Polyoma virus late RNA processing provides a convenient model system in which to study the mechanics of splicing in vivo. In order to understand further the role of the untranslated "late leader" unit in late RNA processing we have constructed a group of polyoma viruses with deletions and substitutions in the leader exon. This has allowed us to determine that there is a minimum exon size required for both pre-mRNA splicing and stability in this system. We show here that the non-viability of a mutant (ALM) with a 9 base late leader unit is due to a general defect in late RNA splicing. In addition, ALM-infected cells show at least 40-fold depression in the accumulation of late nuclear RNA (spliced or unspliced). The ALM late promoter, however, functions nearly normally. Substituted leader variants with 51- to 96-base long exons of unrelated sequence are viable (G. Adami and G. Carmichael, J. Virol. 58, 417-425, 1986). We show here that late RNA from one of these substituted leader mutants (containing a 51-base leader exon) is spliced at wild type levels, with virtually no defect in accumulation. Thus, in the polyoma system, splice sites separated by only 9 bases can inhibit each others usage, presumably by steric interference. We suggest that this type of inhibition leads to extreme RNA instability.