Uriel Weinreich's pioneering monograph on Languages in Contact is now more than fifty years old.1 It is so justly famous that I have heard a speaker at a conference upbraided, after a talk that merely happened to have words like ‘languages’ and ‘contact’ in its title, for not referring to it. But like any other book, it is of its day, and it merits critical re-reading in the light both of the problems that have arisen since the 1950s and of what was known, or was believed, then. It is not enough to point out that each individual is a battle-field for conflicting linguistic types and habits […]. What we heedlessly and somewhat rashly call ‘a language’ is the aggregate of millions of such microcosms many of which evince such aberrant linguistic comportment that the question arises whether they should not be grouped into other ‘languages’. It is not unusual, of course, for brilliant doctoral students to have preoccupations that are not quite — sometimes not at all — those of their supervisors. Nevertheless, it is instructive to read Weinreich's work in the spirit in which Martinet introduced it. The ‘useful assumption’, as Martinet put it, is one that Bloomfield had presented as such in the 1920s. Crucially, what is assumed is that there are communities of people within which ‘successive utterances are alike or partly alike’. Such a community is by definition a ‘speech community’, and its ‘language’ is defined in turn as the totality of the utterances that can be made within it.3 For Saussure, too, a language (‘langue’) had its existence in the community whose language it would be. But it was conceived directly as a system in abstraction from speech, and the processes of speaking, on specific occasions (‘parole’). If it changed, it changed also as a system, from one ‘état de langue’4 to another. These, too, were assumptions; and Firth, in England, was one older contemporary of Martinet who rejected Saussure's reification of a language, as he saw it.5 But for many structural linguists such assumptions were scarcely controversial, and in New York, when Weinreich was a student, Martinet and Jakobson were leading structuralists of the Saussurean school. If a language, then, is an abstraction of this order, contact between languages does not simply consist in interaction among individuals and change in individuals' speech. It involves relations between systems; and, if it leads to change in either language, influence of one system on another. How can such influence be exerted? A natural answer is to posit two distinct and logically successive stages. In the first, the influence is simply at the level of ‘parole’. Thus, for example, a speaker of language A, who as such is a member of one speech community, may, in speaking it on a specific occasion, use a word belonging to another language B, which is spoken by the members of another speech community. But this is merely, in a strict sense, ‘borrowing’. The word does not belong to the system of A. Even if the same speaker uses the word again on another occasion, or other speakers of A also use it, these are different, independent instances where the behaviour of an individual is influenced by that person's knowledge, as an individual, of the language B of which it is a unit. At this first stage, therefore, only speech in A will be affected. A itself — the system of A — is not affected. Such speakers may be heard, however, by other speakers of A, who themselves have no knowledge of B. Therefore, they may use this word too; and, in the second stage, it may come to be ‘borrowed’, in the more familiar, strictly catachrestic sense of this term. Thus, A itself — the ‘langue’ in the Saussurean sense — changes from an earlier system or ‘state’ of its system, of which the ‘borrowed’ word was not an element, to a new ‘state’ or new system, of which it will be. Forty years on, other theorists were to talk of changes in ‘externalized language’ (that is, in the way some group of people speak) that ‘trigger’ changes in ‘internalized languages’ developed in the minds of a new generation.6 Classic illustrations are of change in syntax, and not due specifically to contact. It is instructive, however, to compare the way in which this two-stage model applies to convergence in phonology. In the first stage, we are again concerned with changes in speech, at an overlap (as we might represent it) of speech communities. At the point where they overlap, there will be individuals whose native language — what Weinreich calls their ‘primary language’ (p. 14) — is A. Nevertheless, such individuals may also speak B; and, when they do so, we are likely to find evidence of interference from the system of their primary language. ‘Phonic interference’, in particular, ‘arises when a bilingual’ (that is, any individual who to any degree has knowledge of two languages or more) ‘identifies a phoneme of the secondary system’ (that of B, in this case) ‘with one in the primary system and, in reproducing it, subjects it to the phonetic rules of the primary language.’ Weinreich's treatment of this topic (Chapter 2.2) was immensely influential, not least on the study of second language learning over the next twenty years. Its attraction, however, was precisely that it seemed to explain the behaviour of individuals, and the errors, as seen from the viewpoint of B, that they make as individuals. We would, of course, expect these errors to be similar from one individual to the next. Thus, for example, any speaker of A might fail to distinguish vowels in B, through identifying both with one vowel in the speaker's own system. But, in principle, they would do so independently. For all of them A is the primary system; hence, whenever they speak B, they will independently make errors of the same kind. Weinreich's chapter on phonology deals more briefly with the process of ‘diffusion’, as he calls it, by which changes may ensue in either language generally (pp. 23–24). For the second stage, however, we can turn to studies by his supervisor. Old Spanish, for example, had a consonant system that included pairs of voiced and voiceless sibilants, which merged in the early modern period, as Martinet explained it, following contact with Basque.7 We assume that in the Middle Ages, as now, the system of Basque did not have this distinction; hence, when speakers of Basque became bilingual in Spanish, they would identify two sibilants in Spanish with just one in their own language. So, in speaking Spanish, they would not distinguish them. But Spanish was gaining ground; the children of bilingual parents would acquire their knowledge of it from them, and in that way would grow up with an ‘internalized language’, as conceived by Chomsky, that did not make this distinction either. Their speech would then become part of the ‘externalized language’, or the language, as Bloomfield defined it, to which other members of the Spanish speech community, including those of a new generation, would be exposed. As that changed, so the ‘langue’ (in the Saussurean sense: conceived as an abstraction from a set of internalized languages) would change with it. Martinet's paper on the Spanish sibilants appeared as Weinreich's dissertation was completed, and it makes clear in principle how languages as systems can be changed by contact with a substrate. But, as it changed, was Spanish at all times a single system? We could say that it was; but ‘knowledge’ of it varied between speakers who, at one extreme, distinguished voiced and voiceless sibilants, and others who did not. ‘La langue’, in Saussure's words, ‘n'existe parfaitement que dans la masse.’8 But the variation would not have been random. Another answer, therefore, is that different groups of speakers would form separate speech communities, each with a different system. Let us return, in this light, to the process of borrowing. ‘Nonce-borrowing’, as Weinreich calls it, is a case of ‘interference in speech’ (p. 11). This is again the use by a particular speaker, on a specific occasion, of a word which is foreign to the language being spoken. ‘In language’, however, ‘we find interference phenomena which […] have become habitualized and established’. ‘Their use’, he goes on, ‘is no longer dependent on bilingualism.’ What, then, if a word is borrowed only by a community of bilinguals? Weinreich's illustrations include many examples of the speech of European immigrants to the United States, which, at the level of ‘parole’ at least, included loans from English that were not part, and have not become part, of their language generally. But many were not simply borrowed, independently, on individual occasions; they had become habitual within the immigrant community. Should we see them, then, as part of a new ‘état de langue’? Or do they belong to a new ‘langue’, of the immigrant community specifically? Or is there indeed another possibility: that changes may become habitualized, yet not at that level? It is obvious how, from Martinet's perspective at least, the integrity of languages is called into question. But so, it seems, is Weinreich's concept of ‘interference’. What exactly, that is, interferes with what? Those instances of deviation from the norms of either language which occur in the speech of bilinguals as a result of their familiarity with more than one language, i.e. as a result of language contact, will be referred to as INTERFERENCE phenomena. […] These opening paragraphs are at first sight admirably crisp and clear, and when they were first read fifty years ago, they mapped out a field of study that was only then about to be investigated systematically. Nevertheless, a central definition has been sidestepped. Phenomena, including ‘interference phenomena’, are again of speech; or, as Weinreich puts it, they ‘occur in’ speech. What is defined, accordingly, is not ‘interference’ itself, although that term alone is put in capitals, but a class or type of such phenomena that, on the simplest reading, will exhibit or be characterized by interference. Alternatively, perhaps, interference is a process underlying them. But, in either reading, it appears that we are talking simply about speech. Phenomena of interference, seen as individual deviations, can then have an on the as Weinreich calls them, of a language. A change in norms could result from them. But such an is not interference or, if it is, no definition of it has been Weinreich then this term has been term interference […] of that is, than or to an Even when might more be spoken of in the or interference, be We are on the first and this was so and in the 1950s that it is not to how the has But interference does not merely change in that, it seems, is what it The term would to more than the of an individual deviation in speech, or to a process that to These, we have been are interference phenomena. then did Weinreich not their on in answer is that, had he so, he would have had to make how such should be primary was with and the many different including those that have to do with the of speakers and with that in it. more than the (pp. deals with language in abstraction from for a structural when two languages come into contact, the of contact are first in the speech of people who use These are in the opening although the term is not itself defined, as But, through languages themselves could then In the as had in the they could to form a this then part of the same In one it could not in the Saussurean sense could be only at a level other than ‘parole’. But in Weinreich's or the answer was as a if we for we will not find it. What exactly, for example, should we by of a The as we have of on the of these then, So, in now of a two-stage first there are individual in speech; then, through their are In one natural in the speech of a community. are that is, in a language (as defined by Bloomfield in the or in what some linguists would now see as an ‘externalized A we might is a to which speech a speech community is in to on a might then be for example, by a change in word A, has a in which or will In language B, however, the is the We expect phenomena of if speakers of A speak B, and if many speak B to or in the of other speakers of A, they may be in B as in A, to put the This was indeed one form of interference, which Weinreich briefly in its due (p. Its in B generally is not But, strictly speaking, the of a an would be a to be in B, if it is only in B spoken by It is therefore, that among such speakers there will be of variation between this and the for B in Let us that bilinguals form a community, whose members as individuals, with speakers of B who are The variation for them, to the that, in instances of their might be if they were not in all speakers of B. might then become more to the point at which they a new which all speakers of B to The term ‘interference’ would then be in the way that Weinreich does quite use it, for the of such a The first in his own words, would be that of interference This is the in which, for example, we The second he is that of interference At this (in a partly ‘we find interference phenomena which, in the speech of have become habitualized and established’. would be that are to linguists to be of interference from another but of the language’ them simply as words, whether ‘borrowed’ or not (p. 11). This is in and in and the distinction has a Weinreich on which the about one or the other (p. But the way it is explained ‘In we are ‘interference is like by a The of are to ‘interference ‘In language’, interference ‘is the on the of a (p. 11). These that have are the ‘interference that, as Weinreich puts it, become habitualized and established’. as we have interference phenomena were defined as individual of are strictly speaking, similar to of which as as a them can as become The to that interference, is a process that behaviour in a group of The for example, of an might be first in the speech of a specific It may then occur in the speech of we might say that, in this it is The use of the same which is to the of may be this again it may be again it may be The process is when it is, its in speech is no longer a of interference, but simply as Weinreich calls it, of interference. Another of could be a of word order, in which, might In what we might then call the each of interference involves changes in the way a group of individuals are first, perhaps, a group of a speech community, speaking what is called a language, as a But is that all that Weinreich by on The would be a two-stage which would distinguish between speech, in which phenomena of interference can be and a system of norms underlying speech, which might what might be called another form of interference, at this underlying its from the speech of a language community by the of in its language, not only change the use of individual but also a new of on the model of another language. (p. answer might a model with stages. in in the early and were from at two different and successive A change in norms might have its in individual of to contact, it would again be in those of This in their speech, as we might it, might again become in a community. But a will change within the of a system, and, in to such a which we might compare to processes of these can themselves from a ‘state’ in which the system had been as in the Saussurean model of a language, to a new At this we could the system of one language, through the first of interference in speech, then of an in as influenced by or, as might have put it, by that of another. however, in a two-stage or a is defined as in the chapter on interference’, is on the of interference that is, of of the types of (p. This would as in one of the first that the term to no more than a process underlying individual types of interference however, of the system’ (p. In the case of such phenomena are again what are in his chapter on interference, Weinreich that the of a into a field or is a part of the as the of which that will be a (p. Should the then, of of the system’ not also as part of the point is that interference in speech is one and interference language’, at least where to is another. But is a to to Martinet's does that the of they are as logically they in what does their as Weinreich it, is In Saussurean it at the level of langue’, as to that of ‘parole’. A which to this again of language and speech’ (p. But that at the level of ‘language’ has to in ‘a in an individual ‘langue’, distinct from Thus, in the to which this is Weinreich a have so been as he it, them in languages than in the speech of (p. The might be as there at the level of ‘language’ languages that we can