Undaunted by the lack of comments, I continue...
In the 1960's, it was discovered that it was possible to increase the amount of glycogen stored in muscles by first depleting it (and this led to about a 5% improvement in marathon times among the sub-3 crowd). The lower one could get one's glycogen stores, the better, and holding it low for a few days helped even more; this supercompensation doesn't last forever, however, as it's been shown at least anecdotally that those who maintain a Zone or Atkins diet accustom to the lower glycogen levels.
In the early 1970's, Dr. David Costill was doing studies on glycogen storage and he did an experiment on himself. Normally a 5 mile a day runner, he ran 10 miles on three consecutive days and then measured the glycogen level in his muscles. He found that it took 10 days to recover ("up to two weeks" some report it) and felt that he must be unusual. It was an experiment of one person done once and he didn't say what exercise he did or what he ate during recovery, so it's not exactly a scientific study, but I think it's the first recorded example of what's become known as the crash training phenomenon.
In the early 1990's, there were 3 studies, one each among cyclists, swimmers and runners, which had athletes dramatically increase the volume or intensity of their training for two weeks. After this, they were completely spent and it took two weeks to recover, but after this recovery, they were able to perform at a higher level in a test. The number of variables involved and the small sample sizes make these studies questionable, but the possibility of significant improvement on two weeks of training was intriguing.
The next question to be asked was: is it a one-time thing, or can one repeat the two weeks on, two weeks off pattern and build upon the previous improvement? This appears to be what was done by the top Chinese women runners Wang Junxia and Qu Yunxia, who set world records in 1993 (which still stand), who reportedly ran as much as 25 miles (40K) every day at times during their training. Because their teammates and coach later received bans for use of illegal substances, because their careers were meteoric (rapid success and then oblivion) and because there were no others who followed in their footsteps, it's generally believed that their efforts were because of illegal drugs, but the training is still valid.
Attempts to increase the length of the hard periods to three weeks in length have generally resulted in injury. Shortening the "load" phase to only a few days has become customary in what has become known as crash training, though there is little science to support it. At the most basic level, almost all of us have had some experience with this; the day after a hard run, one often feels surprisingly good (the fatigue sometimes sets in only the next day) and one is tempted to run hard again; actually running hard again often causes fatigue that takes a very long time from which to recover. These back-to-back hard workouts are the hallmark of weekend warriors, who race Saturday and Sunday and then recover the rest of the week.
Next up: periodization history