Systematicity and Variability in Interlanguage Systems

1998 Abderrahim Agnaou

Table of Contents Introduction 1. Interlanguage study: a theoretical background 2. Systematicity in IL 3. Variability in IL 4. Systematicity and variability: a case study 5. General conclusion Bibliography

0. Introduction Ever since the recognition of applied linguistics, both as a distinct and interdisciplinary field of study, increasing interest has been paid to the linguistic production of second language learners. Such linguistic production has been described differently by different applied linguists according to their perspectives. Systematicity and variability have been recognized as two distinctive properties of learner language, thus constituting the locus of many research studies in second language acquisition (SLA). This research paper addresses two key features of the learner's second language development process, namely that it is systematic and variable. It falls into four parts. In the first one, learner language is examined in the way it comes about, thereby attempting to synthesize the major theoretical claims existing in the literature. The purpose of this part is to lay the foundation for the treatment of systematicity and variability as two distinctive features of learner language. In the second part, systematicity is explored while trying to ascertain the factors that come into play in the realization of such a property in the learner language. The third part is concerned with variability and the theoretical solutions that have been proposed to account for variation in the language of second-language learners. These three parts are intended to provide insights into the issues under investigation and to put the approach adopted in the case study provided in this research paper into proper perspective. They are meant to provide the necessary background for the data analysis that is undertaken in the fourth and last part, which presents the case study of a young Moroccan learner of English as a foreign language.
1

1. Interlanguage study: a theoretical background 1.1 What is interlanguage It goes without saying that all learners, in the course of learning a second language, produce utterances that can be evaluated or judged as erroneous by the native speakers of the language being learned. In other terms, the fact that second language learners produce ungrammatical or ill-formed utterances in the course of their learning career is an established truth. This is at least true from the point of view of the native speakers of the target language (TL). The language learner, however, sees his or her own linguistic production as being normal since it emanates from a full-fledged and self-contained linguistic system of which he or she is aware and of which he or she has a full command; otherwise, he or she would not have taken the plunge to produce what would be reckoned erroneous by others. What this means is that before utterances are produced, they have to be processed by the human mind and that there is a system that is responsible for the production of such utterances. Therefore, we can extrapolate the existence and operation of a linguistic mechanism that is responsible for editing rule-governed (learner) language. This, by extension, leads us to say that what native speakers of the TL consider erroneous is in fact correct from the standpoint of the learner. This amounts to taking language learners not for speakers of the TL, but, rather, speakers of a transitional nameless/nationless language. The latter has been given several denominations by a number of leading figures in SLA research for the purpose of referential convenience. From these we can cite Harris's (1954) "in-between language", Corder's (1967) "transitional competence", Corder's (1971) "idiosyncratic dialect", Nemser's (1971) "approximative system", James's (1971) "interlingua", van Buren's (1972) "grafting component", and Selinker's (1972) "interlanguage". Of all these appellations, the latter (i.e. interlanguage) has gained ground among applied linguists. This is so because it is believed to encapsulate the most important aspects of the learner language: that it is transitional and that it is a system in its own right. In fact, Selinker (1992 : 220) points out that the term he proposed "has filled a semantic gap". Moreover, the term "interlanguage" (henceforth IL) is neutral in perspective, as opposed to the other proposed terms which are either suggestive of a rather sided attitude towards the TL or too vague. Given the dissimilarity between the learner language (IL) and the TL, the native speakers of the latter system would consider the utterances of the language learner as erroneous simply because they violate the TL norms. It so happens that ILs are not given conventional names as is the case for native languages (NLs) and TLs which have institutionalized labels such as English, French, Arabic, Japanese etc., leaving aside pidgins and creoles. The language learner, then, can be regarded, in a sense, as "a native speaker" of that transitional nameless/nationless language named IL, perhaps the only native speaker (Corder 1971). But since the purpose of any language learner is ideally to gain near-native speaker competence of the TL, in which he or she wishes to function socially, economically or otherwise, the language learner will hardly stop halfway in the process because the level attained at that stage may not qualify him or her to function fully in the TL society or any context where the TL being learned is required. This is so because, as a speaker of the transitional language, he or she misses a large population of TL speakers with whom he or she may want to interact, and so remains secluded with very few speakers of the transitional language if there happen to be some; however, it is rare to find a homogeneous group of speakers of the same IL.

Now the question that is worth asking is: what processes are responsible for the creation of interlanguage systems? In other words, how does the language learner develop IL? The suggested answers to this question are discussed in the following section. 1.2 IL development It is far beyond the bounds of possibility that anyone can acquire the TL system all at once. Rather, a learner who is in the process of learning a second language has to go through several stages before he or she can be said to have acquired that language. So, one way of answering the question "how does a language learner develop IL?" is to say that he or she develops IL gradually by taking in bits or pieces of the TL incrementally into the existing linguistic store at various stages in the attempted acquisition of that TL. Selinker (1972) posits the existence of a psychological structure which is "latent in the brain, [re]activated when one attempts to learn a second language." This latent psychological structure (LPS) is a genetically determined mechanism (perhaps believed to be part of Chomsky's Language Acquisition Device; cf. Dulay and Burt, 1974a) reset or remade operative for the purpose of acquiring a new language, with the proviso that the learner has constant exposure to the TL environment (formal or informal) and enough motivation to learn the language. Once activated, the psychological structure becomes involved in three situations of IL use: (1) when the language learner attempts to express meanings, which he or she may already have, in the TL, (2) when he or she attempts to understand TL utterances, and (3) when he or she actually interacts with speakers of the TL. When the latter situation is taken to the satisfaction of the learner, he or she runs the risk of fossilizing some part of IL. In a research study conducted by Perdue (1984), it was found that immigrant workers in the host TL society "acquire what is most urgently needed, and some even attain a certain fluency. But normally their acquisition slows down and even stops at a level that is far removed from the language of the world they have to live in" (cf. Perdue, op. cit.:19) Fossilization is a natural phenomenon characteristic of IL. It refers to the situation where deviant linguistic structures arise and the learning process comes to a halt, thus causing erroneous linguistic features to linger in the speech of the language learner. Fossilization takes place irrespective of the level of linguistic attainment achieved. Therefore, fossilizable structures may at any time surface in the linguistic production of IL speakers even when thought to have been eradicated. Since fossilization is a phenomenon characteristic of IL, then it is inevitable that language learners fossilize as it is inconceivable that they can learn the TL at once. So, while their version of the TL is still in embryo, there is a great chance for fossilization to occur in their IL. In trying to account for this phenomenon, Selinker (1972) claims that it owes its existence to an LPS. In fact, in his intriguing 1992 book, Rediscovering Interlanguage, he maintains that "if all, or even most IL speakers fossilize, then it is clear that we must assume that they are preprogrammed to do so and this can be explained, as far as I can tell, only by assuming a latent psychological structure of some sort." He goes on to say that "it is necessary to postulate this latent psychological structure, primarily because no matter what learners do, they fossilize some part of an IL, often far from TL norms" (p. : 33 emphasis in the original). Following Weinreich (1953), Selinker (1992) recognizes "interlingual identifications" as central in IL study. These are what language learners identify as equivalent across linguistic systems. Weinreich and Selinker postulate that units of interlingual identifications exist in the LPS. This being the case,
3

these units purport to be latent in the cerebral system. Moreover, the LPS helps the learner establish equivalencies across three linguistic systems, that is, the two languages in contact, i.e. the NL and (the learner's view of) the TL, and the ensuing IL. This has largely to do with the question of how the language learner gets started. Selinker (op. cit. : 34) holds that "within this latent psychological structure, the NL both is and is not the starting point for IL learning", the NL, on the one hand, not being the line of departure because the first-day exposure to the TL helps to reactivate the LPS, which, in turn, helps the learner to set up interlingual identifications, and it, on the other hand, not being the starting point because of the attested occurrence of early fossilization. Selinker (1972) further claims that the LPS encompasses five of what he considers to be central processes for second language learning. Such processes, which he describes, it should be noted, as psycholinguistic, are hypothesized to be responsible for the development of IL. Seliger (1988) sums up those five psycholinguistic processes and presents them succinctly thus: 1- Language transfer, that is, the transference of rules from the learner's first language (L1) are used to produce utterances in the second language (L2). 2- Transfer of training that resulted from the learner's being overdrilled in a particular form in the second language class. 3- Strategies of second language learning. 4- Strategies of second language communication. 5- Overgeneralization of target language linguistic material. (From Beebe 1988, p: 21) Selinker (op. cit.) points out that each one of these processes "forces fossilizable material upon surface IL utterances, controlling to a very large extent the surface structures of these utterances." More strikingly, processes operating in tandem may engender what Selinker terms an "entirely fossilized IL competence." Seliger (1988) has a different view of the processes involved in the development of IL. He contends that Selinker's proposed processes are not in fact processes, nor are they central as Selinker claims. Seliger (op. cit.) argues that all those processes can be encapsulated in one major process: ...language transfer, transfer of training and overgeneralization can all be categorized as types of some generic form of generalization, [which] (..) can itself be considered a form of hypothesis testing. Selinker's process, called "strategies for second language communication", may or may not be concerned with actual acquisition and may represent rather specialized behaviors for functioning in second language contexts, Selinker's "second language learning strategies" could be fit into the basic categories of hypothesis testing and generalization. Thus, for Seliger (op. cit.), there is only one central process responsible for second language acquisition. This process, which he terms hypothesis testing, comprises five steps: 1- Identifying the characteristics of a particular target concept or observing a relationship between a new form and one already learned. 2- Forming a hypothesis based on that identification or observation. 3- Testing the hypothesis by producing an utterance or listening for a similar example. 4- Receiving feedback on the hypothesis.
4

