by Rod Ellis
Complex Adaptive Systems Theory has implications for how L2 acquisition needs to be investigated. The focus needs to be on individual learners rather than groups. Learners need to be studied longitudinally so that the path of development can be carefully plotted. Attention needs to be given to the fine details—i.e. variation, including chaotic variation, cannot be ignored as messy ‘noise’. Also—ideally—the L2 learner needs to be investigated holistically to capture the nature of the interactions between the components of the system. A simplistic linear relationship between a cause and an effect is not possible.
To some extent, the research to date adheres to these research ideals. However, researchers have not yet found a way of investigating the totality of a system by taking into account the social, psychological, and all the linguistic components of the system. Rather, they have focused quite narrowly on specific linguistic properties of systems in much the same way as researchers with different theoretical leanings have done. Larsen-Freeman (2006) readily admitted that ‘the issue of finding a suitable methodology with which to capture the “fuzziness” … and dynamism of language development is a challenging one’ (p. 594). It would seem, then, that researchers are stuck with looking at relatively discrete parts of an L2 complex adaptive system.
Researching cognitive processes in SLA
Finally, I turn to briefly consider some of the ways in which SLA researchers have investigated cognitive processes—which are of course not available to direct observation. We cannot see in the minds of the L2 learners, but only observe their behaviours and infer the processes involved. In Chapter 7, we considered the methods used by researchers to investigate what learners take from the interactions they participate in—stimulated recall, for example. Here I will focus on three other ways of investigating cognitive processes.
Modelling L2 acquisition
Modelling is a tool that has figured strongly in cognitive SLA. In some studies, computer models have been built to simulate what occurs during actual learning tasks and to examine whether the learning outputs of these models accord with those reported in descriptive L2 studies. A good example of this approach can be found in Nelson (2013). He devised a computer simulation of lexical learning to test the theoretical claim that there is a clear distinction between the storage of the phonological/orthographic form of a word and its meaning—as proposed in Jiang’s (2000) model, which we considered earlier in the chapter. The simulation involved feeding in translation pairs (English–Spanish)—along with 48 other words that were related either formally or semantically—in ‘training runs’. The results of the simulation gave support to the separation of form and meaning representations. Nelson argued that computer-simulations of this kind can be used to test the claims of symbolic models of L2 representation.
Neuropsychological research
Neuropsychological SLA aims to identify the neurological mechanisms and circuits in the brain that are responsible for regularly occurring linguistic behaviour. It raises the possibility of matching cognitive constructs and neurological activity. Two different approaches can be distinguished: neurolinguistic studies examine how damage to different parts of the human brain affects a person’s ability to use an L2, whilst neurobiological studies use techniques for measuring brain imaging or electrical activity in the brain to establish which parts of the brain are activated when L2 learners undertake different tasks.
Neurolinguistic studies have shown that implicit and explicit memory function separately—as claimed in cognitive SLA. Ullman (2001) reported that neurological impairment to different parts of the brain results in differential loss of these two memory systems. For example, the implicit/procedural memory system is damaged in the case of Parkinson Disease, while the explicit/declarative memory system is impaired in Alzheimer’s Disease. Damage to one system does not lead to a loss of functions associated with the other system. Paradis (2004) reviewed studies of bilingual aphasic patients and concluded that loss of implicit memory did not prevent access to explicit knowledge. Interestingly, such people generally struggled to communicate in their L1—which he suggested was because they had minimal explicit knowledge of their L1. In contrast, they were better able to use their L2 because of their more developed explicit knowledge in this language.
Neurobiological studies also support a two-system account of language functioning. Ullman (2001) provided evidence to show that the brain consists of two largely separate systems—the lexicon and the grammar, each with distinct neural bases. He illustrated this model with reference to the processing of English regular and irregular past tense verbs forms. He proposed that procedural memory is needed for the computation of regular morphological features—for example, ‘moved’—while declarative memory handles irregular forms—for example, ‘ate’.
Neurobiological studies have also investigated the mechanisms involved in attention. In this case, however, it has proved difficult to establish the specific mechanisms involved. Suchert (2004), for example, concluded that the psychological distinctions between attention, perception, cognition, and action have no biological correlates and that attention is ‘an ongoing, interactive process that defies singular definition’ (p. 173). The distributed nature of attention allows for an interface between explicit/declarative and implicit/procedural memory. Recently activated neurons in explicit memory may remain primed and become active again when exposed to subsequent stimuli that are the same or related. In this way, the learning of a grammatical rule that is held in explicit/declarative memory can exert an influence over the processing of L2 input and output.
Neuropsychological SLA research is still in its infancy and in many cases the evidence it has provided is mixed—making it difficult to reach clear conclusions. However, the research does demonstrate quite clearly that there is no biological correlate of Universal Grammar. Language functions are highly distributed in the brain. It also lends clear support to the neurological separation of explicit/declarative memory and implicit/procedural memory. As Schumann (2004b) pointed out, the value of the neuroscientific approach to SLA lies in the constraints it imposes on the metaphors used to explain L2 acquisition.
