VIenna English Working paperS

Language change as evolution: looking for linguistic `genes'

Nikolaus Ritt


I would be extremely grateful for comments and criticism. Please send me an eMail.

0.0. Prolegomena

This paper argues for the view that language change and acquisition represent evolutionary processes in a sense that we nowadays tend to associate with Darwinism. This view is not exactly new, but has never really established itself within the community of historical linguists. It goes back, at least, to August Schleicher, to whom we owe the notion of linguistic `family trees' and who viewed languages as similar to organisms with life-cycles in the sense that they are first `born', then grow into fully-developed, or maximally complex `adults', and may finally degenerate, lose their complexity and possibly `die'. His conception of `evolution' was essentially pre-Darwinian, though, and his application of the concepts to linguistic study highly metaphorical. Since Saussure and the formalist linguistic tradition he more or less established were opposed to the borrowing of metaphors from other sciences on grounds of principle, it is no big surprise that approaches like Schleicher's ceased to be taken seriously. Evolutionary concepts only sneaked back in disguise, so to speak, into linguistics via the functionalist approaches developed by Jakobson and the Prague School. There, language change was regarded as functional, `goal directed' and, one could say, `adaptive' in a sense similar to the one the term has in biology. However, the functionalists didn't explicitly base their explanations on a theoretical framework that could be called truly evolutionary in the Darwinian sense - probably because they were too cautious themselves to borrow metaphors which, they must have felt, belonged essentially to a different science and had no place in theirs. Retrospectively, I feel that by being so cautious, they made things unnecessarily difficult for themselves, as it may have been due to the very lack of an explicitly evolutionary framework, that their approaches to language change were vulnerable to attacks from structuralists (such as Lass, see for example 1980: 64ff.) for being teleological and unable to offer truly causal explanations. - Whatever the exact reasons for the development may have been, however, it is apt to say that nowadays, as April Macmahon puts it, „Evolution [...] has become a `dirty word' in modern linguistic theory" (1994: 314).

It is little surprising, therefore, that more recent attempts at dealing with linguistic phenomena from an explicitly evolutionary point-of-view have originated at the margins, or even outside the linguistic community itself. Here is a survey of books and articles that I have found particularly inspiring.

First, I would like to mention Cavalli-Sforza and Feldman's, and Lumsden and Wilson's volumes on cultural evolution. Both include linguistic evolution without giving the latter much space in their arguments or presenting any specifically linguistic cases, though. Similarly, Richard Dawkins' (1982 and 1989) proposal that cultural evolution might be based on the existence of `mental replicators', which he calls `memes', seems to be straightforwardly applicable to language evolution, while Dawkins himself does not develop his arguments in that direction. Interesting comments on Dawkins' proposal as well as on the frameworks developed by the other authors just mentioned can be found in Maynard-Smith (1989).

More recently, a group of scientists based at the Sta. Fe Institute have been developing an elaborate theory of `Complex Adaptive Systems', which presents itself as a metatheory for the study of a large range of phenomena that display `adaptive' behaviour and which are difficult to deal with on the basis of classical scientific approaches. These include national economies, immune systems, artificial intelligence, cognitive development and also, quite explicitly, language evolution. Due to the bad reputation which `evolution' seems to have within the linguistic community, however, the workshop on the evolution of human languages, which the Institute organised in 1991, has not lead to the establishment of a permanent research group.

While the failure of the Sta. Fe initiative shows how alien evolutionary concepts might still be to the community of mainstream linguists, contemporary cognitive psychology seems to be adopting evolutionism more enthusiastically. This becomes particularly obvious in Henry Plotkin's recent volume `Darwin Machines and the Nature of Knowledge', for example. Since the links between cognitive psychology and linguistics are getting continually stronger, however, it seems that before long the linguistic community will have to give up its established biases as well. In other words, I feel that there are quite good chances that evolutionary theory might, after a long period of neglect, find itself to become the majority paradigm in (historical) linguistics within the not-so distant future. Evolutionism is in the air, so to speak, and will soon manifest itself strongly. And I don't seem to be the only one who thinks so. Thus, I have recently kept running into or hearing of colleagues who openly admit to working on possible applications of evolutionary theory to linguistics, although publications are still missing. Among these colleagues are April Macmahon from Cambridge, Guy Cook from London, or Steve McGill from Exeter. Also, Robert de Beaugrande's forthcoming volume on Discourse Analysis adopts many notions from the theories developed at the Sta. Fe Institute.

