Previous studies have shown that the mitochondrial coxII.i3 intron is absent in all rosids examined, in Philadelphus (Hydrangeaceae, Cornales), and in Catharanthus and Vinca (Apocynaceae, Gentianales). We surveyed for the presence or absence of this intron in 177 species representing all orders of angiosperms, where it is primitively present. The intron appears to have been lost independently in Gnetales, Laurales, Zingiberales, Ranunculales, Saxifragales, rosids, Santalales, Caryophyllales, Ericales, Cornales, Gentianales, Lamiales, Boraginales, Aquifoliales, Asterales, Dipsacales, as well as in the genus Escallonia. Depending upon the phylogenies used to interpret the losses, and due to lack of resolution in some groups, the intron could have been lost up to 27 times in the angiosperms (excluding the loss in Gnetales). The losses sometimes corroborate the monophyly of groups (rosids and Lamiales) or of subgroups within orders (in the Ranunculales, Caryophyllales, Cornales, Gentianales, and Asterales). In other groups, such as the Saxifragales, Ericales, and Dipsacales, the patterns of losses are more complex and would require further study. The presence or absence of the coxII.i3 mitochondrial intron seems a useful phylogenetic marker in some groups, but caution in interpretation is needed as multiple parallel losses have occurred throughout the angiosperms.