conclusions drawn from cross-cultural studies eg. of expressive movement and from study of the innate abilities of new-born children. Members of the most varied cultures greet by raising the open hand. "Although the work is still incomplete, we have filmed enough to say that" some of the more complex human expressions can be traced back to the superposition of a few fixed action patterns which do not seem to be culturally determined...The association between facial muscular movements and discrete primary emotions is evidently the same cross-culturally. Observation of newborn infants has confirmed that this must be so; the facial expressions following the perception of sweet, sour, and bitter tastes are present in the newborn. Similar gustofacial responses were observed in anencephalic neonates and the movement pattern therefore can be considered as being inborn. (Eibl-Eibesfeldt 1975: 485, 449)
Developmental 'blueprints' are passed on from generation to generation in the genome. They determine the prospective potentials, not all of which may be actually realised. What develops can be influenced by the environment up to a certain degree but not in all directions with equal magnitude. Instead ranges of modifiability are inherited. In many cases. learning is genetically programmed so that the animals are able to learn in specific sensitive periods and possess specific learning capacities only at this time.(Eibl-Eibesfeldt 1975: 210)
The inborn motor programs can be contrasted with motor patterns formed by learning within a restricted period; there is selection between an array of possible patterns and the building up of more complex patterns using a hierarchical principle, where the elementary inborn patterns are like the elements of chemistry. In higher mammals. we can generally observe that fixed action patterns consist of very short movement sequences, which are combined by learning into acquired coordinations. The rat possesses all nestbuilding behavior patterns innately but learns the appropriate sequence of the individual components.(Eibl-Eibesfeldt 1975: 247-8)
The motor theory suggests that the development of language was the outcome of a gradual process of mosaic evolution, the coming together in the human neural system of homologues or analogues of behavioural elements required for language, which are known to exist individually in other animals, together with the neural structures required to support them - but how exactly should we conceive of this process of mosaic evolution taking place? How in terms of genetic and epigenetic processes did the changes in the connectivity of the human brain occur?
One approach is to consider what is known or hypothesised about the relation between ontogenetic development and evolution. In one of his early notes, Darwin said: "I see no reason why structure of brain should not be born, with tendency to make animal perform some action as well as gain it by habit."(Darwin 1974: 159) The question is in every case what is the balance between behaviour that is neurally programmed at birth and behaviour that is acquired by neural maturation after birth, under the influence of experience of the environment? Maynard Smith comments that "it is difficult to discuss the relationship between evolution and development because, although we have a clear and highly articulated theory of evolution, we have no comparable theory of development".(Maynard Smith 1963: 33) Nevertheless, in general terms evolution acts by altering development. Development is controlled epigenetically. Embryology and evolution are therefore fundamentally interdependent.
In the last 10-15 years neuroembryology has grown to be a vigorous major subfield of neuroscience and though it may be true, as Maynard Smith suggests, that there is no comprehensive and satisfying overall theory of development, a great deal of work has been done which opens the way for thinking about how evolution, the progressive change of the genome, and ontogenetic development may be related. The essence of the problem in ontogenetic development is how the neural pattern of connections is determined: how neurons find their targets and how the patterns of synaptic connections are laid down in embryonic development or mature after birth. "Work on synapses dominates present-day neurobiology.. the consensus that connections between specific groups of nerve cells are the basis of the immense computational powers of nervous systems.. changes in synapses or synaptic connections are almost certainly the basis for storing information and models of behavior".((Purves and Lichtman 1985: 205)
At first sight, one might look for the answer in Sperry's paper: "How a Developing Brain gets itself properly wired for adaptive function" (1971). Sperry said that for many years the dominant belief among neuro-physiologists had been that there existed a practically unlimited functional plasticity of the vertebrate brain:
The new picture that we emerged with implied a functionally specified system of wired-in behavioral nerve circuits, relatively implastic to re-arrangement by function. The whole question of developmental organisation and the question of how a brain gets itself wired for adaptive function was, of course, markedly changed. We were now confronted with the question of how the proper nerve connections get established correctly.... Nerve growth and connections seemed to proceed with the utmost precision and selectivity..It is now assumed that the precisely and prefunctionally ordered growth of brain circuits takes place on the bases of chemical codes under genetic control..The growing fibers are selectively matched by inherent chemical affinities.. In general outline at least, one could now see how it would be entirely possible for behavioral nerve circuits of extreme intricacy and precision to be inherited and organised prefunctionally solely by the mechanisms of embryonic growth and differentiation.(Sperry 1971: 30-32)
However, since then the picture has changed significantly. The relation between the genome and neural circuitry has been seen to be more complex both on the side of the genome and on the side of neural structure at birth. On the side of the genome, there have been recent surprises: "split genes and transposable repetitive elements. These have been added to the earlier surprises that repeated DNA sequences are scattered throughout the genome. None of these phenomena were predicted.. An image of a fluid genome is developing..." opening the way to "change by orders of magnitude the potentialities for useful organised novelty" (Bonner 1982: 35, 31) and also opening the way for theory emphasising the possible role of neutral mutations in evolution. On the side of neuroembryology, there is what some see as "The inescapable gap between genes and evolution... Genetics has been disappointing in its contribution to the understanding of embryology" ((Horder 1983: 345, 338)