Finally, I would like to refer to two papers by myself, namely forthc. (a) and (b), in which I have thrown much caution overboard and been as explicitly evolutionary as I dared.

But back to this paper, then. It introduces some of my personal ideas about evolutionary theory and its relation to language change, acquisition and use. It does so in an informal, associative manner without giving much heed to potential advantages of the advocated approach over other, more conventional ones. It is intended to stimulate the search for such advantages, which I feel can easily be found. The main reason why I am not presenting them in any detail here is that they deserve, in my mind, an elaborate discussion that would necessarily go beyond the scope of this contribution. So, if this paper succeeds in arousing some interest and the desire to play around with some of the concepts and perspectives that it develops, I shall be happy enough. Otherwise, I hope that the reader will at least find the mental gymnastics my attempts at reasoning will force her to perform refreshing and worth the time spent on reading all this.

0. Introduction

Since `tis Nature's law to change.
Constancy alone is strange.

John Wilmot. Earl of Rochester
A dialogue between Strephon and Daphne

This motto opens the first chapter of Jean Aitchison's classic volume on language change (1991). A weird choice, isn't it? After all, the book is about language change, so why are we told that it's constancy that we should really worry about? OK, I'm sorry. Of course, this is just a cheap move of mine, and, of course, Jean Aitchison was just trying to tell us that language change was less strange and unnatural than one might be tempted to think as a novice to the subject. - But still, if one takes the motto seriously, one cannot help but wonder. Why is it, quite generally, that historical linguists have always focused their attention on instances of linguistic change, and have not found it worth the trouble to deal with those elements in the world's languages that have maintained their identities and shapes over longer periods of time? Why is it, to mention a concrete case, that people in our field have mostly been trying to explain why Middle English long /i:/ shows up as /aI/ in Modern English, while the fact that Middle English short /I/ is still short /I/ in many Modern English words has been taken more or less for granted? At least in the light of the Earl of Rochester's insightful remark, this is almost as strange as constancy itself. In this paper, I'll do what strikes me as the obvious, therefore, and approach language history via constancy rather than change.

So, why is it then that Middle English short /I/ is still short /I/ in many Modern English words? Good question. Or is it? In what sense can we say at all that the /I/ in items such as ModE middle, it, children is the same as the - assumed - /I/ in ME middle, it, children? Although it is typically assumed that the two are in some sense identical, even a little bit of reflection tells one that things are not so self-evident at all.

On the one hand, we could say of course that the two /I/s are, in some sense, counterparts, because they fulfil similar communicative functions: middle, it and children mean pretty much the same in Modern and in Middle English, so there is good reason to assume that the purposes for and the situations in which Modern English speakers will use the words are at least roughly comparable to those which Middle English speakers had in mind when they employed them. If, by the same rationale, we then also say that both Middle and Modern English speakers used /I/ to distinguish words such as middle, it or children from others, it follows that the role which Modern English /I/ plays within Modern English speech communities is indeed similar to the one which Middle English /I/ played within Middle English speech communities.

In order for the question why Middle English /I/ is still /I/ in Modern English to be meaningful, though, the mere observation that the two /I/s are communicative counterparts is not enough. What makes the relationship between Middle English and Modern English /I/ special is that the two do not only play similar roles but that, in some sense, there also exists a kind of genetic relationship between them. In some way, Modern English /I/ seems to be a distant offspring of Middle English /I/, and we feel that Modern English /I/ has not only inherited many of the jobs of Middle English /I/ but is its direct descendant at the same time. It is both its communicative, or `functional', and its genetic counterpart. When we want to know why `Middle English /I/ is still /I/ in Modern English, we are therefore dealing with two questions rather than one. The first would be: how has Middle English /I/ managed to produce offspring that have survived over the centuries? And the second question would be: why and how have the offspring of /I/ managed to take over the communicative functions of their forebears? In the following I shall deal with each of these questions in turn.

