Colloidal $^{32}P$ chromic phosphate is used to prevent hepatic metastasis from colorectal cancer. It is speculated that the intravenous injection of colloidal $^{32}P$ chromic phosphate can cause genotoxicity. To evaluate the genotoxicity of intravenously injected colloidal $^{32}P$ chromic phosphate, authors performed a micronuclei test in mice bone marrow. Mice (ICR strain, $25sim30g$) were divided to 4 groups: control, group 1 (19.166 KBq/g, usual therapeutic dose in human), group 2 (191.66 KBq/g), and group 3 (1916.6 KBq/g). Five mice of each group were sacrificed at days 1, 2, 3, 5, 7 and 14. Bone marrow were smeared and stained with Wright-Giemsa method. One thousand polychromatic erythrocytes (PCE) and normochromatic erythrocytes (NCE) were counted under the light microscope, and the number of micronucleated PCEs and NCEs were recorded. The frequency of micronuclei in PCE and NCE in the control group was $0.3{pm}0.06%;and;0.45{pm}0.10%$, respectively. At group 1, frequency of micronuclei is not different from the control. However, frequencies of micronuclei in PCE at groups 2 and 3 were significantly increased from day 1 and persisted to day 14. The frequency of micronuclei in NCE was increased only at group 3. In conclusion, the frequency of micronuclei increases as the dose of colloidal $^{32}P$chromic phosphate increases, while micronuclei was not induced at the usual therapeutic dose. And the frequency of micronuclei persistently elevated for 14 days in the cases of higer doses.