Edit: Wow; I've essentially changed my mind completely about this thanks to the comment by Nornagest and the link provided by icebrand. See my more recent comment for my current take.
.....
My understanding is that this is not a good idea. You may want to read some of Piotr Wozniak's work. He's the author of SuperMemo which had a long article in Wired HERE. This program's algorithms are at the root of those used in Anki and Mnemosyne
In any case, while Wozniak isn't a neurologist or sleep expert, per se, my reading of his site seems to present quite a compelling case for his wide familiarity with sleep models, it's function, and practical applications at the very least.
Here is his main page for sleep.
Note the link at the above to myths and facts. Here is one of them verbatim (#13):
Myth: We can adapt to polyphasic sleep. Looking at the life of sailors, many people believe they can adopt polyphasic sleep and save many hours per day. In polyphasic sleep you take only 4-5 short naps during the day totaling less than 4 hours. There are many "systems" differing in the arrangement of naps. There are also many young people ready to suffer the pains to see it work. Although a vast majority will drop out, a small circle of the most stubborn ones who survive a few months and will perpetuate the myth with a detriment to public health. Fact: We are basically biphasic and all attempts to change the inbuilt rhythm will result in loss of health, time, and mental capacity. A simple rule is: when sleepy, go to sleep; while asleep, continue uninterrupted. See: The myth of polyphasic sleep.
At the link about the myth of polyphasic sleep, the Uberman sleep cycle is discussed at length. Wozniak does not seem to concede any validity to the various write ups about this method that have led to people wanting to try it.
Even glancing at the wiki article on this does not present convincing evidence that this is a desirable experiment. Stampi seems to have adopted a sleep pattern like you suggest, but he along with the other two military groups seem to prefer monophasic sleep to polyphasic and only advocate this pattern to relieve deprivation effects under extreme conditions.
Perhaps we should back up: what are you using to support this as a good idea? Are you just interested in the results (positive or negative) for experimentation's sake, or are you trying to implement something you have been led to think is a method of self-improvement (efficiency, learning capacity/speed, etc.)?
Wozniak seems to suggest that there are at best misconceptions about this method's supposed ability to give you more time (and therefore learning/productivity) and at worst serious detriments to sleep and life quality as well as health.
Then again, like I said, I'm not sure about Wozniak's position as any kind of expert... I just know he's written quite a bit on sleep. I could not tell you if any studies had been directly carried out about this, what they found, or what the collective body of those who have attempted it report.
I'd probably be more inclined to believe that Uberman and other polyphasic sleep patterns have serious drawbacks than that they don't, but Wozniak's page is distinctly long on rhetoric and anecdote and short on data.
The section about the biological basis of sleep timing is about the only thing that I'm not inclined to throw out as obviously biased, although even that isn't good news for Uberman.
My friend and I are starting the Uberman sleep schedule (six 20-minute naps spread evenly throughout each day) tonight. Have other lesswrongians experimented with alternate sleep schedules? Are any of you qualified medical experts who can give input or advice? Success stories and failure stories would both be appreciated, and I'll keep you guys posted on our progress.