Substitutions that result in amino acid replacements are said to be nonsynonymous while substitutions that do not cause an amino acid replacement (such as a GGG to GGC change - both codons still encode glycine) are said to be synonymous substitutions. Because of the difference in their effects on the physiology of the organism, synonymous and nonsynonymous substitutions can have quite different dynamics. For example, synonymous substitutions usually occur at a much faster rate than do nonsynonymous substitutions. Hence, for coding sequence it is often desirable to separate these two.
The most common method to estimate these parameters separately is via an algorithm set out by Li, Wu & Luo (1985; Mol. Biol. Evol. 2:150-174). It is somewhat complicated and I refer you to their paper for a complete description. Basically it counts the number of sites that are potentially 4-way, 2-way or 0-way degenerate (the third position of a glycine codon being 4-way degenerate, any second codon position being 0-way degenerate). It then counts the number of differences at each site of each category keeping tract of transversions and transitions. It then calculates
the rate of synonymous and nonsynonymous substitutions. It has been
found that 
can have large variation and great changes
between/within specific organisms. On the other hand, 
is
generally less variable (though still shows more variation than would
otherwise be predicted) and shows less changes between/within
organisms.