species much cultivated in gardens, is very variable in the number of the different parts of its flowers, and may in some instances be seen even with six stamens. If studied in the same way as Heinricher's iris, it no doubt will yield highly interesting and confirmatory results. Many other instances of such systematic atavism could be given, and every botanist can easily add some from memory. Many anomalies, occurring spontaneously, are evidently due to the same principle, but it would take too long to describe them. Reversion may occur either by buds or by seeds. It is highly probable that it occurs more readily by sexual than by asexual propagation. But if we restrict the discussion to the limits [175] hitherto observed, seed-reversions must be said to be extremely rare. Or rather cases which are sufficiently certain to be relied upon, are very rare, and perhaps wholly lacking. Most of the instances, recorded by various writers, are open to question. Doubts exist as to the purity of the seeds and the possibility of some unobserved cross disturbing the results. In the next lecture we shall deal in general with the ordinary causes and results of such crosses. We shall then see that they are so common and occur so regularly under ordinary circumstances that we can never rely on the absolute purity of any seeds, if the impossibility of an occasional cross has not been wholly excluded, either by the circumstances themselves, or by experimental precautions taken during the flowering period. For these reasons cases of atavism given without recording the circumstances, or the precautions that guarantee the purity of the fertilization, should always be disregarded. And moreover another proof should always be demanded. The parent which yielded the seeds might be itself a hybrid and liable to reversions by the ordinary laws of the splitting up of hybrids. Such cases should likewise be discarded, since they bring in confusing elements. If we review the long list of recorded cases by these [176] strict methods of criticism very few instances will be found that satisfy legitimate demands. On this ground it is by far safer in the present state of our knowledge, to accept bud-variations only as direct proofs of true atavism. And even these may not always be relied on, as some hybrids are liable to split up in a vegetative way, and in doing so to give rise to bud-variations that are in many respects apparently similar to cases of atavism. But fortunately such instances are as yet very rare. After this discussion it would be bold indeed to give instances of seed-atavism, and I believe that it will be better to refrain wholly from doing so.