As humans, we show striking adaptability in our behavioural repertoires. Despite this apparent ability to adapt and assimilate a large variety of behaviours to suit our current contexts, we show some striking limitations in how many tasks we can perform at any given moment. Although some behaviours can appear to be performed concurrently with ease, such as talking while walking, we show a stark inability to combine other tasks and skills; imagine compiling a grocery list while recalling your partner’s phone number. The observation that attempting to perform multiple tasks concurrently results in at least one of those tasks being performed more slowly, and with less accuracy, was among the first to be made when cognitive psychology was emerging as a scientific discipline (Telford 1931; Welford 1959). Interestingly, among these early observations, it was also noted that such performance costs can to some extent be reduced with practice (Telford 1931), suggesting malleability in how tasks are performed that carries consequences for multitasking operations. Since then, efforts have been made into understanding the nature of the putative operations that give rise to multitasking costs, the neural computations and architectures that instantiate those operations, and the neural and functional changes that drive improvements in multitasking performance. The aim of this chapter is to synthesise the in-roads that have been made into understanding the neural basis of multitasking costs and their practice-induced remediation, and to use that knowledge to propose the next steps forward in our understanding.