Before I go on, however, I guess I owe you a little break to reflect on the course which my argument (or rather my loose associations) seems to have taken. Thus, you will have observed that at the beginning my paper seemed to be about a speech sound, a phoneme, something well defined, you may have remembered, as an element of `langue' in the Saussurean sense, and now I have come to talk about /I/'s offspring and about /I/'s jobs - just as if I were talking about a person, a human being, or at least a living thing of some sort, capable of reproducing and of doing things. - Obviously, you will think, my discourse has become metaphorical, and whatever the rhetoric or didactic advantages of metaphors might be, you will be aware that one must not let oneself be carried away by them and that, in particular, one must not tacitly endow them with a technical sense, because that way one might wind up in a completely fictional world and solve merely fictional problems that have no bearing on the world out there and thus no truly scientific value.

1. Phonemes (and other constituents of natural languages, for that matter) as active replicators

So, what about /I/ and its offspring then? Is there a technical sense in which such a statement can be read? In what way, if at all, can phonemes be assumed to increase and multiply? This question may sound weird to anybody whose mind has been framed to think in the categories established within the linguistic community - I agree -, but might it not still pay to take it seriously if only to see what happens? How can phonemes be assumed to `propagate' then? I guess the first type of answer that will probably spring to the mind of most of us is: `through language acquisition.' Children learn phonemes through listening to (more or less) grown-up speakers communicating. The general idea is that some part of the human brain works as a device for language acquisition, into which some of the more basic principles concerning the way human languages work are hard wired, so to speak, and which, when exposed to actual utterances will filter out and store in its more flexible components those pieces of information that a speaker needs to produce such linguistic utterances as are likely to serve her communicative needs in the community she grows up in. Those pieces of information will either be more like elements or more like rules/processes, but the distinction doesn't really matter here, because we are interested in /I/, and /I/ is most probably an element rather than a rule. In any case, when we think of /I/ as being learned by humans and stored within some part of their brains, the picture seems to emerge of /I/ as being transmitted passively, while the active agents in its replication seem to be the speakers. It would seem, therefore, that when trying to answer the question why Middle English /I/ is still /I/ in Modern English, we ought to focus our attention on speakers and rephrase our question as something like `Why have English speakers over the generations successfully acquired /I/?'

But is this necessarily so? In what way are speakers really more active than, say, phonemes in the replication of the latter? Obviously, speakers do not control language acquisition consciously and actively. This has been a home truth ever since Saussure. They are not normally in a position to decide whether they like to acquire a particular phoneme, for example, or not. In this sense `langue' is beyond the control of individual speakers. So, if we wanted to investigate the speakers' role in language acquisition/replication we wouldn't be talking about speakers as autonomous subjects freely determining their own actions, but we would be looking inside them and disregard their personal integrities, so to speak. What we'd be interested in would be the ways their auditive apparatuses and their articulatory organs work, and also, of course, the way in which these interact with their brains and the way in which linguistic elements and processes are mentally stored. (It is assumed here that language ultimately does have - even though it may not be reducible to - physical reality.) The speakers we would be looking at then, would look like a system of muscles, membranes, teeth, assemblies of nerve cells and other such elements - not much like individuals at all, really.

And where would our phoneme /I/ reside in this mass? Well, although there are as yet no ways of verifying this, there is in fact only one reasonably plausible possibility. It must be located within the central nervous system, i.e. the brain. Without knowing much about the way the brain handles information, most neurologists would subscribe to the notion that an element such as /I/ might be located within an assembly of nerve cells that are linked - however remotely or indirectly- to both articulatory and auditory organs. That assembly, which `represents', or, actually, `is' /I/ will be excited when it receives input in the form of sounds that are `recognised' as /I/ or when other parts of the nervous system get excited in such a way that a realisation of /I/ is pronounced. Another way of putting this would be to say that, depending on the actual state of other relevant parts of the brain, the excitement, or `firing', of a nerve cell assembly `/I/' will trigger either the firing of such other nerve cell assemblies as eventually amount to appropriate movement of the articulators, and/or the firing of such assemblies which `encode' or `are' word forms, morphemes, concepts or socially relevant information. A brain can thus be said to host an /I/ assembly, if there exists a set of nerve cells within it that get indeed excited more or less simultaneously when they receive electrochemical input under such conditions as specified above. The existence of /I/ is thus established when the channels through which electrochemical energy flows during a brain's activity come to be set up in such a way that an assembly of them fires in quasi-unison. The acquisition of /I/ can consequently be thought of as a process that results in the establishment of appropriate links among a set of relevant nerve cells.

