In the third collaborative study organized by the Collaborative Study Group for the Micronucleus Test (CSGMT), a task group belonging to the Mammalian Mutagenesis Study subgroup of the Environmental Mutagen Society of Japan (JEMS.MMS), intraperitoneal (i.p.) injection and oral (p.o.) gavage were compared as routes of administration of test chemicals. Two mouse strains, MS/Ae and CD-1, and 17 chemicals with various modes of action were used. The chemicals were 1-beta-D-arabinofuranosylcytosine, 6-mercaptopurine monohydrate, benzo[a]pyrene, 7,12-dimethylbenz[a]anthracene, 2-acetylaminofluorene, phenacetin, cyclophosphamide, ethyl methanesulfonate, N-ethyl-N-nitrosourea, methyl methanesulfonate, mitomycin C, colchicine, vincristine sulfate, potassium bromate, potassium chromate(VI), benzene, and procarbazine hydrochloride. On the basis of the findings of an acute toxicity test and a pilot experiment for dose and sampling time, a full-scale micronucleus test was performed on each chemical. Almost all the chemicals showed a positive response in micronucleus induction by both routes of administration in both mouse strains. Contradictory outcomes were obtained between the i.p. and p.o. routes on potassium chromate in both strains (i.p.: positive, p.o.: negative). In the CD-1 mice, benzene potently induced micronuclei when administered p.o., but gave only a marginal response when administered i.p. Generally, the chemicals induced micronuclei at lower dose levels (mg/kg) when administered i.p. This tendency, however, was decreased or even reversed when the dose was expressed as a percentage of the LD50. Although the i.p. route, an artificial exposure route, is useful to detect the inducibility of micronuclei of a test chemical per se at a small dose, the p.o. route seemed sensitive and valuable enough to evaluate the test chemicals. When the dose levels of chemicals are adjusted on the basis of the LD50, both i.p. injection and p.o. gavage are acceptable as routes of administration in the micronucleus test.