THE INHERITANCE OF LANGUAGE Richard Coates School of Cognitive and Computing Sciences University of Sussex Brighton BN1 9QH CSRP 342 Revised text of a lecture delivered at the Day-School on Inheritance organized by the Society of Genealogists, London, 6 November 1993. The word ___________inheritance is commonly used in two different senses. One has to do with the way in which material goods are passed from generation to generation ("heirlooms"), and this sense is often extended to cover the transmission of cultural activities and beliefs ("our heritage"). It is obviously appropriate to think in this way when talking of the reasons why our speech resembles that of our ancestors. Even if we do not speak exactly like them, they have been a significant force in moulding our own behaviour; or their example may be one that we have consciously abandoned in favour of a different one. In either case, they have bequeathed and we have disposed. The other sense of inheritance has to do with genetic inheritance, the involuntary acquisition of characteristics of our parents. It must be far less obvious that such inheritance has anything to do with languages; the fact that I speak English can hardly be due to its having been encoded in my genes. Many a white South African child (even if only temporarily) has had Zulu as its native or dominant language through being brought up by an African nursemaid - hardly evidence for a language gene. But we shall see that there are good reasons for believing that the transmission of languages to the following generation proceeds in both of these ways. We must first devote a short time to the question of whether all languages have a common origin, and thus whether ________language is a common inheritance of all human beings in the sense of being an heirloom. If so, the case for any genetic inheritance is weakened markedly. It is obvious that all physically and mentally normal humans possess at least one natural language, and so do very many who are abnormal in one respect or the other. Many mentally handicapped people's language-use is indistinguishable in type or complexity (except in the relatively trivial matter of vocabulary size) from that of people in the normal range of intelligence. People physically unable to speak may have a language despite their handicap - they may understand fully, and perhaps also write. Deaf people may use a system of signing of a type which is now commonly accepted by theoretical linguists as being in all relevant respects comparable with spoken natural languages (SNLs). Let's leave aside signed languages for a moment. Spoken languages could, in principle, represent a common heritage. If human beings emerged at one particular time and place, it is equally conceivable that the emergence of language belongs at some single identifiable point in the emergence of the species, before the irreversible dispersal whose result is their appropriation of the entire land surface of the world. There has been well-publicized research during the last ten years or so which has opened up the question of whether all SNLs have a single origin. It is beyond doubt that the languages we hear today belong within identifiable family groupings. Nobody can look for long at French, Italian and Spanish without believing them to be related; nor at Russian, Czech and Polish. These intuitive judgements can be soundly backed by scientific demonstrations of precisely ___how they are related (i.e. what systematic changes have made them different from each other), and the methods which have been evolved to validate these relationships can be pushed further to show that languages and language-families which do not obviously resemble each other are in fact related. For instance, the two groups of languages I just mentioned can be shown to be related to each other, and to the group including German, Dutch and English; to Greek; to the modern languages of northern India (e.g. Hindi, Gujerati, Bengali); and so on. These techniques, collectively known as ___the ___________comparative ______method, do not, however, allow the grouping of all the languages of the world into one large superfamily. Those mentioned so far are a sample of the Indo-European family, the relationships among whose members are better understood than those of probably any other family. Other families established by the comparative method include Afroasiatic (including for instance Aramaic, Hebrew and Arabic), Sino-Tibetan (Mandarin, Cantonese, Burmese), and Austronesian (Malay, Maori, Fijian). Some languages have no relationship with any other language on earth. Basque is a case in point, despite numerous wild claims that it has distant cousins - it has none that can be established by known methods. From time to time, claims are made about possible relations between families; between Indo-European and Afroasiatic, for instance, or between the Altaic languages of western and central Asia (themselves not a family accepted by all linguists without hesitation) and Japanese. Rarely are such suggestions taken up into the canon of scientific truth, and there is often a good deal of bloodletting in the academic and popular periodicals about them. Two cases