Sixty-four isolates of Bean yellow mosaic virus (BYMV) from cultivated and naturalized gladioli were divided into two pathogenic groups, necrotic spot (NS) and chlorotic spot (CS) groups on Chenopodium quinoa. NS-type isolates (S-22N and E-24N), CS-type isolates (S-22C and E-92C), and broad bean isolates (Sb-50C and Sb-12C) differed in their pathogenicity on Antirrhinum majus, Nicotiana benthamiana, Phaseolus vulgaris, Spinacia oleracea and Vigna unguiculata. The four gladiolus isolates were different from BYMV-B, -P, -O and CIYVV-N in their pathogenicity on these plants, while the two broad bean isolates were similar to BYMV-B, originally from broad bean. The nucleotide (nt) sequences of the 3'-terminal region of the BYMV RNA genome of the two NS-type isolates, the two CS-type isolates, the two broad bean isolates and BYMV-B, -P and -O were determined. In a phylogenetic tree based on the CP amino acid (aa) sequences, the two NS-type isolates clustered together (identity 98.4% and 98.2% at the nt and aa level, respectively). The two CS-type isolates clustered with BYMV-O (93.2 to 99.3% nt identity and 95.6 to 98.5% aa identity). The two broad bean isolates clustered with BYMV-B (99.0 to 99.5% nt identity and 98.9 to 99.6% aa identity). BYMV-P clustered with BYMV-CS (identity 97.7% and 99.3% at the nt and aa level, respectively). The obtained sequences were compared with those of the 3'-terminal regions of seven published BYMV isolates. In a phylogenetic tree based on deduced aa sequences, BYMV isolates were divided into four clusters.