Three New South American Species of Randia (Rubiaceae, Gardenieae) Claes G. R. GustafssonBotanical Institute, Systematic Botany, Box 461, 405 30 Goiteborg, Sweden. claes.gustafs[email protected] Three new species of South AmericanRandia (Rubiaceae, Gardenieae) are described andillustrated: Randia wigginsii Standley ex Gustafs-son from montane tropical forests in Ecuador andnorthern Peru is mainly recognized by its tomentoseto velutinous corolla tube and the lanate-velutinouspedicels and fruits. The possible features of beingunarmed and having single male flowers are dis-cussed. Randia pubistyla Gustafsson from lowlandtropical dry forests in western Ecuador and Colom-bia is recognized by its puberulous style, the to-mentose to puberulous fruits that usually arecrowned by a calyx with reflexed lobes, and by thesmall thorns. Randia longifolia Gustafsson fromlowland tropical humid forests in western Ecuadoris mainly recognized by its relatively large, gla-brous corolla, the linear calyx lobes, and by thereticulate finer vasculature on the abaxial surfaceof the leaves. The affinities for each species arediscussed. Randia is a neotropical genus of approximately90 species ranging from ca. 300N to 30�S. In SouthAmerica and the Caribbean there are approximate-ly 45 species. The genus is represented by shrubs,trees, and lianas in deciduous and evergreen veg-etation from sea level to about 3300 m elevation.Randia can be recognized from other members ofthe Gardenieae by the following combination ofcharacters: dioecious (female flowers with nonfunc-tional stamens, male flowers with a nonfunctionalstigma and rudimentary ovary), pollen in permanenttetrads, a unilocular ovary with two parietal pla-centas, fruits with many discoid seeds embeddedin a sweet pulp that turns dark when dry, thorns,and conspicuous short-shoots with clustered stip-ules and leaves. There are, however, exceptions.Monoecious and hermaphroditic species have beenreported (Lorence & Dwyer, 1987; Burger & Taylor,1993), pollen in dyads or monads have been re-ported (Burger & Taylor, 1993), some species areunarmed (Burger & Taylor, 1993; pers. obs.), andsometimes the short-shoots are less conspicuous(Burger & Taylor, 1993; pers. obs.). Within the ge-nus there is also variation in the structure of theinflorescence. The inflorescences are usually ter-minal but sometimes axillary or cauliflorous (Burger& Taylor, 1993). Female flowers are usually solitarybut sometimes in fascicles of 2 to 8 flowers (Burger& Taylor, 1993). The male flowers are usually infascicles with a few to several flowers but some-times form cymes with many flowers (Lorence &Nee, 1987; Lorence & Dwyer, 1987) and are some-times solitary. This large variation makes it difficultto understand which genus has the closest affinitiesto Randia. Robbrecht and Puff (1986) discussed that allgenera in Gardenieae with pollen shed in perma-nent tetrads may represent a natural group. Thisgained support in a phylogenetic analysis by Pers-son (1996) but was contradicted by Andreasen andBremer (1996, in press) and Persson (in press), whodiscussed that pollen in tetrads may have arisenseveral times in Gardenieae. In Persson's (1996)analysis Casasia gained support as sister to Ran-dia, but the later mentioned analyses did not pointout any strongly supported sister to Randia. Otherneotropical genera that in these analyses (Andreas-en & Bremer, 1996, in press; Persson, in press)grouped together with Randia and Casasia as pos-sible relatives were Rosenbergiodendron, Sphinc-tanthus, and Tocoyena. Sphinctanthus and Tocoyenawere not included in the analyses by Andreasenand Bremer (1996, in press). In this group the ge-nus with the morphologically nearest affinities toRandia is Casasia by being dioecious and havingpollen in tetrads. Lorence (1986) and Lorence andDwyer (1987) discussed the morphological featuresin Casasia and Randia and also expressed doubt(Lorence & Dwyer, 1987) whether Casasia can bemaintained as a valid genus. More phylogeneticstudies in Gardenieae need to be done in order tofind the monophyletic group to which Randia be-longs. The taxonomical work in Randia is complicatedby the intraspecific variation in size and shape ofleaves, calyx lobes, degree of pubescence, persis-tence of stipules and calyx, and number of thorns. NovoN 10: 201-208. 2000.