in point recently have been the _________Nostratic __________hypothesis, which seeks to group Indo-European with the Afroasiatic, Dravidian (e.g. Tamil), Altaic (e.g. Mongolian and Turkish) and Uralic (e.g. Finnish) languages, and the __________Amerindian __________hypothesis, which claims the common ancestry of all the languages spoken in the Americas before Columbus except the group including Eskimo and the relatively small group known as Na-Dn. The reason for the outrage that such ideas tend to generate is that they rely on methods which would be regarded as insufficient to establish the well-understood groupings; for instance by using small amounts of data for comparison whose similarity might be due to chance rather than common ancestry, and by not establishing consistent correspondences between sounds across the entire vocabularies of the languages being compared. Many linguists might well accept, in the well-hidden depths of their being, that SNLs do have a common origin, but also declare rather more publicly that by known methods this can never be demonstrated. Believing in a common origin is an act of faith with rather less substance in it than belief in a mitochondrial Eve in Africa some four million years ago. Signed languages, which I mentioned a short while ago, provide an intriguing sidelight on this issue. They are not obviously manual versions of SNLs; they share some properties with SNLs, but have some of their own deriving from the medium in which they are used, for instance the displacement of a message in space, away from the "normal" signing- space in front of the signer, to encode tense (e.g. reference to past time prior to, or embracing, the time being referred to in the currently-signed main clause). The first signed languages were invented by deaf people in deaf communities, which have only existed since institutions for the deaf were first established in the nineteenth century. They are therefore a recent invention. Another clear invention is the special style (known as Damin) of the Australian language Lardil, used by male initiates (and probably now extinct). It had a pronunciation system based on conventional Lardil, but containing all manner of strange sounds in addition, not found in any other Australian language (and in some cases in no other language at all). These two very different instances suggest that language invention is possible under the right circumstances (and perhaps inevitable in cases where no alternative means of communication exists), but that any such invention rests on a linguistic foundation which is already given. It is obvious, but intriguing, that such a foundation cannot be the same in all instances of invention; Damin rests on Lardil, but signed languages were invented, as far as can be ascertained, by people with no experience of any SNL and relying entirely on their genetic endowment and their general communication resources. In that sense, signed languages to some extent resemble _______creoles (languages developed as full languages with native speakers but originating as _______pidgins or reduced languages used for limited types of inter-group contact). The Lardil case shows how it might be possible for a common ancestry of languages to become obscured; though no instance is known in historic times of a language being totally replaced, in all its functions, by an invented variety. A view once fashionable in linguistics was that languages could differ without limit, and had nothing in common except what they had inherited from common ancestors. At its most extreme, this viewpoint suggested that terms traditionally used in describing the properties of languages, such as ______phrase, ____verb, _____tense or ________sentence-_______subject, could not reliably be applied with precisely the same significance in the description of two different languages. Each language was a system in its own right. This view was founded on the idea that languages were built up from scratch in childhood solely on the basis of evidence gleaned from hearing mature users of the language speaking. This idea, at first sight looking like common sense, has been displaced since about 1965 by a viewpoint (the "Chomskyan" viewpoint) which has stressed the supposed naturalness, ease, and procedural similarity with which children acquire their first language - any language - and the relative poverty and ill-formedness of the data with which their ancestors present them. On this account, children must have a genetic endowment which helps them to master most of the central structures of their language within five short years or so. In other words, they come equipped with a brain already prepared to receive and process data about any language. Further, on this view, since the genetic endowment of all humans is broadly similar, it is unsurprising that languages - the products of the brain's grapplings with the speech which is heard or overheard - resemble each other in striking ways. The way is open, as it was not before 1965, to search for __________linguistic __________universals, i.e. characteristics common to all languages irrespective of their historical relationships with each other. This may look