Dose selection in testing chemicals for possible carcinogenicity in rodents continues to be an area of scientific debate. In this paper the definition of "maximum tolerated dose" (MTD) is considered, and the advantages and disadvantages of using MTDs are given. There is no universally accepted definition of an MTD, and as a result, objections to utilizing high doses in carcinogenicity testing may reflect differing definitions of an MTD rather than basic disagreements in dose selection philosophy. Data from 52 National Toxicology Program (NTP) carcinogenicity studies indicate that while dose selection has caused difficulties in certain studies using the gavage route of chemical administration, there is little evidence that this has been a problem in NTP studies using the dietary (feed) route of exposure. These data also indicate that more than two-thirds of the carcinogenic effects detected in feeding studies would have been missed had the high dose been reduced from the estimated MTD to 1/2 MTD. The inherent insensitivity of laboratory animal studies for detecting weak-to-moderate carcinogenic responses also argues against reducing the highest dose level. The addition of a third, lower-dosed group provides for a margin of safety against the possibility of over-estimating the MTD. Primary emphasis should be given to improving procedures for estimating the MTD, particularly for gavage studies. Efforts should also be increased to obtain pharmacokinetic and metabolism data for the test chemical that might be factored into the dose selection and study evaluation processes.