The relation between genome and phenotype is far from simple and not necessarily determinate. Genetic control of development involves a huge array of genes; embryogenesis involves at least 10 to the 4 diverse mRNA sequences with thousands of structural genes being utilised (Bonner 1982: 31); but even this number of genes is small compared with the complexities of neural organisation that have to be produced in development. There is a general problem of economical use of the relatively limited number of genes available for coding synaptic connectivities. Recent discussion of neuroembryology has laid emphasis on the interactive nature of development:
Development is not strictly governed by an unalterable plan locked away in the genome...unravelling the relative contribution of genetic and epigenetic factors to particular developmental events is inevitably frustrated by the fact that genes influence the developmental environment - and that the developmental environment influences the genes... In higher animals.. neurons and their synaptic connections behave more like organisms in a biological system than elements in an electrical circuit. What happens to an individual nerve cell cannot be foretold with any exactness because its role in life depends on competitive interactions with peers striving to succeed in the same niche..Exploring this aspect of nerve cell behavior -neuroecology- may provide the best hope of understanding the remarkable properties of the human nervous system. (Purves and Lichtman 1985: 357, 363)
A dramatic aspect of the complex genetic and epigenetic interaction is constituted by selective cell death. In many parts of the nervous system. more neurons are initially generated than are found in the nature animal. Only those neurons that form appropriate synaptic matches survive into adulthood. The initial excess of neurons provides a reservoir that can buffer against some of the postgenomic variations not explicitly anticipated in the genome and ensures that the mature nervous system is composed of complete and appropriately matched neural populations. "The specific number of surviving neurons is neither explicitly nor rigidly programmed into the genome (e.g. to match dramatic postgenomic changes, amputations and transplantations)../These processes/ account for much of the well-documented plasticity of the nervous system."(Katz 1982: 208, 210)
But where in the context of discussion of the evolution of language do these formulations and characterisations of the process of neural development lead us? The theory has been that language was the result of the gradual accumulation of a range of behavioural aspects which together ultimately went to form the language capacity, and that this involved an increase in brain connectivity, establishing links between brain sectors and neural systems that had appeared separately in other animals and birds. There is current evidence for combination of the preformation of neural systems before birth with malleability after birth, within a limited period and in specified areas of the neural systems, both in humans and in animals. From neuroembryology there now comes evidence of scope for plasticity and malleability in the formation of neural systems before birth, with the possibility that extremely extensive and significant changes in neural connectivity could result from small changes in the genome or from changes in the epigenetic factors affecting neural development.
The evolutionary process which led to language could indeed have proceeded by way of mutations producing minor phenotypic changes with some increase in individual or group fitness or by way of neutral mutations along the lines proposed by Kimura and others. (The neutral theory claims that the great majority of evolutionary mutant substitutions are not caused by positive Darwinian selection but by random fixation of evolutionarily neutral or nearly neutral mutants. Some of the mutations will then turn out to be useful under new environmental conditions. Thus the neutral theory has a bearing on the problem of progressive evolution. (Kimura 1983: 306, 316)
Applying this to the development of neural motor control programs, during infancy a great deal of motor learning takes place, but the basic programs for many patterns of movements do seem to be present in the primate central nervous systems at birth. (Taub et al. 1973: 960). Animals are fitted out with a repertoire of comparatively stereotyped motor patterns that mature during the ontogeny of the individual, the typical patterning not being a result of individual learning. In early development experience effects appear to be encoded in the selective preservation of a subset of preformed connections.(Greenough 1984: 229)
What then one may reasonably ask was the nature of the change which resulted in the pre-wired motor programs in the human establishing special connections with the articulatory apparatus? Of course, it is impossible to give a specific answer but a plausible one is that it was associated in some way with the transition to bipedalism, and took effect by the impact of this on the epigenetic development of the neural systems for motor control and respiration. In an interesting article "Ontogenetic mechanisms: The middle ground of evolution" Katz discusses 'supra-genomic' mechanisms:
It is apparent that extant genomes and extant genomic mechanisms play fundamental roles in channeling evolution. But beyond the genome there appears another level of ontogenetic components that guide evolution. These components are the supragenomic mechanisms that are distant from the genome and that produce detailed phenotypes neither explicitly nor rigidly encoded in the DNA.. Many of the normally occurring postgenomic variations are buffered by ontogenetic buffer mechanisms, most notably selective cell death... It is likely that the preexistence of a variety of ontogenetic buffer mechanisms has made it possible for dramatic neural changes (such as the sudden appearance of the corpus callosum in placental mammals) to develop from a very small number of genomic mutations. For example, even with a single mutation to increase the size of a limb, the excess of motor neurons already present in the spinal cord could immediately form larger functional neuromuscular populations. Such a limb mutation need not wait for other fortuitous concordant mutations in the nervous system.(Katz 1982: 210)
