Although several human studies suggest that sleep facilitates insight learning, the means by which this could occur is unknown. One hypothesis suggests that key elements in the environment necessary for insight are encoded in pieces during waking, then replayed during sleep, allowing the pieces to self assemble into insight. However, it is not clear which memories are selected for reactivation and processed in sleep versus allowed to be forgotten. We have developed a novel computational hypothesis called temporal scaffolding, which can serve as a platform to shed light on both questions. This hypothesis suggests that sleep replay should especially aid in gaining insight into temporal hidden patterns due to the unique compressed dynamics of memory replay occurring during non-rapid-eye- movement sleep (NREMS) and the learned sequences that are replayed in NREMS are those that were accompanied by bursts of activation from the Locus Coeruleus (LC) during the learning session, increasing the probability for replay. We developed a new task for animals that measures whether sleep facilitates insight into a hidden temporal order. We will use this task in combination with a measurements of activity from hundreds of neurons in the hippocampus simultaneously while animals are learning the task and assess activity from task-relevant neurons while animals train and assess activity from the same neurons again while they sleep, and once more when they are tested following sleep.
In Aim 1 we will correlate the number of task-relevant hippocampal replay bursts during NREMS with measures of performance on the subsequent wake period that indicate whether the animals have gained insight into the hidden temporal order. Preliminary data show that a subset of unmanipulated animals can achieve insight into this task and that sleep boosts this gain of insight.
In Aim 2 we will use an optogenetic approach to stimulate or silence LC neurons at critical choice points during initial task exposure and see whether such manipulation alters the density or type of replay events in subsequent sleep and influences the gain of insight. Sleep deprived and LC-silenced groups will serve as controls for the LC-stimulated animals.
In Aim 3, the data gathered from the first two Aims will be entered into a computational model of hippocampal-neocortical networks to better estimate the parameters determining how temporal scaffolding occurs, which, in turn, will inform future mechanistic studies. Positive results in each aim will underscore the importance of memory replay in insight learning, the contribution of temporal scaffolding to this learning, and a learning mechanism (LC activity) modulating it. It will also provide a new handle by which to boost memory processing during sleep.
We propose to test how patterns of activity during sleep provide insights into tasks with hidden temporal rules and see if we can boost the gain of insight during sleep by stimulating a small structure in the brainstem, called the locus coeruleus, during the waking learning period. Preliminary data support the possibility that insight is achieved through the sleep mechanisms we have modeled. Positive results would give us specific ways by which we could influence insight learning, although alternative results would also be useful in developing our model outlining the role of sleep for learning.