Understanding Second Language Acquisition (2nd ed)

by Rod Ellis

  Less clear, however, is whether detection also involves form-function mapping. On this point, Schmidt appears to have changed his position. Schmidt (1994) argued that noticing refers ‘only to the registration of the occurrence of a stimulus … not the detection of form meaning relationships’ (p. 179). However, Schmidt (2001) clearly sees noticing as involving form-meaning mapping. He comments: ‘to acquire morphology … one must attend to both the forms of the morphemes and their meanings and in order to acquire syntax one must attend to the order of words and the meanings they are associated with’ (p. 31).

  The Noticing Hypothesis is not without its critics. Truscott and Sharwood Smith (2011) pointed out the contradiction between Schmidt’s early and later view about whether noticing entails form-meaning mapping. They also disputed his claim that detection—the registration of some feature of a stimulus—does not involve any abstract understanding of the feature. They commented that ‘a pure case of noticing or apperception, with no understanding, is difficult to imagine’ (p. 503).

  Executive control

  To a large extent, the problems that Truscott and Sharwood-Smith identify with Schmidt’s account of noticing can be resolved by distinguishing between ‘detection’—the registration of some part of a stimulus—and ‘cognitive manipulation’. Detection may indeed be a matter of just perceiving tokens in the input. If detection is a transitory phenomenon—dependent on phonological-short term memory—it is difficult to see how it can also entail symbol-recognition, which requires greater depth of processing. However—as Schmidt makes clear—noticing also entails executive control processes which, it can be argued, make possible form-meaning mapping, symbolic formation, and ‘understanding’. Detection may occur consciously or unconsciously—as proposed by Tomlin and Villa—but it involves only shallow processing. In contrast, the complex processing involving the central executive will more clearly involve consciousness.

  Information that enters short-term memory—i.e. is detected—may connect directly with long-term memory. However—as Robinson (2003) pointed out—this involves unaware recognition as it entails only matching what has been attended to with pre-existing knowledge. Thus, its function is to strengthen existing categories, but it is unlikely to lead to modification of these categories. However, when detected information is subjected to further processing in working memory, changes in long-term memory can occur. Robinson suggests that this involves rehearsal processes. He distinguishes two kinds: maintenance rehearsal and elaborative rehearsal.

  Maintenance rehearsal involves data-driven, instance-based processing. Robinson does not explicitly say so, but potentially this allows for noticing-the-gap. By maintaining a chunk—for example, ‘made me go’—in working memory and comparing it to a pre-existing chunk—’made me to go’—in long-term memory, the learner has the opportunity to notice the difference in the two chunks. This can lead to a weakening of the association between ‘made me’ and ‘to go’ in long-term memory and the formation of new connection between ‘made me’ and ‘go’. In this way, learning takes place. Maintenance rehearsal, therefore, provides an opportunity for the cognitive comparison, which—as I have argued—allows for new learning to occur (R. Ellis 1994). Clearly, too, if the two chunks are juxtaposed in interaction—for example, when the learner’s erroneous utterance is recast—maintenance rehearsal is made easier and cognitive comparison is more likely to occur.

  Elaborative rehearsal involves conceptually-driven, schema-based processing. This is what leads to ‘understanding’ and explicit learning. The learner does more than simply attend to the difference between ‘made me go’ and the pre-existing construction ‘made me to go’. Symbolic knowledge comes into play leading the learner to understand that the verb ‘make’ takes a bare infinitive not a to infinitive. Schmidt (1994) was hesitant to claim that noticing leads to awareness at the level of understanding in this way, but he did not rule it out.

  Factors affecting attention

  Why do learners sometimes fail to attend to things in the input and their own output? Why do they attend to some elements in the input/ output and ignore others? What factors are likely to cause learners to pay attention?

  In part, the answer to the first question lies in Tomlin and Villas’ theory of attention. Attention is unlikely if learners are not alerted to attend. In other words, if—like Schmidt’s Wes (see Chapter 4)—learners are not interested in becoming more grammatically correct and are adept at using existing linguistic resources to meet communicative needs, they will not be alerted to grammatical tokens in the input and—even if they do attend to them—will not engage in the rehearsal needed for learning. Orientation is also needed. Learners—like Wes—who are focused more or less exclusively on meaning, may extend their vocabulary, but are unlikely to pay attention to grammatical features. Working memory is limited in capacity and—as VanPatten (1990) showed—if learners attend to one aspect of language (for example, vocabulary) it is likely to be at the expense of some other aspect (for example, grammar).

  Furthermore—as the above discussion of detection and executive control processes makes clear—what learners register and what they process in their working memory is influenced by their existing linguistic schemata. We have seen that L1 schemata block attention to L2 forms preventing detection and processing from taking place. L2 sounds—for example—are filtered through the L1 phonological system. Existing L2 schemata can also block attention to form–meaning mapping. In accordance with the One-to-One principle, (Andersen 1984) learners typically map a particular meaning onto a single invariant linguistic form (for example, they select a single pronoun form to express possession, irrespective of gender or number). This can block attention to variation in possessive pronoun forms in the input.

