Abstracts

Rationale: Humans primarily learn about the world through vision. While the awake brain continuously processes information, behavioral and neurophysiological evidence suggests that memories are formed as discrete events. While looking is experienced as a continuous process, it is chunked by individual ballistic eye movements called saccades. We hypothesized that saccades organize the encoding and retrieval of memory in the human hippocampus.


Methods: We recorded from intracranial hippocampal electrodes in surgical epilepsy patients (N=5) with eye tracking during a visual associative memory task. Subjects were exposed to paired object associations (encoding phase) and were later cued to distinguish paired objects from lures. To isolate memory-related neural activity associated with eye movements, we performed several control experiments in which subjects (N=3) performed similar saccadic eye movements without a cognitive task requirement.


Results: During both encoding and retrieval, saccades to visual objects were time-locked with a hippocampal evoked response consisting of a brief negative discharge followed by ~200 ms of coherent (p< 0.05, cluster mass test) theta-range (10-13 Hz) activity. The initial 30 ms of theta-range evoked activity was associated with phase-amplitude coupling (PAC) to gamma activity.



This saccade-evoked activity distinguished between the behavioral valence of visual objects and predicted memory performance. During encoding, post-saccade hippocampal theta predicted whether the objects would be later paired correctly. During retrieval, theta-range activity and theta-gamma PAC were amplified after saccades to paired objects (cue and target) relative to saccades to the lure. This distinction between paired objects and the lure was not present on incorrect trials. Post-saccadic changes seen during correct retrieval resembled changes seen during correct encoding trials.



In control tasks where subjects looked at objects on the screen, without a memory demand, saccades to visual targets located at the same positions as during encoding or retrieval were not associated with theta-range evoked responses in the 10-13 Hz range. Instead, these saccades were followed by lower frequency activity (5-7 Hz). This was also observed after saccades to individually presented objects in different positions.


Conclusions: Thus, the accurate encoding and retrieval of paired associations is accompanied by a specific pattern of saccades and post-saccadic hippocampal activity that is amplified after saccades to paired objects relative lures. This pattern of activity is not present after saccades without memory-relevance. These data support a saccade-organized process for encoding and retrieval of visual associations in the human hippocampus. Saccade-locked analysis reveals how well-described theta-gamma mechanisms in rodent navigation translate to human memory function and its disruption in temporal lobe epilepsy.


Funding: This work was supported by NINDS K23 (NS127954-01A1) and R01 (NS104252-05) awards.