Published analyses of the sequences of three genes from the 1918 Spanish influenza virus have cast doubt on the theory that it came from birds immediately before the pandemic. They showed that the virus was of the H1N1 subtype lineage but more closely related to mammal–infecting strains than any known bird–infecting strain. They provided no evidence that the virus originated by gene reassortment nor that the virus was the direct ancestor of the two lineages of H1N1 viruses currently found in mammals; one that mostly infects human beings, the other pigs. The unusual virulence of the virus and why it produced a pandemic have remained unsolved. We have reanalysed the sequences of the three 1918 genes and found conflicting patterns of relatedness in all three. Various tests showed that the patterns in its haemagglutinin (HA) gene were produced by true recombination between two different parental HA H1 subtype genes, but that the conflicting patterns in its neuraminidase and non–structural–nuclear export proteins genes resulted from selection. The recombination event that produced the 1918 HA gene probably coincided with the start of the pandemic, and may have triggered it.