Rediscovering Interlanguage

Larry Selinker, William E. Rutherford

• Publisher: Routledge
• Year: 2013
• ISBN:
  9781317898603
• Edition: 1
• Topic: Languages & Linguistics
• Subtopic: Linguistics
• Index: Languages & Linguistics

An account of the development of research and thinking in the field of learner language. Draws on wide-ranging research into contrastive analysis, bilingualism, theoretical linguistics and experimental psychology.

1 Beginnings: Fries/Lado

Practical teaching concers

One can say that the Contrastive Analysis (CA)/Interlanguage (IL) experience began with the following insight. In a much-quoted sentence, C.C. Fries wrote in 1945:

The most efficient materials are those that are based upon a scientific description of the language to be learned, carefully compared with a parallel description of the native language of the learner.

(Fries 1945, p.9)

In this chapter we describe briefly some of Fries's contributions and then consider the work of his colleague, R. Lado. One link we will explore is the important connection between Lado and scholars in a neighbouring field, bilingualism. Thus, in Chapter 2 we will discuss in some detail the relevant contributions to CA/IL thought of a pioneer in several branches of linguistic study - Uriel Weinreich. Then we will move on to a study of the question of what is 'comparable' across languages, more precisely across linguistic systems. We will do this because, we will argue, this latter insight is important to IL and thus to second language acquisition (SLA) thought. Later in the book we will pause for a detailed look at the continuing discovery of IL and the forces that shape it. We will move towards a consideration of the ongoing process of comparison in applied linguistics and SLA, specifically at several recent attempts to understand IL and IL learning through one or another comparative endeavours.

Beginning with Fries, then, the above quotation often appeared in the 1950s and 1960s at the beginning of MA and PhD theses to justify the writing of a particular CA study. The intellectual link between this quotation and CA was made explicit by Lado when he pointed out that for effective teaching materials, implied in Fries's statement is 'the fundamental assumption' of CA work that 'individuals tend to transfer the forms and meanings, and the distribution of forms and meanings of their native language and culture to the foreign language and culture' (Lado 1957, p.2). The direct link, then, between Fries's thought and CA - and thus IL and SLA - is established. The strongest motivation for doing CA from its earliest days involves applied work, namely, in considering what the best teaching materials might be, one has to look carefully at the learner from the point of view of possible transfer. This is still, I would maintain, a useful current position. So one strong motivation for doing a CA in applied linguistics has always been to learn something important about transfer, 'negative transfer' (in this view of things) being traditionally associated with the making of 'errors'. Later in this chapter we will come back to Lado's seminal work and we will be returning to considerations of language transfer (in Sharwood Smith's phrase (1982): 'cross-linguistic influence' or CLI) at various points throughout the volume.

Fries, in the 1945 classic Teaching and Learning English as a Foreign Language, was primarily concerned with the 'first stage of language learning' and the 'end to be attained' in that stage (p.8, emphasis in original). 'That end,' in Fries's words, 'is the building up of a set of habits for the oral production of a language and for the receptive understanding of the language when it is spoken.' On the next page, Fries states that to gain that end 'within a reasonable time' one of the key ingredients is 'proper' teaching materials, the 'most efficient' ends being those gained by the method described in the quote in the first paragraph above. Fries's explicit aim was to develop teaching materials which would lead the adult learner towards making the sound system and structural system of the language to be learned 'automatic habits' and 'unconscious habits' (p.3), using only enough vocabulary to make the systems work. (For further discussion of this point, see Morley et al. 1984.)

The teaching materials contemplated by Fries/Lado and colleagues, since they are based on a careful comparison of the native language (NL) and the target language (TL), obviously have to be language specific. In light of this, Fries framed an important truism, one which was to have substantive import on the development of our field, especially since enchantment and disenchantment with CA have been dominant themes in SLA and IL thought: 'Foreign language teaching is always a matter of teaching a specific "foreign" language to students who have specific "native" language background' (Fries 1945, Preface). How this approach fits in with current attempts to integrate SLA thought into considerations of universal grammar (UG) will be discussed subsequently, but here suffice it to restate that the applied/practical impetus for doing CA over the years has to be clearly emphasized.

In a later discussion in the Japanese context, Fries and Fries (1961) made all of this even more explicit. In order for the Japanese student to 'learn English well', he 'must have' English teaching materials which are 'especially adapted to his linguistic needs'. They then make a statement which has important ramifications for SLA prediction:

The proper selection and contrastive arrangement of the English patterns to meet the special needs of Japanese pupils at the very beginning will determine whether the teaching will be effective and the learning easy or the pupils will be so confused that the teaching will be very difficult and a satisfactory learning almost impossible.

(Fries and Fries 1961, p.1, emphasis added.)

Today, although we would not accept such claims as to ease and difficulty of learning without empirical evidence from an input study, we can nevertheless clearly see the link between practical concerns and SLA thought inherent in a Fries approach to CA. It is important here to preview the following: a conclusion such as that just quoted - whatever its current lack of empirical evidence - now becomes, for the record, an important hypothesis available for empirical testing on some learners under some (as yet unspecified) conditions.

