Recently there has been much interest in assessing the role of alternative splicing in evolution. We have sought to measure functional selection pressure on alternatively spliced single-exon skips, by calculating the fraction that are an exact multiple of 3 nt in length and therefore preserve protein reading-frame in both the exon-inclusion and exon-skip splice forms. The frame-preservation ratio (defined as the number of exons that are an exact multiple of three in length, divided by the number of exons that are not) was slightly above random for both constitutive exons and alternatively spliced exons as a whole in human and mouse. However, orthologous exons that were observed to be alternatively spliced in the expressed sequence tag data from two or more organisms showed a substantially increased bias to be frame-preserving. This effect held true only for exons within the protein coding region, and not the untranslated region. In five animal genomes (human, mouse, rat, zebrafish, Drosophila), we observed an association between these conserved alternative splicing events and increased selection pressure for frame-preservation. Surprisingly, this effect became stronger as a function of decreasing exon inclusion level: for alternatively spliced exons that were included in a majority of the gene's transcripts, the frame-preservation bias was no higher than that of constitutive exons, whereas for alternatively spliced exons that were included in only a minority of the gene's transcripts, the frame-preservation bias increased nearly 20-fold. These data indicate that a subpopulation of modern alternative splicing events was present in the common ancestors of these genomes, and was under functional selection pressure to preserve the protein reading frame.