Isolates of Fusarium oxysporum from lily were screened for pathogenicity, vegetative compatibility and DNA restriction fragment length polymorphisms, and compared to reference isolates of F. oxysporum f.sp. gladioli and F. oxysporum f.sp. tulipae to justify the distinction of F. oxysporum f.sp. lilii. Twenty-four isolates from different locations in The Netherlands (18 isolates), Italy (4 isolates), Poland and the United States (1 isolate each) shared unique RFLP patterns with probes D4 and pFOM7, while hybridization did not occur with a third probe (F9). Except for a self-incompatible isolate, these 24 isolates all belonged to a single vegetative compatibility group (VCG 0190). Isolates belonging to VCG 0190 were highly pathogenic to lily, but not to gladiolus or tulip, except for a single nonpathogenic isolate. Six saprophytic isolates of F. oxysporum from lily were nonpathogenic or only slightly aggressive to lily, gladiolus and tulip, belonged to unique VCGs and had distinct RFLP patterns. Three pathogenic isolates previously considered to belong to F. oxysporum f.sp. lilii were identified as F. proliferatum var. minus; all three belonged to the same VCG and shared unique RFLP patterns. These three isolates were moderately pathogenic to lily and nonpathogenic to gladiolus and tulip. The reference isolates of F. oxysporum f.sp. tulipae were pathogenic to tulip, but not to lily and gladiolus; they shared a distinct RFLP pattern, different from those encountered among pathogenic and saprophytic isolates from lily, and formed a separate new VCG (VCG 0230). Reference isolates of F. oxysporum f.sp. gladioli belonging to VCG 0340 proved pathogenic to both gladiolus and lily, but not to tulip. These isolates, as well as isolates belonging to VCGs 0341, 0342 and 0343 of F. oxysporum f.sp. gladioli, had RFLP patterns different from those encountered among the isolates from lily or tulip. These findings identify F. oxysporum f.sp. lilii as a single clonal lineage, distinct from F. oxysporum f.sp. gladioli and f.sp. tulipae.