During it's lifetime, then, an /I/ assembly may be in either of two states: when it fires, it is `on', when it doesn't, it's `off'. Due to the fact that the /I/ assembly is linked to articulators it may then happen that the firing of /I/ causes the latter to perform a gesture `expressing' /I/, typically the allophone [I]. During a lifetime, an /I/ assembly can thus be assumed to give rise to a relatively large number of [I]s, and the [I]s that can be observed in actual utterances can thus be thought of as consequences, or expressions of /I/ in a similar way as the phenotypic characteristics and some behaviour patterns of organisms can be regarded as consequences or `expressions' of genes. It is exposition to such [I]s in appropriate contexts, then, that allows children to acquire /I/s, or that - to stay within the descriptive framework I have begun to sketch - causes such links between certain nerve cells to establish themselves within children's central nervous systems that may be looked at - from the linguist's point-of-view - as (a representation of) /I/s. Through their inherent ability to produce [I]s under appropriate conditions, /I/s can thus place new copies of themselves within other nervous systems. - Looked at from this perspective, then, an /I/ - or indeed any phoneme - can be thought of as an entity capable of its own reproduction - a true active replicator and perfect mental counterpart of a gene, which expresses itself through creating organisms capable of spreading copies of the gene through reproduction. In the same way as the story of gene reproduction can be told without invoking organisms as central agents (they can be referred to as the `vehicles' or `interactors' of genes), the story of the life cycle of a phoneme such as /I/ can be told without referring to `speakers' as the primary agents in that process. For the life and reproduction of /I/s `speakers' constitute only the necessary environment and the necessary tools, they play no active role in it - and they cannot, normally, influence it.

The first part of our question can now be partly answered. Middle English /I/ has managed to pass its offspring down to Modern times, because it was turned on often enough in the right way to produce [I]s, which in turn placed new copies of /I/ in the brains of new generations of speakers.

Obviously, this is only a rough approximation to an answer, and raises a lot of questions in itself. In particular, it is obvious that a variety of further conditions must be met if the production of [I]s is really to engender new /I/s. Exposure to [I]s will only lead to the acquisition of /I/ if the [I]s form part of communicatively effective linguistic messages. I will turn to the question how those additional conditions will look within the framework I'm just sketching presently. I would like to do this, however, by taking up the second part of our initial question, namely why it is that Modern English /I/ still does many of the jobs that ME /I/ did? As it seems to me, this question is more closely related to the problem of the conditions under which /I/ will get acquired than might first be suspected.

What are those jobs I am talking about anyway? Let me give a few examples. Well, first of all, there is the primary function of /I/ as a phoneme, which is to distinguish the morphemes in which it occurs from one another. Secondly, there are the functions of making pronunciation and perception possible, which /I/ shares with all other elements and processes that `inhabit' human phonologies, of course. Finally, then, there are other potential functions which a phoneme such as /I/ may have, including morphological ones such as indicating a particular morphological environment, or social ones such as identifying the social adherence of a speaker, and probably many others as well. Now, how do these functions translate into the framework that I have been sketching so far? If phonemes are regarded as nerve cell assemblies, then - interestingly, but actually quite obviously - so can the so called functions of a phoneme. If recognition of /I/ is viewed as the firing of an assembly, it is equally plausible to assume that if /I/ and /t/ fire one after the other, this event will in turn excite an assembly {/It/} which will automatically excite assemblies for {3rd person} {neuter} and {singular}. Similarly, both articulation and perception crucially involve the excitement of nerve cells along certain `pathways'. Thus, the perceptibility as well as the pronouncability of /I/ result from /I/'s association with the respective neuronal pathways. In other words and quite generally speaking, then, /I/ fulfils its functions by virtue of its associations with other cell assemblies. In some cases, as in morpheme recognition or in articulation, the firing of /I/ will cause a firing of associated assemblies, while in others the firing of appropriate `function' assemblies will cause a firing of /I/. The fact that /I/ does certain jobs can thus be viewed, quite simply, as its occupying a certain designated place within a network of associated cell assemblies and standing in mutual triggering relationships with those. Viewed this way, however, the jobs which /I/ does are at the same time the clue to its very existence as an assembly, since only through being triggered to fire in unison a set of nerve cells emerges as an identifiable assembly in its own right. A number of connected cells that never come to fire in unison is by definition no assembly in the sense that we have established above.

