learning during sleep

Posted comment on ´Inducing Task-Relevant Responses to Speech in the Sleeping Brain`
By Sid Kouider, Thomas Andrillon, Leonardo S. Barbosa, Louise Goupil and Tristan A. Bekinschtein and published in Curr Biol. Sept 22nd, 2014; 24(18): 2208–2214.
This research by Kouider and his colleagues shows that the brain can even in sleep recognize certain spoken words, categorise them by cognitive processing and then respond in the appropriate manner previously learnt whilst in the awake state. In their first experiment, Kouider et al. trained volunteers to categorise spoken words as either animal or object by pressing buttons with either their right or left hand. In the second experiment, the decision was whether the presented word was a proper word, or a pseudo-word. Lists of words were repeated until the responses were automatic and then continued whilst the subject fell asleep. When asleep the subjects were presented with a new set of words and the brain activity continually monitored. Novel words were used so that the responses would not merely be a reactivation of known memory traces. Awake and sleep states were gauged by a variety of means and results considered in the NREM1 sleep state (no alpha rhythm) and NREM2 for Experiment 1 and only NREM2 for Experiment 2. Results in periods of brief awakenings and microarousals were discarded as it was assumed that in these states, awareness and conscious cognitive processing could be occurring. Brain activity was assessed by lateralized readiness potential (LRP) measurements which gave an electrophysiological indication of response preparation which in this case the readiness to move a particular hand on presentation and processing of the auditory stimulus.
Results showed that in Experiment 1 by collapsing sleep and awake experiments and computing stimulus-locked LRPs there was first a negative deflection with two significant peaks at 660 and 1,620 ms primarily over central electrodes C3/C4 and central posterior electrodes CP3/CP4, but after 2,000 ms, the LRP returned to the baseline for several seconds until another negative deflection peaking around 5,570 ms occurred. To test the difference between the awake and sleep states, the awake state was substracted from the sleep state and this brought about no significant difference for the first LRP deflection, but significant differences later during the opposite deflection, around 2,920 ms for C3/C4 and 3,800 ms for CP3/CP4. The initial LRP are attributed to the preparation of the motor plan occurring in both sleep and awake states, whilst the inversion of potential was linked to the manual responses. The results were interpreted as the recognition and the categorization of the spoken word with then preparation for movement taking place in both sleep and awake states shown by the first LRP with no significant differences between states, and then only the manual response for the awake state with the second deflection.
Similar results were obtained with Experiment 2 with the word/pseudoword categorization with two LRP clusters: an early effect transient peaking at 1,276 ms (mostly driven by the awake state) and a later and more sustained effect peaking at 5,016 ms and extending from 3,508 ms until the end of the study period at 8,000 ms (primarily driven by the sleep state). The slower response was attributed to the consideration of only the NREM2 state, thus indicating that evidence accumulation is slower when only the single sleep state is evaluated.
In both experiments the subjects had no memory of the novel words processed in the sleep state when they awoke.
From a brain memory perspective this experiment shows an example of learning and subsequent use of a sequence not only in the awake state, but also in the light sleep state. There is well-established research on this type of experiment using single cues such as tones with spatial memory recall (Bendor and Wilson) or tones and odours (Arzi et al.) and these observations can be explained by the cues instigating reactivation of established memory traces or formation of novel but theme-associated memory traces requiring synchrony of hippocampal and cortical areas. The experiment described here requires incoming information to be processed and the result of that processing initiating a motor action and this is carried out not only in the awake state, but also in the NREM1 and NREM2 sleep states. What makes it interesting is that although information processing appears to be required, a process involving normally the higher executive functions of the brain eg the prefrontal cortex, in this case the activity of this area is suppressed due to the subject being in the sleep state. An explanation for this lack of requirement may be the nature of the experiment itself.
In this experiment the initial communication of what is required of the subject primes the subject to the nature of the sequence he requires to be successful in the study. He needs to work out if the word presented to him represents an animal or not and based on that assessment he instigates the same motor action, but just with different hands. In the awake state the test is relatively simple. Animal words are normally highly familiar, well-defined and such categorization has been carried out from an early age so is well-established and the decision is either yes (it is an animal), or no (it is something else). It is so simple that repetition of it in the waking state can lead to automaticity without conscious awareness. Kouider and colleagues then took this learnt sequence and looked at what happened in the light sleep state and showed that the same sequence was possible up to the point that manual movements are made.
They assumed that since the function of the prefrontal cortex (an area linked with executive functions) is suppressed in the sleep state then it is not part of the sequence required in their study, but other areas can still function independent of the wake/sleep state. However, the prefrontal cortex is known to play a role in decision-making. This experiment then implies that there are two types of decision-making: conscious requiring prefrontal cortex participation and executive processing and non-conscious that can be carried out when prefrontal cortex activity is suppressed or where the decision is obvious. The experiment begins with the learning phase where conscious decision-making has to be made, but then repetition leads to an example of the latter decision-making type, where the real-time activated memory trace indicates one obvious favoured option in preference to others. Hence, in the later stages further processing requiring higher brain functions is not necessary to satisfy the experiment`s demands once the initial sequence learning has been carried out. The non-requirement of the prefrontal cortex after the initial learning period is supported by looking at the study from the perspective of consciousness and conscious awareness. Since self-awareness is described as being linked to activity in the prefrontal cortex as well as areas such as the anterior cingulate cortex, if the prefrontal cortex activity is suppressed in sleep then no awareness can exist.
Although tempting to suggest that processing and learning of novel information could be carried out in light sleep states, this experiment demonstrates there are difficulties and restrictions associated with it.
Since we`re chatting about the topic…..
….. can we assume that if less well-known categories of objects eg types of vehicles, books etc. that automaticity of the processing would be more difficult and that in the sleep state there would be an increased level of errors and indecision that would force the subject to waken more if at all registered.
…..in this experiment although in 14% of trials subjects pressed their buttons either spontaneously or in response to auditory stimulation it appears that the activity of the anterior cingulate cortex that would normally monitor for errors is also suppressed.
….does this experiment raise the question that learning, processing and other brain activity keeps the subject in the lower sleep states and that the higher NREM3 state is unlikely to be reached whilst this mental activity continues.
…situations that result in abnormal brain functioning eg removal of serotonin by drugs leading to decreased NREM sleep, treatment with the stimulant modafinil that increases the periods of wakefulness and has effects on glutamate, patients with chronic mental fatigue that suffer with deficits in spatial working memory and sustained attention, would lead to differences in the performance of this experiment.
… as the sequence is learnt, the first LRP peak indicating readiness for movement would appear sooner and be maintained at that level throughout the experiment since automatic unconscious processing is considered faster than processing involving working memory and prefrontal areas.

This entry was posted in learning, sleep and tagged . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s