There is much disagreement about how many species should be recognised in this genus which is of exceptional difficulty; nothing is known of the reproductive biology so it is not possible to suggest a reason for the poor segregation of the taxa. Once the reason is discovered the course of action to be followed may become clearer. Suggestions that the difficulties are due to frequent hybridisation are unsubstantiated and annotation labels giving precise hybrid identifications are naive in the extreme. Usually 150–200 species are recognised, widely distributed in the tropics and subtropics of both Old and New Worlds but Leenhouts took a radical step and considered there to be only one species A. cobbe (L.) Räusch. and suggested that in each area races could be recognised without formal taxonomic status. Mabberley (The Plant Book, ed. 2: 25 (1997)) accepts that there is one very polymorphic species “which at a local level can be subdivided into c. 175 apparently distinct species”. Capuron has also followed this in his account of the Madagascan Sapindaceae (Mém. Mus. Nat. Hist. Nat., Paris, 19: 53–68 (1969)) using the name A. cobbe followed by the usually accepted species name in brackets to produce an informal system of dealing with the entities. To many botanists, however, Leenhouts’s suggestion is quite unacceptable. Exell & Sousa (in Fl. Moçamb. 51: 3–16 (1973)), Vollesen (Fl. Ethiopia 3: 498–502 (1990)), Beentje (K.T.S.L.: 412–413 (1994)) and Pennington (as yet unpublished account of the South American species) have ignored it. One can understand Leenhouts’s reasons. I found taxa of Allophylus in Sri Lanka and Taiwan which seem virtually indistinguishable from bushland species in East Africa and when shown to someone dealing with South American species were thought to be a Bolivian taxon. Clearly this is not easy to accept. On the other hand, anyone shown the unifoliolate coastal shrub A. pervillei and the trifoliolate upland forest tree A. abyssinicus is unlikely to accept them as the same species and biologically I am sure they are not. I have accepted 18 species in East Africa. Certainly it would be infinitely easier to follow Leenhouts but it does not seem very useful. Undoubtedly the forest species are easier to delimit than the dry country and bushland species. I have resisted treating any of the East African bushland taxa as variants of A. cobbe sensu stricto even if ± identical since this would imply that they were closer to this Sri Lanka species than to the East African forest taxa.Joseph Njung’e (Chemical and pollen study of genus Allophylus (Sapindaceae) M.Sc. Dissertation, Reading University) examined 31 specimens representing only 12 species (the names on the herbarium sheet labels being accepted at face value). Information from the chemical data was insufficient to help with specific delimitation. Pollen grains were found to be homogenous in type and aperture number but the exine sculpture was found to be variable and falls into 7 types varying from poorly developed reticulate to retirugulate thence to well-developed rugulate or rugulate-striate. This is somewhat unexpected and needs expanding using some hundreds of specimens.Several collectors mention scented flowers and their attraction for flies; most of the species could be fly-pollinated.