It seems to follow, then, that whether or not an /I/ will emerge as a stably connected nerve cell assembly within a particular brain depends on whether it gets excited sufficiently often by the assemblies in its environment - and this, incidentally, represents an answer to the question concerning the conditions under which the production of [I]s can be expected to create new copies of /I/. As we observed, it cannot be enough for some sets of cells to be excited through input from the sensory pathway, because in order for /I/ to be acquired, it is not enough to hear [I] sufficiently often. Rather, [I] must be received and `interpreted' as part of a meaningful message. In terms of the model I am sketching here, then, this can only be done, if there exist the relevant other `linguistic' cell assemblies in the mental environment of /I/

Of course, if the acquisition of any linguistic element depends on the presence of other elements, the notorious chicken-egg question seems to raise itself. It can be solved relatively easily, though, on the assumption of repeated boot-strapping as well as of auto-catalytic self-organisation processes. One only needs to assume a very tiny set of linguistic `universals' to be hard-wired (or `inherited'), so that these then provide the environment in which further linguistic elements can establish themselves: an assumption that is perfectly compatible, if not equivalent, to the well established notion of a genetically provided `language acquisition device'. Whatever the detailed mechanisms behind the emergence of fully functional language systems, though, I feel that the picture that emerges from what has been said so far offers a rather cute re-interpretation of Saussure's view of language as a system `ou tout se tient'. At the same time, the dependence of /I/ on the presence of certain other assemblies represents another beautiful analogy to DNA based evolution, where the evolutionary stability of genes also depends on the presence of other genes within the genome. In the same way - to give just one example - as an /I/ assembly makes sense only within a more or less complete system of other `language assemblies', there can also not exist a gene for eye colour without the system of genes that manufacture the rest of the eye and the organism around the eye.

Another important aspect of the acquisition of such linguistic elements as the /I/ that this paper focuses on, is that they can be assumed to establish themselves within a new brain only, if their establishment and the activity it causes in associated assemblies gets `rewarded' and thus `reinforced'. In other words, the effects of an emerging assembly on its environment must be such that they feed back on the assembly by making its future firing more likely, so that the assembly acquires the necessary stability. `Reward' and `reinforcement' must ultimately come from neuronal activities whose effect is that they make people feel good, or, in other words, cause psychosomatic states which represent positive emotions and which are probably hardwired (i.e. genetically determined) into the nervous systems of human beings.) This is just a technical way of saying that the acquisition of appropriate linguistic cell assemblies is probably rewarded through the positive emotions that go along with successful communicative acts, or - in yet simpler terms - that a brain will acquire such elements as help its bearer to get himself across to and understand other members of the speech community. The acquisition of an /I/ phoneme can thus be viewed as just one series among the adaptive process that a brain goes through as it evolves towards a state in which it enables its owner to interact/communicate with her environment in a sufficiently harmonious manner. At first, an assembly which eventually turns out as an /I/ will be just one of a larger set of many possible neuronal configurations a brain may assume. As its activity, that is its firing, turns out to have rewarding effects more often than the firing of other neuronal configurations, it will probably `attract' more electrochemical energy and get fired more often than the latter. By virtue of that fact, then, it will establish itself as a stable assembly, while its competitors won't. The emergence of an /I/ assembly within an individual brain can thus be viewed as a process that involves chance variation among brain states and selection of some states over others. The relative fitness of the competing states is determined by the degree to which they effect positive `emotional' responses, while the general `tastes' of emotions have probably been shaped through selectional processes acting on the genetic level.

