Mutations in the human survival motor neuron 1 (SMN) gene are the primary cause of spinal muscular atrophy (SMA), a devastating neuromuscular disorder. SMN protein has a well-characterized role in the biogenesis of small nuclear ribonucleoproteins (snRNPs), core components of the spliceosome. Additional tissue-specific and global functions have been ascribed to SMN; however, their relevance to SMA pathology is poorly understood and controversial. Using Drosophila as a model system, we created an allelic series of twelve Smn missense mutations, originally identified in human SMA patients. We show that animals expressing these SMA-causing mutations display a broad range of phenotypic severities, similar to the human disease. Furthermore, specific interactions with other proteins known to be important for SMN's role in RNP assembly are conserved. Intragenic complementation analyses revealed that the three most severe mutations, all of which map to the YG box self-oligomerization domain of SMN, display a stronger phenotype than the null allele and behave in a dominant fashion. In support of this finding, the severe YG box mutants are defective in self-interaction assays, yet maintain their ability to heterodimerize with wild-type SMN. When expressed at high levels, wild-type SMN is able to suppress the activity of the mutant protein. These results suggest that certain SMN mutants can sequester the wild-type protein into inactive complexes. Molecular modeling of the SMN YG box dimer provides a structural basis for this dominant phenotype. These data demonstrate that important structural and functional features of the SMN YG box are conserved between vertebrates and invertebrates, emphasizing the importance of self-interaction to the proper functioning of SMN.