Key points Sleep spindles have recently been shown to occur not only across multiple neocortical regions but also locally in restricted cortical areas. Here we show that local spindles are indeed present in the human posterior thalamus. Thalamic local spindles had lower spectral power than non‐local ones. While non‐local thalamic spindles had equal local and non‐local cortical counterparts, local thalamic spindles had significantly more local cortical counterparts (i.e. occurring in a single cortical site). The preferential association of local thalamic and cortical spindles supports the notion of thalamocortical loops functioning in a modular way. Abstract Sleep spindles are believed to subserve many sleep‐related functions, from memory consolidation to cortical development. Recent data using intracerebral recordings in humans have shown that they occur across multiple neocortical regions but may also be spatially restricted to specific brain areas (local spindles). The aim of this study was to characterize spindles at the level of the human posterior thalamus, with the hypothesis that, besides the global thalamic spindling activity usually observed, local spindles could also be present in the thalamus. Using intracranial, time‐frequency EEG recordings in 17 epileptic patients, we assessed the distribution of thalamic spindles during natural sleep stages N2 and N3 in six thalamic nuclei. Local spindles (i.e. spindles present in a single pair of recording contacts) were observed in all the thalamic regions explored, and compared with non‐local spindles in terms of intrinsic properties and cortical counterparts. Thalamic local and non‐local spindles did not differ in density, frequency or duration, but local spindles had lower spectral power than non‐local ones. Each thalamic spindle had a cortical counterpart. While non‐local thalamic spindles had equal cortical local and non‐local counterparts, local thalamic spindles had significantly more local cortical counterparts (i.e. occurring in a single cortical site). The preferential association of local thalamic and cortical spindles supports the notion of thalamocortical loops functioning in a modular way.

The question of a possible link between dream content and memory consolidation remains open. After a comprehensive review of the literature, we present novel findings from an experiment testing whether the incorporation of recently learned stimuli into dream reports is associated with improved post-sleep memory performance. Thirty-two high dream recallers freely explored new visuo-olfactory episodes for 3 consecutive days. During the nights following each non-explicit encoding, participants wore a wrist actimeter, and woke up at 5am and their usual waking time to record their dreams (intensity of all oneiric sensory perception was assessed using scales). A total of 120 dreams were reported and elements related to the encoding phase were identified in 37 of them, either learning-related (mainly visual- and rarely olfactory-related elements), or experiment-related (lab- or experimenters-related elements). On the 4th day, we found that participants with learning-related (n = 16) and participants with learning-related and/or experiment-related dreams (n = 21) had similar odor recognition and odor-evoked episodic memory with the other participants. However, they had significantly better visuo-spatial memory of the episodes in comparison to the other participants. Our results support the hypothesis that the learning phase is loosely incorporated into dreams and that this incorporation is associated with sleep related memory consolidation.