intuitively very unsatisfying at first - after all, the most obvious things about languages which are not one's own are that they are radically different, and that those differences present massive difficulties for the learner in the classroom. The differences, and not any similarities, are what is stressed in the process of second-language learning. Nonetheless it is clear that if such universals can be found, across and despite family groupings, they can't with confidence be attributed to a unique common ancestor language, but need to be ascribed to ________language __as _a p____________sychological __or __________biological _______faculty. On one influential view (Chomsky's own), any such universals can be interpreted as defining properties of ________language in this sense. That is, they are specific to language and part of the human mind, and presumably therefore ultimately biological rather than cultural. They are wired in as part of our mental machinery. We share such universals because we are human, and they are inherited genetically. Other ways of interpreting any such universals include the following: (a) they are not a purely __________linguistic faculty, but are given as aspects of general intelligence which can be applied to learning languages just as they can be applied in other complex learning tasks; they are aspects of ___non-________specific ____________intelligence (b) they are not psychological, not wired in, but are due to our common experiences as humans and could be constantly reinvented or renewed by different peoples and generations; they exist because of the universality of the of the _____human ________cultural _________condition (c) they derive from the fact that efficiently-organized systems of ___any sort share certain mathematical properties, and this is true of everything from computer databases to international commercial companies; they exist because languages are _________efficient _______systems (though this does not preclude the possibility that these properties are genetically inherited) All these views acknowledge the existence of common properties, but for most people it is precisely this premiss which needs discussion and exemplification. Let's look at some supposed instances of universals as a prelude to trying to established whether, and if so in what sense, they are likely to be inherited. (1) Common categories (i) All languages have ways of indicating the roles played by participants in a conversation, so-called _____first, ______second and _____third ______person (speaker (_I), addressee (___you), bystander (__he/___she)). This can plausibly be ascribed to the fact that all human beings engage in conversation. (ii) All languages appear to have categories broadly similar to _____nouns and _____verbs as we understand them in English. Common human experience alerts us to the existence of things (serving as a basis for the category ____noun) and actions, events and processes (a basis for ____verb). It does not follow from this that a concept expressed by one of these categories in one language will necessarily be so represented in another. Prepositions and articles, by contrast, are by no means universal - Finnish lacks the former and most of the Slavic languages the latter. (iii) At the level of pronunciation, all languages have vowels and consonants, that is sounds which differ in sonority and hence in perceptibility, and which are produced with a greater or lesser degree of opening of the organs in the vocal tract. This has to do with the optimal balance between the speaker's requirement to speak without undue effort and the hearer's requirement to hear (and hence understand) without undue effort. (2) Common functions expressed by particular structures All languages have structures dedicated to expressing commands. This presumably responds to inequalities of status, and to inequalities that arise during natural interaction (you have information that I don't >> ____Tell __me ..., I need your help >> ____Help __me ..., I have personal needs which exclude you >> __Go ____away, and so on). (3) Common integration of units into larger units (i) All languages have mechanisms for indicating close grammatical relationships between certain words in sentences; either by proximity, making integrated strings called _______phrases, or in languages like most of those of Australia or (literary) Latin which have a practically free order of words, by agreement, where the form of words indicates their relation to other words. This universal appears to be based on the need for efficient recovery of information. (ii) Speech consists of an alternating string of vowels and consonants forming syllables, the exact restrictions on which are specific to individual languages. This presumably has to do with maximally efficient use of resources again (cf. (1 (iii)). (4) Common types of relation between structures (i) All languages have sentences for expressing propositions (descriptions which are either true or false about the world); all languages have devices for questioning such propositions or elements of them. _____Marie-_________Christine ______taught ____John ______French. ___Did _____Marie-_________Christine _____teach ____John ______French? ___Who ______taught ____John ______French? ___Who(_m) ___did _____Marie-_________Christine _____teach ______French (__to)? ____What ___did _____Marie-_________Christine _____teach ____John? This is surely related to the unequal distribution of factual knowledge and to the presumption that at least some factual knowledge is in principle public knowledge rather than being reserved for initiates. (ii) All languages have words which are less specific in meaning than full phrases, but which can carry the full weight of meaning of phrases in a suitable context. ___The _____great _____liner ____sank __in ___the _____storm. <> __It ____sank __in ___the _____storm. ___She ______enjoys ________watching ________football ___and _I _____enjoy ________watching ________football. <> ___She ______enjoys ________watching ________football ___and __so __do _I. _A: __Do ___you _______promise __to ____give __me ___the _____earth? <> _B: _I __do. Again, this seems to be grounded in the need for efficient processing (smaller amount of effort required by the producer); but it compromises the efficient recovery of information by the addressee. On the other hand, there appears to be widespread acceptance that wholesale repetition of information (which ought to be efficient for the addressee) is stylistically bad, resulting in the overloading of the capacity of the processor. (5) Common constraints No languages have the possibility of questioning (or doing anything else to) just one member of a set of phrases conjoined into a single structure. (The double asterisk indicates an ungrammatical sentence.) __He _____likes _______Vaughan ________Williams ___and _____Elgar. **___Who ____does __he ____like _______Vaughan ________Williams ___and? **___Who ____does __he ____like ___and _____Elgar? __He _____hunts _________springbok __or ______impala. **__It'_s ______impala ____that __he _____hunts _________springbok __or. __He ________promised __to ____wash ___and __to _____clean ___his _____teeth. **__To _____clean ___his _____teeth ___was ____what __he ________promised __to ____wash ___and. **__To ____wash ___was ____what __he ________promised ___and __to _____clean ___his _____teeth. (6) Implicational relations Some of these relations are quite complicated and need to be expressed very carefully. (i) In all languages which have only prefixes (not suffixes), prepositions are found (not postpositions), and in all languages which have only suffixes, postpositions are found. (A ____________postposition is a "preposition" that comes after the phrase it governs, like ___ago in English or _____gegenu___ber in German. English is not a relevant language, of course, as it has both prefixes and suffixes.) (ii) In all languages in which the subject or object of a sentence agrees with the verb in gender, adjectives agree with nouns in gender and verbs agree with subjects/objects in number (singular/plural, etc.). (iii) Consider this data. The bullet (o) indicates the position where the relative pronoun _____which has to be understood as fitting into the clause beginning with the word that follows _____which (e.g. (b) _I ____tore ___the ______stamps ___off _____which (understood as `(the) parcels')). (_a) _______parcels _____which _____these _____weigh ____more ____than o (_b) _______parcels _____which _I ____tore ___the ______stamps ___off o (_c) _________companies _____which _I _______offered __________incentives __to o (_d) _________companies _____which _I ______admire o (_e) _______parcels _____which o _______arrived ____this _______morning In all languages in which an object of comparison (as in (a)) can be referred to by a relative pronoun, so can oblique objects (as in (b)), indirect objects (as in (c)), direct objects (as in (d)) and subjects (as in (e)) too. In fact, if a language can tolerate a relative clause of some structure in this list, then it can reliably be predicted that it can have all the structures lower down; but not vice versa. My English is as liberal as possible in this regard; other languages are more restrictive. Of these categories (1)-(6), the examples I have chosen do not seem in all cases to be purely linguistic universals, in the sense of being facts about language which have no justification external to language. Some are justified by considerations of information processing (which of course ranges wider than language) or of human interaction (of which language is a significant part but not the whole). All I wish to conclude is that a linguistic universal NEED not be evidence for a genetic predisposition to language as such. On the other hand, it is hard to see what (5) is if not a purely linguistic universal, since it has to do with a constraint on linguistic structures which, so far as I can tell, have no analogues in other aspects of human behaviour. Maybe, therefore, the ability to form conjoined structures is a wired-in human linguistic capacity, and the constraint discussed is a wired-in limitation on that capacity. The strange and beautiful facts in (6) appear to be empirical - they are false, they cease to be facts, if a counterexample is ever found. Again, since they appear to be about linguistic structures ___per __se, they are good candidates for being truly part of a linguistic genetic endowment. Caution with this claim may be warranted, however, because it is possible that the psychological dimension of __________processing __________difficulty might have a role to play, in some of the instances mentioned, that has not yet been properly explored. A fairly large number of such universals have been suggested. The ones I have mentioned in (6) are, so far as I know at the moment of writing, genuinely exceptionless, when expressed in a technically precise way, and may therefore be put forward as candidates for defining features of or constraints on language. However, the approach within which these universals have been suggested has also given many examples of implicational rules which are valid with only a high degree of statistical significance, and are therefore not defining features of language; to put it simply, these other implicational rules have counterexamples. This gives rise to the suspicion that even where absolute universals have been discovered they may simply be accidental facts, extreme cases of the kind of preference or bias which can often be pointed out across the languages of the world: preferences that currently have no exceptions but which in principle might have. In this light, their significance becomes less clear, but not the less intriguing. Another approach to linguistic universals has been to explore those dimensions which might be used to define any system of communication, and to isolate those properties which are specific to language, or more prominent in language than in other systems. It is uncontroversial that a natural language must display the property of _______duality, that is having arrangements of units at two levels. Languages use sound-units (or letters, in their written form) arranged so as to form words (which may have their own internal structure, but we'll overlook that here). The words, in their turn, are arranged into the patterns which form the syntax. The lower-level units (sound-units, letters) are meaningless in themselves, but form meaningful units (words). Words are the constant association of a form (sound-units, letters) with a meaning. The nature of that association is arbitrary to a very high degree, which is why `rain' can be ____rain in English, ____dozd' in Russian, _____pluie in French, ___ame in Japanese, _____matar in Egyptian Arabic, ____euri in Basque, and so on. The sound and meaning are intrinsically unrelated, which is a property not unique to languages among systems of signs, but which is developed in languages to a uniquely refined degree. These words can be arranged, as mentioned above, in the patterns offered by the syntax. But the great flexibility of language, and hence its applicability in any conceivable situation, comes from the relation between words and syntax. Put simply, each language has a finite number of words. (It is possible to invent words, but this is not something which is done with great freedom in ordinary usage.) However, the set of sentences that can be constructed out of them is infinite. There is no upper bound on the number of sentences in English or any other language. Formally, this is guaranteed by the existence of grammatical rules which allow indefinitely repeated application: ________Atherton _____threw __it __to _____Smith, ___who _____threw __it __to __De _______Freitas, ___who _____threw __it __to _______Stewart, ___who _____threw __it __to ____Hick, ___who _____threw __it __to _______Tufnell, ___who ..... __He'_s _a ______stupid, ________arrogant, _____cloth-_____eared, ____self-_______serving, ______jumped-__up, ____foul-_______mouthed, ___________insensitive, .... ________deadhead. __We ____went __to ___the _______station ___and _______checked ___the _________timetable ___and ___got _a _________newspaper ___and _______crossed ___the ______bridge ___and ____went ____down __on ___the ________platform .... ___You'__re ____very, ____very, ____very, ____very, ____very, ..... ____kind. Structures exist which allow the arrangement of words into larger units which have not been exploited by the speaker previously, and probably never been heard by him, in precisely that arrangement. Systems which have finite vocabularies and astronomically large potential for combining the words in them in ways for which the speaker has no direct model are called productive; the features of _______duality, _____________arbitrariness and ____________productivity are central characteristics of all natural languages, and hence of the faculty of language itself. It is their intersection which makes language what it is, rather than each feature separately; certainly arbitrariness and productivity are not unique to language. This approach to linguistic universals has the advantage of setting up a framework within which language can be compared with other signalling systems, and which therefore offers the chance of an evolutionary perspective when animal and human signalling behaviour are compared. We can ask, for instance, whether the set of calls of the vervet monkey or the chimp in the wild shows any of the characteristics which are found to such a degree