Fossils of 134 taxa attributed to the Rubiaceae are described or mentioned in 115 publications dating from 1850 and from deposits as old as the Cretaceous and Paleocene. Close scrutiny of these records indicates, however, that the oldest and most likely (accepted) representatives of the family are four genera, Emmenopterys Oliv. from the Middle Eocene of Oregon and Washington, U.S.A., Faramea Aubl. from the Late Eocene of Panama, and Guettarda L. (cf. as †Guettardidites; †= fossil taxon) and Canthium Lam. (as †Rubipollis oblatus) from the Late Eocene of Australia, and a probable fifth genus, the alternate-leaved †Paleorubiaceophyllum eocenicum from the Middle Eocene of Tennessee/Kentucky, U.S.A. The record represents three subfamilies (Rubioideae, Ixoroideae, Cinchonoideae) from three widely separated geographic regions, implying an earlier origin in the Late Cretaceous or Paleocene. From the Oligocene, there are six accepted genera, Coprosma J. R. Forst. & G. Forst., Coprosma–Opercularia, Faramea Aubl., Macrosphyra Hook. f. (as †Triporotetradites hoekeni), Mitragyna Korth. (as †Retitricolporites annulatus), and Pinckneya Michx. from Africa (Cameroon), Australia and New Zealand, U.S.A. (Oregon), and Puerto Rico. The period of greatest diversification and radiation was in the Miocene, with 20 accepted genera reported from North America, Central America, South America, Southeast Pacific–Asia, Africa, and Europe. Stages in the evolution of three characters are further suggested by the fossil record. The relatively uncommon occurrence of alternate leaves among modern taxa is typical of advanced genera (e.g., Didymochlamys Hook. f., Sabicea Aubl., Theligonum L.), but this feature may have already developed by the Middle Eocene (†Paleorubiaceophyllum). Polyploidy is suggested in the Middle Eocene P. eocenicum var. lawrensis by epidermal cells 32 µm in diameter or nearly twice the size of the other varieties. Pollen polymorphism, possibly coordinated with or as a prelude to heterostyly, is represented by the diporate pollen of Faramea in the Late Eocene, and triporate and tetraporate forms in the Miocene and Pliocene. Currently, the principal needs are: (1) to clarify pending Paleocene records of †Cinchonidium (Cinchona L.), North Dakota, U.S.A.; Galium L., Greenland; and †Psilatricolpites coprosmoides (Coprosma, Chile); and (2) to examine the several large Cretaceous megafossil floras now under study (e.g., Anfiteatro de Ticó, Argentina; Crato, Brazil; Turonian-age floras, New Jersey and Sweden) for specimens with features that suggest a rubiaceous complex or with an aggregation of features suggesting presence of the family.