Matt Davis of the Medical Research Council Cognition and Brain Sciences Unit in Cambridge, the UK, believes that the same may be true for people as they doze off, without the inducement of sleep by drugs.
Twelve volunteers were studied when they were under the influence of varying amounts of an anaesthetic called propofol, which induced varying levels of drowsiness.
The researchers played them recordings of speech or other sounds, and monitored their brains using a technique called functional magnetic resonance imaging.
It was found that the brains of the volunteers were more active in response to speech than to generic noise, suggesting that they still recognised spoken words.
However, the researchers noticed among the drowsiest volunteers that the part of the brain involved with the more subtle job of untangling words that can have alternative meanings depending on context or spelling (such as 'bark', or 'pear/pair') was inactive in them.
This part had also stopped forming memories of speeches when the volunteers were drowsy, said the researchers.
According to them, the findings indicate that the brain shuts down higher-level aspects of speech recognition as sleep starts to set in, making it hard to remember or understand what was said in the moments before sleep.
Anaesthetist David Menon of the University of Cambridge says that the results, published in the journal Proceedings of the National Academy of Sciences, also suggest that speech comprehension can suffer even when people are only lightly sedated, or when they are slightly sleepy.
The researchers are of the opinion that their findings may help better understand the degree of awareness experienced by patients in operating theatres.
"We don't want to overdose but we want to provide a measure of how much is 'enough' anaesthetic," Nature magazine quoted Menon as saying.
Although he admits that it is difficult to know exactly what the patients was experiencing simply on the basis of brain scanning, he hopes that more studies of lightly sedated healthy volunteers may provide accurate descriptions of their cognitive experience to go with their brain readings.
"This has to be seen as a first step, where we try and calibrate brain responses," Menon says.