In Drosophila, the spliceosomal protein SNF fulfills the functions of two vertebrate proteins, U1 snRNP-UlA and U2 snRNP-U2B". The structure and sequence of SNF, U1A, and U2B" are nearly identical with two RNA recognition motifs (RRM) separated by a short linker region, yet they have different RNA-binding properties: U1A binds U1 snRNA, U2B" binds U2 snRNA, and SNF binds both snRNAs. Structure/function studies on the human proteins have identified motifs in the N-terminal RRM that are critical for RNA-binding specificity but have failed to identify a function for the C-terminal RRM. Interestingly, SNF is chimeric in these motifs, suggesting a basis for its dual specificity. Here, we test the importance of these motifs by introducing site-directed mutations in the snf coding region and examining the effects of these mutations on assembly into the snRNP and on snf function in vivo. We found that an N-terminal RRM mutant protein predicted to eliminate RNA binding still assembles into snRNPs and is capable of rescuing snf's lethal phenotype only if the normally dispensable C-terminal RRM is present. We also found that the mixed motif in the "RNA-specificity" domain is necessary for SNF's dual function whereas the mixed motif in the U2A'-protein-binding region is not. Finally, we demonstrate that animals carrying a snf mutation that converts SNF from a bifunctional protein to a U1 snRNP-specific protein are viable. This unexpected result suggests that SNF's presence within the U2 snRNP is not essential for splicing.