Attention and task performance

Attention to linguistic input is considered to be essential for L2 learning as a key process underlying selection and noticing (Robinson, 2003b; Schmidt, 2001). Similarly, attentional mechanisms are also at work during L2 production. In this sense attention can be viewed as a process serving the information-processing sequence that we call task performance, from perception to the execution and monitoring of a response (Eysenck & Keane, 2000). In line with the purpose of this study, this section presents a cognitive-level description of the attentional processes involved in L2 performance. In this context task performance in a second language can be understood as “the skilled deployment of existing knowledge to achieve task goals” (Robinson, 2003b, p. 632). The theoretical models of attention presented in the below attempt to describe how attentional resources are structured and drawn on during the completion of a task.

Information-processing involves three stages: perceptual encoding, central processing and responding (Eysenck & Keane, 2000). First, incoming information is perceived and selected for further processing, then, once the selected information is processed in working memory, a response is formulated and executed. These stages of the information-processing sequence are served by two aspects of the concept of attention:selection of what we would like to further process of all the incoming information and the allocation of attentional resources during information processing and response formulation.

According to several lines of research attention is limited in capacity (Kahnemann, 1973; Wickens, 2007).

Attentional resources are also responsible for sustained attention, usually described as the effort one invests into the execution of a particular task (Robinson, 2003b; Wickens, 2007).

L2 performance and learning involve both the selective or ‘filter’ and the ‘resource’ or ‘fuel’

aspects of attention (Wickens, 2007). Selective attention in the SLA research community is considered to be necessary “as a means of action control [...] Actions are responses to task demands, and allocation of attention to input with the goal of meeting these

demands is the result of control processes operationalized in short-term/working memory.”

(Robinson, 2003b, p. 634).

Research studies comparing attention to the concepts of ‘capacity’, ‘resource’ or ‘fuel’ implies two important assumptions: 1) attention serves and enables the full sequence of information-processing activities, and 2) attention is limited and can be depleted by the demands of the task or tasks being executed (Robinson, 2003b; Eysenck & Keane, 2000; Wickens, 2007).

As section 2.1.3 on task characteristics has shown, tasks vary in the cognitive and interactive demands they impose on our attention. Complex tasks drain our attentional resources to a greater extent than simple tasks as performing two tasks simultaneously and dividing our attention between them can be more difficult than performing one single task. Therefore, the same person may perform differently on a task depending on its degree of complexity and/or whether more than one task needs to be attended to at once. That is why, in the SLA context as well, attentional task demands are considered to be a factor contributing to within-learner performance and interlanguage variation (Robinson, 2003a). There is general agreement between researchers in both cognitive psychology and SLA that tasks demands vary and draw differently on our attentional resources accordingly, but conflicting propositions have been advanced regarding the concept of attentional resource pools and how they work.

The single capacity view (Kahneman, 1973), as its name suggests, argues that there is a single volume of attention, limited in capacity, which can be drawn on during task performance. Although attentional capacity is limited, its limits can be extended if the level of effort or motivation is high because of the physiological arousal effort these may induce. To account for breakdowns and success in dual task performance, this theory argues that two tasks can be done simultaneously if the combined demands of the two tasks do not exceed the total resources of the available attentional capacity. Single capacity theories have been criticized on grounds that they cannot explain how attention is divided during dual or multiple task performance as they rely solely on the difficulty level of tasks in their arguments.

Based on an analysis of dual task interference data Wickens (2007) proposed a multiple resource model of attention, which attempts to account for the differences between the ways tasks interfere with one another by analyzing their inherent structure along the various dimensions of the information-processing sequence. For example, driving a car requires visual/spatial processing and a manual/spatial response while maintaining a conversation calls for auditory processing and a vocal/verbal response. The multiple resource model suggests that attentional capacity comprises several dichotomous resource pools along the dimensions of the information-processing sequence. Three of the proposed four dimensions are relevant for language behaviour, namely 1) stages of processing which includes perception/cognition and response, 2)modalities of processing including visual and auditory, and 3)codes of processing and responding referring to verbal, vocal or spatial. According to Wickens (2007) each of the dimensions of task structure draws on its corresponding attentional resource pool, and two tasks executed simultaneously and drawing on the same resource pool enter into competition for attention. Therefore, performance on each would probably deteriorate, although it has been noted that practice and automatization processes may significantly improve task performance by reducing the amount of attention necessary for the execution of the tasks. However, when two tasks use attention from different resource pools, they do not enter into

competition, which explains why driving and singing along with the radio may be easier to do than maintaining a conversation and listening to the news at the same time.

A key element in Wickens’ model is the identification of the dimensions differentiating between the various resource pools since this is what his explanation of divided attention during dual task performance hinges on. Research into task interference has provided evidence which shows that the identification and definition of Wickens’ dimensions would have to be made more specific to be able accommodate all the findings. For example performing spelling and arithmetic tasks orally would both be cast in the verbal/auditory perception - vocal/verbal response pools respectively, and yet performing a spelling and an arithmetic task versus performing two arithmetic tasks simultaneously yielded different performance results (Hirst & Kalmar, 1987).

Another issue raised in connection with both Kahneman’s and Wickens’ models concerns the basic assumption in both models that attentional capacity is limited. Subsequent ‘interference’ models (e.g., Navon, 1989; Sanders, 1998) do not resort to the argument of attentional capacity limitations when attempting to explain deteriorating performance in dual- or multi-tasking situations. Instead, picking up on Wickens’ suggestion of multiple resource pools, interference models propose that reduced performance is a result of interference and confusion caused by shifts of attention between tasks. Higher task demands, characterized by a greater amount of stimuli/input and many or similar response alternatives leads to a competition for attention between the same types of codes or to cross-talk between them (Sanders, 1998).

Thus the current trend in attentional theories is to move away from the idea of attention as limited in capacity towards interference during the process of resource allocation, all the while maintaining Wickens’ idea of multiple attentional resource pools.

The next section 2.1.4.2 will describe how SLA research has drawn on theories of attention to account for performance limitations, and what predictions the different models of task difficulty/complexity make as regards L2 task performance.