Président du jury Destrebecqz, Arnaud
Promoteur Peigneux, Philippe
Publication Non publié, 2013-04-27
| Résumé : | Although a relative consensus exists about the contribution of post-learning sleep in the consolidation of novel information in long term memory, the definition of the respective contributions of sleep stages in memory consolidation processes remains a matter of debates. Scrima (1982) proposed the hypothesis that Slow Waves Sleep (SWS) contributes preventing retroactive interference on recently acquired information, whereas Rapid Eyes Movement sleep (REM) contributes consolidating this information. This interesting hypothesis was never validated by others studies and rapidly forgotten in the sleep literature. In this framework, our doctoral thesis explored the role of sleep stages in consolidation and protection of lexical and emotional retroactive interference. Regarding lexical interference, we use a classical interference paradigm in which participants learn a list of unrelated word pairs (A). After a retention interval varying according to the study, a novel list of word pairs (B) is learned just before delayed recall of list A. List B is composed of 50% word pairs in which the initial word of the pair is also presented in list A, hence creating interference. In a first study, we showed an interference effect after a retention interval containing 3 nights of sleep (sleep condition) but not after a first night of sleep deprivation and two recovery nights (sleep deprivation condition). Our findings may be in line with the reconsolidation theory in that after a night of sleep the reactivation of consolidated memory traces puts them back in a labile form, hence again sensitive to interference. By contrast, in the sleep deprivation condition, subjects would create a dual trace (A and B) allowing them to fend off the negative impact of interference: the second list does not modify the first but the two lists coexist. The same results were observed in a second study after a 45-min nap vs. a wake episode. In this study, the controlled morning nap paradigm aiming at producing predominant post-training SWS or REM sleep episodes, allows us to investigate whether a nap containing SWS or REM sleep increases nevertheless the protection against interference. Our results partially support Scrima’s hypothesis. Indeed, if SWS seems to have a protective effect against interference for novels memories, we failed to evidence REM sleep-related consolidation effects on word pairs not subjected to interference. In a second section of my doctoral thesis, we investigated the protective role of sleep against emotional interference. Previous studies have showed that a change of mood from learning to recall induced a retroactive emotional interference and consequently impaired recall capacities, whereas a similar mood at both encoding and recall sessions facilitated retrieval (a phenomenon named ‘Mood-state Dependent Memory’, MDM). In a first study, we showed that sleep reduces the MDM effect by unbinding memories (i.e. word pairs) from their emotional context (i.e. the emotional mood context). In this study, subjects learned a list of word pairs after a mood induction procedure, then slept or stayed awake during the postlearning night. After two recovery nights, subjects recall 50% word pairs after the same mood induction procedure and 50% in a different mood state. MDM effect was observed in the sleep-deprived condition whereas it has disappeared in the sleep condition. This study validated the “Sleep to Forget and Sleep to Remember” model (van der Helm & Walker, 2010) stating that sleep (especially REM sleep) facilitates the decoupling of declarative memories from their emotional context, hence reducing the MDM phenomena. However, in a second study comparing the effects of "early" and "late" sleep periods (dominated by SWS and REM sleep, respectively) on resistance to emotional interference, we failed to evidence a specific role of REM sleep in this emotional unbinding process. We surmise that the demodulation process is initiated during the first post-learning night (potentially during REM sleep) but may need several nights or several successions of NREM-REM cycles across a whole night to be achieved. A third study evidenced that MDM is not reduced after one night of sleep. These results do not allow us to confirm that emotional unbinding process needs several successions of NREM-REM cycles across a whole night to be achieved but rather need several nights. To sum up, sleep seems to protect memories against both lexical (especially SWS), and emotional interference; the latter being achieved by unbinding memories from their emotional context. |
