The appearance of scandent biserial graptolites in the Middle Ordovician is one of the most important events in graptolite evolution. Such graptolites rapidly achieved worldwide distribution; moreover, the diplograptids included the basal members of Virgellina, which dominate the post-Ordovician history of the group. We review the range of scenarios, some of them contradictory, that have been suggested to account for the origin and relationships of glossograptids and diplograptids. New information on the proximal end structures of early biserials and their putative relatives is described from well-preserved relief material, mostly from the Ningkuo Shale of south-eastern China. This information, together with previous descriptions, provides the basis for a cladistic analysis of those 25 taxa which have been involved in discussions of biserial graptolite origins. Forty-six characters are discussed briefly, especially those relating to the proximal end. Detailed discussion of the coding of each species is posted on the web. The results support separation of diplograptid and glossograptid clades. Pseudisograptus is the sister taxon of the former, and Isograptus of the latter. Theories that related diplograptid origins to either Phyllograptus, Cardiograptus or Maeandrograptus are rejected. The tree supports a series of steps running from Pseudisograptus through Exigraptus and Undulograptus sinodentatus to U. austrodentatus. This is identical to a stratigraphical sequence of species known from China, and consistent with a previous qualitatively based cladogram. The U mode of proximal development is primitive for the diplograptid biserials. A branch-and-bound analysis of the diplograptid part of the tree supports the monophyly of the Diplograptidae and suggests that Undulograptus may not be a natural taxon. The relationships of glossograptids and diplograptid clades to the broader dichograptoid grade remains to be investigated.