Human beta-globin mRNA precursors (pre-mRNAs) synthesized in vitro from a bacteriophage SP6 promoter/beta-globin gene fusion are accurately and efficiently spliced when added to a HeLa cell nuclear extract. Under optimal conditions, the first intervening sequence (IVS 1) is removed by splicing in up to 90% of the input pre-mRNA. Splicing requires ATP and in its absence the pre-mRNA is neither spliced nor cleaved at splice junctions. Splicing does not require that the pre-mRNA contain a correct 5' or 3' end, a 3' poly A tail, or a 5'-terminal cap structure. However, capping of the pre-mRNA significantly affects the specificity of in vitro processing. In the absence of a cap approximately 30%-40% of the pre-mRNA is accurately spliced, and a number of aberrantly cleaved RNAs are also detected. In contrast, capped pre-mRNAs are spliced more efficiently and produce fewer aberrant RNA species. The specificity of splice-site selection in vitro was tested by analyzing pre-mRNAs that contain beta-thalassemia splicing mutations in IVS 1. Remarkably, these mutations cause the same abnormal splicing events in vitro and in vivo. The ability to synthesize mutant pre-mRNAs and study their splicing in a faithful in vitro system provides a powerful approach to determine the mechanisms of RNA splice-site selection.