- Most cultivars of Cyclamen persicum are autotetraploids. This implies that true breeding cultivars cannot be obtained by ordinary pedigree selection. Theoretically it could be done by repeated pedigree selection in groups of F4-1ines, descended from one F3-1ine and from one F2-plant, or, expressed in other words, by testing part of the F~-plants not only in F a, but also in F 4. It is demonstrated that such a procedure is practically impossible by the large numbers of necessary plants. Therefore, the building up of a complete collection of diploid cultivars is necessary.
.2. The possibilities to reach this aim and which have been discussed in historical order, are mainly: (a) crosses of the existing diploids with wild C. persicum and crested white; (b) crosses between the existing diploids with the red 'Firefly'; (c) the study of an international collection of cultivars, which has resulted in an unexpectedly large number of diploids. .3. In principle the problem of building up a complete collection of diploid cultivars has been solved. New diploids have been developed and the starting-material for numerous others has become available. The methods have been quite elementary: collection of existing material, cytological investigation, selection of diploids, increasing the variability by crossing followed by selection.
.4. In the breeding of other plants, perhaps primarily of ornamentals, an unprejudiced attitude towards the value of the existing material is recommendable.