In scientific selection-experiments such crosses are of course avoided, and the process of purification is unnecessary, even as in the _Chrysanthemum_ culture. The first generation succeeding the original plant with disk-rays was in this respect wholly uniform and true to the new type. In practice the work does not start from such slight indications, and is done with no other purpose in view than to produce double flowers in species in which they did not already exist. Therefore it is of the highest importance to know the methods used and the chances of success. Unfortunately the evidence is very scanty on both points. Lindley and other writers, on horticultural theory and practice assert that a large amount of nourishment tends to produce double flowers, while a culture under normal conditions, [507] even if the plants are very strong and healthy, has no such effect. But even here it remains doubtful whether it applies to the period before or after the internal mutation. On the other hand success is not at all to be relied upon, nor is the work to be regarded as easy. The instances of double flowers said to be obtainable at will, are too rare in comparison with the number of cases, where the first indication of them was found accidentally. Leaving all these doubtful points, which will have to be cleared up by further scientific investigation, the high degree of variability requires further discussion. It may be considered from three different points of view according to the limit of the deviation from the average, to the dependency on external conditions and to periodicity. It seems best to take up the last two points first. On a visit to a nursery at Erfurt I once inspected an experiment with a new double variety of the common blue-bottle or blue corn-flower. The plants were dependent on the weather to a high degree. Bad weather increased the number of poorly filled flower-heads, while warm and sunny days were productive of beautiful double flowers. The heads that are borne by strong branches have a greater tendency to become double than those of the weaker ones, [508] and towards the autumn, when all those of the first group are faded away, and only a weak though large section of the heads is still flowering, the whole aspect of the variety gradually retrogrades. The same law of dependency and periodicity is prevalent everywhere. In my own cultures of the improved field-marigold I have observed it frequently. The number of the ray-florets may be considered as a direct response to nourishment, both when this is determined by external circumstances, and when it depends on the particular strength of the branch, which bears the head in question. It is a case exactly similar to that of the supernumerary carpels of the pistilloid poppy, and the deductions arrived at with that variety may be applied directly to double flowers.