Sleep Promotes Extraction of Grammatical Rules

Busy Clever Brain

Did you know that even while you are sleeping your brain stays busy? Sleep is actually very important for memory. It is known that the brain reactivates memories during sleep, and that this helps you remember things better after sleep. And not only memory for facts and events (which we call declarative memory) benefits from sleep, sleep also helps you learn motor tasks (procedural memory).


We did not know, however, whether sleep also plays a role in unraveling highly complex relations in information we are exposed to. Can sleep also integrate information that is present over a whole set of related memories, extract the overarching information only present when the memories are combined? This is what we set out to investigate.

A grammar is a complex set of relations between words that supports us in comprehending how concepts are interrelated in sentences; who did what to whom. To acquire a grammar, something we automatically do when we learn a language, your brain has to unravel highly complex relations in information you are exposed to. Information must be integrated from a whole set of related memories (sentences you heard). Only by looking at all the memories together, your brain can figure out the rules of the grammar, and you can produce and recognize grammatically correct sentences. Therefore artificial grammar learning was the perfect task for us.

In our artificial language with our artificial grammar, there are no words and sentences, only letters and letter strings. The order of the letters in the letter strings is determined by a complex set of rules we call a Reber Grammar. You can see the rules in the picture below. With this grammar you can make hundreds of strings like VXSVRXRRM or MVRXSVRXRSSV. And we already knew that after you show a whole bunch of these strings to people, they automatically get a feel for the rules. Without even knowing it. We know this because if you show a new letter string, one they never saw before, and you ask them to guess whether it follows the rules or not, they can do this! Our clever brain!



But the big questions was, does sleep help in this process? Can we extract grammar rules while we are in Morpheus arms?

A little lie

Our participants came to the lab thinking that did would participate in a working memory study. A little lie for science, because we were not really interested in their working memory performance. We just wanted to expose them to the artificial grammar. We showed them 100 letter strings, which appeared on their computer screen one by one. After each string, they had to type in the string from working memory. They did this for one and a half hour, seeing 100 grammatical strings two times.


All the participants came back to the lab for a second task. Only then we told them that the letter strings they had seen during the previous task were made according to a complex set of rules. We did not tell which rules. In the second task we showed them brand new letter strings and asked to guess for each sting if they thought it was grammatical or not. Just based on their gut feeling. We called this second task the classification task, since they had to classify the new strings as grammatical or not. We had several groups of participants that differed in the amount of time between the first working memory cover task, and the second classification task. Some did the second task after a mere 15 minute break, some after 12 hours of wake time, some after a full night of sleep.

And the cool thing? The participants that slept between exposure to the grammar and seeing the new strings performed much better! They better recognized the grammatical strings between the non grammatical ones. Importantly, it did not matter whether they did the working memory task and the classification in the morning or the evening, the only thing that mattered was whether they slept or not. We had a whole bunch of groups to make sure of that, one of the nasty things you have to do when being a sleep researcher, running many control groups.


1 + 1 = 3

An other control that we did is that we tested why people were better in classification. Did they recognize short letter combinations (like SVR or XR) that appeared often in the strings that they saw during the working memory? Or did they also recognize strings that were grammatical but did not look very much like the ones they originally saw? We tested this by also putting strings in the classification set that had these frequent letter combinations like SVR or XR, but did not follow the rules (so were non grammatical), and adding strings that did not have these frequent substrings, but did follow the rules. Sleep especially helped people to recognize these latter unsimilar grammatical strings, and to reject the similar but non grammatical strings.

This shows that sleep does not just strengthen the memory for frequently appearing subparts of information. Our study implies that there is a complex integration process going on in our brain during sleep. The brain is hard at work, piecing information together and extracting new information out of the individual bits. Memories of individual sequences are combined, resulting in abstract knowledge reaching beyond the previously encountered exemplars.

So next time when you have to integrate complex information, don't pull an all nighter. Expose your brain to the information, sleep, and behold, you might be surprised.

Want to read the whole story? See the paper here.