The purpose of this general overview article is to outline how research into second language acquisition (SLA) over the last few decades has fed into our understanding of learning and teaching in foreign language classrooms. After a very brief overview of SLA research findings concerning both route and rate of L2 development, theoretical models attempting to explain these findings are presented, ranging from purely linguistic to cognitive models and social/interactionist models. The relationship between SLA research and second language pedagogy is then explored. Finally, recent developments investigating specifically the relationship between instruction and L2 development are outlined. 1. Introduction
The two main, well documented findings of SLA research of the past few decades are as follows: 1. second language acquisition is highly systematic
2. second language acquisition is highly variable
Although these two statements might appear contradictory at first sight, they are not. The first one primarily refers to what has been called the route of development (the nature of the stages all learners go through when acquiring the second language – L2). This route remains largely independent of both the learner’s mother tongue (L1) and the context of learning (e.g. whether instructed in a classroom or acquired naturally by exposure). The second statement usually refers to either the rate of the learning process (the speed at which learners are learning the L2), or theoutcome of the learning process (how proficient learners become), or both. We all know that both speed of learning and range of outcomes are highly variable from learner to learner: some do much better much more quickly than others. Before we expand on these findings a little more, it is important to note that, traditionally, the concern for rate of learning has been the centre of teachers’ and learners’ attention.
This is because it has obvious pedagogical implications: if we understand what makes learners learn faster and progress further, then maybe we can be better teachers or learners. However, these two lines of enquiry are both part and parcel of the same endeavour, which is to understand thoroughly how learners learn. In fact, understanding the route learners follow, and therefore having clear expectations of what learners can achieve at given points on the developmental continuum, is crucially important for both learners and teachers. Such study leads us, for example, to a better understanding of the significance of errors in the learning process. Producing them need not be seen as necessarily problematic (in fact, some errors can be evidence of a more advanced linguistic system than the equivalent correct form: for example, learners will usually produce rote-learned formulaic questions such as ‘where’s X?’, e.g. ‘where’s the ball?’, in which ‘where’s’ is an unanalysed chunk, before producing the developmentally more advanced ‘where the ball is?’, the second stage in the development of the interrogative system before the final stage in which ‘where is the ball?’ is produced correctly; See e.g. Myles et al1998 and 1999 for a discussion.
This is often referred to as the ‘U shape of learning’, typical also of L1 learners, by which learners start with the correct rote-learned form, e.g. took, before over-applying the past tense rule and producing taked, prior to learning the exception to the rule and producing took again, creatively rather than rote-learned this time. Teachers will also be less frustrated, and their learners too, when they become aware that teaching will not cause skilful control of a linguistic structure if it is offered before a learner is developmentally ready to acquire it. Now, of course, if we can speed up progression along the route that research has identified we need to understand how to do so. But understanding this route is inseparably bound up with clarifying the question of rapid and effective teaching. The robust research findings regarding the systematicity of the route followed by L2 learners do not have straightforward implications for language teaching, however.
One logical possibility might be that curricula should closely follow developmental routes; this is not sensible however, given (a) the incomplete nature of our knowledge of these routes, (b) the fact that classrooms are typically made up of learners who are not neatly located at a single developmental stage, and (c) the fact that developmental stages typically contain non-target forms. (For example, typical stages in the acquisition of negation will be: 1. ‘no want pudding’; 2. ‘me no want pudding’ 3. ‘I don’t want pudding’, with forms 1 and 2 representing normal developmental stages, therefore to be expected in early L2 productions, but which will not be taught).