A similar process, depending on the excess of neurons initially found in ontogenetic development, could account for the cerebral reorganisation linking the motor system and the articulatory system. Richards has recently (1986) proposed an important change in the way we view the transition to bipedalism in his paper "Freed hands or Enslaved Feet?". He quotes Wood Jones, who was Professor of Anatomy at Melbourne before the war, who pointed out the exceptional character of the human foot: "Man's foot is all his own. It is unlike any other foot.. his only real distinction and... his only valid claim to human status." Though ironical, there is an important point here. In becoming bipedal, we lost our ability to exercise complex control of the foot and toes; it ceased to be equivalent to the hand, as it is in other primates. The musculature of the foot required less complex neural control: The net neurological effect of enslavement of the foot then would be to bring about reorganisation, rather than enlargement, at the cerebral level: the expansion of manual control areas into those hitherto used for foot control, a need, now two instead of four hands are available, to monitor their functions, and differentiate between them, more effectively, perhaps contributing to lateralisation, and so on. The 'freed hand' did not immediately generate any obviously discernible gross neurological change of a radical nature" (Richards 1986: 148) but given the account of the neuroecology of the neurones in development, and the surplus of neurons that would have become available, this could well have been the occasion for the enrichment of other systems of connection in the central nervous system, including the vital connections between the motor system, the articulatory system and the perceptual system, which serve as the basis for the human language capacity.
One of the significant finding in Kuhl's work on infant perception of speech was the ability of very young infants to perceive a categorical relation between speech sounds and facial expressions associated with particular speech sounds. Other experiments have demonstrated analogous effects. So Bower measured surprise reactions of infants (increase in pulse frequency) while testing them with various optical illusions. For example, he exposed his subjects to objects projected on a screen. The infant reached out to grasp the object, whereupon it failed and showed surprise by an increased pulse rate. The infant expects to be able to touch an object it sees. The fact that children as young as two weeks of age react this way to the experimental situation suggests that expectation of tactile consequences from visual impressions is inborn..these results were surprising and interesting. They show that at least one aspect of the eye and hand interaction is built into the nervous system... "We can only conclude that in man there is a primitive unity of the senses, with visual variables specifying tactile consequences, and that this primitive unity is built into the structures of the nervous system". (Bower 1971: 32-33) Experiments with infants have given support to the view that the processing of data in spatial perception is, to a significant degree, inborn.
If, on the theory advanced in this paper, speech organisation is founded on motor control organisation, and there is a relation mediated by the motor system between visual perception and the formation of words associated with visual percepts, then experiments like these suggesting a very early and possibly innate relation between infant visual perception and tactile/limb movement are directly relevant.
Another of the points at which neural control of vision and movement come together is research into eye movement,where the questions recently considered have been very similar: "An issue not discussed in the traditional eye movement literature...When an observer looks at an object, how does the observer's hand know where to move in order to touch the object?" (Grossberg and Kuperstein 1986: 4,291, vii) Research with infants suggests that the neural program linking vision, arm movement and grasping is already present at birth. In other animals, co-ordinated vision and action is clearly instinctive: in the very quick catching movements of the praying mantis, the movement is not under the control of the eyes, as any corrective order given after the initial release of the action would be too late to have any effect, because of the speed with which the movement is carried out. The mechanism that regulates the orientation of the striking legs must be informed about the position of the fly in relation to the head as well as about the position of the head in relation to the body" (Eibl-Eibesfeldt 1975: 426)
There has been very intensive study of eye movements. "Although ballistic eye movements seem to be a relatively simple type of motor behaviour, a large number of brain regions are used to control them.. The fact that such a simple type of behaviour requires such a massive control structure has made the discovery of quantitative theories of brain dynamics difficult." (Grossberg and Kuperstein 1986: 1) The question arises whether there are universal principles of sensory-motor control. Grossberg and Kuperstein suggest that the answer to the question of whether similar neural designs may be used to control other sensory-motor systems other than the saccadic system in a general sense "clearly seems to be 'yes'. Many of the circuits which we have suggested are naturally decomposed into functionally specialised macrostages... " though one could expect differences between" the control of motor organs that are not regularly perturbed by unexpected loads vs motor organs that are regularly perturbed by unexpected loads." (Grossberg and Kuperstein 1986: 241) ie. between the control of eye movements and arm movements (though one might suggest that control of tongue and articulatory movements might be closer to the case for control of eye movements since the loads are small or in any case highly predictable).
