Examining Reactivation of Hippocampal Representations in Sleeping Toddlers

• In Event: Breaking Barriers to Better Understand Early Cognitive Development: Novel Task-Based fMRI Paradigms in Young Children

Abstract

A growing body of research has demonstrated functional contributions of the hippocampus to early memory abilities using conventional univariate fMRI analysis. These studies have demonstrated that when auditory cues associated with past events are presented during sleep, resulting activation of the hippocampus is correlated with memory for those events. Moreover, there is initial evidence from multivariate analyses approaches that the hippocampus represents past experiences more similarly to relevant past experiences than novel experiences. However, little is known about the how similarity changes as a function of time since the presentation of auditory memory cues This would be helpful to elucidate how readily the hippocampus responds to cues at this age and to optimize experimental designs. Here, we begin to fill this gap.
Forty-six 2-year-olds who completed a tablet-based game. The game involved remembering of associations between unique characters, the places they visited, and the temporal order of these visits. A novel song was played at the end of the game. Toddlers’ functional scan was acquired during sleep after participating. The scan included 9 20-seconds blocks of song, interleaved by 20-seconds blocks of silence (Baseline). Three different conditions of songs were presented: Target (Heard song played forward as learned), Reverse (Heard song played backward), and Novel (completely novel song. We extracted voxel-wise time-course data and conducted a voxel-wise similarity analysis within and between Target, Reverse, and Novel conditions. Within condition similarities defined as the average of inter-voxel time-course activation correlation in the same condition separately for first and last half of a block. In the first half, inter-voxel similarity was significantly higher than zero in Target (t(45) = 2.76, p = 0.004), Reverse (t(45) = 2.22, p = 0.007), and Baseline (t(45) = 2.46, p = 0.009) with Target condition being significantly higher than Baseline (t(43) = 2.25, p = 0.030). In the second half, only Baseline condition was significantly higher than zero (t(45) = 3.11, p = 0.002) suggesting that the first half of each block included more relevant activations for old songs compared to the last half. Critically, novel song resulted in uncorrelated inter-voxel activation patterns in the hippocampus across the whole block. Between condition similarities are defined as the average correlation between time-course activation of the same voxels in two different conditions. In the first half, similarity score for Target-Reverse (t(45) = 2.52, p = 0.008), and Target-Novel (t(45) = 2.51, p = 0.008) were significantly higher than zero, while in the second half only Target-Reverse (t(45) = 2.31, p = 0.013) remained significant suggesting that processing of a forward playing song may produce similar patterns of activation as soon as the cue is presented. However, old songs regardless the playback order consistently produce similar activations in course of processing. Overall, time-dependent changes occur in memory reactivation and our discussion will focus on how effectively memories can be triggered during sleep.

Authors

• Alireza Kazemi, University of California - Davis
  Presenting Author

• Lindsey Mooney, University of California - Davis

• Simona Ghetti, University of California - Davis