!
One primary way in which learners are purported to ‘pick up’ language is through interaction during a task. For Long (1985, 1996), this interaction is pivotal to language acquisition as interactants work to achieve comprehensible input and output through negotiation of meaning. As Foster and Ohta (2005) summarise,
!
“In these negotiations, problem utterances are checked, repeated, clarified, or modified in some way (lexically, phonologically, morphosyntactically) so that they are brought within the optimum i+1 level. The value in these negotiations, especially in group work, is that they can provide i + 1 input which is made-to-measure for individual learners and their current interlanguage level.” (p.405)
!
To a great extent, pedagogic tasks in TBLT are envisioned with this meaning-focused interaction between participants in mind. However, Long’s construct of negotiation for
meaning (NfM) is not the sole source of classroom interaction available during work on a
task. Ellis et al (2001) discuss teacher led meaning-focused interaction that occurs without the identification of a ‘problem utterance’. In these cases, the motivation is rooted in a teacher, or other interlocutor, desiring to direct the listener’s attention more towards specific forms without this shift in attention relying on a communication
breakdown as a prompt. This more strategic employment of meaning negotiation can also be initiated by a native speaker or, potentially, a more proficient L2 learner. Foster and Ohta (2005) similarly point to collaboration itself as a valuable means for learners to address knowledge gaps.
Irrespective of the source and motivation for the shift in focus, this focus-on-form (Long, 1991) is an essential part of TBLT (Ellis, 2013). The crucial point Ellis stresses is that the task itself maintains a primary focus on meaning while allowing for learners to attend to form within the context of task performance. In this way, form is mapped to meaning. To put it more simply: in TBLT, grammar instruction is contextualised. It is not treated as separate knowledge (or a skill) to be first acquired and then later
proceduralised (i.e., in the manner and order described in an early conceptualisation of skill theory by Anderson, 1983).
This view of interaction discussed above carries a marked cognitive focus, as interaction is chiefly conceptualised as the acquisition of knowledge (or schema) and the ways in which the brain processes this information. Thus attention, in terms of a learner’s
propensity for noticing certain features of the language (Schimdt, 1994, 2001), is seen as a crucial part of acquisition. Much research in SLA regarding tasks (e.g. Skehan, 1996; Skehan and Foster, 1997; Robinson, 2001) has investigated the ways in which different features of tasks can facilitate attention to different aspects of information processing and their resultant effects on language production.
This attentional aspect of language processing, involving the relation between working memory and attention, has received considerable focus within cognitive approaches to SLA research. In a consideration of task design, some researchers (e.g. Skehan, 1998; Robinson, 2001) point to the ways in which tasks, by the features of their design, influence how learners variably allocate attentional resources in order to meet the particular demands of different tasks. Motivating this research is an understanding that the ways in which different tasks predispose different outcomes would provide support for deliberate task selection and task sequencing in order to meet given pedagogical objectives. For this area of research, an influential model of language production and comprehension is the serial processing model proposed by Levelt (1989, 1999). In this model, language production is described as a serial process of three components: the conceptualiser, the formulator, and the articulator. The conceptualiser initiates an utterance by establishing of a goal for communication and then, through first macro- planning, and then micro-planning, it sends subdivided portions of this message to the
formulator. This formulator first morphosyntactically, and then phonetically, encodes each portion of the message as it receives it from the conceptualiser. When this dual encoding is completed for a given portion of the message, it is sent to the articulator which renders the phonetically encoded plan into actual speech. This process works in the given order, and once the conceptualiser sends information to the formulator, that, in turn, is sent to the articulator, the various phases of the process can work in parallel. Thus as a speaker is articulating part of a message, his or her mind is already processing both the content and then form of the next part of the message for articulation.
!
3.2.1 Task design features and their effects on language production
!
Levelt’s model of L1 oral language production informs two prominent hypotheses for L2 language processing that differ in their interpretations of how this processing model controls language production. The first hypothesis is a model of L2 language processing known as the Limited Attentional Capacity (LAC) model proposed by Peter Skehan (Skehan, 1998, 2003, 2009; Skehan et al, 2012). In this model, Skehan proposed that increased task difficulty, which refers to various design features of tasks that make greater processing demands, will require more attentional resources. The need to allocate
attentional resources in order to complete more difficult tasks is the result of L2 learners lacking true parallel processing ability to support a dual mode system that employs both rule-based and exemplar-based systems of language processing (Skehan, 1998).
