Abstract Introduction Learning in a variety of tasks is known to be enhanced by post-training sleep, but the underlying neural mechanisms remain unclear. One leading hypothesis is that the re-emergence of training-related neural activity during sleep, known as reactivation, is essential for sleep-dependent learning. According to this hypothesis, task improvement should reflect reactivated representations, and learning should not generalize to representations that are not reactivated. Here, we tested a novel model proposing that the specificity of neural reactivation reflects the specificity of learning. Methods In the experiment, participants were trained to detect a Gabor patch with a single orientation. During a post-training nap session, we recorded fMRI activity from early visual areas simultaneously with polysomnography to determine sleep stages. We applied a previously constructed fMRI decoder capable of classifying four Gabor orientations, including the trained orientation, to the fMRI data collected during sleep Results In the event that learning occurred, the decoder classified the neural activity observed during NREM sleep as similar to the trained orientation above chance, suggesting selective reactivation of the trained orientation during sleep. Furthermore, higher levels of classification of the trained orientation appeared to accompany greater performance improvement after sleep for the trained orientation only. Conclusion This pattern suggests that neural reactivation in visual cortex during NREM sleep reflects the trained stimuli, and it supports the proposal that reactivation plays a key role in mediating the specificity of learning. Support (if any) Acknowledgements: NIH R01EY019466, R01EY027841, R01EY031705, NSF-BSF BCS2241417
LaBonte-Clark et al. (Fri,) studied this question.