THE proposal by Dr. Miller1 for quantitatively defined terms for various degrees of probability seems to me to be excellent. I would like to suggest, however, that the word “conclusive” for the case where P <0.001 is undesirable. The term “conclusive” suggests that no further work need be done and that no alternative interpretation is possible. In fact, an industrious person who does a few thousand experiments on the effects of a similar number of different chemicals on a particular biological process will expect to find a few of these apparently giving effects at the P <0.001 level owing to random chance even if none of the chemicals has any real effect at all.