1. Introduction
The term "native speaker" is pivotal in a number of areas. Firstly, even in generative linguistics, the concept of an "ideal speaker-listener, in a completely homogenous speech community" (Chomsky, 1965:3) is crucial. While idealising from the data is both permissible and necessary, vagueness about the terms "speaker-listener" and "speech community" may leave us uncertain as to exactly what data we are idealising, which then casts doubts on such generative shibboleths as "competence and performance" and "core and periphery". Secondly, it goes without saying that the term "native speaker" and its associated terms "mother tongue" and "member of a speech community" are of primary importance for sociolinguistics. On a more practical note, in the language teaching profession, being classed as a native speaker is the key to status, expanded job opportunities and higher pay, which naturally creates a heated debate (one I do not intend to venture into) Finally, the question has important cultural and political implications, particularly for ethnic minorities, emerging nations and speakers of English as an "international language" (Pennycook, 1994).
Uncritical use of the term "native speaker" begs two questions. The first is the question of what it is one may be a native speaker of. Words like "language" and "dialect" are themselves ill-defined. This has led sociolinguists to prefer the term "speech community", but, as I shall argue, this simply moves the vagueness into a different area. Secondly, even if there is no ambiguity about the language, dialect or whatever, the word "native" is not only vague, but has non-linguistic connotations which are by no means culturally or politically neutral.
Problems in defining these terms do not arise from mere lack of rigour; they are inherent in the very concepts we are attempting to define. In this essay, I shall thus adopt more of a philosophical than an ethnographic approach to the problem, and, rather than reviewing the literature and then attempting to show who has the "best" definitions, shall examine the various concepts in turn, picking out examples which illustrate the problems involved, and then offering suggestions for an alternative approach to their categorisation.
2. What does a native speak?,p> As stated earlier, there is no point in describing someone as a native speaker unless we are sure what it is they are a native speaker of. If we say "Susan Chang is a native speaker of English", are we referring to British, American or Singapore English, for example? In the case of Singapore English, is this Standard Singapore English, Vernacular Singapore English or both? Can she also be regarded as a native speaker of, say, Cantonese, and if so, is Cantonese a language or a dialect of Chinese? These questions, though somewhat hoary, are by no means insignificant when attempting to understand the native speaker concept.
2.1. Language versus dialect
One the face of it, Saussure's (1915) famous distinction of langue and parole looks uncomplicated. If parole is the totality of utterances by the speakers of a language, then langue is "a social product of the faculty of speech and a collection of necessary conventions that have been adapted by a social body to permit individuals to exercise that faculty" (1915:61), which involves abstraction from parole (Ellis, 1993:104). However, isolating a langue from a confusing plethora of language examples requires first that we be selective about which utterances we abstract from. If we wish to describe, say, French, we would not include data from Breton, but would we include the French spoken by people who also speak Breton? Would we also include speakers of French outside France, for example French-speaking Swiss or Canadians? Conversely, would we only accept as French, language approved by the Académie Française?
It seems that even determining the most basic object of study in linguistics - a language - involves sociolinguistic, and even political, considerations. It is perhaps surprising, then, that even many sociolinguists accept "language" as a given (Le Page & Tabouret-Keller, 1985:1). Languages are notoriously hard to define, hence Max Heinrich's oft-quoted (e.g. Chambers, 1995:214) tongue-in-cheek definition: "A language is a dialect with its own army and navy." Outside linguistics departments, there is a pervasive attitude that "nation = language = territory = state" (Lunt, 1986:729, in Rudin & Eminov, 1993), and the assumptions of nationalism have profoundly influenced our thinking about what is and is not a language (Fishman, 1989, Williams, 1992).
A case study of the controversial status of "national language" is provided by Kurdish. The classical taxonomy is set out by Hassanpour (1993:107): "Kurdish is a member of the Iranian group of the Indo-Iranian subfamily of the Indo-European family of languages." On the ground, however, the situation is much more complex. In the Southeast of Turkey, there are at least two "languages" that could reasonably be classed as Kurdish: Kürtçe and Zazaca. Of these, the former is regarded by most Kurdish nationalists in Turkey as "Kurdish", whereas there are good grounds for regarding Zazaca as closer to the Northern Iranian "Kurdish" of the philologists [1].
The test of mutual intelligibility is also suspect [2]. While Kürtçe and Zazaca share a similar grammar, a speaker of the former would have extreme difficulty in understanding the latter. On the other hand, Serbian and Croatian are prominent recent examples of mutually intelligible "dialects" of the same "language" emerging as separate "languages", with archaisms and provincialisms being adopted to "purify" the language (Gee, 1997; Woodard, 1996)[3].