The contrastive approach is presented by Fries for use with Latin-American students, the earliest students dealt with in a systematic fashion (for more detail see Morley et al. 1984). On p.25 of Fries (1945), for example, one of the steps used in considering each TL-English sound is to 'isolate' for the Spanish-speaking learner 'the most similar sounds in Spanish words, as "perro" for

or "mismo" for [z] or "dedo" for

'. Or on p.33, in discussing 'word order patterns that English uses to express fundamental relationships', Fries refers to the fact that Spanish speakers must not, as they 'frequendy do', say 'station bus' when dtey mean 'bus station', and 'must develop a habit of placing single word modifiers of substantives before the words they modify'. Neglecting for now the troublesome word 'habit', this point is then linked to an extensive discussion of the selection of patterns to be taught and learned productively versus those to be learned receptively. The relationships to IL of considerations of perception vs production involve a host of issues that are still with us (see, e.g., Gass and Selinker 1983a, passim, and Selinker 1984, fn 2).

On pp.39 and 40 of Fries (1945) he provides detailed examples of the fact that 'practically no words of one language ... ever cover exacdy the same areas of meaning and use as those of another language'. He contrasts, for example, the English word 'table' with the Spanish word 'mesa', pointing out that 'mesa' is not used for a 'table of figures', nor for a 'table of contents' nor for a 'time table'. And then, elegantly I would claim, he gives examples where mesa 'is used in many connections for which we do not use the word table'. He then gives seventeen examples of the use of the English word 'time' for which there is 'no simple Spanish word'. And there is much more in this vein in Fries's writings. What we have are useful CA facts which, in terms of the essence of our discussion here, become useful hypotheses for understanding and testing IL in various SLA contexts.

One point to note is that Fries is not known for having undertaken detailed CAs himself and that is most likely why histories of CA and SLA usually fail to mention him. In hindsight it is easy to criticize Fries's approach to CA and SLA (and we do criticize Fries - see Morley et al. 1984), but here I wish only to reveal for the reader something of Fries's crucial place in the CA/IL chain, as well as to lend at least some of his conclusions the respect of being reasonable CA hypotheses that can be empirically studied.

A technology of contrastive analysis

It was clearly Lado (1957), however, who carried this approach into a 'technology of CA', which, as we shall show below, provides us with useful current tools and, where certain key issues are first carefully explored in the literature, issues that are still with us today. In Lado's work the comparison of the learner's NL to the language to be learned was explicitly hypothesized as a predictor of learner errors. Importantly for SLA thought, this is sometimes true, and from subsequent attempts to validate this hypothesis arose error analysis (EA), one of the immediate precursors to IL and SLA. It appears then that from Fries's insight that the best and most efficient teaching materials arise out of a comparison of the NL and TL, one derives the notion from Lado that such a comparison is able to predict transfer and thus errors which learners will make, which for Lado equal the 'problems' or 'hurdles' learners will encounter.

We find in Lado the conclusion that before one can prepare the best and most efficient teaching materials, one has to discover the learning difficulties through such a comparison: 'The most important new thing in the preparation of teaching materials is the comparison of native and foreign language and culture in order to find the hurdles that really have to be surmounted in the teaching' (Lado 1957, p.2, emphasis added). Lado goes on to state that 'it will soon be considered quite out of date' to do such things as begin to write a textbook without having 'previously compared' the NL and TL, i.e. in classical CA terms, the 'two systems involved'. Lado (1957, p.2) further states that 'the linguistic comparison is basic and really inescapable if we wish to make progress and not merely reshuffle the same old [teaching] materials'. As we shall see in succeeding chapters, the proposing of IL comes directly out of the empirical attempt to undertake such comparisons and to test these comparisons in light of the phenomenon of language transfer. Researchers now realize (cf. Gass and Selinker 1983a, passim, and Kellerman and Sharwood Smith 1986, passim) that language transfer as a 'selection process' should not necessarily be equated either with the making of errors in a second language nor with learning problems in attempting to acquire that particular TL. The reader should also note that neither of the latter two concepts, errors and learning problems, should necessarily be equated with the other.

We note that Lado (1957) is considered a seminal work. For example, James (1980, p.8) states, 'For me, modern CA starts with' this book. Importantly, Lado is not only historically significant but is quite relevant in many ways to continuing concerns in SLA and IL studies. We have noted this before: '... we have been able to infer from the Toronto French-immersion data we have looked at, that these ... individuals are using in their IL, grammatical structures ... [which] are, to some degree, similar to the "rules" described by Lado [1957]' (Tarone et al. 1976, p.99). My point here is that in empirically studying IL and the influences that shape it, SLA thought has never abandoned some fundamental insights inherent in CA. One of the reasons we have to go 'back again' to classical CA for insights appears in sentence two of the Preface to Lado (1957); Lado's point is that the results of CA prove of 'fundamental value' to 'language learning experiments' (see also his p.7, 'Significance for Research'). We are sure that this is the case and have also noted this before: in the Introduction (Chapter 1) to Gass and Selinker (1983a), after a careful review of what is known about the phenomenon of language transfer in language learning, we conclude that: 'For us, one important preliminary step to understanding language transfer is, at the very least, a native language-target language comparison, which often leads to insightful hypotheses concerning language transfer phenomena' (Gass and Selinker 1983a, p.6). Several chapters in that volume and in that of Kellerman and Sharwood Smith (1986) demonstrate, in a very current way, the truth of Lado's insight.

