Examining the Transient Neural Dynamics Underlying Working Memory Maintenance for Complex Visual Stimuli

Working memory (WM) is the temporary storage of information to accomplish a future goal. The WM delay period is the time after encoding but before retrieval when information is being maintained, typically in the absence of relevant stimuli. Understanding how the brain supports maintenance during the delay period, and how neural activity and connectivity are related to memory is critical for advancing both basic knowledge as well as informing declines in memory and cognition related to neurodegenerative diseases and healthy aging. An open question in the field of WM research is how information is stored during this delay period. One theory suggests persistent neural activity supports the storage of information while another theory suggests rapid synaptic weight changes (i.e., an activity-silent mechanism). While support exists for both theories, numerous confounds complicate the experiments designed to distinguish between these two theories. Most notably, studies typically use simple stimuli with short, predictable, unfilled delay periods, with few studies examining this open question using complex visual stimuli. For this dissertation four EEG experiments were conducted to answer three questions: 1) Does the type of maintenance technique used during the delay period modulate the neural activity for complex visual stimuli? 2) What are the load-dependent delay activity and connectivity patterns associated with maintenance of complex visual stimuli? And 3) How do patterns of delay activity and connectivity change when attention is sustained during unpredictable delay period durations? In the first set of experiments, we examined the role of rehearsal in maintaining complex visual stimuli. We looked at the impact of rehearsal versus suppression of rehearsal using novel naturalistic scenes that contained semantic content and phase-scrambled scenes that lacked semantic content. The benefit of rehearsal was associated with increases in theta and alpha band amplitude, but only when the stimuli lacked semantic content. The overall pattern of change in amplitude was similar for rehearsal and suppression of rehearsal, regardless of the type of stimulus. In the second set of experiments, we examined the role of attention in maintaining complex visual stimuli. We used a similar set of stimuli to the first experiments, employing a load manipulation (low load-2 images, high load-5 images) while perceptual interference was present. While increasing the WM load, particularly with complex stimuli, places a greater demand on attentional resources, interfering stimuli may compete for the available resources. This was confirmed in the examination of theta and alpha amplitude, as amplitude was reduced for the high WM load as compared with the low WM load. We also analyzed functional connectivity to identify the underlying brain networks that facilitated performance for the low load condition and identified three supporting networks, including a frontal- posterior temporal network that is responsible for filtering the interfering stimuli. Additionally, in a separate experiment using similar stimuli, we randomly varied the delay interval (short-2 sec, medium-5 sec, and long-9 sec) from trial to trial, which ensured sustained attention throughout the delay period because participants did not know when the delay period would end and the probe would appear. The delay activity associated with complex visual stimuli suggests a pattern of transient delay activity for medium and long delay periods, regardless of load, with an early increase in event-related synchronous activity (ERS) in alpha and lower beta activity (until 2-3 secs) followed by an extended period of event-related desynchronous (ERD) in alpha and beta band activity, in parietal and parieto-occipital regions. Sustained delay activity (i.e., ERS in alpha activity followed by a return to baseline) was only observed for the short delay interval (~2 sec). Our results suggest that the pattern of ERS reflects an early period of goal setting, in which attention is focused inward to prevent interference. As the delay interval increases, the pattern of ERD reflects ongoing maintenance and associations with stored semantic knowledge. Finally, we compared the underlying brain networks that supported maintenance across all experiments using connectivity measures. This comparison identified a different frontoparietal network that is implicated in cognitive control and was found to be involved in both effortful maintenance (i.e., for stimuli lacking semantic content) as well as maintenance that places an increased demand on attentional resources (e.g., interference present or increased intervals). These results provide some support for the persistent delay activity hypothesis, as there were changes in delay activity from baseline throughout the delay period in all experiments, regardless of maintenance technique, WM load or delay period interval. Furthermore, the delay period connectivity analyses for all experiments implicate fronto-parietal and temporal networks in supporting maintenance and suggest a flexible role of attention (e.g., filtering of interfering stimuli, control over attention) that varies based on task demands. Together, these delay activity and connectivity findings inform the ongoing debate about the neural dynamics that support visual WM