2.2. Language loyalty and standardisation
A language is thus, to a large extent, whatever it is perceived to be[4]. A significant factor is "language loyalty" (Gumperz, 1972), which normally links regional dialects to a national language of which they are seen as variant forms. However, sometimes a group may regard itself as speaking one language, when their dialect is linguistically closer to another, and language loyalty may also confuse the issue of mutual intelligibility (Gumperz, 1972:228-229).
An influence on language loyalty is standardisation; a dialect is often perceived as a variant form of a language because that language possesses a standard form which is accepted by speakers of the dialect in question. Standardisation occurs through a number of processes, such as urbanisation, increasing use of a "court language" (e.g. "the King's English"), increasing commercial, legal and literary use, and planned language reforms. None of these automatically produce standardisation, however, as can be seen in the case of Turkish. The court and literary language of the Ottoman Empire was essentially a sophisticated creole which never became fully standardised because an educated speaker of Ottoman Turkish would also be fluent in Arabic and Persian; code-switching was the norm rather than the exception. While language reforms later purged Turkish of thousands of loan-words, it is only with the spread of the mass media that anything approaching a "Standard Turkish" has become a reality rather than an ideal[5]. We might rephrase Heinrich by saying that these days a language is a dialect with its own newspaper and TV channel.
2.3. Speech community
In an attempt to avoid the terminological confusion surrounding such words as "dialect" and "language", most sociolinguists prefer to adopt the idea of "speech community". However, this is, if anything, harder to define. Saville-Troike (1989:16) states that "since the focus of the ethnography of communication is on the speech community, and on the way communication is patterned and organizes within that unit, clearly its definition is of central importance", then goes on to list no less than five different definitions[6]. This then means that, for example, "questions arise in deciding if speakers of English from Britain and the United States (or Canada and Australia, or India and Nigeria) are members of the same speech community" (Saville-Troike, 1989:17). In others words, we seem to have a similar problem with "speech community" as we did with "language" and "dialect"[7]. The concept of speech community may be a useful research tool [8], but it does not remove the problem of "language". "Many attempts have been made to define a language in social terms, such as "The language of Community X"; but when we have to devise a test for membership of Community X it frequently includes ... speaking Language X, and the definition therefore becomes circular" (Le Page & Keller, 1985:191). [9]
3. Who is a native?
I have gone into the subject of "language" before discussing the term "native speaker" because, as I have stated, the two are mutually interdependent. A native speaker is a type of speaker of a particular language, but our idea of what constitutes a language is dependent on assumptions about who is a native speaker of that language [10]. The term "native", however, seems to be about as clear as the terms "language" or "dialect"; like St. Augustine's "time", "we understand it until we start to think about it" (Ellis, 1993:78).
3.1. Native and non-native
To talk about native and non-native speakers is to possess an assumption (or assumptions) about natives and non-natives per se. The terms themselves are perhaps unfortunate in a discussion of language, since the word "native", as its etymology suggests, implies birth into a specific community, or in a particular place, the two usually being regarded as identical. In fact, cases where being born in a particular place automatically entails membership of a specific community and knowledge of a particular language may well be the exception rather than the norm. It is fair to say that, at least up to the 1930's, someone born in a particular spot in the New Guinea Highlands was virtually certain to be a member of, say, the Mundugumor tribe, and would thus speak only their language, so it would be equally reasonable to call him or her "a native speaker of Mundugumor". Margaret Mead, on the other hand, would be a "non-native speaker of Mundugumor" (Mead, 1980). However, if we assume for the sake of argument that none of the Mundugumor learnt more than a few words of the neighbouring Arapesh or Tchambuli languages (and vice versa), the native/non-native issue only arrives on the scene with the first anthropologist.
In most of the world, however, community does not equal territory and language does not equal either of these 11. Someone born in Wales may or may not regard him- or herself as Welsh, and may speak Welsh as a "first", "second" or sole language, or not at all (Grillo,1989:55-60,93-96). "Nativeness" is only a very rough and ready guide to "native speakerness".
3.2. "Hearth and home"; the "mother-tongue" debate
In bilingual or "diglossic" communities, the question of "mother-tongue" arises (for discussion of the terms "bilingual" and "diglossic", see Ferguson, 1959, Saville-Troicke, 1989:54, or Williams, 1996:104). In these contexts, "mother tongue" is used as a convenient reference for determining who is a native speaker of a particular language or dialect; it is usually taken to be the language a speaker heard as a child, normally uses at home, or both. However, problems arise here as elsewhere. Language loyalty is again a factor in perception of "mother-tongue", as Le Page and Tabouret-Keller point out:
In Singapore, because of the enormous importance attached by Chinese to the written language, "mother tongue" is used - quite arbitrarily - to refer to that spoken language (Mandarin) which is culturally most closely connected with written Chinese, regardless of the fact that most Singapore Chinese speak Hokkien or Teochew or Hakka or Cantonese or Hainanese, i.e. languages from the south of China. Even these are likely to be spoken in a mixed, Singaporean way (1985:189).