5- Deciding whether to continue accepting this hypothesis or to reject it on the basis of the feedback. As afore-mentioned IL is a transitional system in constant state of flux. Thereupon, we should expect the language learner to keep modifying his or her linguistic store every now and then in the light of newly learned material. Ausubel (1968) and Levine (1975) have also made reference to hypothesis testing and characterized it as the major learning process for second language acquisition. As reported in Seliger (op. cit.), Ausubel pertinently asks the question: "what happens to this newly acquired knowledge once it becomes attached to already existing knowledge?" In an attempt to grapple with this issue, Ausubel (op. cit.) proposes that once the language learner receives positive feedback on the hypotheses he or she has constructed, the ensuing linguistic knowledge becomes subsumed under already existing relevant knowledge. This may result, as the process of hypothesis testing unfolds, in either modifying or obliterating what the learner already knows about the linguistic feature under consideration. More will be said about this later on when we deal with variation in IL. In the next section, we will touch upon elicitation procedures for the study of IL. 1.3 The Elicitation of IL: a methodological approach: The purpose of eliciting IL being to study some aspects of the learner language, the IL investigation may choose one of two elicitation procedures: (i) getting the respondent to produce linguistic data in his or her IL, or (ii) guiding the respondent to producing specific IL features. The former procedure is mainly meant to acquaint the researcher with some raw data and it is barely intended to form an overall picture of IL by letting the learner free in choosing his or her words. The latter procedure, however, is so controlled that it constrains the learner and so selective that it helps the investigator to test the hypotheses he or she has made about particular aspects of an IL. Corder (1981) refers to these two elicitation procedures as clinical and experimental respectively and these "form the basic techniques used by any linguist investigating particular manifestations of language child language, interlanguage, or institutionalized language." Be that as it may, as Corder (1981) points out that Elicitation procedures are used to find out something specific about the learner's language, not just to get him [or her] to talk freely. To do this, constraints must be placed on the learner so that he [or she] is forced to make choices within a severely restricted area of his [or her] phonological lexical, or syntactic competence. (Corder 1981, p: 61) Whether the investigator follows this or that procedure (which may take the form of tests, though not always), the outcome is merely textual data, which will obviously not constitute a substantive reflection of the learner language, as such data will hardly provide a glimpse of the learners' intuitions about their ILs. The researcher then need not content himself or herself with textual (or performance) data as they cannot by any means provide a representative sample of the learner language. Therefore, to remedy this state of affairs, the investigator will have to find ways of supplementing textual corpuses by intuitional (or retrospective) data. As a native speaker of IL, the language learner should be able to make judgments about the grammatical acceptability of a linguistic form. Such judgments are inevitably based on IL grammar; and the latter, by extension, must be in compliance with the intuitions of the language learner. The question to be relevantly
5

asked here is: how can a researcher have recourse to the learner's intuitions about the IL under study? Usually, the researcher who has once been involved in teaching a second language is in a fairly good position to make use of his or her own intuitions about IL as he or she is likely to have the same NL as his or her students and so he or she has been sometimes in the past a native speaker of his or her students' IL(s). Another important way of making the learner express his or her intuition of IL is to get him or her to use IL to talk about IL. This metalanguage, Corder (1981) holds, is needed not only to "understand the instructions given in the elicitation exercises" but also, and more importantly, to report the respondent's "introspection or intuitions about the nature of his [or her] language, its categories and systems." It should be clear by now that any study of the learner language has to incorporate both textual and intuitional data for it to be descriptively adequate. This is so because textual data alone are not reliable to give a comprehensive account of the learner language as it may not constitute a representative sample. The textual data themselves are not spontaneously produced in natural conditions or natural communicative situations. Perdue (1984) describes the momentous part that introspective/intuitional data play in enhancing the picture of the factors shaping IL and refers to it as a kind of triangulation." Up to this point, we have been considering the kinds of data that are relevant to IL study. This could be enough were it not that IL is characterized by constant change over time. This very state of affairs calls for another way of looking at IL data elicitation. Research on IL should be conducted in such a way that all factors that come into play in the construction of IL are captured in the IL data analysis. So, one question that should be asked by the researcher is whether to study IL as it develops through time in the learning career of one or a group of respondents, or to study it at a particular point in the continuum of IL evolution. Putting this another way, a researcher should decide whether he or she is interested in studying IL diachronically or synchronically. Such a decision will directly influence the way data are to be collected. Therefore, the researcher can gather IL data longitudinally (i.e. over a lapse of time from the same respondents) or crosssectionally (i.e. at a particular point in the IL continuum). Researchers doing longitudinal studies take a small group of language learners as respondents whereas those doing cross-sectional studies take many learners as respondents in order to get a wide array of data. After all, it seems that a longitudinal study of IL is more comprehensive and more invaluable than a cross-sectional study in that it provides an overall coherent picture of IL development, thus highlighting crucial characteristics of the learner language such as variation, systematicity, error making etc. Meisel et al. (1981) make a comparison between these two instruments of data collection, emphasizing the value of a longitudinal study of IL: the procedure generally followed [in a cross-sectional study] (...) is to analyze the linguistic performance of a number of L2 learners at a certain point in time. The results are then interpreted in one of two (...) ways: (a) either one tries to find out which areas of the grammar of the target language present the most difficulties of learner the through counting the number of errors, or (b) one attempt to give the learner's distance from the standard norm in terms of grammatical complexity. [On the other hand, a longitudinal study] gives a description of linguistic performance at several points of time, analyzing changes which occur between [time one] (...) [time two] ... (Meisel et al. 1981, pp: 111, 113)

For the purposes of arriving at the optimal picture of the forces shaping IL, we have to consider, in our task of data elicitation, all the factors involved in the variability and systematicity of IL. This we are going to discuss in the subsequent sections. 1.4 Summary and Conclusion The main thrust of this section has been to pay a look at the learner language. We have seen that IL comes as a result of a language contact situation where the learner, by choice or otherwise, has to acquire a second language for him or her to interact with the speakers of the TL. We have also touched upon the factors that come into play in the realization of IL. We have considered Selinker's (1972) five psycholinguistic processes, which he claimed are part of a latent psychological structure, and Seliger's (1988) hypothesis testing as two ways of accounting for IL development. We then moved on to touch upon some methodological considerations in IL data elicitation, stressing the importance of incorporating both introspective and performance data and the preponderance of a longitudinal study of IL over a cross-sectional one. This section has examined only some of the research findings considered to be important background for dealing with the issues under investigation, viz. Systematicity and variability. It is hoped that they provide insights into second language acquisition as a whole and IL study in particular. Our next concern then in the following section will focus mainly on the systematic aspects of IL, leaving aside the variable aspects until the third part of this research paper. 2 Systematicity in IL 2.0 introduction In this section, we shall examine systematicity and the way in which IL can be said to be systematic. Do identifications of regularities in the learner language show that it is systematic? Given the evolving nature of IL, is it enough to say that the learner language exhibits certain regularities and, therefore, it is systematic? Is systematicity characteristic of IL at a given point in the learning career of a second language learner or is it a distinctive feature of the whole learning process? It is hoped that our attempts to answer these questions will contribute to the crystallization of our views about IL. 2.1 Native Speaker Language Systematicity Before proceeding to consider systematicity in IL, it might be fruitful to discuss how researchers have dealt with systematicity in institutionalized language. By this means, we might eventually find out the way in which external languages such as NL and TL impact on the learning process of second language. Tarone (1988) considers three approaches to identifying systematicity in language: an ethnographic, a rationalist, and a function-form, each of which looks at systematicity in different ways. 2.1.1 The Ethnographic Approach The ethnographic approach, which recognizes Gumperz and Hymes (1972) as its mentors, tries to account for all aspects of linguistic data to the effect that every item is assumed to be systematic unless proven otherwise. Pauses, incomplete sentences, fillers, interruptions etc. would be relevant parts of data. Moreover, an ethnographic approach to studying systematicity in language would consider some paralinguistic factors such as facial expression, eye contact, gesture and
7