Other possibilities are that curricula should be recursive with inbuilt redundancy, and that teachers should not expect immediate accuracy when teaching a new structure, or that they should give up on closely prescribed grammar curricula and opt instead for functional and/or task-based syllabus models. Many teachers/language educators have actively welcomed the role of ‘facilitator’ rather than ‘shaper’ of development, implied by such models. I will now briefly summarise research findings relating to both systematicity and variability, drawing implications for teaching methodology as I go along. 2. Systematicity
A substantial part of the SLA research community has concentrated on documenting and trying to understand the discovery that language learning is highly systematic. A defining moment for the field was in the late 70s / early 80s when it became evident that L2 learners follow a fairly rigid developmental route, in the same way as children learning their L1 do, and not dissimilar in many respects from the L1 route. Moreover, this developmental route, crudely represented below as a series of interlocking linguistic systems (or interlanguages: La, Lb, … Ln … ), sometimes bore little resemblance to either the L1 of the learner, or the L2 being learnt. Developmental route
Crucially, these interlanguages are linguistic systems in their own right, with their own set of rules. For example,Hernández-Chávez (1972) showed that although the plural is realised in almost exactly the same way in Spanish and in English, Spanish children learning English still went through a phase of omitting plural marking. It had been assumed prior to this that second language learners’ productions were a mixture of both L1 and L2, with the L1 either helping or hindering the process depending on whether structures are similar or different in the two languages. This was clearly shown not to be the case, even if the L1 of learners does of course play some role, especially in early stages and more persistently at the level of pronunciation (more about this later). For example, the developmental stages in the acquisition of German word-order, in both naturalistic and instructed learning contexts and irrespective of the L1 of the learners, are claimed to be as follows (Pienemann 1998): Stage 1:| Canonical Order (SVO)
Die kinder spielen mim ball (= the children play with the ball) Learners initially hypothesise that German is SVO, with adverbials in sentence-final position.| Stage 2:| Adverb preposing
Da kinder spielen (= there children play)
Learners now place the adverb in sentence initial position, but keep the SVO order (no verb-subject inversion yet).| Stage 3:| Verb separation
Aller kinder muss die pause machen (= all children must the pause make) Learners place the non-finite verbal element in clause-final position.| Stage 4:| Verb-second
Dann hat sie wieder die knoch gebringt (= then has she again the bone brought) Learners now place the verb in sentence-second position, resulting in verb-subject inversion.| Stage 5:| Verb-final in subordinate clauses
Er sagte dass er nach hause kommt (= he said that he to home comes) Learners place the finite verb in clause-final position in subordinate clauses.| Similar sequences of acquisition have been found for a wide range of structures in a range of languages (see e.g. Ellis 1994; Mitchell & Myles 1998). After the 1980s the SLA research agenda focused on (a) documenting the route followed by learners in a range of structures and languages – although English remains by far the most studied L2, and increasingly (b) explaining this route which, if it is for the most part independent of both the L1 and the context of learning, must be due to learner-internal processes. This still remains today a crucial part of the SLA research agenda. 2.1 Learning Development Models (Universal Grammar, Cognitive models, Interactionist / Sociocultural models) The theoretical approaches which have been used in order to investigate L2 development fall into three broad categories: Universal Grammar (UG)
The UG approach, following in the footsteps of L1 acquisition research, applies the Chomskyan paradigm (Cook & Newson 1996; White 1989; 1996; 2000) to the study of L2 development. See papers by Adger and Sorace in this guide. In a nutshell this linguistic theory claims that humans inherit a mental language faculty which highly constrains the shape that human languages can take and therefore severely limits the kind of hypotheses that children can entertain regarding the structure of the language they are exposed to. This is why children acquire their first language easily and speedily, in spite of its complexity and abstractness, at an age when they are not cognitively equipped to deal with abstract concepts generally. In this view, the core of language is separate from other aspects of cognition, although it operates in close interaction with them of course. If the L2 developmental route is similar in many respects to the L1 route, then it must also be because the innate UG constrains L2 development. This approach has given rise to a wealth of studies (see for example White 1989, 1996, 2000; Flynn, Martohardjono & O’Neil 1998; Schwartz 1998;Archibald 2000; Herschensohn 2000; Balcom 2001; Hawkins 2001a, 2001b). Cognitive models
The cognitive and information processing models generally, which originate from psychology (and neurolinguistics), claim, on the other hand, that language learning is no different from other types of learning, and is the result of the human brain building up networks of associations on the basis of input. Information processing models see learning as the shift from controlled processes (dealt with in the short term or working memory and under attentional control) to automatised processes stored in the long term memory (retrieved quickly and effortlessly). Through this process, what starts as declarative knowledge (knowing ‘that’) becomes procedural knowledge (knowing ‘how’) which becomes automatic through repeated practice. Recently, connectionist models have further assumed that all learning takes place through the building of patterns which become strengthened through practice.
