Barrouillet et al. (2004) describe a new model of working memory proposing shared-resources with special emphasis on time-related decay of temporary memory traces. This Time-Based Resource-Sharing Model has received a large amount of attention, with numerous exploratory and confirmatory studies resulting in a fast-developing body of supporting literature. Many of the empirical findings reported by Barrouillet and colleagues are incompatible with the structure of working memory posited by the Multiple-Component Model (Baddeley \& Hitch, 1974), and appear to refute many of the domain-specific interference phenomena cited as evidence for this model. This dissertation focusses on one such paper by Portrat et al. (2008) in which a spatial processing task is shown to disrupt concurrent verbal memory performance (a pattern inconsistent with multiple-component theory). We present three experiments investigating possible methodological reasons behind Portrat et al.'s findings, and argue that certain flaws in their procedure (and in much of the TBRS literature) has led to an insensitivity to true WM processes. Evidence from these experiments demonstrate that when the cognitive demand of tasks in a dual-task paradigm are adjusted according to the individual abilities of participants the cross-modal effects reported by Portrat et al. disappear, but when task demand is set the same levels for all participants the effect returns. A fourth experiment demonstrates how the domain-specific interference can be observed even when task demand is adjusted according to participants' abilities. An analysis of these results is argued in support of a multiple-component structure of memory, with considerations relating to Logie's (2011) hypothesis concerning dynamic reallocation of domain-general resources when the working memory system is under high load.