posture as an integral part of the data to be analyzed. In this approach, every bit of information that could help in clarifying the content of the data or that is an any way relevant to the phenomena associated with the linguistic data is to be taken into account. 2.1.2 The Rationalist Approach The rationalist approach, which recognizes Chomsky (1968) as its mentor, is more restrictive in selecting data. "By-products" of linguistic performance such as slips of the tongue, false starts, hesitations etc. are not considered relevant because they do not reflect the speaker's competence. Chomsky and his followers argue that a good theory of language is the one that accounts for the data that are produced in an ideal speaker-hearer situation or the data that native speakers of the language would judge as being grammatical or not. The purpose is to attempt to describe the native speaker's competence which underlies his or her performance. Moreover, no matter what a native speaker does, his or her linguistic production, especially oral, may not portray his or her competence. Therefore, the latter can only be accessible through introspection, that is, through judgments of grammaticality. In short, the linguist, who may also be a native speaker of the language under consideration, uses the sentences that have been judged/proven to be grammatical to devise a model of his or her competence. It should be noted that this approach would predict more coherently systematic language data, especially in grammaticality judgments rather than in the actual linguistic performance. 2.1.3 The 'Function-Form' Approach The 'function-form' approach, which recognizes Sampson (1980) as its mentor, contrasts the rationalist view that introspection is the only means by which we can have access to the linguistic competence underlying language performance. The purpose of this approach is to account for the systematic relationship holding between language form and function in actual communicative situations. Language forms are believed to be used to serve certain functions. While from the sociolinguistic perspective, language form functions to signal social class, gender, social identity, race etc., from the pragmatic perspective, it functions to signal agency, topic, shared knowledge, old vs. new information etc. As for systematicity, it is claimed to be found in variation itself. Labov (1971), for example, holds that language variation involves both systematicity and unsystematicity. This aspect of variation will be discussed further later on when we deal with variability. 2.2 IL Speaker Language Systematicity Having examined some theoretical claims made by researchers in explaining systematicity in language, let us now focus on how systematicity has been identified in the linguistic production of second language learners. The latter have been observed to be influenced by their NL so much so that they occasionally make use of NL items, features, structure, rules etc. in their attempt to express certain meanings in heir IL. This, second language learners do in a systematic manner. Even the errors that they make are systematic if not, ad nauseam, predictable (Lado, 1957), the causes of error-making being attributed to, inter alia, the transfer of linguistic knowledge from the mother-tongue (Richards, 1971). From the TL perspective, IL production should be composed of two types of utterances: well-formed ones and ill-formed ones. (Recall that we have argued for the correctness of all IL production from the point of view of the IL speaker.) Therefore, when we say that IL is systematic, systematicity should characterize both types of utterances. Corder (1981) further subdivides the ill-formed type into two major sub-component: errors of performance,
8

which are unsystematic, and the errors of competence, which are systematic. As was noted earlier, a number of procedures have been proposed to define what is meant by systematicity in TL. Among these, we can cite three: (1) the cognitive/psycholinguistic approach, influenced by Chomsky and expounded by Corder (1971; 1973), Dulay and Burt (1973; 1974a, b and competence), Krashen (1981; 1982) among others; (2) the sociolinguistic approach, influenced by Labov (1971; 1972) and expounded by Tarone, Frauenfelder and Selinker (1976); and (3) the 'function-form' approach expounded by Hakuta (1975; 1976) and Huebner (1983a and b; 1985) among others. 2.2.1 The Cognitive/Psycholinguistic Approach Corder (1971) holds that IL is systematic in the sense that the learner's linguistic performance, which he terms "an idiosyncratic dialect", presupposes the existence of a system. For him, IL is "systematic because it has its own rules." To put it differently, the language produced by learners in the course of acquiring a second language is in fact a full-fledged linguistic system which has to be studied in its own right, not through the lens of the TL or NL. As Nemser (1971) puts it, the learner language, or the "approximative system" as he calls it, is "structurally organized, manifesting the order and cohesiveness of a system." Likewise, Selinker (1972) maintains that IL is systematic in the sense that it is internally consistent, self-contained and rule governed. Therefore, to account for the regularity and organization found in the language produced by second language learners, we need to posit the existence of a system independent of that of the NL or TL. In fact, IL is, as a result of its unremitting dynamism, an innovative or creative system in that it has rules unique to itself, not just borrowed from the NL. Nemser (op. cit.) advocates that there is "frequent and systematic occurrence in non-native speech of elements not directly attributable to either NL or TL." The corollary of this is that those elements have been introduced by the learner into his or her IL system. In other words, items or rules which cannot be traced to either NL or TL are to be extrapolated as being created by the language learner. Accordingly, IL is a creative system by virtue of its structural independence. Dulay and Burt (1973; 1974a, b & c), following the Chomskian approach to child language acquisition, claim that children who are in the course of acquiring a second language reactivate what Chomsky hypothesized as the "Language Acquisition Device" (LAD). In plain words, children internalize their first and second languages in a similar fashion. For Dulay and Burt (op. cit.), "learner language was systematic because large groups of learners could be shown to acquire elements of the target language in a statistically verifiable sequence." They contend that no NL transfer to IL has been attested and that children in the process of acquiring their L1 or L2 internalize bits of language incrementally by following a well-organized schedule. As far as data elicitation is concerned, the cognitive/psycholinguistic approach recommends that data be composed of both the learner's actual linguistic performance and his or her judgments of grammaticality. Such a mixture of data is believed to pay dividends inasmuch as it enhances the depiction of the learner's linguistic competence. 2.2.2 The Sociolinguistic Approach Following Labov (1971; 1972), Tarone et al. (1976) claim that IL is systematic by virtue of its demonstration of "internal consistency in the use of forms." The quantitative method of the Labovian sociolinguistic approach is based on the principle of accountability. The latter principle
9

has been adopted by Tarone et al. (op. cit.) as a yardstick to measure the degree of internal consistency (i.e. systematicity) in IL. Labov sums up this notion and concludes that: ...any variable form (a member of a set of alternative ways of "saying the same thing") should be reported with the proposition of cases in which the form did occur in the relevant environment, compared to the total number of cases in which it might have occurred. (Labov 1973 : 94; emphasis in the original) Tarone et al. (op cit.) argue that all predictable forms are systematic, regardless of the degree of accuracy. In other words, whether a rule predicts a given form 90 percent or 20 percent accurately, the form is said to be systematic. Therefore, systematicity in this approach is a rulegoverned phenomenon characteristic of the speech of second language learners. 2.2.3 The 'Function-Form' Approach In his 1976 study of IL development, Hakuta succeeded in obtaining a clear picture of the relationship between language form and function. This work reports a longitudinal study of a fiveyear-old Japanese child learning English as a second language. Among the findings of this study, we can mention the proven relevance of language function to the identification of what Hakuta describes as "prefabricated patterns" in the child's IL, the preponderant treatment of IL as a whole continuum rather than a fragmented body, and the occupation of new forms in IL the place of old forms whose use seems to be increasingly dwindling, the new forms "taking over" some of the old form's functions (cf. Hakuta 1976). Huebner (1985), who also adheres to the 'function-form approach', makes three assumptions in setting up IL systematicity: (1) learners must make a set of functional distinctions (e.g. agent, object, old vs. new information etc.), (2) learners must make use of linguistic devices (forms) in a systematic way, and (3) IL is likely to exhibit variation at a given point in time since it is in constant diachronic change. As it can be noticed, treatment of systematicity is inextricably bound up with variation in IL to be reviewed in the following section. The reason for this is that systematicity and variation are two distinct features of one entity. The latter (i.e. IL) is a system which has posed a problem for applied linguistic theory in that it has to account for IL as an evolving continuum and at the same time as "stable states". Corder (1981) points out that data culled from IL in such a state of affairs are likely to lack the sought internal consistency. For this reason, the notion of systematicity has been called into question by many an investigator. They contend that the term "systematic" is inappropriate when applied to IL. Corder suggests that ...much of the apparent inconsistency is an artifact of the theoretical models we are forced to use to describe the [IL] data. In other words, the learner's approximative systems merge gradually into each other rather than switch from one discrete state to the next. (Corder 1981 : 68) If what is meant by systematicity has to do with the fact that IL is a system independent from NL and TL, then the coexistence of systematicity and variability seems to be justified. In fact, James (1974) pertinently asks: "How can a system remain a system if it is in flux?" In the next, section we shall deal with variation in IL and, not surprisingly, the systematicity of this variation.
10

2.3 Summary and Conclusion The upshot of this part has been to characterize IL as being systematic in the way it is at a single point in time and in the way it evolves. Its prime concern was to review how different theoretical approaches have defined systematicity. Prior to this, we also touched upon some research considered to be important background for dealing with this particular issue of systematicity. So, we surveyed the theoretical claims and assumptions made by investigators in institutionalized languages. The motivation for having this subsection was to pave the way for our examination of IL systematicity. The most salient approaches to the study of IL systematicity were then sketched. It should be clear from the discussion in this part that we are still a long way from understanding the phenomenon of systematicity in IL. This is because, as was mentioned earlier, this phenomenon is closely linked with the other distinctive feature of IL, namely variation, on which we elaborate next. 3 Variability in IL 3.0 Introduction That IL is consistent in certain respects and inconsistent in others has now become an established fact. The consistent aspects of IL we have called systematicity and the inconsistent aspects variability. The fluctuation between the former and the latter is referred to as variation. When we talk about variability, we deal with variables, but when we talk about variation, we deal with variants. Because of the overlap between variability and variation, most researchers tend to use these two terms interchangeably (e.g. Ellis and Roberts 1987, Tarone 1988, among others). For Ellis and Roberts (1987), IL variability can be studied from two distinct but slightly overlapping perspectives: one, the synchronic perspective whose focus is the variability existing in the learner language at a given point in time; two, the diachronic perspective whose focus is the evolving nature of the learner language between two or more points in time. The task of the investigator, thereafter, is to pin down the relationship obtaining between "variability on the horizontal axis (i.e. diachronic variability) and variability on the vertical axis (i.e. synchronic variability)." 3.2 Types of IL Variability Ellis (1985b) proposes that there are two major types of variability in language-learner language, namely systematic and nonsystematic. The latter type, also referred to as free variation, is "haphazard" (Ellis 1987), "random" (Tarone 1988) and "unpredictable" (Ellis and Roberts 1987). If we were to accept the view of Adjmian (1976), which assumes that IL is to be treated as a natural language, then we should give credit to Downes's (1984) claim that "heterogeneity is the normal state of language." As for systematic variability, it is further subdivided into individual variability, which is characteristic of every individual learner, and contextual variability, which is sensitive to certain linguistic and paralinguistic variables. The former type of systematic variability (i.e. individual variability) refers to the different tasks (linguistic or otherwise) that different learners carry out differently. This kind of variation arises, according to Ellis (1985b), as a result of individual learner factors such as attitude and motivation. The contextual variation is, in turn, shaped by either one of two contexts, situational or linguistic. Situational context has to do with social variables that are pervasive in the speech situation such as the interlocutor, with whom the learner interacts, the topic of interaction between the learner and the interlocutor, the setting where the interaction takes place, the nature of the task which the learner is asked to perform,
11

and the social norms internal to the learner like attitudes and external to the learner like codes of behavior. Linguistic context, on the other hand, has to do with language variables which determine the language forms (linguistic structures with more than one variant) used by learners.