On balance, however, Lado surely promises too much when he states in his Preface that language teachers 'who understand this field [i.e. CA] will acquire insights and tools for ... diagnosing student difficulties accurately'. This is where classical CA began having problems. These problems unfortunately led to the ten-year effort (called the 'baby and bathwater syndrome') of attempting to discard the entire enterprise of CA. This effort has failed and, as we shall show in these pages, CA has made a comeback. But here one has to ask: Does 'diagnosing student difficulties accurately' mean accurately diagnosing 'all difficulties' or only the major ones? And if the latter, which are they and how does one decide? How does this conclusion then relate to individual differences which Lado himself brings up (p.72)? In the final analysis, one has to ask: Why is Lado's claim stated so strongly? This is not nit-picking; this sort of claim led many teachers and researchers totally to stop paying attention to the insights of CA.

In order to understand IL and the forces that create it, we have to strive for some balance, for it is a fact that CA predictions sometimes work. In this regard I have pondered long and hard the following assumption, that by doing CA '... we can predict and describe the patterns that will cause difficulty in learning, and those that will not cause difficulty ...' (Lado 1957, Preface). My concerns with this statement mirror those above. I wonder why this strong assumption was made in the first place. I have asked people who were there and who cannot really explain the strong categorical tone, even in hindsight. One has to ask: When, and especially why, did the notion of CA as predictor of errors, learning problems, difficulties, interference, hurdles, what have you, begin to take over the research endeavour? Where did these scholars go wrong? Why would one have believed in the first place (as I surely did for a time) that by comparing abstract linguistic structures one would necessarily learn about 'the patterns that will cause difficulty in learning' a second language. This is not a trivial point. Why would a whole generation of linguists and applied linguists, actively involved in second language learning and teaching in many parts of the world, have believed this as accepted fact? The only conclusion I can come to is sociological, related to a dominant pseudo-issue of the times, of whether or not linguistics is a 'science'. My conclusion is that the hope was there that had Lado's categorical statements been right, i.e. had CA predictions been unequivocally borne out by the empirical CA study for which Lado himself was calling (see below), then linguistics would have become a 'real science'. We must continually be aware of this pseudo-goal, for unfortunately one sees this same trend now with some experimental SLA. I can recall a discussion with Lado, about 1964, where the section of CA structuralist studies called 'residue' (= unexplained phenomena) was apparently growing with each study. This was certainly puzzling since some CA predictions were indeed being empirically borne out. As we shall see throughout this volume, this continues to be a central area for understanding IL and IL learning in SLA.

We must note again that it is unfortunate that the extreme claims of CA as SLA prediction led many to abandon CA entirely because of those cases when the predictions of errors, especially, did not come true. But in the context of the discussion here, we can easily proceed since our mission is to try to show why, in order to understand IL one must understand CA. If it is true, as we will argue throughout this volume, that in order to understand IL one must understand CA, then Lado deserves much credit for looking at CA as a technology for the SLA issues it continues to raise.

Techniques for comparison

In the Preface to Lado (1957), 'techniques for comparison' are mentioned and much of the book is devoted to spelling these out. We are fortunate here for it is a related point of this discussion that most SLA scholars do in fact compare linguistic systems as a central part of their SLA work. For example, they compare either ILs with NLs or with TLs, or Native/Non-Native (N/NN) interaction with Native/Native (N/N) interaction, or in fact make some other important SLA comparison. If this is so, then it is a good thing to have this technology in place. We will look at Lado's techniques of comparison in a moment.

On p.1, Lado introduces the work of scholars in the neighbouring field of bilingualism, by stating that a 'practical confirmation' of 'the fundamental assumption' of his book (quoted above) has come from such scholars. He specifically states that 'practical confirmation of the validity' of this basic assumption has come from the work of Weinreich (1953) and Haugen (1953). Is this a fair reading of the evidence? Lado states that researchers in bilingualism 'report that many linguistic distortions heard among bilinguals correspond to describable differences in the languages involved' (Lado 1957, p.1, emphasis added). They did so report, but the emphasized words raise two central issues: a) the hedging issue of many leads to the conclusion that classical CA statements were in general statistical in nature without statistical controls (see Selinker (1966) and the discussion below of hedging used in classical CA), and b) the correspond issue leads to the conclusion that domains of inquiry were confused in classical CA. This latter issue was raised by Weinreich himself and is discussed in Chapter 2.

In brief this issue is best understood in terms of the Saussurean distinction (Saussure 1922) of 'langue/parole' in that (as we shall see in the next chapter) Weinreich points out that quite often the speech of bilinguals (one domain of research) relates well to abstract linguistic categories (a different domain of research), but quite often it does not. As pointed out above, this lack of correspondence between language transfer and linguistic categories remains an issue unresolved in SLA (and in general linguistic inquiry as well). But even so, in this context we must specifically ask: Is the work...