Résumé : The in vitro syncytium induction capacity of human immunodeficiency virus type 1 (HIV1) isolates is an important marker in the progression of the disease. Two methods have been widely used to determine the biological phenotype of HIV1. These two methods, the direct MT-2 assay and the supernatant assay, were compared for the detection of syncytium-inducing (SI) variants on 275 blood samples obtained from 87 HIV infected patients during a 13 month follow-up period. A SI virus was detected in 152 blood samples. In 44 blood samples, the HIV isolate was found to be SI by only one method, but was SI by both methods in another blood sample of the follow up. Among SI carriers discordant results between the methods were more frequent when the patient was on antiretroviral therapy, and a transient reversion to a non syncytium-inducing (NSI) strain confirmed by both assays was sometimes observed. The supernatant assay has a 93% sensitivity and the direct MT-2 assay has a 78% sensitivity for detection of the SI phenotype. The supernatant assay is as rapid as and less tedious than the MT-2 assay. Antiretroviral therapy could have some effects in decreasing or even suppressing the SI part of the virus population of patients with SI phenotype.