Types of variability in language-learner language (from Ellis 1985b:76) Given this multiplicity in types of variability, it becomes difficult to give one monolithic definition to variation. Variation can mean a number of different things: (1) variation across learners: i.e. learners go about learning a language in different ways; (2) variation across individual styles, registers or tasks; or (3) variation within the same sample of interlanguage of an individual speaker as well as over time. (Anderson 1989:46) Leaving aside the unresolved question of how to subsume all these three lines of investigation under the umbrella of a single integrated line of research, we examine in this research paper IL as has been defined by Labov (1969), namely that it refers to "different ways of saying the same thing." In the following subsection, we shall consider the major theories set forth to account for variability in IL. 3.3 Theories of IL Variability In her 1988 study, Tarone provides a critical account of the current theoretical models that have been proposed to account for variation in IL. Tarone suggests that this myriad of models can be classified into two major categories. The first group, which she calls "inner processing theories", consists of approaches with a psycholinguistic orientation. These theories ascribe the causes of variability to various psychological processes. The second group of theories, referred to as "sociolinguistic and discourse theories", consists of approaches which draw upon sociolinguistics, social psychology and pragmatics. Such theories, in contrast to the former, tend to attribute the causes of variability to factors external to the learner.
12

Before setting about the task of examining the theories per se, it would be worthwhile to start with the consideration of the criteria set forth by Tarone (1989) for evaluating the theories of variation in IL to be sketched here. She maintains that three major criteria must be met for a theory to be described as scientific, hence objective. First, the "proposed causes of systematic variation must be empirically verifiable." Second, known facts and causes of IL variability must be thoroughly explained and ultimately predicted. Third, in the absence of empirical data, a theory of IL variation must meet some "aesthetic" (Schaumann, 1983) standards like being "internally consistent, parsimonious, and elegant." For further details on the criteria for evaluating IL variation theories, see Tarone (1988; 1989). 3.3.1 Inner Processing Theories As afore-mentioned, theories coming under the heading of this group impute the causes of variation in IL to various psychological processes. Among these we can cite parameter setting of core grammar, attention paid to language form rather than content, automaticity and monitoring. Tarone (1988) claims that all these theories will face much difficulty in meeting the first criterion set for evaluating IL variation theories. The difficulty has to do with the fact that it is almost impossible to prove empirically that mental processes operate in a way or the other: We have no direct empirical window on the working of the mind, and thus cannot correlate inherently unmeasurable mental processes with observable subjective or objective data on varying phenomena in interlanguage production. Be that as it may, unobservable mental processes can be extrapolated on the basis of their observable outcome. In what follows, we deal succinctly with four major "inner processing" theories, namely the Monitor Model of Krashen (1981; 1982), the Chomskian models of Adjmian (1982) and Liceras (1985), the cognitive models of Bialistok (1982), Bialistok and Sharwood-Smith (1985) and Ellis (1985a), and the Labovian model of Dickerson (1974; 1975) and Tarone (1983). 3.3.1.1 The Monitor Model We will not attempt here to survey, however briefly, Krashen's Monitor Model. We feel that treatment of second-language acquisition and teaching, for which it has been designed, is not a profitable use of space in a research paper on IL variation. This space can be used better to check on the adequacy of the model. In lieu of such discussion, we will also concentrate on the predictions that the Monitor Model makes about the nature of IL variability. The Monitor Model is based, among other things, on the premise that the linguistic knowledge that a learner internalizes consists of two parts: an 'acquired' part, which comprises linguistic knowledge internalized unconsciously, and a 'learned' part, which comprises linguistic knowledge internalized consciously. Additionally, the Monitor is explicit, learned knowledge of the TL rules. A second-language learner operates the Monitor under three major conditions: (1) if there is enough time for Monitor intervention, (2) if the learner focuses on the formal aspect of their linguistic production and (3) if the learner is aware of the rule governing the structure being used and able to state it. Only when these three conditions are met may the learner make use of the Monitor to modify the output of the acquired knowledge system. It is this modification that induces us to talk about the Monitor because it is at the hub of IL variation.

13

Central to this Model is a dichotomy to be established between Monitoring (with a big 'M'), where the language learner corrects himself or herself by means of the rules of which he or she is aware (learned, conscious linguistic knowledge), and 'monitoring' (with a small 'm'), where the learner corrects himself or herself by means of intuition or feel (acquired, unconscious linguistic knowledge). The latter kind of self-correction is peripheral to the focal point of the Monitor Model. Variation characterizes the linguistic output of the learner depending on Monitoring. In other words, the learner's performance varies according to the use or non)use of the Monitor. Then, the primary concern of an IL variation theory which draws upon the Monitor Model has to be drawing a distinction between "learner performance under very constrained circumstances, and learner performance under all other conditions." (Tarone 1988, emphasis in the original). To conclude, then, the Monitor Model traces IL variability to learner's use or non-use of the Monitor. 3.3.1.2 The Chomskian Models As we have already pointed out in the previous part of this research paper when dealing with systematicity, introspective data on grammaticality judgments of the language learner are rather central in the study of IL as a whole. The purpose of such models as those couched within the Chomskian slant is to account for the linguistic performance of the ideal speaker-hearer and to map such performance onto his or her competence which is assumed to be composed of homogeneous linguistic knowledge. Initially, Adjmian (1976; 1982) and Liceras (1981) thought that variation in IL is primarily due to performance error. For them, the learner's linguistic competence becomes subject to the intrusion of rules alien to its internal make-up, "or the overgeneralization or distortion of an IL rule" (Adjmian, 1976). This phenomenon is called permeability. The assumption of the original formulation of this model was that the introspective data lacked variability and that only performance data could be described as variable. However, Tarone (1983) discredits this, for she finds out that there is more permeability in learner's introspective data than in their performance. In a revamped version of this model, Liceras (1987) recognizes that variability characterizes both IL competence (intuition) and IL performance (textual and verbal production). What this means is that the Chomskian assumption that competence is a homogenous entity is no longer valid. "Variable intuitions result from the fact that there may be rules or parameters of core grammar that will be fixed in a variety of ways or not fixed at all." (Liceras, op. cit.) Liceras provides a definition of "permeability" in terms of Universal Grammar: ...permeability can be defined as the abstract property of all grammars that accounts for speakers' variable intuitions ... It is a consequence of the fact that universal grammar contains a system of rules which presents pragmatic variation. (Liceras, 1987 : 357) Even though Liceras (op. cit.) has shown that competence is permeable, she attributes differences in accuracy of performance on different tasks to performance, not to competence. However, this seems to contradict the criterion we set at the outset for evaluating the theories, namely that this model lacks internal consistency in terms of competence-performance distinction (cf. Tarone 1988).