Computer models of such processes have had some success in replicating the L1 and L2 acquisition of some linguistic patterns (e.g. past tense, gender; Sokolik & Smith 1992; Ellis & Schmidt 1997). The view of language encapsulated withinconnectionism, as this view of cognition is called, is fundamentally different from linguistic models, where language is seen as a system of rules rather than as patterned behaviour. In both the UG and cognitive models, the focus is on explaining learner-internal mechanisms, and how they interact with the input in order to give rise to learning. The emphasis on the role played by the input however, varies, with the UG approach assuming that as long as input is present learning will take place, and the other models placing a larger burden on how the input is decoded by learners, paying particular attention to concepts such as noticing or attention. Interactionist/sociocultural models
In contrast to these models, the interactionist approach has paid particular attention to the nature of the interactions L2 learners typically engage in. It has focused on investigating, for example, the role of negotiation for meaning in the context of NS-NNS (Native Speaker – Non-Native Speaker) conversations (Gallaway & Richards 1994; Gass 1997;Gass & Varonis 1994; Pica 1994; Oliver 1995; Long 1996), in order to see how interactions are modified by both NSs and NNSs to ensure that the input the latter receive is comprehensible. The role of feedback given to learners when they make mistakes has also been the object of attention (Aljaafreh & Lantolf 1994; Lyster & Ranta 1997; Long, Inagaki & Ortega 1998). For example, Lyster & Ranta (1997) found that the most common feedback given to learners when they produce incorrect forms are recasts, i.e. a repetition of the learner’s utterance minus the error; however, they also found that recasts were the kind of negative feedback learners were most likely to ignore.
Researchers adopting a socio-cultural framework, following in the footsteps of Vygotsky (1978; 1986), who believed that all learning was essentially social, have explored the way in which L2s are learned through a process of co-construction between ‘experts’ and ‘novices’. Language learning is seen as the appropriation of a tool through the shift from inter-mental to intra-mental processes. Learners first need the help of experts in order to ‘scaffold’ them into the next developmental stages before they can appropriate the newly acquired knowledge. This is seen as a quintessentially social process, in which interaction plays a central role, not as a source of input, but as a shaper of development (Lantolf & Appel 1994; Lantolf 2000). 2.2 Teaching implications
The implications of these models of learning for teaching methodologies are essentially as follows: UG If the development of the L2 linguistic system is primarily driven by learner-internal mechanisms, requiring the learner to map the L2 input onto an innate highly constrained linguistic blueprint, then all the classroom needs to provide is linguistic input, and learning will take care of itself. In this view, the L2 acquisition process is seen as very similar to L1 acquisition, and children do not need to be taught grammar in order to become fluent native speakers. The UG view of language learning is consistent with the communicative language teaching approach, in the sense that both believe that learning will take place if rich natural input is present. It is important to stress though, that the two approaches developed independently of one another, with UG evolving out of the need to understand how children acquire their mother tongue, and then being applied to L2 acquisition, and communicative language teaching being the result of the perceived failure of grammar-translation or audiolingual methodologies by teachers, who felt that they did not prepare learners for real life communication needs.
Krashen (1982, 1985) was influential in articulating the first model putting together these views of learning and teaching, and the subsequent work on the role of input and interaction helped us better to understand how different kinds of interactions may contribute to providing usable input for the learner (Gass 1997; Pica 1994; Long 1996; Swain 1995). Cognitivism The information processing or connectionist models, on the other hand, which see learning as the strengthening of associations and the automatisation of routines, lead to much more behaviourist views of learning. Thus learners are seen as central to the acquisition process, in the sense that they have to practise until patterns are well established, and external variables take on a much greater role. For example, the role of input, interaction and feedback, and how they can speed up development, is seen as much more crucial, as is the role of practice in the development of fluency and control of the L2 system.
