Theories that Support Multiple Components

The relative lack of dual-task disruption seen with particular task combinations has been linked to the operation of a multiple-component working memory system thought to offer online processing and temporary storage of information by means of a number of specialized cognitive functions (Baddeley, 1986; Baddeley & Logie, 1999). Within a multiple component model, each task is believed to draw specialized resources for visuospatial and verbal processing and storage (Baddeley et al., 1991). This perspective proposes that it is the types of tasks combined rather than the overall cognitive demand of the dual-task requirement that determines performance reduction in working memory (Cocchini et al., 2002).

Several studies have used combinations of processing and storage tasks to examine the influence of dual-tasking on working memory capacity (Cocchini et al., 2002; Duff & Logie, 2001; Friedman & Miyake, 2004; Maehara & Saito, 2007; Saults

& Cowan, 2007; Towse, Hitch, & Hutton, 2000). Cocchini et al. (2002) conducted two experiments, in which participants were required to perform pairwise combinations of a verbal memory task or a visual memory task with either a perceptuomotor tracking task (Experiment 1) or an articulatory suppression task (Experiment 2). They found minimal disruptions when two tasks were combined. Interestingly, there was a drop in accuracy when articulatory suppression was combined with a verbal memory task but not with the visual memory task. Based on these findings, Cocchini et al. concluded that a multiple component working memory model provided a better account of performance in concurrent immediate tasks than either models that assume a single capacity system for processing and storage or a limited capacity attentional system combined with active memory traces.

Duff and Logie (2001) conducted two experiments that examined the nature of working memory involved in span tasks using a version of the sentence span task (Daneman & Carpenter, 1980). In Experiment 1, participants verified sentences to determine whether the sentence was plausible and remember the final word of each sentence for later serial recall. Before the combined task, participants were required to perform each component of the task on its own. They found an overall decrement in both memory span and verification span when these tasks were performed

together but suggested that this was due to an increased load on the central executive. In Experiment 2, Duff and Logie (2001) decreased the degree of integration within the general procedure: the processing task used arithmetic verification, whereas the memory task involved the recall of unrelated words. They argued that if performance was supported by a single flexible resource, then a less integrated task should require more resources for time sharing. Again they found only

a small drop in performance in the combined task condition. They suggested that overall their results supported the multiple component models and argued that the central executive coordinates combined task performance as well as dealing with processing. The overall drop in performance on the two tasks can be attributed to an increase of load imposed on the executive by coordination. The alternative, from the single resource perspective, is that if both the processing and storage demands stretch working memory to its limits, then combining the task elements should result in a substantial drop in performance, greater than that observed in the experiments.

The storage-processing relationship has been examined by manipulating the processing requirements on storage activities and the role of storage on processing activities (Friedman & Miyake, 2004; Maehara & Saito, 2007). These have used the stimulus order effect, whereby the set of processing operations is held constant, and the completion order of individual activities is varied (Towse et al., 2000). Typically in these studies a trial comprised of either a short-duration task as the first activity and a long duration task as the last activity, or vice versa. This allowed the retention interval to be varied while the overall processing requirements of the task were kept constant. Towse et al. (2000) prepared mixed lists of short and long sentences and made short-final and long-final lists for their working memory span tasks. They found lower span scores for the long-final lists than for the short-final lists. They suggested that participants switch attention flexibly from the processing to the storage task requirements and then back again. They argued that a single-capacity model (like the one proposed by Just & Carpenter, 1992) could not fully explain the stimulus order effect because the same amount of processing was required in the short-final and

long-final list. Overall, both lists had the same processing requirements, thus the only difference between the two was the temporal structure.

More recent research has contradicted the findings of Towse et al. (2000) by distinguishing between mechanisms involved in forgetting and storage/maintenance. Maehara and Saito (2007) used the stimulus order effect in addition to the processing time effect to examine the specific predictions from the representation-based

interference account of working memory span. The representation-based account suggests that the nature of the material used in processing and storage episodes is critical in determining the magnitude of the stimulus order effect due to the

interference of similarly attributed representations. They found that when items were from the same domain, the representations generated during the processing tasks interfered with the memory representations. Further, they found that the processing time effect in verbal-verbal and spatial-spatial working memory span tasks was consistent with the prediction of the representation-based interference account. However, inconsistent with the representation-based interference account they also found a processing-time effect in the verbal-spatial and spatial-verbal working memory span task. These results suggested that forgetting mechanisms in the working memory span tasks were domain specific. They also argued that storage or maintenance activities require attentional control and seem to involve domain-

general properties as demonstrated by their findings on the processing time effect. Such activities may interrupt or interfere with processing activities. In contrast to Towse et al. (2000) who suggested a multiple component theory, Mahera and Saito (2007) proposed a more hybrid theory of working memory which incorporates both

single and multiple capacities. Thus, the ability to recall information is domain specific but there is a central processor that is influenced by processing time.

Despite extensive research into the relationship between processing and storage there is little consensus on whether both activities are performed by a single mechanism or multiple mechanisms. Although most studies acknowledge that there is at least a small dual-task decrement, the underlying reasons may be due to the difficulty in coordinating processing and storage requirements (multiple component theories) or due to additional processing demands on a single capacity system

(single component theories). What is clear regardless of theoretical orientation, is the importance of attention. Attention is central to both central resource and multiple capacity theories as is the concept of cognitive control. This will be discussed in detail in the upcoming section.

1.5 Task-Switching and Working Memory and its Relationship to Cognitive