A brief history is given on the circumscription and subdivision of Lecythidaceae and the position of Barringtonia. All aspects, also of auxiliary disciplines point to affinity with Myrtaceae but warrant separate family status. The conviction is that only one family is concerned, not two, or even three, as Knuth proposed. The subdivision of Lecythidaceae appears to meet with great difficulty. It does not seem to be possible to divide the family in tribes which would be restricted to the New and Old Worlds respectively, affinities running transoceanic. In the present revision of the genus Barringtonia 39 species are distinguished, 3 of which have one or a few new subspecies, varieties, or forms; three species are newly described. A survey is given of the morphological characters, their variation, and systematical value. Great attention is given to the seed structure and blastogeny which could be studied in two species. Various dubious or erroneous former statements on the peculiar embryo could be straightened out. The mature embryo is almost unique among Angiosperms because it is a solid body, provided with a spiral of small scales, but without cotyledons as usually understood. After germination this spiral is continued, first with cataphylls and finally with normal leaves. A similar structure is possibly found in other genera of Lecythidaceae; it appears highly desirable that embryogeny of all genera be studied, to ascertain its systematical importance. Two sections based on the calyx structure can be distinguished. This coincides with former generic distinction. Barringtonia sens. str. has initially a closed calyx which later disrupts into pseudolobes or opens circumscissile and Stravadium has a calyx which is open from the beginning. Within these sections eleven smaller groups of mutually allied species are present but they can only vaguely be defined and do not represent taxa. All species were examined palynologically and appeared to belong to eight pollen types, each comprising one or more species. These types can be arranged into two main types which do not coincide with the two taxonomical sections. On the whole it appeared that at the species level the groups are satisfactorily in accord with the pollen types.