Studies in humans and rodents have consistently shown that sleep supports memory consolidation. This research provided the basis for the concept of active system consolidation during sleep. Simplified, the concept assumes a two-stage process of long-term memory formation, where episodic information is first encoded primarily into an initial store represented by the hippocampus. At a second stage, some of the newly encoded representations are reactivated and redistributed such that representations become predominantly stored in an extra-hippocampal circuitry which serves as long-term store. The redistribution of representations from initial to longer-term-storage sites goes along with a qualitative transformation of the representations, specifically the formation of more generalized and abstract representations, which can be flexibly used independent from the context in which they were originally acquired. Sleep is thought to support this second stage of memory formation in particular through the slow oscillations of SWS.
Here, we have discussed research on the memory function of sleep from a comparative biological perspective in different taxonomic groups, in order to clarify whether this function represents an evolutionarily conserved core function of sleep. Indeed, we found that sleep produced an enhancing effect on memory for all species that were examined.
Although there is still a lack of research on sleep and memory formation in non-mammalian groups, it appears that in birds SWS coevolved to serve the same function. Strikingly, slow oscillatory activity is required during sleep after filial imprinting in chicks for the stable formation of imprinting memory, linking long-term memory with SWS. Furthermore, studies of filial imprinting in chicks and of song learning in song birds suggest the view that sleep supports the redistribution of memories from networks serving as the initial storage of information to different networks serving as a long-term store (summarized in Fig. 2). Additionally, electrophysiological recordings in song birds have suggested that the putative redistribution originates from neuronal reactivations of memory representations during sleep. Although there is no evidence for a separate SWS stage in invertebrates, spatial learning in bees revealed hints that sleep might favor qualitative transformations of newly acquired memory representation. Bees after sleep had a higher probability of returning to their hive from a newly learned place suggesting that sleep enhances integrative spatial mapping. Furthermore, studies in bees demonstrated the selectivity of sleep-dependent memory formation, with sleep strengthening extinction memory but leaving unaffected the original classical conditioned proboscis extension response.
Collectively, these observations indicate that cornerstones of the active system consolidation process during sleep – i.e., neuronal reactivations, redistribution of representations, qualitative transformations, and the selectivity of consolidation – can be identified not only in mammals but also in birds and invertebrates. Thus, the concept might indeed either describe an evolutionarily conserved process of memory formation associated with sleep, or be an example of evolutionary convergence pointing at a fundamental need. If so, then the central question arises: What is the adaptive advantage of using sleep for consolidating memory? Presently, this question cannot be clearly answered. If we assume that the storage capacities of the nervous system are limited and thus do not suffice for representing the complexity of environmental conditions as a whole, then there is a basic need to reduce information. The transformation of memory during active system consolidation might solve this problem by producing more general, abstract, and schema-like representations that lack contextual detail, and by favoring the formation of memories for the gist of an experience that is relevant to advance future adaptive behavior (Inostroza and Born, 2013, Kumaran and McClelland, 2012 and Wilhelm et al., 2011a). Shifting system consolidation to the offline period of sleep basically enables the abstraction of gist memory in conditions unbiased by external stimulus inputs. Because the same networks that are used for system consolidation in a two stage memory system are also used for acute processing of environmental stimuli during wakefulness, ongoing processing of environmental inputs would bear the risk that external stimuli interfere and disturb the effective abstraction of gist from previous experience.