What "language" is actually spoken at home is by no means obvious, since code-switching applies at home as much as anywhere else; language domains are not rigid [12].
The mother-tongue debate also raises questions about the status of standardised languages. If the standard language (or dialect, if you prefer) is rarely spoken at home, can anyone really be said to be a native speaker of it? Even in cases where a particular vernacular has been taken as the standard for a nation (as in the case of Putonghua, where the Beijing dialect was taken as the standard for the whole of the PRC), this still leaves a relatively small proportion of the population having the standard version as their "mother-tongue", and once the dialect has been standardised, the local dialect may drift away from the standard version.
In short, the "hearth and home" attitude exalts one domain at the expense of others, ignores code-switching, and confuses the issue of standard languages.
3.2. The question of competence
One way to avoid the "native speaker" trap is to speak of "native speaker competence" (a common alternative in advertisements for teaching posts). Natives of a community have native speaker competence, more or less by definition (Hymes, 1972c; Fishman, 1972a:49). On the other hand, non-natives may also acquire native, or near-native competence.
The first problem with the competence idea is the by now familiar one of circular definition. Native speaker competence can be broadly defined as the ability to conform to the set of linguistic and sociolinguistic expectations of a particular speech community. As we have seen, the notion of a speech community implies members by whose linguistic behaviour the community is defined, so unless we possess a definition of "native speaker" which is not related to competence, we are back where we started. If, on the other hand, we do possess such a definition, sociolinguistic competence is, to use Aristotelian terms, an accidental rather than an essential property of a native speaker.
The second problem is that, unlike Chomskyan "competence", communicative competence is not an all or nothing affair, and in some areas of competence "non-natives", such as immigrants, may actually be more competent. If we are talking about competence in the standard language, what would we make of the sentence "Someone should learn them Pakis to speak proper" (Leeds taxi driver, personal communication)? The speaker here is demonstrating both linguistic and sociolinguistic competence for a working class Leeds speech community, but not with regard to the standard language, where, say, an Eton and Oxford-educated Pakistani would be regarded as more competent [13].
3.3. The critical period hypothesis
A more promising approach comes, perhaps surprisingly, from the generative school. If we ignore communicative competence and concentrate on the narrower notion of linguistic competence proposed by Chomsky (1965, 1986), we can make use of the critical period hypothesis, first proposed by Lenneberg (1976). According to this view, the syntax of a language is acquired rapidly and effortlessly by young children, but not by adults; hence there is a "critical period" for language acquisition. A reasonable "rule of thumb" definition of "native speaker" might thus be someone who acquired the language during this critical period; a language acquired later would thus be a second, non-native language, irrespective of the ethnic or speech community the speaker is a nominal member of.
This still leaves us with some unanswered questions, however. The first is the validity of the hypothesis itself. While there is strong empirical evidence to suggest a critical period for language acquisition (Bickerton, 1990:115-122), it would still be premature to base a definition upon a hypothesis, especially when that hypothesis deals with acquisition of language in general, rather than acquisition of a particular language or language variety. Secondly, it still skirts the question of what it is that is acquired; once we leave the rarefied area of a supposedly independent syntax, the problem of defining a language or dialect remains [14].
3.4. The status of "second languages"
If communities employ more than one language or dialect (as most do), at what point does one of these codes become a non-native language for the speakers concerned? The terms "first" and "second" languages are misleading to the extent that they imply that a bilingual speaker has a first and a second language, that they are acquired in that order, and used in that order of ease and frequency. In fact it is perfectly possible to use the language acquired later with greater frequency and fluency than the "first" language, and a speaker may even forget much of their "first" language if they stop speaking it after early childhood. This in turn may be complicated by a conscious attempt to relearn the first language. Let us say that a child in South-East Anatolia spends the first few years of her life in a village where the predominant language is a variety of Kurdish, but then goes to a boarding school where only Turkish is spoken, and forgets most of her Kurdish. On leaving school she decides to get back to her roots and makes a conscious attempt to relearn Kurdish (maybe a different variety). What, then, is her "first language"?
Where second languages used as a lingua franca are adopted by an educated elite, they may produce their own "native" literature. As stated before, Seljuk and Ottoman Turks frequently wrote in Turkish, Arabic, Persian or a mixture of all three. There is certainly nothing "non-native" about Celaladdin Rumi's Mesnevi, dictated rapidly and spontaneously in Persian. As Kachru (1986:12, in Saville-Troike, 1989:105) says of Indian English literature, "The medium is non-native, but the message is not" [15].