of elaboration in SNLs; whether, therefore, we can be said to have an inheritance whose basis predates the emergence of our species. We can conclude that all SNLs do share certain abstract characteristics; but the extent to which these are unique to languages is in many cases controversial and some may be found to some degree in less developed signalling systems, both of other species and of our own. The extent to which languages share more concrete properties which are specifically linguistic, and the extent to which such properties are inherited in a genetic sense, are controversial matters. In a general sense, it would not be surprising if our brains came equipped to handle a symbolic signalling system; after all, our livers and hearts come pre-wired to do certain jobs. The key issue is whether what the brain is equipped to do is specifically linguistic, or whether it comes with intelligent strategies available to it which happen to be handy in the great enterprise of constructing a language from the bits the child hears around it, amongst a pile of other things. Let's now turn to inheritance in the sense of cultural or material inheritance. There is an obvious sense in which we inherit a specific language or dialect from those who bring us up; or, in many cases, a multiplicity of languages or dialects. The metaphor is not perfect; what we "inherit" is still left with those from whom we "inherit", and what we "inherit" may not be precisely what was "bequeathed". But the overtones of sentimental attachment to what we inherit are useful for anthropologists and linguists when they view how beliefs and behaviours are passed on to the rising generation. It's widely accepted in social and historical linguistics that closed communities are very conservative. People raised in such communities tend to preserve linguistic and other aspects of their cultural heritage virtually intact. It is not surprising that the last haunts of dying languages (such as Irish) are in remote places with relatively little outside contact, nor that when outside contacts increase, the situation of these languages becomes more perilous still, as speakers become more willing to use languages with wider currency and without the stigma of the backwoods attached to them. There are probably no truly hermetically sealed communities, impenetrable by any outside influence, but there have been some which allowed only a little leakage. Some which have been suggested include: (i) Icelanders in general (ii) the English aristocracy (iii) working-class male groups in Northern Ireland All these have been held either to speak extremely conservative languages, or to have linguistic repertoires relatively uninfluenced by other languages. If no community is truly totally isolated, then the members of any are exposed to the possibility of diversifying linguistically, and therefore to differing degrees to abandon their inherited varieties of language. As a result, they may become able to exploit the arbitrary relation between form and meaning that was discussed earlier by allowing competition between different forms (words or pronunciations) representing the same meaning, or by allowing "foreign" meanings for existing forms. The inheritance of the members of such communities is diverse. They may choose among alternatives. Such choices are never neutral, but political in the broadest sense; we choose among them in such a way that we signal our identity and allegiances. My own upbringing exposed me to northern and southern varieties of English English, and I am able to slip from one to the other according to the way I perceive the requirements of the situation. (Normally this is done completely subconsciously, so perhaps the use of the word `perceive' may be misleading.) Selection pressures in the encounters typical of my lifestyle push my behaviour towards southern forms on a majority of occasions, with the result that my children hear little of my northern repertoire and they have not actively taken up what they have heard into their own. They have inherited restrictedly from me. The mechanisms by means of which behaviours are taken up from other people are called _____________accommodation. A speaker adjusts his/her behaviour towards that of someone else for one or more of a variety of reasons including the expression of solidarity, the desire for approval or the expectation of advantage. (There is a downside to this: since it is normally done subconsciously, it is hard to guard against the possibility that addressees will think the speaker is taking the mickey. I have caused embarrassment on more than one occasion by doing this.) Accommodation may take a variety of forms, ranging from the wholesale adoption of another language to a subtle adjustment of a small number of sounds in the inherited language or dialect. If such a shift is supported by general social norms and values, and if therefore lots of people are doing the same, long-term language change may be observed. Simple examples of this include the gradual northward progress in the vernacular of the more generally favoured southern pronunciation of words like ___cut, at the expense of the northern, on the eastern side