Nothing shows more clearly the extraordinary complexity of the reproductive system of this plant, than the necessity of making eighteen distinct unions in order to ascertain the relative fertilising power of the three forms. Thus the long-styled form has to be fertilised with pollen from its own two kinds of anthers, from the two in the mid-styled, and from the two in the short-styled form. The same process has to be repeated with the mid-styled and short-styled forms. It might have been thought sufficient to have tried on each stigma the green pollen, for instance, from either the mid- or short-styled longest stamens, and not from both; but the result proves that this would have been insufficient, and that it was necessary to try all six kinds of pollen on each stigma. As in fertilising flowers there will always be some failures, it would have been advisable to have repeated each of the eighteen unions a score of times; but the labour would have been too great; as it was, I made 223 unions, i.e. on an average I fertilised above a dozen flowers in the eighteen different methods. Each flower was castrated; the adjoining buds had to be removed, so that the flowers might be safely marked with thread, wool, etc.; and after each fertilisation the stigma was examined with a lens to see that there was sufficient pollen on it. Plants of all three forms were protected during two years by large nets on a framework; two plants were used during one or both years, in order to avoid any individual peculiarity in a particular plant. As soon as the flowers had withered, the nets were removed; and in the autumn the capsules were daily inspected and gathered, the ripe seeds being counted under the microscope. I have given these details that confidence may be placed in the following tables, and as some excuse for two blunders which, I believe, were made. These blunders are referred to, with their probable cause, in two footnotes to the tables. The erroneous numbers, however, are entered in the tables, that it may not be supposed that I have in any one instance tampered with the results.