In this thesis I have described a number of studies investigating the relation between sleep and memory from a cognitive neuroscience perspective. More specifically, I looked into the mechanisms underlying the reactivation and consolidation of declarative, mostly episodic, memory traces in humans. My main focus was on brain rhythms during non-rapid eye movement (NREM) sleep, and whether they could support systems-level consolidation. Additionally, but certainly inspired by the notion of memory reprocessing during sleep, I explored whether and how brain waves during sleep influence, first, the processing of incoming information, and second, the possibility to learn new information.

In Chapter 2 we asked whether the importance of sleep spindles for memory consolidation is dependent on the presence of slow oscillations (SOs) during deep sleep. Indeed, we observed that the rate of spindle occurrence during deep sleep was associated with declarative memory retention from pre-nap to post-nap. Such a relation was completely absent during light sleep, even though this stage is known for its sleep spindle activity. In addition, we noticed higher spindle density in deep sleep than during light sleep, while other spindle characteristics were unaffected. These findings suggest that the presence of SOs, which occur only during deep sleep, is required for spindles to mediate memory consolidation.

Exploring the apparent governing role of SOs further, in Chapter 3 we assessed the macroscopic organizing role of SOs on brain activity. We saw that SOs coordinate quicker brain rhythms, such as spindle and gamma activity, in a spatially restricted manner. Furthermore, interregional communication in the spindle frequency band, and cross-frequency spindle-gamma network interactions, were also governed by the slow oscillatory phase. These data suggest widespread neocortical memory networks may communicate effectively during deep sleep, depending on the SO phase. Such network activity is indispensable if stable cortical memory representations are to be formed.

Returning to the role of sleep spindles in Chapter 4, we investigated the hypothesis that local spindles reflect the reactivation of local memory networks. We asked subjects to memorize blocks of word-location associations that were tied to a specific visual hemifield. In addition, each block was studied in the presence of a distinct odor. During a subsequent nap, we administered one of the previously used odors to selectively cue the associated memories. Because hemifield-specific visuospatial processing is known to rely on contralateral posterior brain regions, these networks were expected to respond to the olfactory cues. Indeed, we observed symmetrically lateralized response patterns of spindle amplitude and density, consistent with the idea that spindles reflect local memory reprocessing.

In Chapter 5, we turned to questions concerning the long-term fate of different aspects of episodic memories and how sleep-related processes are involved. In particular, we asked how the importance of context for memory retrieval was affected by time and sleep. We instructed participants to study words against unique background photos, and later asked them to recall these words while either the same or a different contextual photo was present. Subjects recalled memorized words on two occasions: immediately following learning, and after an extended interval containing nighttime sleep, daytime waking, or both. We found that the presence of the same context as during learning had a large beneficial effect on memory performance, but that this contextual benefit diminished over time. Sleep, however, did not affect the amount of decontextualization. On the other hand, sleep did impact overall memory performance, regardless of contextual congruence, such that sleep followed by waking resulted in better memory than wakefulness followed by sleep. Thus, sleep stabilized memories, but did not aid in the forgetting of the contextual aspects of memories.

Finally, in Chapter 6 a novel phase prediction algorithm was introduced for targeting the delivery of stimuli at specified phases of the SO. Using this approach, we explored the possibility that sleep-learning may be attainable for materials dependent on neocortical networks. While we witnessed differential neural processing in response to sound presentation aimed at up and down states, disparate neural responses to these stimuli were not apparent during subsequent wakefulness, nor were behavioral responses concerning stimulus familiarity affected. Thus, while SOs are of definite importance for the way external stimuli are processed, sleep-learning does not appear to be easily accomplished for stimuli relying on neocortical circuitry.