Sleep promotes illusory word compositions, a distinct form of false memory.

Extensive evidence supports the beneficial effects of sleep on memory and learning, including the consolidation and reorganization of memories and the extraction of regularities from encoded experiences. Nevertheless, some studies suggest that sleep may also increase false memories, potentially as a byproduct of regularities extraction. Physiologically, time-compressed memory replay in the hippocampus during non-rapid-eye-movement (nREM) sleep is believed to contribute to the consolidation process, although the functional significance of time compression remains elusive. Recently, we proposed that compressed replay might allow associating events that happened at disparate times, thus supporting the extraction of regularities with a temporal nature. This model predicted that sleep might also facilitate a distinct kind of false memories, in which two separate events occurring consecutively are encoded as a single composite event. Here, we tested this prediction by exposing male and female adults to separate word pairs (e.g., car, pet) that could form a new composite word if combined (carpet). We then tested their memory for composite words following a period of sleep or wake. Confirming our main prediction, we found that sleep actively facilitated false composite memories. Furthermore, EEG recordings indicated the involvement of nREM sleep in the process, albeit in a nuanced manner: While some slow-wave or spindle-related parameters predicted increase in false memories, others were associated with fewer false memories and a decline in veridical memories. The latter result resembles previous findings from non-composite false memory studies and could suggest a competitive mechanism between semantic and episodic consolidation during sleep.