We have examined the RNAs involved in the heterologous protection of adult mice from otherwise lethal intranasal infection with mouse-adapted human A/WSN (H1N1) by defective interfering (DI) equine A/equine/Newmarket/7339/79 (H3N8: EQV) influenza virus, as well as the RNAs involved in the protection of WSN- or EQV-infected mice by their homologous DI viruses. The aim of this study was to describe the types of defective RNAs present in protected mice in order to guide the design of potentially protective DI RNAs. The interfering and mouse-protecting activity of DI virus was destroyed by prolonged UV irradiation (iDI virus) demonstrating that protection correlated with an active DI genome, and not viral antigen. Protected mice were all infected but suffered a lower degree of morbidity than those given iDI virus. The DI EQV inoculum contained defective segment 1–8 RNAs while DI WSN inoculum contained only defective segment 1–6 RNAs. However lungs of mice given EQV + DI EQV contained only defective segments 1–4 or 1–6 RNAs (mouse-to-mouse variation), while control mice given EQV or EQV + iDI EQV contained few very defective RNAs. Thus prevention of death was the result of quantitative and/or qualitative differences in defective RNAs administered to the mice. Only defective segments 1–3 RNAs were isolated from the lungs of mice given WSN + DI WSN, confirming the earlier report of Noble and Dimmock (1995). A detailed analysis showed that most defective RNAs isolated from the lungs of mice protected from a lethal WSN infection by DI EQV were EQV in origin. Thus, as no infectious EQV was present, these defective RNAs from the DI EQV inoculum must have been heterologously replicated in mouse lung by WSN. All defective segment 3–6 RNAs isolated were of EQV origin, indicating that they were replicated by WSN in preference to its own. Defective segments 1 and 2 were a mixture of EQV and WSN RNAs. Of 17 defective EQV segment 1–3 sequences from mouse lung, all but three differed in their primary central deletion from 20 defective RNAs isolated from the inoculum. No bias in the break points was evident. A number of minor deletions of 2 or more nts were also present in defective EQV and WSN RNAs in segments 1 and 2, but none in segment 3. Their 5′, but not 3′, breakpoints were heterogeneous, suggesting that defective RNAs were generated during positive strand synthesis. Two cloned EQV-defective segment 3 RNAs were chimeras containing a 30 nt insert from segment 1. Most defective RNAs possessed at least 178 nts from the 5′ end of vRNA. The amount of 5′ sequence present in those RNAs correlated with the segment of origin, suggesting that this was the minimum required for propagation of viral RNA in mouse lung and hence possibly for protection also.