of England especially - the transitional zone currently wriggles across the Fens in a broad band from about Kettering to Ely to King's Lynn; and the spread of that notorious indicator of successful counter-culture, the working-class London _f pronunciation for the __th of words like _____think and _____three, which has hopped from large town to large town, often skipping the intervening countryside. This is essentially what has already happened in the levelling of the most parochial local dialect into more broadly-based regional forms of English, and in the marked degree of resemblance between educated varieties of English over the entire country. You don't have to print newspapers in different editions for readers with different regional dialects. You can no longer place a person with confidence to within two or three miles of their birthplace just by listening to their accent, as you once might have done. The most local of dialects are in the process of disappearing, and becoming "heritage"; it is no accident that there are often dialect sections in museums dedicated to rural life and folk customs, as in Cardiff and Reading. We can conclude that, as far as languages are concerned, there is a pool of words and pronunciations in constant flux; speakers select and develop their use of prestigious forms, though what counts as prestige may not always be obvious to educated observers (as in the case of the street credibility afforded by the use of London vernacular pronunciation far beyond London). The backwaters of the pool may be more stagnant, and less liable to change from outside. It's also true that the transmission of parental language(s), where the caregiver might appear to be reasonably in control of both the content of what is said and its form, may not proceed smoothly for the developing baby. Some features fall down the generation gap. (I've already mentioned some aspects of my own usage that my children have not picked up.) The most cursory observation shows that children approximate their parents' speech in stages. Their early efforts may result in the construction of linguistic systems which are very different from the target system, even if eventually a reasonable degree of convergence is likely to be attained. That convergence is then likely to be subverted by the pressures to accommodate to peergroup behaviour which all parents will know something about. But there may be ways in which the child never gets to the point of total convergence with the parental model. Consider those English verbs that have irregular past tense and past participle forms: DO: (past tense) RUN: SING: COME: LIE: did (past participle) ran sang came lay (has) done (has) run (has) sung (has) come (has) lain The general rule for English verbs is that their past tense and past participle forms are identical (LIKE: liked, (has) liked; STAND: stood, (has) stood; HAVE: had, (has) had). Children learning Standard English may successfully master this generalization without fully mastering the exceptions and embedding them into their linguistic system. They may partly learn the irregularities and perhaps use such forms as ___ran/___run interchangeably both for past tense and past participle. Through contact with other English dialects where the irregular past tense forms are absent, such as Cockney, the use of the participial form for both meanings may be reinforced, so that the child finishes up with unitary ____done, ___run, ____sung and so on. In such a case, structural pressure from the dominant pattern, psychological considerations having to do with ease of learning, and social considerations having to do with accommodation to a different dialect may conspire to eliminate the irregular past tense forms from a child's developing grammar. These kinds of pressure are especially strong against features that are already recessive and therefore poorly represented in the parents' speech that a child is exposed to (such as the Norfolk short /u/ sound in words like ____road, _____roast, that I never acquired from my mother). They are also especially strong where structural ambiguity gives rise to analytic uncertainty, as in the case of constructions like _I ______object __to ___him/___his ______coming. In such cases, is ___him or ___his preferred? In the feminine equivalent, _I ______object __to ___her ______coming, the word ___her presents a structural ambiguity about whether it is the possessive determiner or the object form of the pronoun. Preference for the second of these interpretations, supported by the clearly pronominal form in _I ______object __to ___her, might well lead to pressure for the pronominal form to be preferred in such constructions generally, and therefore load the dice in favour of _I ______object __to ___him ______coming, rather than _I ______object __to ___his ______coming. Ultimately, such pressures tend to favour the use of the pronoun in its object form with any gerund, as in ___Him ______coming ___was _a ___bit __of _a _____shock (as opposed to ___His ....). I have just been illustrating ways in which the form of a linguistic system itself may be capable, in the right conditions, of inducing changes in its inherited form. Other, rather