Combining the models These two apparently conflicting approaches are not the only ones that have been applied to the study of second language learning and teaching, but they have received most interest and generated most empirical work. These models might appear contradictory at first sight, but in fact they can be reconciled in so far as they are concerned with different aspects of SLA, which is, after all, a highly complex process. Even if one accepts the view that language development is highly constrained, possibly by UG (and, after all, the robust developmental routes that learners follow, as illustrated earlier, seem to be a strong argument in favour of this view), it is not the whole picture. We also need to understand many aspects of the SLA process other than the acquisition of syntax and morphology, such as lexical acquisition or the development of pragmatic and sociolinguistic repertoires.
Moreover, if developmental sequences show how learners construct the L2 linguistic system, they do not tell us anything about how learners develop their ability to access in real time the system they have constructed. In other words, if we believe UG constrains the mental grammars constructed by L2 learners, we still need to understand how learners become more fluent. In order to understand SLA, we need to know not only what the system constructed by learners looks like, but also the procedures which enable efficient use of this system, and how the two interact in real time, as well as develop over time. The fact that these two endeavours are independent is clearly evident when we think of learners who are good system builders, i.e. they are accurate in their productions, but not necessarily good at accessing this system in real time, i.e. they are very non-fluent.
The reverse is also true, with some learners developing high levels of fluency quickly, but remaining very inaccurate in their productions. Similarly, if we are to find out what can facilitate the learning process, we need to gain a much better understanding of the kinds of interactions and social settings which promote learner development. Gass (1997), for example, argues that task-based methodologies (in which learners have to negotiate with one another in order to perform a meaning-focused activity) force learners to notice ‘gaps’ in their L2, a prerequisite for filling such gaps. Swain (1995), in her ‘pushed output hypothesis’, argues that it is when learners’ own productions fail to meet their communicative goals that they are forced to revise their linguistic system.
Some recent teaching methodologies have recognised the important role played by the setting of learning, and by the quality of interactions therein. Although not necessarily well-informed either theoretically or empirically, a number of humanistic teaching methodologies such as ‘suggestopaedia’ (which aims to relax the student through e.g. listening to music), or ‘the silent way’ (making use of coloured rods to express meaning), which believe that L2 learning is facilitated if the learner’s inner-self is set free from inhibitions by providing a stress-free learning environment, have been very popular in some parts of the world. 3. Variability
The variability that occurs in L2 development, in terms of rate of acquisition and outcome, have received much less attention in the SLA literature until relatively recently. This was because of the very robust general findings showing that, in key respects, learners develop in similar ways no matter what their age is, whether they are learning the L2 in a classroom or in a country where the language is spoken, no matter what their L1 is, and no matter what they were actually taught. As more and more empirical research has been carried out, however, a number of important points have emerged which have meant qualifying these statements somewhat. 3.1 Variability in route
Despite the relative rigidity of the L2 learning route, transfer does occur in so far as the L1 has an impact upon L2 learning, even if it remains true that it is primarily in the sense of speeding up the learning process in the case of closely related languages or similar linguistic structures, rather than changing the route of development itself (i.e. learners still follow the same stages, but at different speeds, depending on their L1). For example, Italian learners of French will acquire the idiosyncratic placement of object pronouns in French more quickly than say English learners because it is similar in both languages, but they will still go through the same stages, when in fact transferring their L1 structure would lead to acquisition of the correct system.
