In many experiments of drugs and poisons on animals differing in size but otherwise similar, the same dose is given to each individual, as if its action were independent of body size, e.g., toxicity test with house flies, or doses are proportioned directly to the body weight, e.g., acute oral toxicity test with mouse. Instead of an arbitrary correction such as the latter, a size factor for equalizing individual differences in body weight can be determined experimentally for each drug by multiple regression. When the response y is a measurement, size factor w^h can be estimated from the regression of y upon the dose per animal (x₁=logd) and its body weight (x₂=logw) by transforming the basic regression equation from Y=a'+b₁x₁+b₂x₂ or Y=a'+b₁(log d+b₂/b₁logw) to Y=a'+b₁log(dw^h) where h=b₂/b₁; b₂ is almost always negative.<br>This approach was illustrated by BLISS (1936, 1961) in details using the experimental data on the rate of toxic action of sodium arsenate in larvae of silkworm moth by CAMPBELL (1926). The writers applied his method of statistical analysis to the experimental data obtained from the rate of toxic action of pentachlorophenol in the “Dojo” fish. Each fish was dipped in aqueous solution of prescribed concentration and the lethal time was measured in minutes. The variates for analysis are the log dose per fish x₁=(log p.p.m.), its body weight x₂=log grams, and the rate of toxic action y=log (100/minutes survival).<br>One of the 100 fishes at a concentration of 25p.p.m. has been omitted after statistical test as outlier, subject to physiological processes present in the other 99 fishes. As computed in the paper, the multiple regression equation is Y=-0.05834+0.58337x₁-0.16823x₂. From the ratio of the two partial regression coefficients, h=b₂/b₁=-0.28837, it is concluded that k times larger fishes required relatively k^0.288 times less PCP than that indicated by the ratio of their body weights to kill them in the same time as the smaller fishes. Here, h=-1 means that the doses should be proportioned directly to the body weight and h=0 means doses are independent of body weight. For a graphic test of linearity, the log dose for each individual was computed as z=x₁-0.28837x₂ in Table 1, and y plotted on the ordinate against z on the abscissa in Fig. 1. It has been fitted with the line Y=0.4936+0.5859(z-0.9452), the slope differing from the original b1 by rounding errors. The fidulcial limits of h and z were calculated and the test method on a suspected outlier was also explained.