14

3.3.1.3 The Cognitive Models This trend of 'inner processing' theories is based on research on human information processing. Basic to this slant is a distinction to be made between knowledge and the process of knowledge use in actual communicative interaction. Concurrently, MacLaughlin (1978) proposes to make a difference between two sorts of processing: 'controlled' processing and 'automatic' processing. This distinction is "based on behavioral acts, not on inner states of consciousness." In its approach to IL variability, this slant attempts to answer the question why second language learners can use a given form in simple tasks (in the classroom for example) but unable to use the same form in communicative tasks. The answer that has been advanced to justify this state of affairs is that the learner first uses a TL form by means of controlled processes, and only when this form is completely internalized through practice may the language learner use it instantaneously by means of automatic processes. A detailed description of how this process takes place is taken verbatim from Tarone: ... interlanguage production would initially take place by means of controlled processes, which requires active attention on the part of the learner. A controlled process activates a temporary sequence of connections in short-term memory, and requires so much attention that only one sequence can be successfully produced at a time. As the sequence is practiced, it becomes familiar and can eventually be done via automatic processes. Automatic processes activate a relatively permanent set of associate connections in longterm memory. These processes do not require attention; are difficult to suppress or alter, once in place; and can be used to activate several sequences at once. (Tarone, 1988 : 34) In addition to this, Bialistok (1982) proposes that one's linguistic knowledge may be 'analyzed' or 'unanalyzed'. Analyzed knowledge is that of which the learner has a mental representation which enables him or her to visualize its structure. Thanks to this mental representation, the learner has immediate and easy access to his or her analyzed linguistic knowledge, which he or she may employ on different tasks and modify in various ways. Unanalyzed knowledge, however, is that overall mental structure of which the learner is unaware such as grammatical judgments. In an attempt to account for IL variability, Bialistok (op. cit.) claims that a learner's linguistic production becomes variable inasmuch as he or she uses either the knowledge system or the control system in different tasks. For example, tasks which require the learner to act on the spot under time pressure necessitate analyzed knowledge and hence more effective use of automatic retrieval procedures than other tasks. Bialistok and Sharwood-Smith (1985), however, explain IL variation differently. They attribute diachronic variation to the change taking place in the linguistic knowledge of the learner in the IL development continuum, and synchronic variation to "the processing constraints that operate on the learner's knowledge of the system." The latter type is called 'control variability'. Nevertheless, this version of the approach to IL variability runs contrary to the stipulation of our third criterion as it makes the distinction between knowledge and control rather blurred when it tries to show that control procedures affect knowledge in some way. The picture becomes more confounded when Bialistok and Sharwood-Smith (op. cit.) claim that control procedures can 'know' what forms to be used and that variation determines pragmatic competence.
15

3.3.1.4 The 'Labovian' Models The 'Labovian' Models, exemplified in the work of Labov (1966; 1969; 1970; 1971; 1972) and the Dickersons (1974; 1975; 1976; 1977), Beebe (1980), and Tarone (1979; 1982; 1983), base their argumentation on the variance exhibited by language learners in their attention to language form when performing different tasks. Labov's aim has been to pin down the systematicity which has been perceived to be at the hub of style shifting in communicative situations. In so doing, he manipulates the communicative situation to the effect that every part relevant to it is likely to affect attention to speech, thus causing attention shifts, which, in turn, give rise to style shifts. Moreover, while he makes use of different elicitation procedures to get different styles, he has to grapple with the issue of the respondents being systematically observed when they talk. This issue he terms the "observer's paradox". He maintains that "the aim of linguistic research in the community must be to find out how people talk when they are not systematically observed; yet we can only obtain this (sic) data by systematic observation." (cf. Labov, 1970). Whereas Labov (op. cit.) attributes IL variation ,or "style shifting" as he refers to it, to change in social situation, topic, setting, the identity of the interlocutor, Tarone (1983) attributes synchronic task-related IL variation to the degree of attention paid to language form. She point out that different tasks engender different styles and that there are two ways in which new forms can be incorporated into the IL continuum: when they are spontaneously produced in informal styles and then spread to formal styles; or the other way round, that is, forms initially appearing in the formal styles of learners paying attention to speech are spread to the formal styles. Dickerson (1974) elaborates on this spread in compliance with the notion of "staging", by means of which newly introduced language forms in the formal style in a given linguistic environment spread gradually to other linguistic environments. Second language learners tend to set up an order for those environments and respect it. As learners move along the IL continuum towards TLlike competence, their formal styles consist of the greatest number of target-like variants while less formal styles become gradually bereft of such variants. By way of conclusion, the Labovian models discussed in the forgoing have contributed to our understanding of IL variation, which, it is claimed, is caused by style-shifting, which, in turn, is caused by the amount of attention paid to speech. In the next subsection, we shall examine the second paradigm in the theories of IL variation, namely the 'sociolinguistic and discourse' paradigm. Theories in this rubric are largely influenced by research findings in sociolinguistics, social psychology and discourse analysis/pragmatics. 3.3.2 'Sociolinguistic and Discourse' Theories As we have mentioned earlier, theories coming under the heading of this group ascribe IL variation to factors external to the learner. These are largely to do with the impact of the social surroundings of the learner. Such factors include the learner's social distance from speakers of the TL, degree of acculturation to the target culture (Schumann, 1978), the learner's attitudes towards and motivation to learn the TL (Gardner, 1985), the identity of the interlocutor, the topic of discussion (Bell, 1984) and so on. In what follows, we try to sketch some of the most salient research orientations which we are going to classify into two major paradigms: 'socialpsychological' models and 'function-form' models.
16

3.3.2.1 'Social-psychological' Models At least four eminent adherents of this trend are worthy of being mentioned: Meisel et al.'s (1981) multi-dimentional model, Littlewood's (1981) model, Beebe's and Giles's (1984) Speech Accommodation Theory, and Selinker's and Douglas's (1985) 'discourse domains' model. Meisel et al. (op. cit.) designed a theory of second language acquisition which takes into consideration the role of the learner's attitude towards and motivation to learn the TL, without overlooking the importance of the social distance of the learner from TL speakers and society. This model claims that the same language structures may be used differently by different learners depending on their socio-psychological characteristics. Such structures are said to be, it is claimed, influenced by linguistic environment. Building on the proposal of Meisel et al. (Op. cit.), Littlewood (1981) comes to the conclusion that there are three factors behind IL variation, namely functions of linguistic forms, influence of linguistic environment, and social-situational factors. For him, three types of language learners can be recognized: learners lacking norms, learners with pedagogical norms only, and learners with social norms. Speech Accommodation Theory, which has been developed by a large number of scholars, the most outstandingly prominent of whom are Beebe and Giles (1984). This theory, which is, it should be noted, more elaborate than the other theories of this paradigm, seeks to explain why the interlanguage production of second language learners is variable and to what extent factors such as feelings, values and motives have a hand in speech variation among learners in interaction. In this approach, style-shifting is ascribed to the adjustments language learners make when interacting with others, be they TL learners or TL native speakers. Those who adjust their speech to become similar to that of their interlocutors are said to converge and those who adjust their speech to become less similar than that of their interlocutors are said to diverge, the two cases being referred to as convergence and divergence respectively. Speakers make such speech adjustments depending on social psychological factors such as intergroup distinctiveness and identity assertion (Beebe, 1982). Language and the type of style used are very crucial in identifying oneself with a social group or another. Therefore, speakers, language-learners included, change their styles in an attempt to assert their identity in a situation where they come into contact with members of other social classes or speech communities. Beebe (op. cit.) holds that language is "both a powerful symbol and a common vehicle for expressing group loyalty." In short, Speech Accommodation Theory provides a principled explanation, though not exhaustive, for the variability that occurs in the speech of second language learners. Beebe and Giles (op. cit.) summarizes the factors affecting such variability as follows: 1 linguistic context (a) environment - the effect of the form of one part of one's speech upon another part, at phonological, grammatical and discourse levels (b) input - the effect of the form of the interlocutor's speech upon that of the learner 2 linguistic proficiency (a) status accorded an interlocutor with higher proficiency in the target language (b) limited repertoire - the presence or absence of the learner's ability in that second language (c) background - the previous language experience of the speaker. (Beeb and Giles, 1984, quoted from Tarone, 1988:51)
17

Nevertheless, not all variation in IL has to do with accommodation. Beebe and Giles do not downplay the importance of factors such as conforming to norms, negotiation of one's identity, and the drive for positive, competent self-representation. The last model under the social-psychological paradigm is referred to as the 'discourse domains' model. Selinker and Douglas (1985), the founders of this approach, propose that IL varies depending on "context", or "discourse domains", defined by the learner, style shifting taking place as he or she moves from one such discourse domain to another. In this approach, two types of data are used: "primary data, consisting of recordings of conversations in contexts selected by the learner, and secondary data, consisting of the learner's comments on the primary data" (Tarone, 1988:52). Given the obscurity of the notion "discourse domain", which, although claimed to be defined by the individual learner, has to be delineated by the investigator, this model did not receive wide support among SLA researchers. 3.3.2.2 'Function-Form' Models In this paradigm, the interplay of language form and function is claimed to be at the root of IL variation. Proponents of this approach include Hakuta (1975 and 1976), Huebner (1983a, b and 1985), Tarone (1985) and Schachter (1986). Since we the assumption of this approach have already been discussed when dealing with IL systematicity, we move on directly to consider the implication of the interaction of form and function in the variability occurring in the speech of second language learners. When learners use language forms (be they IL forms or TL forms), they may run the risk of using certain forms to signal certain functions inappropriately in the TL communicative situations. Tarone (1988) points out that "the learner may in fact be relating TL forms to functions using a different system from that used by speakers of the TL." (p. 54) One of the assumptions in this approach, it is worth recalling, has to do with the learner's inevitable use of language forms in a systematic manner. Such systematicity in using forms is likely to contribute to the understanding of the phenomenon of variation in IL performance since the latter, and hence language in general, is used for various purposes. Thereupon, we should expect function to influence form more than the other way round. However, it is possible to find functions being misused linguistically as well as linguistic forms being appropriately used to serve the wrong function (Hakuta, 1976). Without belaboring the point, and for the sake of brevity, the function-form approach to the IL variation has it that the accuracy of a given linguistic form varies as it is used to perform different pragmatic functions in discourse. 3.4 Evaluation and Conclusion In this section, the types of variation affecting interlanguage were first enumerated, then the theories that have been proposed to account for variability in IL were considered. They were classified into two major groups, namely "inner processing" theories, which ascribe IL variation to psychological processes of various kinds, and "sociolinguistic and discourse theories", which attribute variability in the speech of language learners to various factors external to the learner such as social norms, attitudes towards the TL etc. Prior to the examination of the IL theories per se, some criteria advanced by Tarone (1988 and 1989) for evaluating those theories were considered. According to these criteria, all 'inner processing' theories are found to be wanting as they fail to specify the nature of the psychological
18

