Fitch and Markowitz (1970)
have proposed that a limited number of codons in a given gene can undergo favorable mutations, or mutations leading to amino acid substitutions of little or no effect. These mutations are widely termed "neutral," and Fitch and Markowitz call the codons "concomitantly variable codons" or eovarions.
When a favorable or neutral mutation takes place, the set of covarions may be altered: some former members of the set may no longer mutate without consequence, and mutation at other codons may now be innocuous or favorable for the first time. This concept seems to reconcile the observation of strong restriction of substitution sites with the hypothesis that most amino acid substitutions in evolution have been random and of no adaptive significance (King and Jukes, 1969;Crow and Kimura, 1970). This hypothesis extends to electrophoretic polymorphism as well (Crow, 1972;Fitch, 1972;Milkman, 1972).
I wish to note one hidden assumption made by Fitch and Markowitz (1970) when they apply the limitations on a single allele (with, say, ten covarions) to an entire species. They assume that an entire species is characterized by a single covarion set at any one time. But in fact the number of codons that can accept a mutation in a species should by now far exceed that in an individual. The average number of such codons (covarions) per individual will of course remain the same, and this is what the authors suggest determines observed substitution rates.