We experience the world as a continuous sequence of events, but we remember the events as segmented episodes (e.g., my sister?s wedding). During encoding, we associate a sequence of relevant events and segment deviant events. At retrieval, episodic memory utilizes the encoded associations to replay the flow of events. The encoded associations lead to remembering the sequence of events that occurred within an episode better than the flow of events across segments. The hippocampus and the prefrontal cortices (PFC) are essential parts of the neural circuit for segmenting, linking, and retrieving memories of associated events. This proposal aims to identify neural dynamics in the hippocampus-PFC circuit that support encoding a naturalistic flow of events, i.e., sequences of words. We will determine these neural dynamics using intracranial encephalography (iEEG) acquired from the hippocampus and PFC of epileptic patients, who have electrodes implemented for pre-surgical seizure monitoring. I will model associations of words using Natural Language Processing algorithms, and I will combine the extracted features with advanced data analysis techniques including multivariate pattern analysis to determine neural dynamics engaged during encoding. I will use speech as a model with an identical flow of events in the speech stimuli across participants. This consistency will allow validation of effects across a group of participants. Algorithms for identifying features of speech are well developed and freely available. I will specifically use elements of speech that distinguish context, word dependencies, and reference points of pronouns for modeling concurrent changes in patterns of activity in the local field potential recorded from the hippocampus and PFC. The central hypotheses are that bidirectional communications between the hippocampus and PFC support the encoding of sequences of events and successful subsequent memory. To address a causal relationship between hippocampal function and event segmentation, I will study speech comprehension and speech memory in developmental amnesic patients who suffer from hippocampal damage and have trouble tracking reference points in a speech. To achieve the proposal?s goals, I will pursue training under the mentorship of Dr. Elizabeth Buffalo (University of Washington) that will focus on the advanced analysis of local field potentials. The advanced study of human iEEG data will include comparable electrophysiology signal analyses that have been applied to the recordings from the hippocampus of non-human primates in Buffalo?s memory lab. This skill-set along with ongoing mentoring from Dr. Robert Knight (University of California, Berkeley), who has an established laboratory for human iEEG, and my previous work on human iEEG will provide a vigorous methodological, conceptual, and analytical basis for developing an independent research program. The combination of iEEG, Natural Language Processing modeling, and patients? behavioral data will provide valuable insights into the neural dynamics of effective speech encoding that predicts subsequent memory, which may inform development into therapeutic interventions.
The ability to segment the world into meaningful episodes engages the human hippocampus and prefrontal cortex. Using direct electrophysiological recording from the human brain, advanced analytical techniques, Natural Language Processing models, and the behavior of patients with hippocampal lesions, this proposal will determine the neural dynamics for efficient encoding of sequences of words that predicts successful memory formation. The findings of this proposal may help inform the development of neural prosthetics for assisting patients with memory deficits.