factors in an empirically verifiable way (criterion 1). The approaches within the 'inner processing' theories such as Adjmian's (1982) and Liceras's (1987) models were not successful in keeping up with the metaphors they set for themselves, thus losing their elegant and aesthetic sides (criterion 3). After all, even though some 'inner processing' theories were found to be faulted for their violation of criterion 1 and 3, a psychologically oriented explanation of IL variation seems to be essential. On the other hand, 'sociolinguistic and discourse' theories are evidenced to meet criterion 1, namely that they are empirically verifiable. Moreover, most of them fail to satisfy the requirement of criterion 3 because they are still at the stage of development, except the Speech Accommodation Theory, which provides a rather principled explanation of IL variation. Nevertheless, neither group is exhaustive in specifying the causes of variability. The approach that is adopted n the case study reported in the following part draws upon the sociolinguistic and discourse theories, which emphasize, among other things, the role of attention to form, linguistic context, topic and function-form relationships. All these factors seem to affect IL variability. In the following section, the case study of a young second language learner of English will be presented. It is hoped that the discussion in the above parts has provided insights into the issues discussed and hence a sound foundation for undertaking this study. 4 IL Systematicity and Variability: a Case Study 4.0 Introduction In this final section, evidence for the occurrence of systematicity and variability in the IL production of Youssef, a young Moroccan student of English as a foreign language at high school, is elicited. In particular, systematicity and variability in the way he asks questions in English are studied. In this respect, focus will be mainly on wh-questions, leaving aside the other types like yes/no questions, echo questions, etc. Before discussing the methodology and elicitation procedure used in setting our task, it seems that some background information about the respondent is necessary for the case study. The reason why this study is devoted only to the linguistic production of one respondent has to do with spatial and temporal considerations. Youssef is a sixteen-year-old learner of English in the first year of secondary school. He is a science stream student who has been studying English for one year at Lyce Hassan II, Rabat. He has been exposed to English for only six hours a week in the classroom. He never practices English outside the classroom and he does not take any back-up courses in any language center. Youssef is a native speaker of Moroccan Arabic and he also speaks Classical Arabic and French. 4.1 Methodology and Elicitation Procedure Given the limited scope of this research paper and mainly the time limitations within which it has to be completed, the methodology that is adopted here is a cross-sectional study of the elicited data. In fact, a longitudinal study would be more rewarding in that it takes into account the evolving nature of IL development and it is more appropriate to a single respondent. But for the above mentioned reasons, only a cross-sectional study, whose advantage lies in the fact that it provides a general pattern of how the learner asks questions at a particular point in his career as a student of English, seems to be possible. One interesting implication of this type of study is that it
19

focuses on the individual factors characteristic of variable IL production of a single respondent/learner, thus skirting around the issue of having to take into account the different factors characteristic of every individual learner/respondent. In this study, a language test was used as a technique of data elicitation. In it, the respondent is given a list of responses which are full affirmative sentences (statements) or elliptical sentences (fragments). The respondent is required to provide potential questions to the sentences on the list. Since there is no way to divert his attention from the purpose of the elicitation, data culled by this means are likely to exhibit a pattern in the way questions are asked. However, one important element that is expected, it is hoped, to pay dividends to the study has to do with the variety of wh-words to be used in the various questions. Therefore, the test has been designed in such a way that Youssef did not have to move from a set of answers requiring a particular wh-word to another. For this reason, we answers of various kinds were mixed. As for the data elicitation instrument used in this study, it is simply a language test, where the respondent is given a list of responses which are full affirmative sentences (statements) or elliptical sentences (fragments). The respondent is required to provide potential questions to the sentences on the list. Since there is no way to divert his attention from the purpose of the elicitation, data culled by this means are likely to exhibit a pattern in the way questions are asked. However, one important element that we hope will pay dividends to our study if not singled out by the respondent had to do with what wh-word to be used in such or such question. Therefore, the test has been designed in such a way that Youssef will not have to move from a set of answers requiring a particular wh-word to another. For this reason, we have decided to mingle answers of various kinds. In addition, the test that has been administered to Youssef consists of two parts. In the first one, he is asked only to provide the appropriate wh-words corresponding to the answers in a limited amount of time. In the second part, he has to provide the whole questions without being restricted by time. The motivation for having these two parts is to see whether the variable of time pressure will or will not affect his performance and, at the same time to see the ways in which accuracy rates change with the change in tasks. While the focus of the first part is on the accuracy of the wh-words provided, in the second part it is the linguistic structure of the full questions. When the respondent works out the questions at his own pace, it is likely that he pays more attention to the form of the question. By this means, it will be possible to check if he uses the right verb tense and whether he uses subject-verb inversion at all. In order to assess his performance within the confines of the adopted data elicitation instrument described above, the classical scoring procedure employed by language teachers in grammar tests is used, that is, one point for every correct item. The respondent's scores on each question are calculated on the basis of complete accuracy. For example no half a point will be given to a question in which the wh-word is supplied appropriately, but without inverting the subject-verb order. 4.2 Data Analysis and Discussion As mentioned earlier, the test consists of two parts in each one of which Youssef performs on a different task. In the first one, he is asked to provide the question words, whereas in the second he is required to provide full questions. Table 1 below indicates the respondent's performance on
20

the two tasks (numbers in the following tables and the ensuing discussion are taken from Youssef's answer sheet (the test); see appendix). Table 1: Youssef's performance on the two-task test on asking questions in English
No. who what when where why how whose 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 (+) (+) + (+) (+) (+) . (+) + (+) (+) _ _ + + . + (+) . _ + (+) + . +? _ _ (+) (+) + (+) No. who what when where why how whose 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54
+: correct wh-word | (+): correct wh-word but deviant question form; _: incorrect question; | .: the alternative wh-word

_ (+) (+) _ (+) _ (+) + (+) . + (+) (+) _ . + (+) (+) (+) _ + + . (+) _ +? (+) + _ _ .

A quick look a the table above reveals that there are more +'s than -'s, which amounts to saying that Youssef has done fairly well in the first part of the test. If we add up the number of +'s and (+)'s, we get more than 77 % of the question words used accurately. However, if we exclude (+)'s 21

here we are dealing with the second task-, we will get less than 23 % of accurately used wh-words. These figures predict that there will be more variability than systematicity in Youssef's production of questions in English. This prediction will be borne out by data to be discussed below. To get more insight into the nature of the variability exhibited on these tasks, it is important that it be approached from the point of view of function and form. While it is enough in the first task to provide just the wh-words appropriate for the answers, in the second task, the accuracy of the provided questions depended not only on the presence of the wh-words in obligatory context but also on the function of those wh-words. Table 1 shows that difference in accuracy rates of question words in obligatory contexts is significantly different on the two tasks. The data demonstrate that Youssef generally chooses the question words correctly (task 1). However, when it comes to the linguistic structure of the questions themselves (task 2), he seems to create patterns of questions depending on whether the answer starts with a preposition or whether it contains a genitive or not. Obviously, language transfer from L1 (Moroccan Arabic) or even L2 (Standard Arabic and/or French) will not help in accounting for such structures because all these languages do not involve such forms. They do not cause confusion that might occur between preposition "to" as in (14) and the infinitive "to" as in (15); nor do they have the genitive "'s" at all. Youssef's performance, by and large, on question words (task 1) is systematic. However, his performance on questions as a whole (task 2) is variable, ranging from no target-like use (examples 7, 28, 51...) to a very close approximation to the English system for quite a few examples (1, 5, 9...). it follows then that the task used to elicit data is, without doubt, highly relevant to the variation being exhibited between performance in the first task and that in the second one. It should also be noted that although Youssef has been asked to provide questions in a limited amount of time in the first task, he has managed to score fairly well. This should not be taken to mean that Youssef asks questions accurately in English, as the second task will unveil his actual performance on how he forms questions. The variable of time, rather, is to be construed as having no bearing on the respondent's performance in the fist task. Results show that time pressure led to fair accuracy rates because the respondent was required only to provide short responses (i.e. one of the seven wh-words provided), whereas focus on form and situation (task 2) culminated in less accurate production of the full questions. It could not have been possible to administer the second part of the test within time constraints because the respondent's degree of knowledge of the language would not allow him to come up with full questions to so many elliptical answers and because pressure on the respondent may block his imaginative power that would help in imagining the situations in which the provided answers could be said. On the basis of the data collected and tabulated above, Youssef, it should be recalled, tends to generally use the question words in a systematic manner. By way of illustration, he uses "who" for person in the subject position (e.g. 13) and the object position (e.g. 50). As for the question word "what", he uses it exclusively for inanimate objects, probably because he has been trained only on that usage in the classroom. This causes him to follow the footsteps of his ideal learner - i.e. the teacher - who did not vary his or her use of questions in different contexts. Although this should not be considered as the only factor producing induced errors, Youssef's IL performance has exhibited various kinds of error that can be attributed in some way or another to the learning situation. This, in fact, invokes the teacher's role in facilitating or hampering the learner's progression along the way of foreign language development. So, the excessive drilling in such a
22

