I've started reading the famous "Lore of Running" recently, and it's been really enlightening. Noakes has done most of the serious medical studies on running physiology, so he is the best person to write a comprehensive book on the subject. (It is filled with citations like "Noakes, 1998" which means he is quoting himself.)
The best part of the book is how he shoots down long-standing beliefs with hard science. His most emphatic attack is on the "dehydration is the root of all evils" theory. That argument goes like this: any weight lost during exercise is due to sweating, which causes dehydration. Dehydration leads to overheating, and hyperthermia leads to at least decreased performance and at most heat stroke. Therefore runners should drink enough to replace all fluid lost during exercise.
There are several huge flaws in this argument, which Noakes is quick to point out. First, not all weight lost during exercise is due to sweating. Some material is used up by metabolic processes. Also, water is a byproduct of some of those metabolic pathways, so some of what you sweat could be that new water that was created by your muscles. Even a perfectly hydrated person could lose as much as 3 kg of body weight through these ways.
Second, dehydration has only a minimal effect on your body's thermoregulatory system. The bigger factors that contribute to heat stroke are the runner's weight, the environmental conditions, and the exercise intensity. A larger person running at high intensity in hot and humid conditions would still easily develop heat stroke despite how much water he or she drinks.
Third, there is no evidence that even moderate dehydration (5-10%) impairs performance. Early studies showed that the fastest runners in any race are usually the most dehydrated; clearly their level of hydration wasn't hindering them because they were the winners.
This exaggeration of the problems of dehydration has lead to the recommendation that you should drink as much as you sweat. I've read this in reputable running books and magazines, and never heard an argument to the contrary until now, so clearly it has become pervasive. But since this advice has been promulgated, cases of hyponatremia (too much water) have spiked, leading to serious injury and in many cases death.
So the best advice is drink as much as you feel you need to, which is usually around 400 mL per hour. Noakes debunks other myths, but the arguments are pretty complicated (even though I'm a scientist I have to stop and reread a paragraph from time to time to follow his logic.) So I will try to digest them as much as possible and share them on the blog.