For /I/'s success as an active replicator, this means, that its capacity of producing [I]s is just one aspect determining the success of its replication. Producing [I]s will probably increase the number of times an assembly corresponding to /I/ gets excited within a new acquisitor's brain but it cannot guarantee that new /I/'s stability. However, the very existence of `adult' /I/s in the speech community also increases the chances for the effects of a potential new /I/ to be `rewarded' and thus greatly increases the chances of a new /I/'s stability. /I/ is thus an active replicator in two ways: it generates new /I/s through emitting [I]s, and it breeds newly generated /I/s by making its host (or: speaker) react to their effects in a way that gives the hosts of the new /I/s the feeling of being `understood' or having reached their communicative goals.

So much for my view as to why Middle English /I/ is still /I/ in Modern English. /I/s are still around in abundance in Modern English in environments similar to those in which Middle English /I/s throve, because the latter have managed to replicate themselves successfully through multiple generations of host brains.

2. Some reasons why I find it so attractive to view linguistic elements as replicators.

What are the advantages, if any, of viewing phonemes as replicators? After all, it would have been equally possible to say that /I/ and its functions have been successfully acquired by successive generations of English speakers and that's why it is still around? And this sounds more familiar as well, doesn't it? It's much more easily digestible, doesn't demand mental gymnastics and seems to capture, basically, the same facts. So why take the trouble?

Well, I can think of a number of good reasons. Here I will present just two.

First, a statement to the effect that speakers have successfully acquired /I/s over successive generations forces one to change perspectives, because phonemes and speakers belong to different ontological domains. Such a change is intellectually unsatisfactory, particularly in lack of a theory that makes the links between the two domains or levels explicit. Thus, saying that a speaker acquires a phoneme begs the question of how she actually does that, and in structural historical accounts this question is typically never taken up seriously or even answered, because speakers are treated as black boxes at best or even regarded as outside the domain of linguistic science proper. - The view sketched here, on the other hand, is perfectly explicit about the way in which language acquisition is supposed to work, while at the same time describing it exclusively from the point of view of the linguistic elements themselves. No recourse to notions such as that of `speakers' as agents in the story of linguistic evolution has to be taken. Speakers figure exclusively as the environments in which the replication of linguistic elements take place, and the story can remain a consistently `linguistic' one.

Second, viewing linguistic elements as replicators opens the possibility of transferring to the study of linguistics many theoretical insights gained in other sciences that have for a longer period and more explicitly concerned themselves with the study of replicating systems and their evolution. First and foremost these include evolutionary biology, of course, but also other sciences applying generalised Darwinian frameworks, such as evolutionary cognitive psychology, artificial intelligence and particularly the study of systems that are capable of `learning' on the basis of genetic algorithms, and so on.

In particular, the view that linguistic elements are replicators affords interesting aspects on questions relating to language change. Remember that I argued that the success of /I/s replication depends, among other things, on whether a new brain can integrate /I/ meaningfully within the network of linguistic elements and processes that are emerging within it. One aspect of an emerging linguistic system or competence, is therefore, that it provides slots that need to be filled, and it is such a slot that an element /I/ needs to fill if it is to replicate successfully. It seems to be a characteristic of such slots that there is not only one way of filling them. As the evidence of linguistic variation and change show, it is possible and happens quite frequently that within one and the same speech community corresponding slots happen to be occupied by different elements within different brains. Thus, as a very trivial example, take the variation between /U/ and /A/ in English, where either of the two may occupy the slot provided by the context of such words as hut, but, butter, and so on. It can be said, therefore, that within the pool of linguistic elements which occupy brains of speakers of English, there is a competition between /U/ and /A/ for a certain defined slot, or that - with regard to that slot /U/ and /A/ are alleles, to borrow a term from genetics (where it refers to different genes that can occupy the same slot on a DNA strand and are thus competitors). Typically, such a competition for slots - on which any linguistic element depends for its existence - will take place in all instances of linguistic variation, so that variation itself can in fact be regarded as competition. Whenever variation leads to actual change, the cause for such an event must be that for some reason a new competitor for a given slot manages to place more copies of it there than its established rival. In the case of phonological elements, ease of perception and production will obviously play a great role there, in the case of other elements semiotic parameters determining the ease with which links between associated elements are established will probably play similar roles. At the same time, however, the fact that existing elements will tend to reward the establishment of what they `recognise' as new copies of themselves will be a powerful barrier against the spread of new variants, and if a new variant is to establish itself successfully then it must have sufficient selective advantages over the established competitor. Such selective pressures which favour the spread of a new variant at the expense of an established one may among other things be due to changes in the environment of the slot, where environment includes both a slot's immediate systemic environment and the wider environment comprising the whole set-up of the speech community in which linguistic elements `live'.