Jeannerod gives an account of the relation between vision and motor control: Vision is not only devoted to building up an internal representation of the external world, it also has a motor function. Visually directed action implies continuous transformation of incoming visual stimuli into motor commands... the motor program module is presumably fed with information from at least two other modules, herein called visual map and proprioceptive map respectively. Visual map represents mechanisms that encode target position in space with respect to the body.. in addition the visual map also receives information about visual hand position in space...in the proposed model.. visual and proprioceptive maps are interconnected...jointly exert a steering influence on the program module..this model remains tentative. (Jeannerod 1986: 41, 76)
A recent study (1987) supports the view that oculomotor and skeletal motor systems share one map of visual space; repeated eye movements in the dark systematically altered the subject's estimate of the position of a target which had to be manually located with a pointer: Because pointing could be changed by manipulating only saccades, we conclude that the two systems share a single map of space..Human visually guided behavior is aided by an internal representation of visuospatial coordinates...How might multiple eye movements change a spatial map? Spatial models of saccade generation suggest that oculomotor commands to the extraocular muscles are generated along with a copy of that signal (... corollary discharge..).. This corollary information may be used by the skeletal motor system as well.. We suggest that the internal coordinate system is recalibrated after every target acquisition. (Nemire and Bridgeman 1987: 393-394, 399)
The concept of corollary discharge was introduced by Sperry who suggested that any excitation pattern that normally results in a movement that will cause a displacement of the visual image on the retina may have a corollary discharge into the visual centres to compensate for the retinal displacement. This implies an anticipatory adjustment in the visual centres specific for each movement with regard to its direction and speed. A central adjustment factor of this kind would aid in maintaining stability of the visual field under normal conditions during the onset of sudden eye, head and body movement. (Sperry 1985: 1) With this important link between the motor and perceptual functions Teuber was led to postulate that a voluntary movement is always characterised by a twofold process: an efferent discharge to the effectors and a simultaneous central discharge to the appropriate sensory system which forewarns them, so to speak, of the impending change. (Teuber 1985:1) The corollary discharge is a very important concept for the cross-modal interaction of motor control and vision. Intended action transfers information about its consequences to the visual system, which then adjusts in advance to it. Equally importantly, intended eye-movements scanning an object by corollary discharge inform the visual system of the intended pattern of movements, so that the visual system in its interpretation can adjust to cancel out spurious effects of the saccades etc. If the structures governing language became integrated neurally with this complex system linking motor control and vision, then it would not be surprising if language acquired the capacity to form into articulatory patterns information about both visual perception and sequences of action.