Importantly, Skehan draws on the views of VanPatten (1990) and proposes that these attentional resources of working memory are limited and spring from a single pool. As a consequence, increased attention to the greater demands of more difficult tasks will be at the detriment of attention to other areas of task performance. As Skehan explains,
!
“Processing-based analyses of tasks are concerned with their information-processing load, and effectively focus on the difficulty of the task. The assumption is that more
demanding tasks consume more attentional resources simply for task transaction, with the result that less attentional resources are available for a focus on form.”
!
As a result, he initially posited that there would be a trade-off between fluency and complexity (Skehan, 1998). However, more recently he predicts that more difficult tasks will result in a trade-off between complexity and accuracy, with one being attended to at the expense of the other (Skehan, 2003; Skehan et al., 2012). This trade-off occurs
because increasingly difficult tasks will tax attentional resources to the point where there will not be enough attentional capacity to attend to both simultaneously.
For Skehan, the implication of this focus is that identifying the difficulty of tasks (in terms of their variant demands on cognitive load) can inform the selection of tasks to match pedagogical objectives. His LAC model favours appropriate task selection that alternates focus between the complexity, accuracy and fluency of output, to help foster interlanguage development by first pushing the limits of that interlanguage, and then pushing control of that interlanguage. Thus, he claims that, through task-based research, we can identify which tasks predispose learner attention towards either features of their output, discourse features, or particular language structures, or a combination thereof. More recently, Skehan (2009) proposes and extends (Skehan et al, 2012) his framework for organising the various influences of task design features on second language
performance. In contrast to his earlier classification system (e.g., Skehan, 1998), this framework is based more deliberately on Levelt’s model of speech production with the stages of “conceptualiser”, “formulator - lemma retrieval”, and “formulator - syntactic encoding” all matched with respective stage-specific influences that Skehan classifies as “complexifying / pressuring influences” and “easing / focusing influences.” (Skehan et al, 2012, p. 184). This means of classifying task design features is empirically motivated and for Skehan better identifies the specific stages of language production that are affected by certain contrasts in task design. This framework essentially identifies influences on task performance and the potential to lead students to either further develop their underlying interlanguage (through increasing task difficulty) or facilitate better control of their current interlanguage (through decreasing, or “easing” task difficulty). However, Robinson (2011a) points out that this newer framework does not provide a metric for
sequencing tasks based on these influences and also notes that Skehan has not offered a means of relating these influences to real-world equivalents of task performance.
Robinson (2001, 2005, 2007, 2011b) himself offers an alternative hypothesis for L2 language processing, one that also draws on Levelt’s model, which he calls the Cognition Hypothesis (CH) for L2 learning. In this hypothesis, Robinson proposes that “…
breakdowns in ‘action control’, not capacity limits, lead to decrements in speech production and learners’ failure to benefit from the learning opportunities attention directing provides” (Robinson, 2011b, p. 12). This reasoning is based off of Cromer’s (1974) earlier Cognition Hypothesis that proposes that L1 development results from cognitive and conceptual development. As a result, Robinson reasons that the degree of complexity involved in a task will have a direct effect on the language used to complete it, so that more cognitively complex tasks will push learners to use language that requires greater monitoring and control. He refers to task design features that facilitate this push as ‘resource-directing’. Rather than being a trade off, detrimental effects from tasks are not due to their difficulty (task complexity) but due to constraints on learner ability to attend to the task. These constraints are called ‘resource-dispersing’ and refer to
performative and procedural demands of the task that can divert learner attention away from language production. An important prediction of this hypothesis is that facets of language production can be attended to simultaneously, as they are proposed to draw from individual pools of attention. This is a counterproposal to Skehan’s perceived ‘trade- off’.
Robinson’s hypothesis also favours a deliberate sequencing of tasks, but bases this sequencing on their increased cognitive complexity, rather than an overt alternation between task effects that favour attention to either rule-based or exemplar-based processing systems (as is the case for Skehan). As he explains,
!