Also, the view of linguistic evolution as a story of items attempting, with varying success, to replicate themselves affords interesting perspectives on the notorious problems concerning the actuation and the implementation of language change. Change is actuated under this view simply through undirected copying mistakes that occur in the replication process. It may have various reasons, some of them pretty straightforward, such as misarticulations and misunderstandings. More crucially, though, variation is already built into the copying process itself, since an item relies on an acquisitor's brain to produce `random' varieties of neuronal configurations among which the replicating item will reinforce those most similar to itself and thus make them stable. In other words, the processes by which linguistic elements get copied involves successive stages, during the first of which the average copying fidelity must necessarily be relatively low. The resulting variability of languages is thus a natural matter. - As far as the implementation of changes is concerned, then, it can be regarded as parallel to the implementation of mutational changes within the gene-pool of living species: it results from the relatively more successful replication of a new variant within the brains of a linguistic community.

Finally, this perspective makes one of the more puzzling aspects of language evolution easy to digest, namely the fact that - on the one hand - language changes often seem to be adaptive in the sense that they seem to make things easier for language users, while - on the other hand - the world's languages are full of elements and qualities that must count as clearly suboptimal with regard to those criteria that allow language change to be regarded as `optimisation' in the first place. As the approach I have sketched here suggests, the items of which languages are made up do not primarily exist `because of' the purposes they serve their speakers, but rather simply because they have managed to replicate sufficiently well. Language evolution takes place, in other words, primarily because the elements that constitute languages are replicators which strive to propagate. The needs of speakers only represent constraints on the propagation of linguistic replicators, determining which of them will survive more easily in an environment where resources (in this case electro-chemical energy supplied within human brains) are as limited as they are anywhere within our world.

References

Aitchison, Jean, 1991: Language change : progress or decay?. 2nd. edition. Cambridge: University Press.

Beaugrande, Robert de, forthc.: New foundations for a science of text and discourse. Norwood; NJ: Ablex.

Bichakjian, Bernard H., 1988: Evolution in language. Ann Arbor: Karoma.

Cavalli-Sforza, Luigi L. and M. W. Feldman, 1981: Cultural Transmission & Evolution: A Quantitative Approach. Princeton: University Press.

Dawkins, Richard, 1982: The extended phenotype. Oxford: University Press.

Dawkins, Richard, 1989: The selfish gene. 2nd edition. Oxford: University Press.

Hawkins, John A. and Murray Gell-Mann (eds.), 1992: The evolution of human languages. Redwood, CA: Addison-Wesley

Hebb, Donald O, 1949: The organisation of behavior. New York: John Wiley.

Hull, D.L., 1988: Interactors vs. vehicles. In: Plotkin, Henry C., ed.: The role of behaviour in evolution. Cambridge, MASS: MIT Press.

Lass, Roger, 1980: On explaining language change. Cambridge: University Press.

Lumsden, Charles J. and Edward O. Wilson, 1981: Genes, mind, and culture : the coevolutionary process. Cambridge, MASS: Harvard University. Press.

Macmahon, April, 1994: Understanding language change. Cambridge: University Press.

Maynard-Smith, John, 1989: Did Darwin get it right? : essays on games, sex and evolution. New York: Chapman and Hall.

Plotkin, Henry C., 1994: Darwin machines and the nature of knowledge. Cambridge, MASS: Harvard University Press.

Ritt, Nikolaus forthc. (a): The spread of Scandinavian 3rd psn.pl. pronouns in English: optimization, adaptation and evolutionary stability. In: Kastovsky, Dieter, ed. Papers of the 1993 conference on Languages and Contact in the History of English, Tulln, Austria.

Ritt, Nikolaus forthc. (b): Mutation, variation and selection in phonological evolution: a sketch based on the case of Late Middle English a > au/ _l{C/#}. In: Fisiak, Jacek, ed. Papers of the 1994 conference on Middle English, Rydzyna, Poland.