In fact there is ample evidence, in the literature, of transfer not taking place when it would help, and conversely of transfer taking place when it leads to errors. Moreover, transfer often occurs one way and not the other, with English learners of French, for example, producing la souris mange le (the mouse eats it) rather than la souris le mange (the mouse it eats), but French learners of English never produce the mouse it eats in their interlanguage, which one would expect if transfer was taking place (Hawkins 2001a). But there are also areas in which the L1 gives rise to structures not found in the language of other L2 learners (see e.g. Odlin 1989; Selinker 1992). The impact of the L1 on interlanguage development needs to be better understood, even if its potential influence on SLA remains limited since we know that only a small subsection of structures from the L1 are likely candidates for transfer. 3.2 Variability in rate and outcome
In contrast to the undeniable systematicity of the route of development (bar the relatively minor differences alluded to previously) the rate of acquisition and the outcome of the acquisition process are highly variable, unlike L1 acquisition in which children seem to progress at roughly similar rates (give or take a few months), and all become native speakers of the language they are exposed to. It is very difficult to predict in second language acquisition what makes some people learn faster and better than others. Some factors have been isolated as playing some part in this. For example, age is one such factor (Singleton & Lengyel 1995). Although the commonly held view that children are better L2 learners is a gross oversimplification if not a complete myth, differences have been found between children and adults, primarily in terms of eventual outcome.
Although teenagers and adults have been found to be generally better and faster L2 learners than young children in the initial stages of the learning process (on a wide range of different measures), children, however, usually carry on progressing until they become indistinguishable from native speakers whereas adults do not, especially as far as pronunciation is concerned. Whether this is due to the process of acquisition having changed fundamentally in adulthood (e.g. because UG is not available anymore once the L1 has been acquired), or for other reasons (e.g. the process remains the same but stops short of native competence), is an issue hotly debated today, and the source of much empirical investigation (Birdsong 1999). The fact remains, though, that the route followed by young and older L2 learners is essentially the same, and is similar in many respects to that followed by children learning that language as a native language. Another salient difference when comparing L1 and L2 outcomes is that whereas native competence is the norm in the L1 context, it is the exception in the case of L2s.
This phenomenon is commonly referred to as fossilisation. Some structures seem very difficult to acquire in the L2, even when there is plenty of input. In immersion programmes in Canada, in which English-speaking children are taught the normal curriculum through French and are therefore exposed to large amounts of input within a communicative focus, end results have been mixed. Although these children become very proficient and fluent in French, their accuracy in some areas (e.g. gender, adverb placement etc.; White 1996; Harley 1998; Hawkins 1998) remains far from native-like, suggesting that some aspects of language resist spontaneous acquisition. In order to explain variability in rate and outcome, SLA researchers have focused primarily on the role of external factors in the acquisition process. As we have seen, one line of research inquiry has addressed questions about the nature of the input and the role of interaction in the learning process. Other lines of inquiry have investigated the role of learner variables, such as intelligence, aptitude, motivation, attitude, as well as the social and sociolinguistic variables which impact on them (Skehan 1989; 1998; Berry 1998; Dörnyei 2001; Sawyer & Ranta 2001).
These variables have been found to play an important role in determining how successful learners are. For example, recent motivation research has witnessed something of a boom since the nineties, with research questions becoming more sophisticated and addressing more directly language teaching issues. Motivation is now seen as situation-dependent as well as a relatively stable learner trait, and much work has been carried out investigating issues such as the role of tasks in motivating learners, the role of the teacher in motivating learners, or the role of learning strategies in enhancing motivation (Dörnyei 2001 and 2002). If motivation, as well as other learner variables, is now widely recognised as playing a determining role in SLA, more research needs to be carried out on its pedagogical implications, i.e. on how to motivate learners. For further discussion, see article in this Guide on Learner Difference by Skehan . 4. Conclusion – SLA research and good practice
The picture emerging from research into second language development is, unsurprisingly, highly complex, and many factors have been identified as playing a role. Here I will outline more systematically the relationship which is emerging between SLA research and language pedagogy at the beginning of this century. As SLA research has matured, and the key constructs which form its theoretical basis have become established, the field has become better able to look outwards and investigate the role of different contexts of learning. Furthermore, there has been renewed interest in grammar pedagogy (Lightbown 2000; Mitchell 2000), partly because of the perceived failure of contexts of learning promoting ‘natural’ communication (immersion, and Communicative Language Teaching) in producing learners who are consistently accurate in their productions. Consequently, the role of instruction and the role of the input in facilitating the L2 learning process have increasingly become foci of interest.