  Some linguistic forms, however, are inherently less ‘noticeable’ than others. Redundant forms such as third-person -s in English are not detected in the input because they play no role in understanding the grammatical meaning of a sentence. Loewen, Erlam, and Ellis (2009) investigated whether learners who were exposed incidentally to an input flood of sentences containing third-person-s demonstrated any acquisition of this feature and found that they did not. Even features such as verb + -ed may remain unattended to if pastness is encoded by a more salient item (for example, an adverbial such as ‘yesterday’).

  To explain why learners are predisposed to attend to some linguistic features rather than others VanPatten (1996) proposed a number of input-processing principles, which he claimed constitute natural processing tendencies that determine how learners allocate attention during online processing. Table 8.3 gives two of these principles. VanPatten has continued to work on these principles over time, modifying them slightly and adding to them. See for example VanPatten (2004b).

  Principle Description

  P1 Learners process input for meaning before they process it for form.

  P1 (a) Learners process content words in the input before anything else.

  P1 (b) Learners prefer processing lexical items to grammatical items (for example, morphological markings) for semantic information.

  P1 (c) Learners prefer processing ‘more meaningful’ morphology before ‘less or non-meaningful morphology’.

  P2 For learners to process form that is not meaningful, they must be able to process informational or communicative content at no (or little) cost to attention.

  Table 8.3 Examples of input processing principles (Van Patten 1996: pp. 14–5)

  One factor that can induce noticing is the frequency with which linguistic forms appear in the input. N. Ellis (2002) argued that learners have a built-in capacity to ‘count’—i.e. to register the frequency with which tokens and types occur in the input—and that every time a form is registered, some change in the learner’s connectionist network takes place. However, frequency—i.e. positive evidence—does not guarantee noticing. As Ellis pointed out, the learner’s L1 prevents learners from registering incidences of a token. The most frequent form in t
he input to English learners is the article ‘the’, but this form is late acquired and—in the case of many learners whose L1 lacks articles—may never be acquired. For some features, then, negative evidence may be needed. Negotiation involving corrective feedback that provides learners with negative evidence (see Chapter 7) may be crucial for ensuring that adult L2 learners pay attention to forms that would otherwise not be salient to them.

  There are other ways of making features salient in the input and thus of increasing the chances they will be attended to. In Chapter 10 and Chapter 11, we will examine the roles played by explicit and implicit instruction in enabling learners to focus their attention on specific linguistic features in both the input and their own output.

  Summing up

  What can reasonably be concluded about the nature of attention and its role in L2 acquisition?

  Attention is a complex construct. It involves both detection—i.e. the registration of attributes of a stimulus—and what Robinson (2003) calls rehearsal. Detection involves short-term memory and may not lead to rehearsal. Rehearsal allows for connections to be made between what has been registered and existing L2 knowledge, as in noticing-the-gap.

  For some—for example, Schmidt—attention is necessarily a conscious process: it involves noticing. For others—for example, Tomlin and Villa—however, it can take place without consciousness. I have tried to reconcile these different positions by suggesting that whether attention is conscious or not depends on whether it entails only detection or also complex processing.

  Learners detect exemplars of symbolic categories, not the categories themselves. However, category formation can arise in working memory as a result of rehearsal.

  Attention can lead to changes in both implicit and explicit knowledge systems. Elaborative rehearsal is needed for the development of explicit representations.

  Attention is involuntary—i.e. it happens without any intention on the part of the learner—but can also be voluntary—i.e. learners can consciously elect to attend.

  In the case of involuntary attention, input frequency—i.e. positive evidence—drives attention but learners’ natural processing tendencies (for example, VanPatten’s Input Processing Principles) and the blocking effect of the L1 can prevent detection occurring.

  Interaction that involves negotiation—especially corrective feedback—provides learners with negative evidence that may be necessary for some linguistic forms—i.e. those that are redundant or non-salient—to be noticed.

  Instruction that directs learners’ conscious attention to linguistic forms can also help to overcome the limitations in detection.

  I have two further points to make—one generally recognized, the other more speculative. First, not everything that is processed in short-term or complex working memory results in observable changes in interlanguage. Second—more speculatively—the different processes involved in detection and rehearsal may account for differences in how children and adults process input. Children detect but are less likely to rehearse. Adults with their more developed working memory are more likely to rehearse what they detect. Perhaps it is children’s rich capacity for detection that enables them to develop native-like competence in an L2. In contrast, the adult’s complex processing of input leads to a faster rate of learning initially, but also interferes with the implicit learning required for achieving full competence.

  Cognitive theories of L2 acquisition

  Cognitive theories of L2 learning address how attention leads to change in L2 representations and therefore draw heavily on the constructs and research I have considered in the previous sections of this chapter. I begin by considering how implicit and explicit learning differ and then move on to skill-learning theory and Processability Theory. I will conclude by examining Complex Adaptive System Theory.

