[553] The same holds good for _albida_, which appeared in 15 specimens, or in 0.1%, of the whole culture. By careful cultivation these plants proved not to be sickly, but to belong to a new, though weak type. It was evident that I had already seen them in former years, but having failed to recognize them had allowed them to be destroyed at an early age, not knowing how to protect them against adverse circumstances. Even this time I did not succeed in getting them strong enough to keep through the winter. Besides these, two new types were observed, completing the range of all that have since been recorded to regularly occur in this family. They were _scintillans_ and _gigas_. The first was obtained in the way just described. The other hardly escaped being destroyed, not having showed itself early enough, and being left in the bed after the end of the selection. But as it was necessary to keep some rosettes through the winter in order to have biennial flowering plants to furnish seeds, I selected in August about 30 of the most vigorous plants, planted them on another bed and gave them sufficient room for their stems and branches in the following summer. Most of them sent up robust shoots, but no difference was noted till the first flowers opened. One plant had a much larger crown of bright blossoms than any of the others. [554] As soon as these flowers faded away, and the young fruits grew out, it became clear that a new type was showing itself. On that indication I removed all the already fertilized flowers and young fruits, and protected the buds from the visits of insects. Thus the isolated flowers were fertilized with their own pollen only, and I could rely upon the purity of the seed saved. This lot of seeds was sown in the spring of 1897 and yielded a uniform crop of nearly 300 young _gigas_ plants. Having found how much depends upon the treatment, I could gradually decrease the size of my cultures. Evidently the chance of discovering new types would be lessened thereby, but the question as to the repeated production of the same new forms could more easily and more clearly be answered in this way. In the following year (1896) I sowed half as many seeds as formerly, and the result proved quite the same. With the exception of _gigas_ all the described forms sprang anew from the purely fertilized ancestry of normal _lamarckiana_s. It was now the fifth generation of my pedigree, and thus I was absolutely sure that the descendants of the mutants of this year had been pure and without deviation for at least four successive generations. Owing partly to improved methods of selection, [555] partly no doubt to chance, even more mutants were found this year than in the former. Out of some 8,000 seedlings I counted 377 deviating ones, or nearly 5%, which is a high proportion. Most of them were _oblonga_ and _lata_, the same types that had constituted the majority in the former year.