Form-focused instruction Many researchers have used current understanding of the relationship between cognition and SLA in order to investigate what kind of instruction is most helpful (Doughty & Williams 1998; Doughty 2001; Ellis 2001; Robinson 2001). In an extensive review of the empirical literature on the effectiveness of instruction,Norris & Ortega (2000, 2001) conclude that explicit form-focused instruction is effective in promoting learning. The object of most of these studies is to test what kind of instruction is most effective, such as ‘input enhancement’ (that is ways of making the input more noticeable for learners, such as e.g. having all object pronouns in bold in a text if the focus of the class is object pronouns, or ‘input flood’ in which learners are exposed to vast quantities of a given structure). Different ways of presenting structural input have also been explored, with explicit form-focused instruction being contrasted with implicit form-focused instruction (with learners having to work out the rule for themselves, or having the rule made explicit to them).
Although the results of this increasingly rich and sophisticated new body of research are tentative at present, it has identified key themes/agendas for further research, such as the role of explicit vs. implicit instruction, the role of negative evidence or the role of noticing. This research is crucial for gaining a better understanding of the relationship between learning and teaching (Mitchell 2000). Concluding observations To conclude, SLA research is an extremely buoyant field of study which has attracted much theoretical and empirical work in the last two or three decades. Much progress has been made in gaining a better understanding of the processes involved in learning second languages, as well as the different external factors which affect this process.
Although these complementary agendas remain less integrated than one might wish, bridges are being built which connect them. Similarly, the implications of SLA research for teaching are now receiving more attention, as is the specificity of the classroom context for understanding learning, but much more work remains to be done in these areas. There is still a huge gap – not unsurprisingly, given the limits of our knowledge – between the complementary agendas of understanding the psycholinguistic processes involved in the construction of L2 linguistic systems, and understanding what makes for effective classroom teaching. Acknowledgements
I wish to thank Rosamond Mitchell and Emma Marsden for useful comments on an earlier version of this article, as well as Christopher Brumfit and David Bickerton for their helpful suggestions as reviewer and editor of this piece. Any remaining errors are my own. 5. Glossary
Audiolingual method The behaviourist teaching method popular in the sixties and seventies, based on the premise that you learn to speak languages through habit-formation, and therefore need to practise drills until the new habit has been learnt. Audiolingual method| The behaviourist teaching method popular in the sixties and seventies, based on the premise that you learn to speak languages through habit-formation, and therefore need to practise drills until the new habit has been learnt.| Communicative Language Teaching| This approach to teaching believes that languages are learnt through communication, and that the focus of the classroom should be on encouraging learners to engage in speaking activities which simulate ‘real life’ communication. This approach de-emphasises the role of a metalinguistic knowledge of the L2 linguistic system.| Fossilisation| The phenomenon by which L2 learners often stop learning even though they might be far short of native-like competence.
The term is also used for specific linguistic structures which remain incorrect for lengthy periods of time in spite of plentiful input (e.g. in immigrant speakers whose fluent L2 still contains non-target like structures).| Grammar-translation method| The traditional teaching method which believed that the best way to teach languages is through the teaching of grammar and the translation of texts.| Immersion| This term refers to educational programs in which children are taught academic subjects (e.g. maths, geography etc…) through the L2. These programs are well established in Canada, where many anglophone children are educated partly through the means of French (especially in the province of Quebec).| Interlanguage| A term used both to refer to the linguistic system of L2 learners at a specific point in time, and to the series of interlocking L2 systems typical of L2 development. The significance of this term is the emphasis it places on the L2 system being a linguistic system in its own right, independently of both L1 and L2.| Transfer| Use of L1 properties in the L2. Transfer can be positive, when the borrowing of an L1 structure leads to a correct form in the L2 (e.g. the German learner producing ‘I am twelve years old’ in English L2 as a direct translation of the German structure), or negative when it leads to an incorrect form (e.g. the French learner producing ‘I have 12 years’).