Herbs or shrubs, rarely trees or vines. Leaves simple or compound, usually alternate or opposite, usually exstipulate. Flowers usually in cymes, panicles, or racemes, rarely solitary, usually bisexual, rarely unisexual, hypogynous or ± epigynous, rarely perigynous, usually biperianthial, rarely monochlamydeous, actinomorphic, rarely zygomorphic, 4- or 5(-10)-merous. Sepals sometimes petal-like. Petals usually free, sometimes absent. Stamens (4 or)5-10 or many; filaments free; anthers 2-loculed; staminodes often present. Carpels 2, rarely 3-5(-10), usually ± connate; ovary superior or semi-inferior to inferior, 2- or 3-5(-10)-loculed with axile placentation, or 1-loculed with parietal placentation, rarely with apical placentation; ovules usually many, 2- to many seriate, crassinucellate or tenuinucellate, sometimes with transitional forms; integument 1- or 2-seriate; styles free or ± connate. Fruit a capsule or berry, rarely a follicle or drupe. Seeds albuminous, rarely not so; albumen of cellular type, rarely of nuclear type; embryo small.

During the past several years, cladistic analyses of morphological, chemical, and DNA data have made it clear that the recognition of the Saxifragaceae sensu lato (Engler, Nat. Pflanzenfam. 18a: 74-226. 1928) is untenable. Among the angiosperm families, Saxifragaceae sensu lato may in fact represent the most extreme example of a polyphyletic assemblage. For example, recent analyses of DNA sequence data indicate that these taxa represent at least ten separate evolutionary lines, many of which are only distantly related to one another (Morgan & Soltis, Ann. Missouri Bot. Gard. 80: 631-660. 1993; Soltis & Soltis, Amer. J. Bot. 84: 504-522. 1997). Furthermore, very large molecular phylogenetic analyses of hundreds of angiosperms indicate that these separate lineages are distributed among four of the six traditionally recognized subclasses of dicotyledons (Savolainen et al., Syst. Biol. 49: 306-362. 2000; Soltis et al., Nature 402: 402-404. 1999). These recent studies have also greatly clarified how this phylogenetically diverse assemblage should be divided into families and treated taxonomically (see The Angiosperm Phylogeny Group (APG), Ann. Missouri Bot. Gard. 85: 531-553. 1998). Recent studies of DNA sequence data have clarified both the circumscription and affinities of a narrowly defined Saxifragaceae (Saxifragaceae sensu stricto) and Hydrangeaceae (Soltis et al., Amer. J. Bot. 82: 504-514. 1995; Savolainen et al., loc. cit.; Soltis et al., loc. cit. 1999). Saxifragaceae sensu stricto should consist only of Saxifragoideae, a group of about 30 herbaceous genera. Members of Saxifragaceae sensu stricto from the Chinese flora include Astilbe, Astilboides, Bergenia, Chrysosplenium, Mitella, Mukdenia, Oresitrophe, Rodgersia, Saxifraga, Tanakaea, Tiarella, and the recently described Saniculiphyllum. Close relatives of Saxifragaceae sensu stricto include Itea, Penthorum, and Ribes. These genera, the sole members of Iteoideae, Penthoroideae, and Ribesioideae, respectively, are also best treated in separate families: Iteaceae, Penthoraceae, and Grossulariaceae (see APG, loc. cit.). These taxa, as well as several others, such as Crassulaceae, are basal to a large assemblage of taxa, most of which were traditionally placed in Rosidae. Sequence data also indicate that Parnassia (the sole member of the Parnassioideae) is a more derived member of the rosid alliance, most closely related to Brexia and Lepuropetalon (also part of Saxifragaceae sensu lato) and Celastraceae. Parnassia and Lepuropetalon should be placed in Parnassiaceae with Brexia part of an expanded Celastraceae (APG, loc. cit.).

Both morphological and molecular data indicate that Hydrangeoideae and Escallonioideae are, in contrast, allied with taxa traditionally placed in Asteridae. Hydrangeoideae are a well-defined, monophyletic lineage that should be treated as Hydrangeaceae. In China they include Cardiandra, Decumaria, Deinanthe, Deutzia, Dichroa, Hydrangea, Kirengeshoma, Philadelphus, Pileostegia, Platycrater, and Schizophragma, and are closely allied with families such as Cornaceae, Loasaceae, and Nyssaceae. Escallonioideae appear to be polyphyletic, and this group of approximately 14 genera is in need of thorough study. Members of this subfamily are allied with several different lineages of higher asterids. Polyosma, the only member of Escallonioideae in China, appears closely allied with Caprifoliaceae (Xiang & Soltis in Boufford & Ohba, Sino-Japanese Flora: its Characteristics and Diversification, 1998).

Nevertheless, in the present account, the Saxifragaceae are retained in the sense of FRPS, using the same sequence of genera and indicating the subfamilies in the key below, in order to facilitate comparison with that flora.

The genus Changiodendron R. H. Miao (Acta Sci. Nat. Univ. Sunyatseni 34: 65. 1995) and its single species, C. guangxiense R. H. Miao (loc. cit.: 66), was described from Guangxi (Napo Xian) and stated to belong to the Iteaceae (i.e., Iteoideae). However, Peng (Acta Bot. Yunnan. 18: 299-300. 1996) demonstrated that C. guangxiense is a synonym of Sabia parviflora Wallich (Sabiaceae).

Pan Jin-tang. 1992. Saxifragaceae (1) [Penthoroideae, Saxifragoideae]. In: Pan Jin-tang, ed., Fl. Reipubl. Popularis Sin. 34(2): 1-309; Hwang Shu-mei, Wei Chao-fen, Lu Ling-ti, Ku Tsue-chih & Jin Shu-ying. 1995. Saxifragaceae (2) [Parnassioideae, Hydrangeoideae, Escallonioideae, Iteoideae, Ribesioideae]. In: Lu Ling-ti & Hwang Shu-mei, eds., Fl. Reipubl. Popularis Sin. 35(1): 1-406.

About 80 genera and 1200 species: worldwide; 29 genera (two endemic), and 545 species (354 endemic, seven introduced) in China.

(Authors: Pan Jintang , Gu Cuizhi (Ku Tsue-chih), Huang Shumei (Hwang Shu-mei), Wei Zhaofen (Wei Chao-fen), Jin Shuying , Lu Lingdi (Lu Ling-ti); Shinobu Akiyama, Crinan Alexander, Bruce Bartholomew, James Cullen, Richard J. Gornall, Ulla-Maj Hultgård, Hideaki Ohba, Douglas E. Soltis)