“…task-based learning, sequenced according to the cognitive complexity…[ ]…leads to progressively greater attention to, “noticing”, and elaborative processing and retention of input (Robinson 1995b; Schmidt 1983, 1990, 2001); progressively more analysis of the input and output occurring during task work (Doughty 2001; Muranoi 2000; Pica 1987),
and also progressively greater amounts of interaction which in part facilitate those attentional and analytic processes (Long 1996; Mackey 1999). That is, I argue both the cognitive processing, and interactive consequences of task sequencing decisions are mutually responsible for subsequent task-based language development.
(Robinson, 2005, p. 3)
!
Thus the reasoning of the Cognition Hypothesis is that the underpinning of any
implementation of tasks should be to establish a progression through tasks of increasing complexity and increasing interactivity. Broadly speaking, Robinson bases this reasoning on the observation that L2 learning ‘involves some recapitulation of a sequence of
conceptual development in childhood’ (Robinson, 2005, p.6). In this light, tasks can be sequenced so that resources are directed towards this function-form mapping of
increasingly complex conceptual demands, leading to a situation, described by Robinson, where, “…forms may be currently known but not well controlled, or if they are unknown then attempts to complete the task may make them more salient and
‘noticeable’” (Robinson, 2011b, p. 15).
To assist this sequencing of tasks for syllabus (and test) design, Robinson proposed a triadic componential framework to provide a taxonomic means of identifying task
features. In this framework, he proposes three dimensions of task design that have effects on performance: task complexity, task conditions, and task difficulty (Robinson, 2005, 2011b). These three areas are further subdivided on the criterion of whether the task design feature directs attentional resources towards more complex processes or disperses attentional resources from them. Task complexity addresses features of the tasks
themselves, while task conditions and task difficulty address variable interactive demands and individual differences between students respectively. All three areas have variables that can be manipulated to push learner attention, and this framework establishes means of controlling for these variables through either an increase along the resource-directing dimension or a decrease along the resource-dispersing dimension. Both options are claimed to push more complex and accurate output. In respects to individual learner differences (task difficulty), the Cognition Hypothesis predicts that greater differentiation
Both the Limited Attentional Capacity model and the Cognition Hypothesis offer similar but competing views for the role and limit of a language learner’s attentional resources during task performance. Both models share the view that consideration of these resources, in terms of a task’s demands on those resources, should inform task selection and the appropriate sequencing of tasks to reach overarching pedagogic goals. Both of these interpretations of Levelt’s model of language processing for L2 production similarly support the importance of noticing and the directing of attentional resources to different aspects (i.e., complexity, accuracy, and fluency) of task performance. The
principle differences between these two hypotheses are: 1) a disagreement over
attentional resources having either a limited single-source capacity or multiple sources without such a limited capacity; and 2) the distinction Robinson makes between features of task complexity, task conditions, and task difficulty that either direct learners’
attentional resources towards certain features or disperse them towards other factors involved with carrying out a task. Both hypotheses also make distinct predictions for the effects that task features will have on task performance. While both Skehan and Robinson agree that an increase in task difficulty will degrade fluency, Skehan proposes that this increase will result in learners prioritising either complexity or accuracy due to the limited capacity of their attentional resources. In contrast, Robinson’s hypothesis proposes that this increase in difficulty, if it is a resource-directing feature of task complexity, will push both accuracy and complexity. Both views are in agreement, however, that increased difficulty in task conditions (the resource-dispersing dimension of task complexity for Robinson’s taxonomy) will degrade all aspects of performance. This study recognises the theoretical justifications of both of these hypotheses, and ostensibly does not seek to provide support for one over the other. However, in order to create hypotheses for quantitative analysis, this study aligns itself with the central tenet of Skehan’s LAC model: that increasingly difficult tasks make greater demands on attentional resources and, given that attentional resources are limited, these increasing demands will result in a trade off between complexity and accuracy in language
production. Additionally, as a provision of this alignment, this study follows Skehan’s more recent proposal (Skehan, 2009, Skehan et al, 2012) that states that either increasing
or easing task difficulty (in this case for the conceptualiser stage) will result in a trade-off, favouring complexity and accuracy respectively.