There is a two-way traffic in research on speech production and motor control; findings in relation to articulation have suggested processes that might apply to motor control generally;conversely, research directed towards motor control generally can be suggestive when transferred to the particular case of control of speech production. As Abbs puts it, the generation of speech is perhaps the most characteristic human motor act. Historically, owing to the complexity of speech motor behavior, most analyses have provided only general information on the underlying neural control mechanisms: However, recent work offers some intriguing neurophysiological insights....These observations appear to offer some useful hypotheses for the neural control of other complex motor behaviors.... the freedom to accomplish the same intended motor objective in many different ways relieves the nervous system of the burden of having to prespecify all the complex details of the motor subgestures. Such details are apparently determined via sensorimotor processes operating dynamically and flexibly among the multiple subactions.(Abbs 1985: 189, 196)
It seems plausible that in nature the controlled variable of many motor behaviors is a functional combination of several interlaced actions.. It may be that the precise combination of motor schemata defines skilled movement and is achieved through a co-ordinated control program, the purpose of which is to control the timing of activation of a number of subsystems.. coordinated control programs 'orchestrate' motor schemata to perform movement.(Abbs 1982: 541-542)
Studies of Bizzi. Kelso and Schmidt suggest that locations of body members in space can be coded, relative to a base position of the body.. helps to solve the problem of context noted, for example, by MacNeilage with respect to tongue movements in speech. The production of a speech phoneme is influenced by context. Does this mean that the program for each phoneme must take into account the preceding and following phonemes? MacNeilage suggests it does not. Instead, each phoneme program can be specified as a target location for the tongue... in principle, programming a particular movement can be largely context-free.(Keele 1981: 1397)
A recent puzzling result: Contrary to expectations, speech was produced faster by ambidextrous subjects than by either strongly left- or strongly right-handed subjects. The superiority of ambidextrous subjects was found to extend to certain manual movements as well. Ponton comments: The superiority is a new result and was not predicted.. explanation can only be speculative /one possibility/ it has been suggested that a single system in the parietal region of the speech-dominant hemisphere may be responsible for programming certain movements of the oral and manual musculature in most people. In this case... might reflect better interhemispheric transfer of such motor programming..An alternative class of explanations... If motor programming is better developed /as ambidexterity suggests/..and if the left hemisphere-based programming system exerts control over both the oral musculature and the manual musculature for both hands, then articulation and manual skills for each hand should benefit. In this case, speech should be produced faster and manual movements should be executed more rapidly. (Ponton 1987: 305, 310)
Another interesting account of the relation between speech production and arm and body movements is given by McNeill: "Into a hierarchy originally established to control manual movements, for example, lower modes can be substituted which control speech articulatory movements. Thus, deeply embedded within the speech process can be manual actions and the schemas of representation which they support." (McNeill 1981: 204-205) More mundanely, much scientific study of film-strips, frame by frame, of people in conversation, in interaction, has shown remarkable synchronisation of body movements and expressions, of gesture and so on, with speech.
One of the assumptions in the papers presented at Cracow and Oxford was that motor programs are constructed from the melding or shingling together of elementary motor sub-programs, and that these elements are pre-wired, probably universal constituents of vertebrate neural organisation for movement. A few additional comments on the evidence for such sub-units and on the manner in which they might be combined to form motor programs.Evarts posed the question "How are individual units of movement built up into integrated behavioral chains? This question has long concerned the neuropsychologist... It is clearly a problem that will come to the forefront as soon as the neurophysiologist studies any but the simplest sorts of movements". (Evarts 1967: 247) "Planning of movements (as opposed to instruction of movements) is presumably a complex activity which depends on ready access to percepts and memories". (Gunilla et al.1985: 285) Pribram suggested that the ready answer to the question how movement becomes transformed into action is that the organisation of behavior, its serial ordering, is due not to the chaining of movements but to the differentiation, the decoding of an already formed spatial configuration. (Pribram 1971: 363)
Hoyle's way of envisaging the construction of a motor program was somewhat similar: "there is another alternative.. that at least more complex nervous systems can use to generate behavioral sequences. This is to utilise information stored as memory of the desired objective to guide the formation of a motor program. The form of a nest is stored as a visual image of what a nest looks like, not as a series of motor instructions on how to make one. Intensive study of these movements shows that each has an element of fixed motor-pattern generation as well as modification in regard to detail. The animal must therefore have a collection of motor programs, one for each behavior, which may be thought of as a series of tape-recorded instructions that I have long referred to as motor tapes to be called upon as occasion demands. Each is played back in a continuous sequence and each sequence can be repeated endlessly." (Hoyle 1983: 584) There seems a broad consensus that there is a hierarchical principle of organisation, holding for all complex fixed action patterns whether they are birdsongs or sucking movements; they can be divided into elements that in turn are fixed action patterns.It is suggested that the commands are universal i.e. the nervous system operates by using them during the performance of any movement. (Feldman 1980: 81) Evidence for the existence of the parts from which the programs are formed can be found in pathological conditions. For example, "when the cerebellum is damaged...This decomposes movements into sequential constituents that are made with errors of force,velocity and timing.../the damage/ degrades but does not abolish movements." (Brooks and Thach 1981: 877)
Examples were given in the Oxford paper of the detailed investigation of the neuronal basis of motor programs in a variety of animals, mostly invertebrate. The specificity of the results is remarkable; "As we already know, the walking movements of an insect's legs are caused by a system of six, self-sustained, mutually coupled oscillators." (Eibl-Eibesfeldt 1975: 35) "individually identifiable neurons have been found to have specific roles in the CPG /central pattern generator/ for leech swimming, with some being involved in initiation and others in modulation of the rhythm..Cell 204 is of special interest because it excites and is excited by the CPG for swimming..However, though activity of a single cell is sufficient to turn on the swimming rhythm, no one cell is necessary for swim initiation." (Kristan 1985: 45)
How far are these results applicable to motor programming in the human? As the study of animal behavior has revealed that it is pre-programmed in well-defined ways by phylogenetic adaptations the question has been asked whether human behavior might not be structured in similar ways. Recent investigations have shown this to be the case. Human beings have motor patterns that developed their adaptations during phylogeny. Some are already functional at birth, others mature during ontogeny, as can be demonstrated by the study of those born deaf and blind, as well as by cross-cultural investigations, (Eibl-Eibesfeldt 1975: 533) as already described in an earlier section of this paper.
