The nucleotide sequence of a nearly full-length cloned cDNA copy of an 18S defective interfering (DI) RNA of Semliki Forest virus has been determined. This corresponded to a major virus-specific cytoplasmic RNA species at the 11th undiluted passage of the virus in BHK cells. The 1652-nucleotide-long sequence consists of a unique 5'-terminal sequence followed by three tandem 484-nucleotide repeat units derived from the 5' two-thirds of the viral genome and a unique sequence of 106 nucleotides preceding the poly(A) of the 3' terminus. One of the tandem 484-nucleotide repeat units contains an extra segment of 60 nucleotides. Hybridization experiments showed that the cloned cDNA was colinear with an 18S DI RNA and that it contained an approximately 320-nucleotide-long segment colinear with the viral genomic RNA. Analysis of 18S DI RNA oligonucleotide fingerprints revealed that the molecule studied and the heterogeneous DI RNA population contain similar repeated sequences. The mechanism by which the DI RNAs are generated is not known, but it seems likely that multiple internal deletions and duplications are involved.