Eric Armstrong
The Role of Sleep Spindles in Declarative Memory Consolidation
Introduction
In the field of cognitive neuroscience, short bursts of activation during NonREM sleep measured using EEG are known as sleep spindles. The function of sleep spindles has been the subject of some debate. A study conducted by Bergmann and colleagues indicates that sleep spindles reactivation is category specific, at least in some instances. The study, conducted using EEG in conjunction with fMRI, yielded some correlations between cortical regions used during a waking state and those activated by sleep spindles during NonREM sleep.
A fair deal of research in the cognitive sciences has dealt with declarative, or episodic, memory. In this particular type of memory, as research indicates, patterns of specific activation in the neocortex are bound via the hippocampus and associated medial temporal lobe regions. These traces created by hippocampal binding can be reactivated until the hippocampus is no longer needed for the accurate retrieval of the declarative memory episode. Sleep is said to be time of memory consolidation via the reactivation of the episodic memory traces. With the reactivation of the hippocampus during NonREM sleep, cortical connections are strengthened. Indeed, the researchers cite evidence of learning-related increases correlating with spindle activation (for both number and amplitude).
Bergmann and colleagues sought to establish some link between sleep spindles and hippocampal and neocortical activation. While the general mechanisms of sleep spindle action may be understood, substantial evidence on the factors that mediate this “sleep-dependent memory consolidation” needs to be found. To do this, the researchers chose to conduct a study using EEG and fMRI in conjunction. In this way, they could get the temporal resolution needed to detect sleep spindles and the spatial resolution that would be needed to show the concurrent area-selective activation in the cortex and activation in the hippocampus. Finally, Bergmann and colleagues juxtaposed the two measures in an effort to identify a correlation between the spindles and the BOLD signal. Specifically, Bergmann and colleagues designed a study that would make a distinction between spindle and localized activation that hinged on whether participants encoded declarative information with effort, and moreover what kind of information was being encoded. Since the learning task involved participants' encoding faces and places, the researchers have the advent of comparing their data to well-established regions, namely the fusiform face area (FFA) and the parahippocampal gyrus/place area (PPA), and their patterns of activation.
Methods
Bergmann and colleagues recruited 24 volunteers for the study, though data was only collected for the nine participants: those who managed to fall asleep on both nights. Participants took part in both the learning and the control conditions, separated by at least a week, and the researchers made sure to balance among the participants which task was conducted first/second.
In the paired associate learning condition, participants were presented with pairs of predominantly faces either followed or preceded by a location. This was, of course conducted using EEG and fMRI. Stimuli were presented in a pseudo-random order, four trials followed by a cued recall of some previous item. Participants were asked to make and effort to link paired faces with locations for the learning component. For the cued recall, participants returned the following evening, and when presented with one picture of a pair associate, they were asked to respond with some further information on the cued associate. For example, when presented with a given face, participants should indicate, for example, that the associate was an urban scene. For the experiment, a color-changing fixation cross was employed to ensure participant attention was maintained.
The control condition was a visuomotor task conducted in a manner parallel to the learning condition. The researchers make a point that the control condition proceeded with identical timing to the learning condition. In the first phase of the control condition, participants were asked to determine whether scrambled images presented were horizontal mirror images. This functioned as a measure of attention, much like the fixation cross used in the learning condition. The next task, comparable to the cued recall of the learning condition, required that participants count the number of boxes in a given scene and press a corresponding button. Participant accuracies in the control condition were not taken into account, and participants were never told their score.
EEG Data
The experimenters chose to restrict the available EEG signals indicating sleep spindles. They selected for “fast centroparietal” spindles in the range of 12 to 14 Hz, the range likely to be involved in hippocampal activity. As noted earlier, thresholding algorithms were used to identify the spindles, and for each participant, the researchers identified not only spindle events but also the amplitudes for each event. The researchers also took a number of measures to smooth the results of the fMRI data collection.
fMRI Data
Non-sleep fMRI data was used to contrast selected areas of the cortex needed in the learning and control tasks that are not activated in non-trial times. Bergmann and colleagues then developed a regression model to compare the “pre-sleep recall run with the post-sleep recall session.” With this, the researchers were able to form a prediction for the recall-related activation. In addition, participants' retention of the paired associates was measured the following day with cue-response testing.
Using the sleep fMRI data, the researchers first looked for brain regions that showed a change in the BOLD signal co-occurring with sleep spindles, whether or not the learning or control condition was being examined. Next, Bergmann and colleagues used the sleep fMRI data to show that activation in the FFA as well as the PPA is stronger after trials involving declarative learning, as evident by the timing and amplitude of spindles. In effect, the researchers isolate what they call spindle-associated activation.
Results and Discussion
The study yielded some valuable results. To begin, all participants considered in analysis were able to sufficiently learn the paired associates, as evident by their performance on cued recall during encoding and recall the next day. The researchers confirmed the use of selective areas of the neocortex, as made evident by the fMRI activation data. The same can be said of activation in the hippocampus. On the whole, the researchers found that spindle-related activation was stronger after sleep, even given the recency effects for recall for just having completed an encoding task. On the other hand, there was “spindle-coupled brain activity independent of the experimental condition.” This is fairly intuitive, given that episodic encoding takes regardless of whether a effort to learn is being made. The bottom line presented was the evidence that, the better a participant performed in the learning task, the more pronounced the spindle response, in terms of both the spindle amplitude and the hippocampal activation.
Despite their meticulous methods, there are a few caveats to the present study. The largest potential concern is the small pool of participants that was selected for inclusion. This is surely mentioned in the discussion of the article, but no real alternatives are offered, but only that the results should be interpreted cautiously. By and large, the study did an excellent job of perking interest in and providing justification for employing dual measures of brain activity.
References
1. Til O. Bergmann, Matthias Mölle, Jens Diedrichs, Jan Born, Hartwig R. Siebner, Sleep spindle-related reactivation of category-specific cortical regions after learning face-scene associations, NeuroImage, Volume 59, Issue 3, 1 February 2012, Pages 2733-2742, ISSN 1053-8119, 10.1016/j.neuroimage.2011.10.036. (http://www.sciencedirect.com/science/article/pii/S1053811911012006)