The major speculation advanced in the previous two papers was that motor control might be organised in terms of a limited set of elementary motor subprograms (comparable to phonemes in speech). This section describes recent research that seems to substantiate the idea of a motor control alphabet of this kind.
The earlier paper prepared for the Oxford meeting described research on neural motor control in a variety of invertebrates (leeches, locusts, molluscs). This research was described because with their relatively simple nervous systems, the results in determining the neuronal patterning producing motor behaviour could be more specific and definitive. Nevertheless, there is a great deal of relevant research concerned with human movement and movement control in physically comparable organisms, eg. monkeys - and related work on robot movement.
One line of research has derived from work in constructing robots which, within limits, are intended to produce comparable movement patterns to those of humans: Considerable progress has been made in the last 15-20 years in understanding control of robot arm movement... Many of the problems associated with the planning and execution of human arm trajectories are illuminated by planning and control strategies which have been developed for robotic manipulators..The application of principles of robot arm control to biological arm movement control rests on the premise that at a certain level of abstraction, since the problems are common to both the artificial and the biological systems, the solutions will be too....A research strategy / in robotics/ has been to presume that the biological motor control system has found simplifications or shortcuts which generate near-general behavior. In this view, the movements that the motor control system can produce are good enough for the vast majority of circumstances. (Hollerbach 1985: 140-142)
Arm movements at stationary or moving targets are common in the motor repertoire of primates. Yet little is known how the brain uses spatiovisual information concerning the locations of objects for the generation of reaching movements and how it controls the different neural, muscular and skeletal, structures involved in the formation of arm trajectories. (Georgopoulos 1984: 147) Research that has been carried out has recognised the immense complexity involved in constrained arm and other bodily movement, but some simplifying assumptions, based on experimental results, have emerged. The first is that "brain motor commands are patterned in terms of movements rather than in terms of muscles.. since their final path is not 'common' but rather involves different connectivities in the motoneurone pools for movements in different directions." (Desmedt 1985: 139) The second is the "concept of hierarchical motor planning.../which/.. suggests also that the same general abstract and internal representation of movements is used each time a movement is generated with temporal and spatial parameters chosen for that specific movement..This would allow the CNS to use the same reference spatial coordinates for coding both visual information and motor actions". (Flash and Hogan 1985: 1697)
Ideas on hierarchical organisation have been drawn from robotic control: A hierarchical movement plan is developed at three levels of abstraction..The top level, where a task command, such as 'pick up the cup' is converted into a planned trajectory for the hand or the object held by the hand. At the joint level the object trajectory is converted to co-ordinated control of the multiple joints of the human or robotic arm. At the actuator level the joint movements are converted to appropriate motor or muscle activations.(Hollerback 1985: 140)
The motor system is viewed "as being divided between higher levels which plan ideal trajectories for the end-effector, and lower level processes which translate them into torques and forces. These theories suggest that at higher levels there exists a kinematic representation of movement which does not take into account the mechanical nature of the actual effectors ". (Flash and Hogan 1985: 1696) The third assumption, which is necessary to and implicit in the hierarchical planning described above is that movement is planned at the object level rather than at the joint or muscle level; "the observation that unconstrained, unperturbed arm movements are coordinated in terms of hand motion shows that motor control is .. organised as though a disembodied hand could be moved in /extra-corporal/space". In experimental research into arm movement, mathematically analysed, "because the common invariant features of these movements were only evident in the extracorporal coordinates of the hand, these results are a strong indication that planning takes place in terms of hand trajectories rather than joint rotations". (Flash and Hogan 1985: 1697, 1688)
There are of course many other complexities in the organisation of movement. "The acceptance of the object level hypothesis places severe requirements on the motor system, which must now translate hand or object trajectories into a complicated joint movement and thence into actuation of the joint musculature." (Hollerbach 1985: 141) "Coordination is not a problem for movement alone; in a multi-articular system even posture requires coordination and control..Controlling dynamic behavior is a far more demanding task than controlling motion... Posture is not merely the outcome of a motor act, it is one of the most important preparatory stages in the production of motor behavior". (Hogan et al. 1986: 32)"Even a simple and small movement of the body is a component of a complex pattern of muscular activity,which involves not only muscles directly producing the observed movement but also other muscles often remotely located from the moving part. This latter group is responsible for the postural component of the motor act...postural adjustment is initiated shortly before the movement.. anticipating the disequilibrium that the movement would have provoked." (Gahery and Massion 1985: 121)
