1. It is possible to split a strain of mice into certain substrains in which the tumor incidence is in some way linked to the color of the mice. Thus we could split off from the English strain, which as a whole and in various substrains with mixed colors (English A and Sable) has a high tumor rate, substrains with light tan color and pink eyes (IOI and Tan) which have a high tumor rate like the large majority of the English mice, and two other apparently recessive strains breeding true, which have a very low tumor incidence (Silver and Silver Fawn). Therefore, certain combinations of factors which determine certain colors of mice determine at the same time the tumor incidence of these strains or substrains. In the majority of cases isolated families bred through several generations separately from the majority of the other substrains give approximately the same tumor rates as the others; in some cases, however, it may perhaps be possible to separate from the main strain a family with a different rate. 2. The tumor incidence and the tumor age found in the earlier periods of our work are approximately the same as in the more recent period. On the whole, the results obtained in successive generations of the same strain also agree well with each other; the results are fairly constant; the deviations which occur are in most cases due to the small number of animals observed in the certain generations. Discarding all the mice dying in the first or in the first and second periods of life usually does not alter essentially the tumor ratio of a certain strain. 3. A certain relationship exists between tumor frequency and tumor age. On the whole, the more frequent the tumors, the earlier they appear in the various strains. This parallelism between tumor frequency and tumor age is, however, not complete. The tumor age seems to be as characteristic for a certain strain as the tumor rate. Certain substrains which differ in tumor frequency may show approximately the same tumor age. Strains with similar tumor frequency may show a different tumor age. We may therefore conclude that in all probability tumor rate and tumor age represent distinct unit factors, which are frequently, but not in all cases, linked in some manner to each other. 4. The age at which the maximum of tumors appears varies in different strains. The maximum may fall into the second or third period of life. On the whole, the maximum is reached at an earlier period of life in those strains which have a high tumor rate. But here also peculiarities exist in different strains.