Eye-tracking
One of the most recent methods employed to investigate cognitive processing is eye-tracking (see Winke, Godfried, and Gass 2013). Eye-tracking machines have been used to investigate the cognitive processes involved in attending to visual information—for example, when reading texts in the L2. This involves examining the fixations—i.e. the words a learner fixes attention on—and saccades—i.e. the movement of the eyes from one location to another. Movement is typically in a forwards direction but sometimes regressions—i.e. backward movements—occur. The length of time that learners spend on a particular word or phrase and the number of regressions to that word indicates the extent of the learners’ attention to it.
Winke (2013) used eye-tracking to investigate what effect input enhancement has on learners’ attention to English passive constructions in a text. Her results showed that the enhanced forms were attended to more than unenhanced forms—i.e. the learners spent more time fixating on them and rereading them—but that this was not found to promote learning. One interpretation of this study is that—although detection of the highlighted forms took place in short-term memory—little complex processing in working memory took place.
Conclusion
In this chapter, I have tried to survey the rich body of theory and research on cognitive aspects of L2 acquisition. It is not easy to reach clear conclusions given the multitude of perspectives evident in SLA and their obvious differences: cognitive SLA is replete with conflicts. Nevertheless, there does appear to be a growing consensus about a number of issues. I will conclude this chapter by summarizing where there is consensus and pointing out where theoretical differences continue to exist.
There is general agreement—and clear evidence—to support the existence of two separate linguistic systems, variably referred to as explicit/declarative and implicit/procedural.
There is also general agreement that it is the implicit/procedural system that is primary and that the main goal of SLA is to elucidate how this system works and how it is acquired.
There are—however—different views concerning the nature of the implicit/procedural system. According to Universal Grammar (a symbolist theory), implicit knowledge of grammar takes the form of highly abstract properties that are innate and domain-specific and that make possible the acquisition of the grammar of specific languages. According to connectionist theories, implicit/procedural knowledge takes the form of an elaborate and highly-interconnected network, the nodes of which do not correspond to any linguistic categories. Connectionist theories claim that language learning draws on the same cognitive apparatus as any other kind of learning. In this chapter, I have emphasized the connectionist perspective as this is clearly gaining ground in SLA. However, the debate about whether language learning involves domain-specific or domain-general mechanisms continues. Processability Theory, for example, assumes a domain-specific set of processing procedures to account for the acquisition of L2 grammar.
Information-processing and connectionist theories recognize the importance of both explicit/declarative systems and implicit/procedural systems. Both acknowledge that receptive and productive language use draw on these systems; both recognize that L2 acquisition involves both systems; and both see the implicit/procedural system as primary.
However, information-processing and connectionist models differ in two key respects. First, Skill-learning Theory and Processability Theory—both information-processing theories—assume skill specificity: that is, they maintain that different systems underlie receptive and productive language use and learning. Connectionist theories, however, assume that learners draw on a single network. Whilst Skill-learning Theory proposes a direct interface between declarative and procedural knowledge, in connectionist models, the interface is only indirect—i.e. explicit knowledge fine-tunes the processes involved in the development of implicit knowledge. If proceduralized explicit knowledge is indeed functionally equivalent to implicit knowledge—as DeKeyser (2003) claims—it is difficult to see how the various positions can be tested empirically.
All cognitive theories—with the exception of Processability Theory—emphasize the importance of attention for language learning. They draw on a model of working memory to explain the role that attention plays. Because working memory capacity is limited, L2 learners (especially those of low L2 proficiency) experience difficulty in attending simultaneously to both form and meaning and are likely to prioritize one over the other. They employ top-down processes when processing for meaning, but need bottom-up processing to process and acquire linguistic forms.
Theories differ in terms of whether they see attention as a conscious process—what Schmidt called ‘noticing’—or as unconscious—as in sublimal learning. In this chapter, I have suggested that both are possible. Detection of exemplars—but not of symbolic categories—takes place in phonological short-term memory. This can occur unconsciously: what is detected can feed directly into the implicit system. However, when complex processing takes place in working memory—when detected information is rehearsed and compared with information in long-term memory—consciousness is involved. Complex processing contributes to both explicit and implicit systems.
Various factors influence whether and what learners attend to in the input: the learner’s alertness and orientation; the learner’s working memory capacity; the intrinsic properties of specific linguistic features—for example, their salience and frequency in input—the learner’s natural default processing strategies; and the correspondence between L2 and L1 forms.