Given the complexities, the question remains: "How is movement control organised? Which variable(s) are controlled?.. on what basis does the nervous system select one specific trajectory from the infinite number possible? In what coordinate frame is the trajectory planned?" (Flash and Hogan 1985: 1688) Any attempt to solve the problem runs up against the degrees of freedom problem, the difficulty of finding a solution when so many possibilities are open. The search has been on for ways in which biologically the problem might be simplified. One suggestion was that a movement might be planned simply in terms of the final position to be achieved - but experiments to test this showed that "the transition from the initial to the final position is implemented by a gradual change in the control signal establishing both a trajectory and a final equilibrium condition...A precise solution to the problem of mapping a desired trajectory into the torques applied at the joints (the calculation of the necessary torques is called the inverse dynamic problem) has been found for the domain of man-made computer-controlled, multi-linked manipulators. Clearly animals must be doing something equivalent to the inverse dynamic computation, but the rules for executing this transformation are unknown." (Bizzi et al. 1985: 43)
Georgopoulos, dealing with the biological rather than the robotic problem, in investigation of reaching, refers to the variety of forces involved (gravitational, inertial and Coriolis) and says that "inverse kinematics for unrestrained three-dimensional (3-D) arm movements are almost intractable" (with up to 10 muscles used in many arm movements). (Georgopoulos 1984: 147) A curious sidelight is thrown on the problem from an apparently rather remote area of research. The difficulties encountered in dealing with human arm movement do not come from the multi-articular, jointed arm, which seems in fact to reduce the severity of the problem; the neural problem for the human being is small compared with that for the 8-armed octopus, with completely flexible arms; octopuses are highly intelligent but are not very good at manipulative tasks. They have, for invertebrates, large brains, but no central control of arm-movements, which are controlled by nerve complexes in each of the arms. They face the near impossibility of computing with any degree of accuracy the whereabouts of their flexible ends. (Wells 1978: 242-243)
How then has the problem of the multiple degrees of freedom, the multiplicity of muscles involved in each movement, the complex relation between posture and movement, the need to integrate motor action with visual perception, the need to allow for dynamic factors, etc etc been solved, or at least simplified? One step was, as the reference to the octopus has shown, to have a very limited number of joints. A second step appears to have been to segment movements.There is research evidence which suggests that a central mechanism "plans trajectories as sequences of movement segments which are then overlapped in time...it has been suggested that curved movements are generated from separate strokes, each stroke characterised by various geometric parameters, such as length, total angular change etc. Different movements may then be composed from different strokes, with different geometric parameters. " (Abend et al. 1982; Morasso and Mussa-Ivaldi 1982) The model "implies a need for storage and retrieval of strokes from a dictionary in memory; this dictionary would have to be very large..it is possible that such a mechanism underlies the planning of long and complicated sequences of movement. " (Flash and Hogan 1985: 1694) "These changes suggest a clear segmentation of the movement into units of action which overlap but do not coincide with the figural units as defined by the discontinuities of the movement (cuspids, points of inflection)." (Viviani and Terzuolo 1982: 431)
In effect. control of movement, on these results, would be simplified by a system of elementary motor units, pre-existing motor modules, which could be fitted together to produce a wide range of movement patterns. What is involved is a hierarchical system of motor programming, drawing on already available sub-programs. "A program is a set of instructions or an algorithm translated into the language of the computing device. It organises a structure, as it were, within the central processing unit.. the instructions detail the operations to be performed upon receipt of specific inputs, whether of peripheral or central origin." (MacKay 1985: 102, 105) "Our results suggest that the system does not preplan a totally different hand equilibrium trajectory each time a movement is about to be generated. Instead, based on the same general motor program, and using specific information concerning successful previous outputs, the system can develop a particular equilibrium trajectory for each individual movement." (Flash 1986: 24). Hollerbach presents a similar view:
Both the effects of simplifying the dynamics computation and the limitations of feedback control in biological arms.. strongly suggest that there must exist substantially correct preprograms in order for humans to make accurate fast arm movements. Experimentally, the importance of preprogramming in the control of movement has been well established. The conclusion is that the motor system must do a very accurate job of solving the inverse dynamics if it is to execute successful arm movements.. /In robotics at high speeds of movement, feedback control fails and finally becomes unstable/ For biological arms, limitations in the feedback system impose even more severe demands on the control system e.g. substantial delays/ (Hollerbach 1985: 144-145)
