NEUROIMAGING, BEHAVIORAL, AND COMPUTATIONAL INVESTIGATIONS OF MEMORY TARGETING Sean Matthew Polyn A DISSERTATION PRESENTED TO THE FACULTY OF PRINCETON UNIVERSITY IN CANDIDACY FOR THE DEGREE OF DOCTOR OF PHILOSOPHY RECOMMENDED FOR ACCEPTANCE BY THE DEPARTMENT OF PSYCHOLOGY

The conceptual focus of this dissertation is the ability of humans to target episodic memories, i.e., to re-access past states of the world encoded by the memory system. I describe a framework for understanding this process that relies on the interaction of three cognitive systems in the brain: semantic memory, episodic memory, and a context maintenance system that acts to probe the episodic memory system. I build upon previous studies of these systems, in which each of these cognitive systems is mapped onto a particular anatomical region of the brain. Respectively, these areas are posterior cortex, medial temporal lobe, and prefrontal cortex. This framework is investigated in a series of studies of the free recall paradigm. First, I describe a neuroimaging experiment in which I use pattern classification methods to track the second-by-second fluctuations of patterns of brain activity during free recall. The results of this experiment provide evidence of contextual reinstatement processes during the recall period. These results highlight the idea that context effects on memory organization can be carried out by any brain area that has both a relatively stable pattern of activity during the encoding process, and connections with medial temporal lobe brain regions. Second, a behavioral study manipulates the accessibility of memories by changing encoding task midway through the study list, causing the sets of items encoded in the context of each task to become isolated from each other in memory. The results of this experiment suggest that patterns of task-related activity act as an effective context, uniting memory traces encoded within-task and isolating traces encoded under different tasks. Finally, I describe a connectionist model of the interactions of these brain regions during free recall. I present a series of simulations that illustrate how the prefrontal system can manipulate the medial temporal memory system to flexibly extract previously encoded memories. The model also captures the “memory isolation” effects seen in the encoding task-shift experiment described above. A final set of simulations attempt to capture the pattern of deficits seen in elderly subjects in free recall (Kahana, Howard, Zaromb, & Wingfield, 2002).