identify as If so, the distinction will be simply on the one languages in that sense and, on the speech’ in individuals and on individual would again be in with what I have called a two-stage Or are there also partly that in some sense at the level of ‘language’ as to speech, but do not or form part a whether the two systems are of the same or of the same The the between the i.e. the more the and in the is the learning and the of interference. But the of interference, from the of interference, would to be the same whether the contact is between and or between two of English by (pp. also In one reading, this is also a term whose sense might have to be It would include all as Weinreich them, that might be in contact, whether or not they are different ‘languages’ that is, in the sense that Or are we simply to the definition of It involves two and it is therefore, whether they are or or of or This may at first sight to be than a Weinreich's topic is the of interference, and all that at this of his is that we should not or on between where the as he to be the The is into what we might as of speech, or of in to the two-stage model these simply speech. If different individuals, all of have A as their primary system, speak B in that are it is are in a similar however, that among them of their speech become that then, this form will itself into a or of B, and we might expect what Weinreich as when whose is different, it. that the use of A is in may be part of a community whose members speak B in this but do not all speak A, or do not all have A as What would such a form of speech as Weinreich is that of speech may as a are such as the and (p. and also what he (in a with the same as an language between the in (p. Their as such may be he to or of they may have ‘a form (p. from either of the languages from whose contact they have This of course, that these were themselves (p. they may have ‘a of form after This as he in which the two languages are by the bilinguals and the of the of (p. they may have other than those of a they may have ‘a among speakers themselves as a separate The are like those by which in might be to from But, as in that they are partly The two in Weinreich's words, the of (p. as by this term which was from that of A way of speaking might the first and second which are within in a structural sense, in particular, no more than a It is too, that structural may be to The of any form of speech is seen as (p. there is no single point at which a set of of and or would become What of speech that have not The term ‘speech is Weinreich's own (p. and, as he there are other whose are in contact and so on, they do not these be new languages in a sense of the they at least, in that The term is not in this But they have from other speech and, in the communities of which those who speak them are part, also be as Their will be in some to that of simply do not as languages as now the sense of this term. a system, however, itself be a process of What it will be speech, at least, which is that of a community. We might that, in it, will have become But, at that stage, they will to more than individual or individual that become what Weinreich to a yet a new of or a new ‘langue’ as a might see it. This would be so, perhaps, the of new alone of systems which might not be ‘languages’. What then, we are is a form of speech or ‘speech of that In the two-stage model it is no more than speech. phenomena are similar in different individuals, but the in each The same may become habitualized in many different but that too, in principle, could independently. Such of speech’ would have no This is however, as Weinreich it. in his words, ‘a group of some two languages into contact, in linguistic to each speech and processes of the group as a become As he such of language as groups not to a between individual and group is the of the chapter from whose opening (p. this is then that the speech of such a group is, in part, The simplest examples where the of we might the in principle between a language or a and of language that are of a community of What is a ‘langue’, as structuralists had seen it, but a way of speaking that is It will not be I that I to find in Weinreich's the I that he was of the can of that his was Martinet's us to read in that The of his however, to both and the had to be on the whether speech’ or language’, is a process of and the of change in was the topic on which Weinreich had been in at such a early his The paper which his is his only in part, and in the part that might have been what we have are at Weinreich's as and by Thus, in is to occur in of the speech It is we are the community as a not to between (p. is, the of a group or change as a unit. This concept of a first appears some in with that an (p. that of in as that a earlier in the same (p. The of an however, was what was a in the that is, knowledge of a language as by an individual What, in was this a speech As a it made sense, and the reification of such is But, if read this was structural any community, speech would be and, as Weinreich and both made clear, its was and not random. such a community, there would be one and the same It included what at least, called and these for variation in the speech of individuals and between individuals. But as in own was within a and this system was for member of each of the ‘speech into which a might be At the level of langue’, each such community by definition as it Weinreich have been with such as I should like to that, years his treatment would have been more