Our three illustrative examples are good and permanent races, producing their peculiar qualities [372] regularly and abundantly. In this respect they are however very variable and dependent on external circumstances. Such a regularity is not met with in other instances. Often pedigree-experiments lead to poor races, betraying their tendency to deviate only from time to time and in rare cases. Such instances constitute what we have called in a former lecture, "half races," and their occurrence indicates that the casual observation of an anomaly is not in itself adequate to give an opinion as to the chance of repetition in sowing experiments. A large number of species seem to belong to this case, and their names may be found in the above mentioned work by Masters and elsewhere. But no effort has yet been made to separate thoroughly the pistilloid half-races from the corresponding ever-sporting varieties. Some plants are recorded as being more liable to this peculiarity than others. Stamens are sometimes replaced by open carpels with naked ovules arising from their edges and even from their whole inner surfaces. This may be seen in distinct strains of the cultivated bulbous Begonia, and more rarely in primroses. Here the apex of the carpellary leaf is sometimes drawn out into a long style, terminated by a flattened spatulate stigma. The pistillody of the stamens is frequently [373] combined with another deviation in the poppies. This is the growing together of some of the altered stamens so as to constitute smaller or larger connate groups. Often two are united, sometimes three, four or more. Flowers with numerous altered stamens are seldom wholly free from this most undesirable secondary anomaly. I call it undesirable with respect to experiments on the variability of the character. For it may easily be seen that while it is feasible to count the stamens even when converted into pistils, it is not possible when groups of them are more or less intimately united into single bodies. This combination makes all enumeration difficult and inaccurate and often wholly unreliable. In such cases the observation is limited to a computation of the degree of the change, rather than to a strict numerical inquiry. Happily the responses to the experimental influences are so marked and distinct that even this method of describing them has proved to be wholly sufficient. In extreme instances I have seen all the changed stamens of a flower of the opium-poppy united into a single body, so as to form a close sheath all around the central ovary. Lesser sheaths, surrounding one-half or one-third of the capsule are of course less rarely met with. Leaving this description of the outer appearance [374] of our anomaly, we may now consider it from the double point of view of inheritance and variability.