  Implicit and explicit L2 learning

  There are two different traditions in cognitive SLA research. One has investigated the difference between incidental and intentional learning and the other between implicit and explicit learning. Incidental acquisition occurs when learners ‘pick up’ linguistic features from input. Much of the research we considered in Chapter 7 examined incidental acquisition. Intentional learning occurs through the learners’ deliberate efforts to learn a specific feature. Implicit learning is generally defined as learning that occurs without intention and without awareness although—as we have already seen—there are differing views about whether any learning without some level of awareness is possible. Explicit learning involves reflection about language that leads to ‘understanding’. Here I will focus on research that has addressed implicit/explicit learning.

  Research on implicit/explicit learning began with a series of studies by Reber (for example, Reber 1976; Reber, Walkenfeld, and Hernstadt 1991). In these studies, the implicit learning condition involved asking people to memorize a set of letter strings generated by an artificial language. In the explicit condition, they were asked to figure out the underlying rules of the same letter strings. Both groups then completed a judgement test where they had to decide if the strings of letters followed the same rules as the strings in the learning conditions. The main findings of these studies were: (1) there was clear evidence of implicit learning—i.e. learning without awareness; (2) there was no difference between the test scores of the implicit and explicit learning groups in the case of simple rules, but implicit learning proved more efficient for complex rules; and (3) the test scores of the explicit group demonstrated much greater individual variation than those of the implicit group—reflecting the mediating role of the learner’s analytical skills. It should be noted, however, that—because the artificial grammars that Reber used in his studies were purely formal in nature—his research only addressed whether people can learn linguistic form implicitly, not whether they can learn how form maps onto meaning.

  Other cognitive psychologists have also disputed Reber’s claim that learning can occur without awareness. The main problem in resolving this issue lies in the way in which learning is measured. Learning without awareness is generally considered to be evident if learners demonstrate learning in a judgement test following training but are unable to report on what they have learned. Shanks (2003), however, argued that a better measure of awareness is reaction time at the point of learning—i.e. during training. He concluded that when this measure is used, there is no clear evidence of null-awareness. Current work in cognitive psychology—for example, Rebuschat 2013—is directed at identifying valid ways of determining whether learning involves awareness. Similar work is taking place in SLA—for example, Leow and Hama 2013. I address methodological issues involved in investigating implicit learning in a later section in this chapter.

  A number of SLA studies claim to provide evidence of learning without awareness. Foremost among these are studies conducted by John Williams and his co-researchers—for example, Williams 2005; Leung and Williams 2011. However, such studies continue to be criticized on the grounds that they do not clearly distinguish between learning as a process and learning as a product and that to convincingly show learning without awareness it is necessary to examine the cognitive processes involved at the point of learning. DeKeyser (2003) reviewed a number of L2 studies of implicit learning and concluded ‘there is very little hard evidence of learning without awareness’ (p. 317). As we have seen, Schmidt (2001) reached a similar conclusion.

  However, if implicit knowledge is seen as consisting of a complex of weighted connections between neurons that do not encode symbolic categories—as in connectionist models of representation—it is difficult to see how else knowledge can develop other than through implicit (unaware) learning. Similarly, if learners consciously set about trying to learn a grammatical rule, explicit learning is clearly involved. In other words, distinguishing implicit and explicit learning at the theoretical level seems unproblematic. The problem arises because implicit and explicit processes are intertwined in any use of language, with some types of use—for example, careful
ly editing a piece of writing—drawing heavily on explicit processes and other types—for example, free conversation—drawing on implicit processes. As learning takes place through language use, implicit and explicit processes are potentially always involved.

  This is the position that N. Ellis (2005) evinces when he talks about the ‘collaborative mind’. He accepts the disassociation of the implicit and explicit processing systems, but also recognizes that they are ‘dynamically involved together in every cognitive task and in every learning episode’ (p. 340). According to N. Ellis, learning commences with the explicit representation of a formula—a holistic form that encodes a particular semantic or pragmatic meaning. At this point, learning is a conscious process. However, once a form–meaning conjunction has been established, noticing is no longer necessary and the pattern-recognition mechanisms of connectionist memory involved in implicit learning automatically take over. Gradually more abstract schema evolve as chunks are subconsciously analysed.

  In Chapter 1, we examined a number of interface positions. The non-interface position (Krashen 1981) claims that ‘acquisition’—i.e. implicit learning—and ‘learning’—i.e. explicit learning—are distinct processes. This is compatible with the N. Ellis’ claim—as he acknowledges. However, the other claim of the non-interface position, namely that ‘acquisition’ and ‘learning’ are unrelated and that the products of ‘learning’ play no role in ‘acquisition’ are clearly incompatible with Ellis’ position, which is more supportive of a weak-interface position. I too have argued for this, claiming that the interface arises when learners bring their explicit knowledge to bear on processing input and monitoring their own output (R. Ellis 1993)NOTE 4. If we accept that the mind is ‘collaborative’ in the way N. Ellis proposes, then a non-interface is not defensible. However, acknowledging a role for explicit processes does not necessitate abandoning the distinction between implicit and explicit learning and implicit and explicit knowledge.