I think what he is describing is talking about quality of sleep rather than quantity.
I'm no expert, so I'm really just regurgitating something I read awhile ago, so please bare in mind that it is probably not 100% correct, but the seed of the idea is there. I'm more or less going to describe what I've done based off what I've read.
The body only needs a couple of hours of REM (rapid eye movement) sleep to feel fully refreshed, anything over that and you're essentially wasting time.
I'm surviving pretty well on about 4-5 hours a night, it really depends but I find that if you exercise first thing in the morning as soon as you wake up, you do not miss that sleep.
I've actually purchased a recumbent exercise bike so that I can read blogs and forums in the morning while I exercise - this helps me compress my time so that I can get more done during the day.
The most important thing I've found is to ensure you have a consistent pattern for sleeping, otherwise my body will invariably want more sleep as it tries to get back into some sort of pattern.
To aid this, I've had my doctor prescribe me 10mg Temazepam which I think is a fantastic drug. I'm not on a high enough dosage to get hooked, but if need too, I take 1 tablet and it knocks me out fairly easily and unlike other drugs it is not one of those ones that will keep you asleep.
I'm no expert on this either; I just studied this section of the book pretty thoroughly :)
What you say is correct, but about SWS ("Slow Wave Sleep", or stages 3 and 4 -- stage 1 being when REM occurs). These are the main recouperative stages.
Eliminating REM sleep (e.g. by waking someone up at the start of a REM cycle then letting them go back to sleep when the cycle would have ended) has little or no effect on alertness the next day, but doing the same for SWS cycles results in symptoms of sleep deprivation.
It has been hypothesized that REM sleep plays a role in strengthening memories, but evidence for this is spotty.
Regardless of how many hours you're asleep, reducing the amount of time spent asleep reduces the amount of REM sleep, but not NREM sleep, making you a "more efficient" sleeper.
The main evidence for long-term sleep reducton comes from studies by Friedman et al. (1977) and Mullaney et al. (1977). In these studies, which covered eight subjects over a period of several months, two reduced their sleep down to 5.5 hours, four down to 5 hours and the remaining two down to 4.5 hours. Below about 6 hours, the subjects reported daytime sleepiness, which increased as sleep was futher reduced. A follow-up study a year later showed that the subjects were sleeping 7-18 hours less per week. With only eight subjects, I'm a bit concerned about the generalizability of this study, but I think it's good enough for my purposes.
I read some material on this subject several years ago and have been putting it into practice ever since, so the specifics get lost in the day to day usage.
