To date, there is extensive evidence showing that sleep is important for memory consolidation. Specifically, a night of sleep seems to either strengthened many forms of memory or protect them against decay. The information overlap to abstract (iOtA) model suggests that when memories are replayed in an overlapping way the combination of potentiation for overlapped areas and downscaling for everything else leads to extraction of gist. In this thesis, I set out to test the above proposal. In order to control the extent to which memories overlap or remain isolated, I developed a novel paradigm. Namely, I morphed facial images along both age and gender, creating a 2D âface spaceâ. I conducted Experiment 1 to determine how far apart images need to be in this face space in order to be remembered as distinct (not merged). I was then able to train participants on selected images from this space. These learned images were either close together (densely packed representations) or far apart (sparsely packed representations). Participants could then be tested for both veridical memory of the learned (old) images, and false memory of unstudied items in any part of the face space. In a series of experiments, I then examined the impact of sleep on both veridical and false memory for items falling in both dense and sparsely populated areas of the face space. As expected, I consistently found that veridical memories were protected by sleep. However, surprisingly, Experiment 2 showed an overnight increase in false recognition for the face images in sparsely populated areas of this space, with no change in false recognition for images in densely populated areas. Experiment 3 replicated this finding even when I controlled spatial features of facial images, except face density. Polysomnography in Experiment 3 also showed a strong correlation between slow wave sleep and the extent of increased false recognition in the âsparseâ area. Next, I conducted Experiment 4 as a control study using inverted faces, and revealed that the sleep effect was not specific to facial images. This suggests that the finding generalises at least as far as these control images. In Experiment 5 I used a finer grained map of the face-space by using smaller morphing steps. This study suggested that, across a night of sleep, compared to wake, the representation of a learned face actually shifts away from other learned face representations, . In Experiment 6, I conducted a neuroimaging study in order to determine the neural correlates of these shifts, and found that activity in face selective areas shifts in parallel with the behavioural shift, while activity in the hippocampus shifts in the opposite direction. Overall, my findings do not support the iOtA model, as memory representations in densely populated areas were not strengthened over sleep. Instead, my work adds to the growing literature suggesting that sleep is not only important for strengthening memories, but also for selectively weakening some aspects of memory. Specifically, my work could suggest that sleep plays a role in pattern separation, as it seems to force the representations apart, even when this increases false alarms.
|Date of Award||31 Dec 2018|
- The University of Manchester
|Supervisor||Karen Lander (Supervisor) & Andrew Stewart (Supervisor)|
- face recognition
- gist-based false memory