There is an interesting discussion to be had about the experimental design. The intervention was essentially: "the subject managed to fall asleep, and then woke

There is an interesting discussion to be had about the experimental design. The intervention was essentially: "the subject managed to fall asleep, and then woke up in the N1 stage." This assignment is not "random" in the same way that a drug trial is random. There are many general-state variables which likely govern our ability to fall asleep (anxiety, hunger, etc) and this study doesn't block any of them in the same way that a drug trial would (via randomization). That being said, the fact that participants who fell asleep and _didn't_ wake during N1 did not see improvement over control makes me more intrigued. (Assuming this is all true: brain scan research is really noisy and small-sample-size and has an atrocious replication record).