form engenders the overgeneralization of the usage of "what", thus making him ignore the use of this question word to ask about somebody's profession. As a result, instead of saying "What are you?" for sentences/answers (3) and (12), and "What is your father?" for sentence/answer (52), where "what" should occur in the subject position (animate), he uses other question forms, which, although with the wh-word "what", require answers other than the ones provided. The reason for this is that he uses "what" tacitly in the object position (inanimate), unaware that it can be used otherwise. Presumably, the amount of exposure to one form at the expense of the other could be a good explanation for the difference in accuracy taking place in those examples. It seems likely that he hardly ever hears "what" being used by the teacher in the classroom to ask about somebody's profession. This is a process of transfer of training whereby the overuse of a given language form at the expense of the other(s) in the classroom engenders the learner's overgeneralization of certain forms, to the extent that other alternatives are ruled out in his or her IL. The data being relatively sparse, no matter how many questions we have, there is only a single example (28) in which "who's" is used wrongly for "whose". Moreover, while we expect the genitive question word to be used in at least four examples (28, 33, 46 and 51), no question with "whose" has been attested in the respondent's performance. It is high probable that Youssef has failed to form a pattern for these four answers because of the confusion that might exist in distinguishing between contracted forms like "My father's at home" (35) and genitive forms like "My father's" (28). A closely related issue has to do with what Lyons (1969) terms "schemata". Forms such as "what's" and "who's", which would fall into this category, seem to be used as frozen structures in questions for both singular (9; 13) and plural (43; 53), although they are just contracted forms of "what is" and "who is". For Faerch and Kasper (1983), this is a "communicative strategy" which they term "formal reduction". So, not only transfer of training in the classroom but also "communicative strategies" seems to have a say in the variability occurring in the respondent's IL production of questions. No variation at all occurred in the respondent's use of the wh-word "when" on both tasks. Therefore, systematic variability is shown to exist in his written production of grammatical forms, although it would be exaggerated to say that all forms are subject to variability. Another example of variability in asking questions is manifested in examples (2), (24), and (25). It seems that although the respondent has made a correct distinction between "how many" and "how much" -probably because examples (24) and (25) are put next to each other, which allows an opportunity for comparison-, this distinction may not be based on countability. This is so because example (2), which is far from the other two examples and which does not have any other similar example next to it, is deviant from TL norm. This point will not be further discussed here because it falls outside the scope of this study on question forms. Youssef also has failed to use the question structure "what + be ... like?" characteristic of questions asking the interlocutor to describe people or objects. In question (20), where he is supposed to use this form so familiar in asking about the weather, he uses instead the question word "how", thinking that it is used for describing states and manners. He also uses the same question word in examples (37), where he is required to ask a question about somebody's appearance. The word "like" is mentioned only once in the whole data but, unfortunately, not in
23

the right place. This is so because although question (53) is meant to ask about the description of the shoes, the form "what + be ... like" can only be used to describe the appearance of human beings, states of events or the quality of something, not color (Thomson and Martinet, 1986). Another important comment to be made about the data elicited in the second task has to do with the variable performance on subject-verb inversion. Five questions out of fifty-four are found without such inversion. Related to this is the omission of the dummy verb "do" which has to be inserted in some environment to play the role of inversion; however, Youssef's performance shows that it is variable in this respect since there are some questions (26, 27...) lacking this grammatically important element. Discussion of verb tense will not be raised here because it requires an independent study with more data elicited by other means. To cut a long story short, Youssef's performance on the seven question words presented to him in the test is variable to the extent that there are those which he masters fairly well ("what" and "when") and there is another which he never uses ("whose"). Table (2) below indicates the wide range of variability in his performance on the wh-word. Table 2. The level of (in)correctness in Youssef's performance on questions
wh-words who what when where why how whose Total whwords number of wh-words 9 17 4 7 5 8 4 54 correct questions 44.5% 23.5% 0% 57% 20% 50% 0% 31.5% correctly chosen wh-word but incorrect questions 22% 64.5% 100% 28.5% 40% 50% 0% 46% incorrect questions 33.5% 12% 0% 14.5% 40% 0% 100% 22.5% incorrectly chosen wh-words 22% 29.5% 25% 14.5% 20% 25% 0% 22%

Statistical analysis of the respondent's performance on the two tasks indicates that accuracy rates in the two parts of the grammar test are considerably different. Only the wh-word "how" seems to have been produced at the same rate of accuracy on the two tasks (50%). Youssef's 100% accuracy in using "when" in the right context (especially task 1) many not mean that the form has been acquired completely as it was required only four times in the language test. It is clear from Table 2 that there is a marked variability ranging from full correctness, as for "when", to full incorrectness, as for "whose". Thus, Youssef's performance would be characterized as variable on two tasks. His performance on the first task is more accurate and more native-like than on the second task as it approximates 80 % of accuracy. In the second task, there is a rather steep decline in accuracy, which is mainly due to linguistic context and attention to form. Thus, it might be that this task-related IL variability was not so much governed by attention to form as by the nature of language form (interrogative sentences) which the second task required. In other words, this complex pattern of variation cannot be predicted or accounted for in terms of attention to form as the sole cause of variation. It should be noted that performance on the written language test did not seem to be a very good predictor of the respondent's ability to use
24

the tested grammatical forms correctly. The test did not provide more than basic information about his selection of question words in the various situations and in the various functions. It becomes clear then that function and attention to form are related when the respondent attempts to encode the functions required by the different situations in the various questions. In this task, he felt that clarity is essential in focusing in the linguistic forms which encode the various functions. 4.3 Conclusion The results of the present study on variability and systematicity in an aspect of IL production of a young Moroccan learner of English indicate that although his performance on question words is in large part systematic, it is also variable to a considerable extent, especially when it comes the overall structure of the questions. The most apparent factors that seem to affect the respondent's IL performance on questions are the amount of exposure to various language forms, linguistic context, familiarity with the vocabulary, the task used to elicit data, to cite but these. Nevertheless, variability in the use of questions as a whole does not necessarily imply variability in the use of wh-words only. Still, variability in this particular aspect of Youssef's performance does not entail variability in other aspects not discussed here. Therefore, prospective research with a large number of respondent from different language background, at different level of proficiency, and in a variety of social contexts is needed to provide more insights into the factors lying behind variation and systematicity in IL. 5. General Conclusion The purpose of this research paper has been to investigate systematicity and variability in IL and to survey the major theoretical claims advanced by different researchers adhering to different trends in the field. In setting about this task, IL as an idiosyncratic feature of second language development was first explored, then the processes that have been claimed to be responsible for the creation of this system were explained. Thereafter, the elicitation procedures used to collect IL data were briefly introduced and the ways of studying them briefly discussed. After that, the central concern of this research paper, namely systematicity and variability in IL systems, was addressed. Each one was dealt with at a time in two independent sections. the most salient theories existing in the literature that approach these two distinctive features of IL performance were also examined. It was found that each theory accounts for systematicity and variability differently and attributes the occurrence of such phenomena in IL to various factors, depending on the orientation adopted by the theory. Once these theories were evaluated in the light of a set of criteria proposed by Tarone (1989), the task of studying Youssef's IL data was then addressed. The findings of this cross-sectional study of a young Moroccan's IL production of questions have not provided more than basic information about the extent to which the performance of a language learner can be systematic and variable. Given the paucity of data that have been collated from the administration of a double-task test, a thorough coverage of all the relevant issues from all possible angles was not attainable. But the case study reported here has quite fairly dealt with some interesting issues that mark second language development. The research paper as a whole has provided a substantive analysis of systematicity and variability. Nonetheless, while systematicity may be relatively easily captured, there are still many lacunas that have to be filled in order to complete our understanding of the phenomenon of variability in IL. It is hoped that
25

future studies will examine more comprehensively this phenomenon in the light of abundant IL data in order to make progress in our understanding of this rather intricate and slippery issue.

A research paper submitted in partial fulfillment of the requirements for the first year of the DESA (MA) degree in general linguistic studies.