Research has shown that a great variety of movement is possible without feedback: "The data on the extent to which somato-sensory feedback is required to learn motor programmes are limited. However, they suggest that a range of motor programmes is available within the nervous system of the infant without the need for learning. Taub concluded that 'our data do not conclusively demonstrate that the motor programmes for a great many movement patterns are hardwired into the primate central nervous system at birth, (but) they do increase the probability that this hypothesis is correct.' The concept emerges that a range of hardwired motor programmes is available within the immature nervous system." (Marsden et al. 1985: 219-222)
Clinical evidence strongly supports these views:
We have had the opportunity to study a man deafferented by a severe peripheral sensory neuropathy of the hand and forearm, which spared motor functions. In the laboratory environment, we were able to show that 'he could perform a remarkable repertoire of manual movements... We can conclude that it is possible to contract a muscle by a prespecified amount, to activate one set of muscles independently of others, and to execute a string of muscle contractions with the correct timing, all in the absence of any sensory feedback. This introduces the important concept of a motor programme, a set of muscle commands which are structured before a movement begins and which can be sent to the muscle with the correct timing so that the entire sequence is carried out in the absence of peripheral feedback. The observations on deafferentiation clearly indicate that the mature nervous system houses a large collection of motor programmes that are quite capable of providing an extensive repertoire of movements....That adults may not require feedback may not be altogether surprising if they have learned a range of motor programmes in their youth. But did they use such feedback to learn the motor programme in the first place? Taub and his colleagues addressed this problem by deafferenting infant and even foetal monkeys. 'Ambulation, climbing and reaching towards objects developed spontaneously in all infants and were as good at three months as in monkeys deafferented in adolescence.' This was true even if the infants were not allowed vision to learn. /The conclusions reached by Marsden et al. were that/ a wide range of relatively gross motor activities may be undertaken in the absence of somatosensory feedback, by the use of stored motor programmes; some of these motor programmes may be hardwired into the infant's nervous system; other motor programmes must be learned, and somatosensory feedback contributes to such motor learning.(Marsden et al. 1985: 219-222)
This seems to be most compelling evidence for the existence and role of motor programs in humans; and fits well with the evidence, equally compelling, for the existence in new born infants of elementary motor programs. One conclusion that might be drawn is that such elementary programs, and of course more elaborated motor programs constructed from them, overcome the difficulties involved in de novo computing the complete requirements for any particular movement or sequence of movements; all the complications eg. of postural adjustment could be precomputed, including the translation into joint and individual muscle values of the requirements of any particular trajectory associated with an elementary motor program.
Research into biological and robot movement control has certainly recently made important progress; the main themes in the research, the hierarchical principle in planning movement, the use of subprograms to form more complex action programs, the construction of movements by the melding together of movement segments, seem compatible with and indeed useful confirmation of the plausibility of the idea that motor control is built up from a limited set of elementary motor units.
Butterworth in Explanations for Language Universals (1984:1) says that one of the most exciting issues in modern linguistics is: 'What do languages have in common?' This paper suggests that what they have in common is the physiological and neurological uniformity of human beings. Humboldt. without being able to give any specific explanation suggested that 'Driven by an inner urge, man extorts the articulated sound, the basis and essence of all speech, from his anatomic apparatus; animals would be capable of the like were they endowed with the same spiritual urge.' For the purposes of this paper, one might replace 'spiritual urge' by 'neural organisation'. Philip Lieberman in his 1984 book advanced a thesis broadly similar to that in this paper that "human linguistic ability derives from a number of innate i.e. genetically transmitted, anatomical and neural mechanisms. The plurality of these mechanisms is consistent with the mosaic nature of evolution..The book.. develops the theory that human language is the result of a mosaic of biological mechanisms." (1986: 1521-22)
Roeder, one of the founders of neuroethology and an authority on the neural mechanisms of insect behaviour, described scientists interested in the brain and behaviour as miners. each constructing his own tunnel into a mountain...his miners as very independent, tunnelling away, studiously ignoring their neighbours, their tunnels intersecting only by accident.I offer a slightly less heroic simile, earthworms which butt blindly forward, avoiding small stones, trying to digest the material through which they travel, occasionally meeting other worms no doubt, a manifestation of what I call the 'vermiform intellect'. There is a lot of ground to cover, the worm travels slowly, its digestive capacity is very limited, but with all the worms active together, eventually the ground is made fertile - and things, other than worms, have better opportunities to grow and survive.