less principled, instances might also be given, based on learners' private perceptions of how the data provided by the previous generation of speakers is structured. I know, for instance, that there is no /h/ sound at the beginning of _______honesty , ________honorary and ______honour. It was possible for me to make a mini-generalization across these words and declare that words having to do with chivalry don't pronounce the /h/. That, of course, put me in great difficulty with ______homage, of which I believed for many years that the _h was silent, and pronounced it accordingly. What, then, can we understand by "the inheritance of language"? What is transmitted to all of us by the previous generation includes some things that may possibly be built in as mental capacities or constraints, and come as part of a genetic endowment for using language; others, for solving complex puzzles, of which language is the most recondite of all. Some "universals" may, however, come to all of us from the fact that the human condition is common to us all, and that this condition may have linguistic repercussions involving renewing some things afresh in each generation. The specifics of particular languages are, of course, learnt afresh by each new speaker, but what is transmitted is not free from the possibility of distortion and interference from psychological or social sources or both. We can truly say that all individuals create the language which they speak anew; they inherit a certain model from their forebears, partly given by endowment and partly learned, which they may not implement in the old way in every detail, and the inheritance is subject to complex and subtle variation and change on the basis of the developing child's total linguistic experience and of the principles of organization that s/he brings to bear on what s/he hears all around. NOTE This paper was delivered as a lecture, and as such contains no references in the ordinary sense. Some of the issues raised are discussed in the following publications (the least dauntingly technical ones, where possible): The comparative method: Bynon, Theodora (1977) __________Historical ___________linguistics. Cambridge University Press, pp. 45-58. More technically: Raimo Anttila (1989) __________Historical ___and ___________comparative ___________linguistics. John Benjamins, pp. 229-63. The possible common origin of all languages (very controversial): Wright, Robert (1991) "Quest for the mother tongue." ________Atlantic _______Monthly (April), pp. 39-68. More generally on the origin of language: John Lyons (1977) _________Semantics, vol. I. Cambridge University Press, pp. 85-94. Recent controversy about the family groupings of languages: Greenberg, Joseph and Merritt Ruhlen (1992) "Linguistic origins of native Americans." __________Scientific ________American (November). This article, and Greenberg's book on which it is based, have aroused outrage among many linguists. Kaiser, M. and V. Shevoroshkin (1988) "Nostratic." ______Annual ______Review __of ____________Anthropology 17, pp. 309-29. The Nostratic hypothesis is sympathetically but critically reviewed by Alexis Manaster-Ramer (1993) "On Illich- Svitych's Nostratic theory." _______Studies __in ________Language 17, pp. 205-50. Signed languages: Deuchar, Margaret (1987) "Sign language research." In John Lyons, Richard Coates, Margaret Deuchar and Gerald Gazdar (eds.) ___New ________horizons __in __________linguistcs __II. Penguin, chapter 14. Language as a psychological faculty: Aitchison, Jean (1989) ___The __________articulate ______mammal (third edition). Unwin Hyman, chapters 3-5. Chomsky, Noam (1987) "Language: Chomsky's theory." In Richard Gregory (ed.) ___The ______Oxford _________companion __to ___the ____mind. Oxford University Press, pp. 419-21. Fromkin, Victoria and Richard Rodman (1988) __An ____________introduction __to ________language (fourth edition). Holt, Rinehart and Winston, chapter 10. Linguistic universals and their interpretation: Hawkins, John A. (1988) "Issues in the explanation of language universals." In John A. Hawkins (ed.) __________Explaining ________language __________universals. Blackwell, pp. 5-23. Properties of communication systems in general: Lyons, John (1987) "Introduction." In John Lyons, Richard Coates, Margaret Deuchar and Gerald Gazdar (eds.) ___New ________horizons __in ___________linguistics __II. Penguin, chapter 1, esp. pp. 12-16. More technically: John Lyons (1977) _________Semantics, vol. I. Cambridge University Press, pp. 70-85. Accommodation: Trudgill, Peter (1986) ________Dialects __in _______contact. Blackwell, chapter 1. More generally on language change in its social contexts: Richard Coates (1987) "Historical linguistics." In John Lyons, Richard Coates, Margaret Deuchar and Gerald Gazdar (eds.) ___New ________horizons __in ___________linguistics __II. Penguin, chapter 8. Children's approximations to adult language: Aitchison, Jean (1989) ___The __________articulate ______mammal (third edition). Unwin Hyman, chapters 6 and 7. Fromkin, Victoria and Richard Rodman (1988) __An ____________introduction __to ________language (fourth edition). Holt, Rinehart and Winston, chapter 10.