From managing memory to multitasking: Exploring the role of cognitive control

Abstract

Humans are remarkably adaptive agents, able to succeed in a variety of environments. Central to human versatility is our ability to rapidly reconfigure our cognitive processes for the performance of new tasks through appropriate adjustments in perceptual selection, response biasing, and maintenance of contextual information. Developing a better understanding of the processes behind such adaptability, referred to as cognitive control, is the focus of this dissertation. A limitation in the capacity for cognitive control has historically been a central assumption of some of the most influential theories of cognition. This dissertation reviews different accounts for why cognitive control may be limited, as well as a different conception of cognitive control, in which restrictions in control-dependent processing represent the purpose of control, rather than a limitation in its function. In this framework, cognitive control demanding processes may be difficult not because we have a limited capacity for cognitive control, but instead because there is a tradeoff between learning efficiency and processing efficiency. Study 1 discusses how studying multitasking can give us key insights into how cognitive control manages the tension between leveraging existing general purpose representations in long term memory to quickly perform a new task and undergoing the costly construction of new representations to more accurately perform specific tasks. Multitasking behavioral data from laboratory and online studies, as well as computational modeling of multitasking behavior, show how existing representations are initially used to perform multitasking, but training can cause the brain to build new, separated representations. Study 2 extends the findings of Study 1 by having participants do the same multitasking experiment while in an fMRI scanner to measure neural representation separation over time. Study 2 presents the fMRI study design and preliminary findings from pilot subjects. Study 3 presents behavioral, computational modeling, and neuroimaging evidence for how old representations in episodic memory are automatically activated and injected into working memory during new information integration. Together, the results from these studies give insights into how limitations in memory, multitasking, perception, and cognitive control may be related to a common mechanism: interference between shared representations.