By molecular analysis of a high number of gammarids, including 29 out-group genera, we could assure the monophyly of Gammaridae. To avoid the paraphyly of the family, we propose the omission of Pontogammaridae, Typhlogammaridae, and all Baikalian families. Similarly, the genera Fontogammarus, Sinogammarus, Lagunogammarus, Pephredo, Neogammarus, and Laurogammarus may be cancelled. But, tens of Baikal genera, nested within Gammarus, are so diverse that they must be retained, although rendering Gammarus paraphyletic. Besides we propose the polyphyletic Echinogammarus–Chaetogammarus group to be divided into monophyletic genera Echinogammarus s. str., Homoeogammarus, Parhomoeogammarus, Marinogammarus, Relictogammarus gen. nov., Chaetogammarus, and Trichogammarus gen. nov. These solutions made it possible to complete the first analysis of the family evolution in light of its phylogeny. Perimarine clades are mainly basally split clades, whereas in some ancient lakes extremely rich endemic faunas had developed polyphyletically. The troglobiotic Typhlogammarus group from Dinarides and Caucasus formed a monophylum, whereas the troglobiotic assemblage of Gammarus species is highly polyphyletic. Reduction of the uropod III endopodite, which classically distinguishes between the genera Gammarus and Echinogammarus, appeared to be highly polyphyletic. Protective dorsal pleonal projections occur scattered across the family and beyond, whereas lateral projections were limited to species of ancient lakes, so both structures were polyphyletic. The evolutionary history of Gammaridae was investigated with ten different calibration schemes, which produced incompatible results; however, the most probable scenario is a late rise of the family, which can only explain the absence of Gammaridae species around the Indo-Pacific. © 2015 The Linnean Society of London