Two types of learning can be distinguished—implicit and explicit. Implicit learning can occur without awareness. Explicit learning always involves awareness. The main area of controversy concerns whether learning without any awareness is possible. This reflects the disagreement about the nature of attention and its role in learning—see point 6. To resolve this controversy, some researchers have argued that it is necessary to demonstrate whether awareness is demonstrated at the point of learning—i.e. during training—and not just in a post-test that measures whether any learning occurred.
Ideally, a complete cognitive theory of L2 acquisition needs to account for both how input is processed and how output is produced. Both play a role in L2 development. Currently, however, different theories deal with these two key aspects of L2 acquisition separately. The most complete theory to date is one that views the L2 system as a dynamic adaptive system. Theories that draw on this concept emphasize the interconnectedness of the components that comprise the system and its emergent nature. Usage-based accounts of L2 acquisition acknowledge the importance of accounting for both how input is processed and output is produced.
All cognitive theories acknowledge the important role played by language transfer. There is general recognition that the L1 can inhibit the development of the L2 system by blocking detection and influencing executive processing.
Understanding L2 acquisition necessarily involves the study of the cognitive processes involved. No theory of L2 acquisition is complete without an explanation of how the L2 is represented in the mind of the learner, the role played by attention in reception and production, and the processes that can account for interlanguage change. Equally, however, no theory can ignore the importance of the social context of learning. In their exposition of complex systems, Larsen-Freeman and Cameron (2008) emphasized that ‘the social dimension of language is indispensable’ (p. 126). However—by and large—the theories I have addressed in this chapter focused on the cognitive aspects of L2 acquisition rather than the social. In the next chapter, I will examine the body of research that has given precedence to the social dimension of L2 learning.
Notes
1 Eysenck (2001) noted that ‘it remains the case that declarative memory resembles explicit memory, in that it involves the integration or linkage of information’ while ‘procedural memory still resembles implicit memory’. (p. 213).
2 Pinker (1999) also argues for a dual-mode system. He claimed that regular morphological features such as past tense -ed in English involve rule-based or symbolic processing, whereas features that are not constrained by readily identifiable rules—such as irregular past tense forms like ‘swam’—are stored as items.
3 In an earlier version of UG, Chomsky referred to the ‘language acquisition device’, which he claimed was necessary to explain children’s ability to acquire the grammar of their mother tongue.
4 The interface hypothesis continues to be debated. Paradis (2009), for example, critiques N. Ellis’s weak-interface position on the grounds that the acquisition of implicit knowledge involves entirely unconscious processes and therefore excludes conscious ‘noticing’.
5 DeKeyser (2007) only makes a limited claim for Skill-Learning Theory, namely that it best accounts for the early stages of acquisition, adult learners with above-average aptitude, and instructed learners.
6 See Chapter 5 for R. Ellis’s (1988) account of learners’ following pronouns and nouns.
9
Social aspects of second language acquisition
Introduction
In Chapter 5, we saw that it was necessary to consider social factors to account for variability in learner language and in Chapter 7, we saw that social interaction plays an important role in acquisition. In general, though, the perspective so far has been a cognitive one. Social factors were only invoked to account for how learners obtained input and to explain the characteristics of their output. No consideration was given to the possibility that language learning itself is essentially a social rather than a cognitive enterprise.
The turn of the last century saw a strongly argued critique of cognitive SLA emerge. Firth and Wagner (1997) criticized the predominant cognitive model on the grounds that it was ‘individualistic and mechanistic’ and ‘failed to account for the interactional and sociolinguistic dimensions of language’ (p. 2
85). They considered such an approach ‘flawed’ and sought to reconceptualize SLA in terms of an integrated account of the social and cognitive dimensions of L2 use and acquisition. Atkinson (2002) pointed out that ‘just as surely as language is social, so is its acquisition’ (p. 527). Block (2003) noted that mainstream studies of L2 acquisition typically failed to provide rich descriptions of the learners under study and there was a ‘certain monolinguistic bias’ (p. 33) in the assumption that the learners have a single L1 which remains intact despite contact with the L2. In short, what Block called ‘the social turn’ had very clearly arrived in SLA.
Two senses of ‘social context’
But what exactly does it mean to talk about the ‘social turn’ in SLA? In what sense is L2 acquisition a social rather than a cognitive phenomenon? To answer this question, it is helpful to distinguish two different senses in which acquisition can be viewed as social. Central to both is the notion of social context. Siegel (2003) distinguished a structural view of social context where social factors such as power and prestige are seen as determining the context in which learners learn and an interactional view where social context is seen as constructed dynamically in each situation. Applied to SLA, this suggests two ways of investigating the ‘social turn’:
By examining how social factors impact on the individual learner’s acquisition of an L2. In such a view, social factors are seen as just affecting the learner’s access to input (see Chapter 7) and the learner’s motivation to learn (see Chapter 3). In this view, social factors do not alter the cognitive/psychological processes responsible for acquisition but merely influence the data made available to the learner and the learner’s propensity to make use of the data.