It's to do with known mutation rates. Simply put, we, as a species, undergo far too many mutations. Think of it like this, allowing for the simplistic language, an individual undergoes a mutation that improves his eyesight in the dark by a fraction, right. Well, by even the most conservative estimates of mutation rates in humans, he's also going to be picking up hundreds of other mutations, all told, and a percentage of those will be in the germ line. When he passes on his genes to his children, he will be passing on not just the odd incredibly rare mutation that benefits an organism in a given environment, he will also be passing on an array of other mutations so long as they aren't so severe that they hinder his reproductive capabilities. This makes them, by the probabilities involved nearly neutral mutations, which is just another term for "slightly harmful". Since selection doesn't work at the level of genes, but at the level of the organism, it cannot select out the good mutation from the many bad ones, so in order to solve this paradox, population geneticists hypothesised that the vast majority of these mutations were occurring in the 98% or so of junk DNA in the human genome, where changes to the DNA are allowed to accrue without affecting the organisms "fitness". These they called "neutral mutations". Simply put, they were hoping that given enough time, a random sorting of these segments of the DNA could result in new segments of coding DNA, resulting in beneficial adaptations in an organism, but that they would also soak up the great number of mutations every individual is currently collecting in his/her genome without degradation to the integrity of the genome.