Infection is certainly the most dramatic complication of reconstructive vascular surgery, even after long periods following the operation. A hematogenous mechanism is often implicated in late infections ; colonization of an arterial graft appears to depend on the material used and cicatrization processes. Animal studies have the disadvantage of requiring repeated operations. We have developed an in vitro infusion system in order to reproduce the bacteremic phenomenon, and to study the conditions necessary for colonization by a staphylococcus of a textile prosthesis (Dacron) in velvet, placed in the "Ex vivo" circulation of a dog for 2 hours. The prosthetic material is colonized very much more easily than a fragment of aorta used a control. The bacterial deposit (y) is a linear function of the inoculum (x) and, by approximation, y = 10(4) x (p 0.01). Colonization is independent of the infusion rate when this varies between 5 and 20 cm/sec. The reproducibility of the results, their quantification, and the ease with which these tests can be employed suggest that such models could be used to assess resistance to bacterial colonization of various materials used for arterial substitution.