Our ability to maintain and later retrieve detailed visual memories is an everyday experience, and completing almost any task requires us to rapidly shuttle information between active and passive memory states. For example, imagine that you are on a hike and you approach an unfamiliar fork in the path. You need to retrieve the mental representation of the trail map and compare it to your recent path in order to decide where to go. Incredibly, all of this can take place in a few moments.
I’m broadly interested in understanding how the brain supports detailed visual memories and how these memories impact future learning. Memory formation and retrieval is tightly connected to a host of cognitive processes and neural mechanisms. Thus, my research program encompasses the study of questions pertaining to perception, attention and memory in addition to how these different processes interact. My methodological expertise is primarily in combining behavioral tasks, human electroencephalogram (EEG) recordings, and machine learning approaches in order to understand how the brain encodes, maintains, and retrieves information about visual features. My ongoing research program is composed of four related lines of inquiry:
Developing techniques to track active representation of visual features.
During my PhD, I helped develop a technique for estimating the precise location that an observer is holding in working memory at a given moment from patterns of alpha band activity (i.e., rhythmic brain activity between 8 and 12 Hz). We have subsequently found that these patterns of alpha-band activity track the precise location participants are attending and retrieving from long-term memory. In collaboration with the Shevell Lab, we have also found that separate patterns of evoked EEG activity track stimulus chromaticity and luminance. In my post-doctoral work at Vanderbilt, we have used computational modeling to establish that patterns of alpha-band activity track the two-dimensional coordinates of remembered stimuli within a visual hemifield, and that these patterns specifically track memory targets and not distractors, suggesting that these patterns reflect a spotlight of attention focused on locations maintained in working memory.
How many active memory representations are maintained at a time?
Working memory is a limited capacity “online” memory system that maintains information in a readily accessible state. Prominent models of working memory maintain that the severely limited capacity of the working memory system arises from limits in the number of active representations that can be maintained at a given moment. We recently tested whether these capacity limits reflect either rapid switching between active representations or the number of representations held in an active neural state at the same time. We found that two simultaneously presented locations could be maintained at once, ruling out the possibility that capacity limits can be explained by memory models that propose only a single item can exist in an active state at once. While we’ve established that multiple memory representations can be concurrently maintained when they are presented at the same time, a related question is how many active representations are maintained at a given moment when we memorize sequences of information. An experiment I recently conducted during my post-doc at Vanderbilt finds that observers can strategically switch between prioritizing individual locations as they are studied and refreshing the location of prioritized items between study events. Together, this work suggests that observers can flexibly switch between concurrent and serial storage depending on task demands.
To what extent do working and long-term memory rely on shared neural mechanisms?
A long-standing question in the field of memory research is whether the same neural mechanisms that allow you to hold something in mind for a few seconds (i.e., in working memory) are reactivated when information is retrieved from long-term memory. To test whether the same frequencies and patterns of brain activity engaged during working memory are reinstated during long-term memory retrieval, we had observers memorize the location of clipart objects while measuring their EEG activity during encoding and memory retrieval. We found that spatial representations during memory retrieval were tracked by the same patterns of alpha-band activity that tracked spatial representations during working memory maintenance. In a follow up fMRI study, we systematically examined the extent that cortical patterns of brain activity in working and long-term memory overlap, while carefully matching the stimuli, timing, and behavioral recall precision. We found that the strength and cortical distribution of feature-specific memory representations in visual and parietal cortex are largely similar between working and long-term memory. Intriguingly, we also found that we could distinguish between working memory and long-term memory trials based on activation patterns in the parietal, but not visual, cortex. In sum, this work revealed that working and long-term memory engage overlapping oscillatory mechanisms and sensory cortical regions, while parietal regions contain information about whether information was maintained in working memory or retrieved from long-term memory.
How does memory retrieval affect subsequent memory performance for precise visual details?
A defining feature of episodic memory is the ability to remember precise details. However, until recently, tests of long-term memory have predominantly relied on categorical words and pictures, leaving our understanding of memory for precise visual details incomplete. To understand how retrieving specific visual details from long-term memory affects our memory for those same details in the future, we use continuous report tasks in which participants study and report the exact color or location of common objects. In one study, we tested how previously memorizing the color of an object affects observers’ subsequent ability to remember the color of the same object in a working memory task. We found that working memory was enhanced (proactive facilitation) for objects presented in the same color across tasks, and that working memory performance was unaffected for items presented in a different color across tasks. This suggests that the contents of working memory are protected from proactive interference but can be facilitated by long-term memory information when it is helpful. In another study, I tested whether retrieving information from long-term memory (retrieval practice) increases the probability that memories can be retrieved or the precision with which they are retrieved, and found that taking a test improves memory access but not memory quality.