Of course there may be exceptions. As yet we cannot detect the causes which control them, though injury, impurity, accidental crossing, mistakes of various kinds, account for many. Mendel himself says, for instance, that unhealthy or badly grown plants give uncertain results. Nevertheless there seems to be a true residuum of exceptions not to be explained away. I will recite some that I have seen. In my own crosses I have seen green × green give yellow four times. This I incline to attribute to conditions or other disturbance, for the natural pods of these plants gave several yellows. At Messrs Suttons’ I saw second-generation seeds got by allowing a cross of _Sutton’s Centenary_ (gr. wr.) × _Eclipse_ (gr. rd.) to go to seed; the resulting seeds were both green and _yellow_, wrinkled and round. But in looking at a sample of _Eclipse_ I found a few _yellow_ seeds, say two per cent., which may perhaps be the explanation. Green wrinkled × green round _may_ give all wrinkled, and again wrinkled × wrinkled may give _round_[73]. Of this I saw a clear case--supposing no mistake to have occurred--at Messrs Suttons’. Lastly we have the fact that in exceptional cases crossing two forms--apparently pure in the strict sense--may give a mixture in the _first_ generation. There are doubtless examples also of unlikeness between reciprocals, and of this too I have seen one putative case[74]. [73] Professor Weldon may take this as a famous blow for Mendel, till he realizes what is meant by Mendel’s “Hybrid-character.” [74] In addition to those spoken of later, where the great difference between reciprocals is due to the _maternal_ characters of the seeds. Such facts thus set out for the first cross-bred generation may without doubt be predicated for subsequent generations. What then is the significance of the facts? _Analysis of exceptions._ Assuming that all these “contradictory” phenomena happened truly as alleged, and were not pathological or due to error--an explanation which seems quite inadequate--there are at least four possible accounts of such diverse results--each valid, without any appeal to ancestry. 1. That dominance may exceptionally fail--or in other words be created on the side which is elsewhere recessive. For this exceptional failure we have to seek exceptional causes. The artificial _creation_ of dominance (in a character usually recessive) has not yet to my knowledge been demonstrated experimentally, but experiments are begun by which such evidence may conceivably be obtained.