We explored the interordinal relationships of mammals using DNA–DNA hybridization, with particular reference to the much–debated problem of whether the megabats and microbats are more closely related to each other than the megabats are to primates. To try to improve resolution when taxa are distantly related and the melting points of hybrids are low and difficult to distinguish, we increased the GC–content of DNA by a fractionation method that used the same melting–point apparatus also used in the hybridization studies. When we used GC–rich DNA as the tracer to make hybrids, the melting point of the self–hybrid shifted to a higher temperature as expected, but the behaviour of heterologous hybrids varied with the taxa being compared. When the melting point of the heterologous hybrid also shifted to a higher temperature so that the two compared taxa maintained the same or proportional distance, we called this ‘following behaviour’, because the heterologous hybrid made with GC– tracer ‘followed’ the GC– self– to higher temperatures. We also commonly saw anomalous behaviour, where the melting point of the heterologous hybrid shifted to a lower temperature when compared with an AT– hybrid. In these anomalous cases, the distance measured between the taxa increased markedly as a result of GC–, indicating that an underestimate of distance may have resulted from AT– in DNA. This inference was supported by the finding that it was rare to observe a decrease in measured distance between taxa using GC– DNA, but very common to find an increase as would be expected from the generally higher AT–contents of eutherian DNAs. Moreover, the most extreme cases, where distances changed most using GC–rich DNA, were usually those involving comparisons between taxa known to have the most extreme AT–biases among mammals, such as the megabats and rhinolophoid (including megadermatid) microbats. Our results show that AT–bias in eutherian DNA leads to consistent underestimates of measured differences between taxa with extreme AT–biases.