Bibliography Adjmian, C. 1976. "On the nature of interlanguage systems." Language Learning, 26, pp. 297320. Adjmian, C. 1982. "La spcificit de l'interlangage et l'idalisation des langues secondes." In J. Gueron and S. Sowley (eds.) Grammaire Transformationnelle: Thorie et Mthodologies. Vincennes: Universit de Paris VIII. Anderson, R. W. 1989. "The theoretical status of variation in interlanguage development." In S. Gass, C. Madden, D. Preston and L. Selinker (eds.) Variation in Second Language Acquisition: Psycholinguistic Issues. Clevedon: Multilingual Matter. Ausubel, D. 1968. Educational Psychology: a cognitive approach. New York: Holdt, Rinehart and Winston. Beebe, L. M. 1980. "Sociolinguistic variation and style shifting in second language acquisition." Language Learning, 30(2), pp. 433-477. Beebe, L. M. 1982. "The social psychological basis of style shifting." Plenary address, Second Language Research Forum, Los Angeles. Beebe, L. M. (Ed.) 1988. Issues in Second Language Acquisition: multiple perspectives. New York: Newbury House. Beebe, L. M. and H. Giles. 1984. "Speech Accommodation Theories: a discussion in terms of second language acquisition." International Journal of the Sociology of Language, 46, pp. 5-32. Bell, A. 1984. "Language style as audience design." Language in Society, 13, pp. 145-204. Bialystok, E. 1982. "On the relationship between knowing and using linguistic forms." Applied Linguistics, 3(3), pp. 181-206. Bialystok, E. and M. Sharwood-Smith. 1985. "Interlanguage is not a state of mind." Applied Linguistics, 6(2), pp. 101-117. Corder, S. P. 1967. "The significance of learners' errors." International Review of Applied Linguistics, 5(4), pp. 161-170.
26

 Corder, S. P. 1971. "Idiosyncratic dialects and error analysis." International Review of Applied Linguistics, 9(2), pp. 147-159. Corder, S. P. 1973. "The elicitation of interlanguage." In J. Svartvik (ed.) Errata: Papers in Error Analysis. Lund: Gleerup. Corder, S. P. 1981. Error Analysis and Interlanguage. Oxford: OUP. Dickerson, L. 1974. Internal and external patterning of phonological variability in the speech of Japanese learners of English: towards a theory of second language acquisition. Ph.D. dissertation. Urbana-Champaign: University of Illinois. Dickerson, L. 1975. "The learner's interlanguage as a system of variable rules." TESOL Quarterly. 9, pp. 401-407. Dickerson, L. and W. Dickerson. 1977. "Interlanguage phonology: current research and future directions." In S. P. Corder and E. Roulet (eds.) Interlanguages and Pidgins and Their Relation to Second Language Pedagogy. Also The notion of Simplification, Interlanguage and Pidgins: Actes de VIme colloque de Linguistique Applique de Neuchtel (Neufchatel). Geneva: Droz et Universit de Neuchtel. Dickerson, W. 1976. "The psychological unity of language learning and language change." Language Learning. 26(2), pp. 215-231. Downes, W. 1984. Language and Society. London: Fontana. Dulay, H. and M. Burt. 1973. "Should we teach children syntax?" Language Learning, 23(2), pp. 245-258. Dulay, H. and M. Burt. 1974a. "Natural sequences in child second language acquisition." Language Learning, 24(1), pp. 37-53. Dulay, H. and M. Burt. 1974b. "Errors and strategies in child second language acquisition." TESOL Quarterly, 8, pp. 129-136. Dulay, H. and M. Burt. 1974c. "A new perspective on the creative construction process in child second language acquisition." Language Learning, 24, pp. 253-278. Ellis, R and C. Roberts. 1987. "Two approaches for investigating second language acquisiti on in context." In R. Ellis (ed.) Second Language Acquisition in Context. New York: Prentice Hall. Ellis, R. 1985a. Understanding Second Language Acquisition. Oxford: OUP. Ellis, R. 1985b. "Sources of variability in interlanguage." Applied Linguistics, vol. 6, no. 2, 118-131. Ellis, R. 1987. "Contextual variability in second language acquisition and the relevancy of language teaching". In R. Ellis (ed.) Second Language acquisition in context. New York: Prentice Hall.
27

 Faerch, C. and G. Kasper (eds.) 1983. Strategies in Interlanguage Communication. London: Longman. Gardner, R. C. 1985. Social Psychology and Second Language Learning: The role of Attitude and Motivation. London: Edward Arnold. Gumperz, J. and D. Hymes (eds.) 1972. Directions in Sociolinguistics: The Ethnography of Communication. New York: Holt, Rinehart and Winston. Hakuta, K. 1975. Becoming bilingual at age five: the story of Uguisu. Honors thesis. Harvard University. Hakuta, K. 1976. "A case study of a Japanese child learning ESL." Language Learning, 26(2), pp. 321-352. Harris, Z. 1954. "Transfer grammar." IRAL, 20, 259-270. Huebner, T. 1983a. A Longitudinal Analysis of the Acquisition of English. Ann Arbor: Karoma Publishers. Huebner, T. 1983b. "Linguistic systems and linguistic change in an interlanguage." Studies in Second Language Acquisition, 6(1), pp. 33-53. Huebner, T. 1985. "System and variability in interlanguage syntax. Language Learning, 35(2), pp. 141-163. James, C. 1971. "The exculpation of contrastive linguistics." In G. Nickel (ed.) Papers in Contrastive Linguistics. Cambridge: CUP. James, C. 1974. "Linguistic measures for error gravity." AVAL Journal, 12(1), pp. 3-9. Krashen, S. 1981. Second Language Acquisition and Learning. Oxford: Pergamon. Krashen, S. 1982. Principles and Practice in Second Language Acquisition. Oxford: Pergamon. Labov, W. 1966. The Social Stratification of English in New York City. Washington, D. C.: Center for Applied Linguistics. Labov, W. 1969. "Contraction, deletion and inherent variability of the English copula." Language, 45(4), pp. 714-762. Labov, W. 1970. "The study of language in its social context." Studium Generale, 23, pp. 30-87. Labov, W. 1971. "The notion of 'system' in creole lan guages." In D. Hymes (ed.) Pidginization and Creolization of Languages. London: CUP. Labov, W. 1972. Sociolinguistic Patterns. Philadelphia: University of Pennsylvania Press. Lado, R. 1957. Linguistics Across Cultures. Ann Arbor: University of Michigan Press.
28

 Levine, M. 1975. A Cognitive Theory of Learning. Hillsdale, NJ: Lawrence Erlbaum Associates. Liceras, J. 1981. "Markedness and permeability in interlanguage systems." Working Papers in Linguistics, 2. Toronto: University of Toronto. Liceras, J. 1987. "The role of intake in the determination of learners' competence." In S. Grass and C. Madden (ed.) Input in Second Language Acquisition. Cambridge, Mass.: Newbury House. Littlewood, W. 1981. "Language variation and second language acquisition theory." Applied Linguistics, 2(2), pp. 150-158. Lyons, J. 1969. Introduction to Theoretical Linguistics. Cambridge: CUP. MacLaughlin, B. 1978. "The Monitor Model: Some methodological considerations." Language Learning, 28(2), pp. 309-332. Meisel, J., H. Clahsen and M. Pienemann. 1981. "On determining developmental stages in natural second language acquisition." Studies in Second Language Acquisition, 3(2), pp. 109-135. Nemser, W. 1971. "Approximative systems of foreign language learners." IRAL, 9(2), pp. 115-123. Reprinted in J. C. Richards (ed.) (1974). Perdue, C. 1984. Second Language Acquisition by Immigrants: a field manual. Rowley: Newbury House. Richards, J. C. 1971. "A non-contrastive approach to error analysis." English Language Teaching, 25, pp. 204-219. Richards, J. C. (ed.). 1974. Error Analysis: perspectives on Second Language Acquisition. London: Longman. Sampson, G. 1980. Schools of Linguistics. Stanford: Stanford University press. Schachter, J. 1986. "In search of systematicity in interlanguage production." Studies in Second Language Acquisition, 8, pp. 119-134. Schumann, J. 1978. The Pidginization process: A Model for second language Acquisition. Rowley, Mass.: Newbury House. Schumann, J. 1983. "Art and science in second language acquisition research." Language Learning, 33(5), pp. 49-76. Seliger, H. W. 1988. "Psychological issues in second language acquisition." In L. M. Beebe (ed.) (1988). Selinker, L. 1972. "Interlanguage." IRAL, 10(3), pp. 209-231. Reprinted in J. C. Richards (ed.) (1974).
29

 Selinker, L. 1992. Rediscovering Interlanguage. London: Longman. Selinker, L. and D. Douglas. 1985. "Wrestling with 'context' in interlanguage theory." Applied Linguistics, 6(2), pp. 190-204. Tarone, E. 1979. "Interlanguage as chameleon." Language Learning, 29(1), pp. 181-191. Tarone, E. 1982. "Systematicity and attention in interlanguage." Language Learning, 32(1), pp. 69-84. Tarone, E. 1983. "On the variability of interlanguage systems." Applied Linguistics, 4(2), pp. 142163. Tarone, E. 1985. "Variability in interlanguage use: a study of style-shifting in morphology and syntax." Language Learning, 35(3), pp. 373-395. Tarone, E. 1988. Variation in interlanguage. London: Edward Arnold. Tarone, E. 1989. "Accounting for style-shifting in interlanguage." In S. Gass, C. Madden, D. Preston and L. Selinker (eds.) Variation in Second Language Acquisition: Psycholinguistic Issues. Clevedon: Multiligual Matters. Tarone, E., U. Frauenfelder and L. Selinker. 1976. "Systematicity/Variability and stability/instability in interlanguage systems: more data from Toronto French immersion." In H. D. Brown, (ed.) Language Learning (Special Issue), 4, pp. 93-104. Thomson, A. J. and A. V. Martinet. 1986. A Practical English Grammar. Oxford: OUP. van Buren, P. 1972. "Contrastive analysis." In J. P. B. Allen and S. P. Corder (eds.) The Edinburgh Course in Applied Linguistics, 3. Weinreich, U. 1953. Language in Context. Publication of the Linguistic Circle of New York, no. 1.

30