Investigation of the assumption of additivity has been hampered by the lack of adequate and appropriate experimental designs and statistical methods. This is particularly true for mixtures of more than two chemicals. This paper describes the threshold additivity model, a flexible experimental design and statistical methodology that is applicable to mixtures of large numbers of chemicals. Advantages of this design and analytic approach are that it allows investigators to focus on particular mixture combination points of interest, decreasing the number of animals required as well as the time and cost of performing the experiments. This approach is illustrated by two samples applicable to drinking water disinfection byproducts formed during either chlorination of water or ozonation of water followed by post-treatment with either chlorine or chloramine. In the first example, the additivity assumption is examined, with hepatotoxicity in female CD-1 mice as the endpoint, for a mixture of the four trihalomethanes, chloroform (CHC13), bromoform (CHBr3), bromodichloromethane and chlorodibromomethane, formed during disinfection of water. The mixture tested was based on the average seasonal proportions of these four chemicals at 35 water treatment facilities. For the particular mixture and dosage tested, the experimental sample mean was within the 95% prediction interval from the threshold additivity model, providing evidence that dose additivity is a reasonable assumption for risk assessment. In the second example, the additivity assumption for developmental toxicity is examined for binary mixtures of CHC13 and CHBr3 in medaka fish. With the statistical power afforded by the present experiment, antagonism was detected at the highest mixture dose tested (25 ppm CHC13:25 ppm CHBr3) and departure from additivity was not detected at the three lower-dose mixture groups. In summary, a threshold additivity model for efficiently detecting departure from additivity is described and illustrated with two biological examples.