SEVERAL syngens or breeding groups of Paramecium aurelia contain endosymbionts which may result in the development of the killer trait(1). The symbionts are of different morphological types and it has been suggested that these are non-randomly distributed between syngens(2). Experiments involving crosses between paramecia with and without symbionts showed that each endosymbiont required a specific nuclear gene for its maintenance. A single gene controlled the presence or absence of endosymbionts as shown by segregation in the F(2) generation. The pattern of loss of symbionts in some of those F(2) clones varies, generally extending over a number of asexual fissions which for different endosymbionts can be from two to over sixty. Infection experiments also demonstrate the requirement for a specific gene. Infection of endosymbionts from homogenates or purified preparations occurs via the medium but only into particular cells possessing a specific gene(1,3). Finally, some endosymbionts growing in vitro retain their infectivity into certain paramecia (author's laboratory).