What is left unresolved by this discussion are the effects that conceptual creativity has on language production. Earlier, Ellis’ (2003) eight principles for task selection and design were discussed and task difficulty was identified as a prime concern for teachers. Skehan and Robinson address this concern by proposing, with alternative views, that task difficulty predisposes learners to focus attention on different aspects of their language. However, there are tasks, such as devising a play, in which the language necessary for a completed script is not the same as the linguistic knowledge necessary to collaborate on that script. As an example, devising a story about two people waiting for a bus when a random accident occurs can be completed with relatively simple language. However, to collaboratively compose that same story, through a process which includes the
introduction and elaboration of ideas as well as the evaluation and selection of competing ideas, involves much higher level language skills. In such a case, while students may possess the linguistic knowledge to devise the play, they may lack the knowledge necessary to undertake such a task with others. The result of this duality is that task difficulty is not always strictly determined by the linguistic demands of the task. Ellis, amongst others, states that tasks are meant to optimise learner interaction during tasks to ensure opportunities for language development. What happens, then, when this interaction is either conducted in a context in which students share an L1 (which can alleviate the cognitive load of conducting a complicated task in the L2), involves
collaboration on a task with heightened demands on conceptual creativity, or both? The next section discusses this question.
3.2.2 Learner interaction during tasks
!
Investigating the effects of task features on task performance can provide empirical support for selecting and sequencing tasks according to the ways in which task demands differentially direct learners’ attention to the complexity, fluency, and accuracy of their output. However, such a focus diverts attention away from the beneficial interaction that
tasks and TBLT are structured to promote (and which Robinson mentions specifically in his CH). By foregrounding the effects that these task features have on the outcome of the task, rather than the effects they have on how participants actually undertake and
accomplish the task, such research neglects how different tasks affect the quality of participant interaction and the potential learning opportunities that this interaction facilitates. Earlier in this chapter, the role of interaction was presented from a
prominently cognitive-interactionist view (e.g., Long, 1985, 1996; Pica, 1994) that sees interaction as a tool for intra-mental processing. Yet as Foster and Ohta (2005) point out, sociocultural approaches to language learning view this same interaction as being
fundamentally social and inter-mental. In this view, the learner is not separable from their environment and as a consequence knowledge is not constructed by an individual but rather is the joint property of both the learner and a given social context. Learners interface with this social context; and thus language is acquired through social interaction. This vein of research has focused on the ways in which peers support, scaffold, or otherwise collaborate with each other in order to create and sustain
interaction. As Swain and Lapkin (2000) put it, drawing on the views of Vygotsky (1978) amongst others, “Language is understood as a mediating tool in all forms of higher order processing (e.g. attending, planning, reasoning) [and] furthermore, language derives its mediating cognitive functions from social activities” (p. 253-254). In essence, this view of language development, and by extension additional language development, claims that more advanced language is indicative of more advanced cognitive processes and,
importantly, these advanced processes (and the language necessary to mediate them) are first accessed by the learner inter-mentally through either social interaction with more capable interlocutors, or through co-construction of knowledge with more level-
equivalent peers.
Socially motivated and mediated collaboration has been researched using the constructs of languaging and language related episodes (LREs) (Swain, 2001; Swain & Lapkin, 1998, 2002). Put in simplest terms, languaging is when language is used to mediate more cognitively demanding concepts (Swain, 2007) whereas an LRE is when “students reflect consciously on the language they are using” (Swain, 2001, p. 53). As some have pointed
out (e.g. Pica, 1996), communication breakdowns are not the only instances in which students negotiate meaning and in this sense LREs represent moments of peer interaction that could conceivably facilitate L2 development in the same manner that Long argues for negotiation for meaning (NfM), but without the narrower focus of communication
breakdowns that is central to Long’s construct (Foster and Ohta, 2005). Swain proposes that these episodes of language mediated cognition demonstrate that language output, and learner self-monitoring of that output, are crucial parts of the process of learning (Swain, 2007).
During collaborative dialogues on consensus building activities, student pairs or small groups use these LREs to access and construct the language knowledge that is necessary for them to arrive at one solution and complete a given task. However, such meta-
linguistic talk does not, of course, constitute the sole topic of discussion for collaborative dialogue between partners. During work on a task, students collaborating with one another must also mediate the requirements of the task and their joint understanding of