What limits dual-tasking in working memory? An investigation of the effect of sub-task demand on maintenance mechanisms employed during dual-tasking

A number of models of working memory have been proposed since the seminal work of Baddeley and Hitch (1974) on the Multiple Component Model (MCM). Subsequent MCM research focussed on developing a theoretical framework based on modality-specific stores that can operate in parallel during dual-tasking. The MCM can be contrasted with theories of working memory that assume an attention-based domain-general shared resource responsible for both short term retention as well as on-line cognition, such as the Time-Based Resource Sharing (TBRS) model (Barrouillet, Bernardin, & Camos, 2004; Barrouillet, Bernardin, Portrat, Vergauwe, & Camos, 2007). The TBRS model assumes that short-term memory is dependent on access to attention, and any diversion of attention results in increased forgetting. The model describes ‘refreshing’ as the process of serially bringing memory items briefly into the focus of attention. Barrouillet and colleagues have demonstrated in numerous studies that memory spans lower as the cognitive demand of the secondary task increases - findings that are incompatible with the MCM. However, Camos, Mora, and Oberauer (2011) found that both sub-vocal rehearsal (the verbal maintenance mechanism described in the MCM) and attention-based refreshing can be selectively employed by participants depending on task demands. Since TBRS methodology compares spans measured under different cognitive load levels that are the same for every participant, we were interested in whether ensuring that secondary task demand was set within each participant’s abilities would avoid ‘over-taxing’ the working memory system and reduce dual-task costs. Our initial investigations re-measured memory and processing spans under dual-task conditions with secondary tasks’ demand titrated according to each individual’s measured ability (Experiments 1 and 2, and Doherty & Logie, 2016). We found that memory span was unaffected when processing demand was titrated, but that processing performance was lower when memory load was set above participants’ span. Subsequent experiments (3-8) investigated the effect of setting memory and processing load ‘below span’, ‘at span’, and ‘above span’ on memory and processing accuracy during dual-tasking. Overall it was found that processing resources can be reallocated to support memory performance but memory resources cannot be reallocated to support processing performance. We interpret the results as evidence for specialised memory resources and rehearsal mechanisms that can be supplemented by attention-based processes once storage capacities are exceeded. Experiments 6-8 aimed to encourage the use of phonological- or attention-based rehearsal mechanisms for verbal short term memory by either introducing articulatory suppression (AS) or shortening available encoding time for memory items. It was found that participants exhibited shared-resource effects when they completed the dual-task under AS, suggesting a shift to attention-based rehearsal. When encoding time was limited participants’ memory performance during dual-tasking was unaffected by concurrent processing load, suggesting the use of a rehearsal method which did not require access to attention. Experiment 9 investigated whether participants could dynamically allocate attention to one task or the other, and found that while ‘priority’ tasks received no benefit, non-priority tasks exhibited a marked decrement in performance. We conclude that the perceived incompatibility between the MCM and attention-based theories of working memory such as the TBRS model may be more apparent than real, and suggest that future research should incorporate procedures and methodological considerations that take into account findings from both literatures