Genetic mutation is a necessary component of biology, because without the generation of genetic diversity there could not be any evolution –– this is true for "deterministic" evolutionary forces such as natural selection as well it is for "stochastic" forces, such as genetic drift. Nevertheless, allowing mutation to occur at an appreciable rate comes at a cost. Genetic mutation is a spontaneous and random process. It cannot be turned on and off at the whim of the organism. Nor can an organism direct the outcome of mutation in response to environmental change; with rare exceptions, life does not obey Lamarckian evolution.
Consequently, mutation –– the source of variants with which evolutionary change is made possible –– is a costly process. Like typographical errors made by a printing press, random change seldom improves an optimized form. The accumulation of errors in the genetic sequence, oft caused by mistakes made during DNA replication, tends to cause a reduction in fitness. That is to say, mutation debilitates an organism's ability to survive and reproduce. Natural selection acts through the death of mutant individuals; the worst printing presses tend to be thrown away more often. Thus, natural selection and mutation act in opposition. One erodes the fitness of a population while the other sustains it.
With opposition, a balance is obtained. Nevertheless, a population that experiences mutation must necessarily be less fit on average than another population that does not –– on the assumption that the latter is already perfectly adapted to the environment. The difference between the two is an evolutionarily significant quantity, and we call it the mutation load. It is the cost incurred by the ability to mutate.
Note that it is implicit in the assumption above that if the environment were to change, the non-mutable population would no longer be perfectly adapted. Indeed, it might then have a fitness lower than the mutating population; because of mutation, it has the genetic variation available to adapt to the new challenge.
Is then the concept of a mutation load a useful one? Certainly, it is. Once the mutating population has defeated the other and become perfectly adapted, its fitness immediately begins to be eroded by mutation. Any individual fortunate enough to have a mutation that shut off its mutation rate would have a fitness advantage, and we are back to square one.
It is important to note that the quantitative definition of mutation load assumes that the effect of genetic drift is negligible (i.e. the population is effectively infinite). This effect is separately addressed in the concept of drift load. Indeed, there are several other "genetic loads" (e.g. recombination load).
Arguably, it is thermodynamically impossible to have a zero rate of mutation. The population with no mutations appearing at all, that has been so useful in our discussion above, is a rather imaginary one. Outside of theoretical models, consequently, the mutation load itself is quite difficult to measure directly. It becomes a matter of measuring the rate that mutations appear, and the effects on fitness that they tend to have. A great deal of effort has been invested towards these ends, for a wide array of organisms (see mutation rate for further discussion). The general outcome is that the mutation load predicted by these studies is quite high –– it roughly translates to being on the order of 10% of our fitness. To put a real face on it, consider that stillborn children and those suffering from congeneric diseases such as cystic fibrosis are expressions of mutation load. It is no trivial matter.
The population genetic literature is rich with material on the mutation load. A recent article, based on a public lecture delivered to the National Academy of Sciences, is an appropriate place to begin:
Crow JF. 1997. The high spontaneous mutation rate: Is it a health risk? Proceedings of the National Academy of Sciences, USA 94: 8380-8386.