The phenomenon of nonsense-associated altered splicing raises the possibility that the recognition of in-frame nonsense codons is used generally for exon identification during pre-mRNA splicing. However, nonsense codon frequencies in pseudo exons and in regions flanking 5 splice sites are no greater than that expected by chance, arguing against the widespread use of this strategy as a means of rejecting potential splice sites.Keywords: NAS; nonsense; pre-mRNA splicing; exon definition; latent splice sites; pseudo exons; exon skipping Two recent papers have bolstered the idea that the translatability of an exon can influence its splicing. Wang et al. (2002) have shown that an exon in the T-cell receptor transcript is not only excluded when it contains a nonsense codon, but that this event is accompanied by the inclusion of an overlapping cryptic alternative exon. The new exon restores an open reading frame. Independently, Li et al. (2002) have shown that a latent 5Ј splice site downstream of an exon can be activated if all in-frame nonsense codons are removed from the region between it and the upstream exon. In these mutated pre-mRNA molecules, a new enlarged exon is chosen for splicing, based on its newfound extended translatability. Explanations offered for these two results include nuclear recognition of the translatability of an exon before it is spliced.The recognition of exons or introns during the splicing process cannot rest on the splice site sequences alone, because similar sequences abound in large pre-mRNA molecules (Senapathy et al. 1990). There is considerable evidence that it is the exon that is the initial element of recognition, providing a size constraint for choosing real splice sites. However, if we define "pseudo exons" as intronic regions of typical exon size (50-250 nt) bounded by sequences that closely match the consensuses for 3Ј and 5Ј splice sites, their abundance still outweighs that of real exons by an order of magnitude (Sun and Chasin 2000). Splicing enhancers, assayed mostly in the context of alternatively spliced exons, are also quite varied and degenerate, and candidate sequences are easily found in pseudo exons (data not shown). How then does the cell distinguish real exons from pseudo exons? Exon translatability could be providing the missing information.To test this idea, we examined the predicted translatability of pseudo exons. If pseudo exons are being generally discriminated against because they contain nonsense codons, then each should contain at least one in-frame nonsense codon. If, on the other hand, nonsense codons play no such role, then nonsense codons should occur at a frequency dictated simply by chance. We culled pseudo exons from a human intron-exon database (Saxonov et al. 2000) after eliminating redundant and predicted genes. We defined these pseudo exons as intronic sequences that have at the upstream end a pseudo splice site with a consensus 3Ј matrix score of 78 and at the downstream end a pseudo splice site with a consensus 5Ј matrix score of 75; u...