But in these and in many other cases, Gartner is obliged carefully to count the seeds, in order to show that there is any degree of sterility.

He always compares the maximum number of seeds produced by two species when first crossed, and the maximum produced by their hybrid offspring, with the average number produced by both pure parent-species in a state of nature.

But causes of serious error here intervene: a plant, to be hybridised, must be castrated, and, what is often more important, must be secluded in order to prevent pollen being brought to it by insects from other plants.

Nearly all the plants experimented on by Gartner were potted, and were kept in a chamber in his house.

That these processes are often injurious to the fertility of a plant cannot be doubted; for Gartner gives in his table about a score of cases of plants which he castrated, and artificially fertilised with their own pollen, and (excluding all cases such as the Leguminosae, in which there is an acknowledged difficulty in the manipulation) half of these twenty plants had their fertility in some degree impaired.