How negative selection, positive selection, and population size contribute to the large variation in nucleotide substitution rates among RNA viruses remains unclear. Here, we studied the ratios of nonsynonymous-to-synonymous substitution rates (d N /d S ) in proteincoding genes of human RNA and DNA viruses and mammals. Among the 21 RNA viruses studied, 18 showed a genome-average d N /d S from 0.01 to 0.10, indicating that over 90% of nonsynonymous mutations are eliminated by negative selection. Only HIV-1 showed a d N /d S (0.31) higher than that (0.22) in mammalian genes. By comparing the d N /d S values among genes in the same genome and among species or strains, we found that both positive selection and population size play significant roles in the d N /d S variation among genes and species. Indeed, even in flaviviruses and picornaviruses, which showed the lowest ratios among the 21 species studied, positive selection appears to have contributed significantly to d N /d S . We found the view that positive selection occurs much more frequently in influenza A subtype H3N2 than subtype H1N1 holds only for the hemagglutinin and neuraminidase genes, but not for other genes. Moreover, we found no support for the view that vector-borne RNA viruses have lower d N /d S ratios than non-vector-borne viruses. In addition, we found a correlation between d N and d S , implying a correlation between d N and the mutation rate. Interestingly, only 2 of the 8 DNA viruses studied showed a d N /d S < 0.10, while 4 showed a d N /d S > 0.22. These observations increase our understanding of the mechanisms of RNA virus evolution. picornaviruses | flaviviruses | influenza A viruses | selective constraints | positive selection R ates of nucleotide substitution can be up to 1 million-fold higher in RNA viruses than in their cellular hosts (1-3). This rapid evolution is mainly due to high mutation rates (4, 5), while natural selection occurs mostly as purifying selection (5, 6). Selection is usually measured by the d N /d S ratio, where d S (d N ) is the number of synonymous (nonsynonymous) substitutions per synonymous (nonsynonymous) site between 2 sequences. Although d N /d S has been studied in many RNA viruses (7), some important issues remain unresolved. One question is the relative contributions of natural selection and effective population size (N e ) to differences in d N /d S among viral species. Positive (Darwinian) selection increases, while negative (purifying) selection decreases, d N /d S . Unfortunately, it is difficult to determine whether an instance of elevated d N /d S is due to positive selection or relaxed negative selection. Positive selection has been found in viruses such as influenza A viruses (8,9) and HIV-1 (10-12). However, the contribution of positive selection to the genomic mean d N /d S has not been evaluated. Because natural selection is more effective in large populations and negative selection predominates (7), an increase in N e would be expected to reduce the mean d N /d S . Unfortunately, N e is usual...