4. An alternative categorisation
Attempts to produce stricter definitions of terms like "language", "dialect" or "native speaker", often founder because of the assumption that there is a set of native speakers (or languages, or dialects) with Aristotelean "necessary and sufficient conditions" for membership. In this ideal set, all members would be equal. Something would either be a language or it would not, and a person would either be a native speaker of this language or would not; no native speaker would be more representative than another. As we have seen, this is not the case, and if we wish to scrutinise these terms, we would be better off using more sophisticated models of categorisation, such as those proposed by Wittgenstein (1968), Rosch (1973), Lakoff (1987) or Ellis (1993) [16].
4.1. The "language" category Traditional language typology is modelled on biological taxonomy. Just as in biology we have a hierarchy with kingdoms at the top and sub-species at the bottom, in linguistics we have a hierarchy as follows:-
FAMILY (Indo-European)
SUB-FAMILY (Indo-Iranian)
GROUP (North Iranian)
LANGUAGE (Kurdish)
DIALECT (Zazaca)
However, while in biology it is possible to set up definite taxonomic criteria (e.g. fertile members of the same species can mate successfully), this is not possible in linguistics[17]. "Mutual intelligibility" can be seen as a failed attempt to provide such a criterion; the problem is that while two animals either can or cannot mate, two humans understand (or claim to understand) each other to a certain degree, and in a particular sociolinguistic context.
If, however, following Lakoff (1987), we regard "language" as a prototypical category, the problems discussed earlier do not arise. A prototypical language (i.e. the most representative member of the set of languages) would have the following characteristics:
Of course very few languages meet all of these criteria in full, but something may be regarded as a language to the extent that it approximates these. Thus Italian is fairly prototypical, English is less so, and Kurdish is on the fuzzy boundary of the set. Weakness in one criterion may be compensated for by strength in another; thus French loses out on criterion 2., but makes up for this with its high degree of standardisation.
A similar prototype could be constructed for "dialect" along the following lines:-
We should also remember that some categories in these criteria, such as "country", themselves display prototype effects.
4.2. The "native speaker" category
The first point to note here is that "native speakers of English" do not comprise the intersection of the set of natives of England and the set of speakers of English. Firstly, it is obvious that there is no correspondence between "England" and "English", except historically and, perhaps, prototypically. Secondly, even if we ignore this, there is a logical problem. "Turkish delight" is not the intersection of the set of Turks and the set of delights; the whole may not only be greater than the sum of the parts, it may be quite different.
Nevertheless, at least in the popular imagination, there is some relation between "being English" and "speaking English", hence the popular (mis)conception that British English (by which is normally meant English English, not Welsh or Scottish) is somehow "better" than other varieties. Again, we are up against a prototypical category. We might set up some tentative criteria for "prototypical native speaker of language L" as follows:-
We may note here that criteria originally dismissed as irrelevant when attempting to construct a "strict" (i.e. Aristotelean) definition, are admissible when describing a prototype.
Obviously not all people who are classed as native speakers have all these characteristics; some (such as 6.) are of minor importance, while others (such as 4), may suffice for a non-native (in terms of birth) to be regarded as a native speaker, depending on the context in, and purpose for which, the native speaker criteria are applied. For peripheral "native speakers" there appears to be a kind of Turing test in operation; if a "central" native speaker (i.e. close to the prototype) cannot tell the difference between the speech of that person and that of another "central" native speaker, that person may be regarded as a native speaker [18].
5. Conclusion
Having shown that terms such as "language" and "native speaker" are somewhat vague and fuzzy around the edges, the question arises of whether, as linguists, sociolinguists or language educators, we should abandon them and search for more precise terms. I would suggest that this is not only unnecessary, but also impracticable. Alternatives to "language" and "native speaker" such as "speech community" and "member of a speech community" are, as we have seen, equally problematic, since they take their terms of reference from the original offending concepts.
What we can do is define terms more precisely for the field of discourse in which we are working. There is nothing wrong in saying "For the purposes of this study I shall take the term "native speaker" to mean X." Thus the native speaker of the grammarian would be different from the native speaker of the sociolinguist or the educationalist. This could lead in each case to various prototypical criteria being elevated to the status of "essential properties", in order to create a clearly-bounded and uniform set. In practice this is what all sciences do to an extent; what a botanist means by "fruit" is close to what the rest of mean by the word, but it is so defined that everything is either "fruit" or "not fruit", and no "fruit" is more "fruity" than another. The question of what is really "fruit" does not arise, and should not arise with "native speaker" either.