In my report of 1906 I called attention to the nature of inheritance of the crest in the first and second generations. The result seemed simple enough on the assumption of imperfect dominance. However, in later generations difficulties appeared, one of which was referred to in a lecture given before the Washington Academy of Sciences in 1907. I stated (1907, p. 182), that "when a crested bird is crossed with a plain-headed one, and the crested hybrids are then crossed inter se, the extracted recessives of the second hybrid generation are plain-headed, to be sure, but they show a disturbance of certain feathers." This was an illustration of the statement that recessives which are supposed to come from two pure recessive gametes show in their soma traces of the dominant type. Dr. W. J. Spillman, who was present, made the suggestion that the crest is composed of two characters, T and t, instead of a simple element, and that when t alone is present the result will be the roughened short feathers on top of the head.

Further studies demonstrate the validity of this suggestion. There are in the crest two and probably three or more factors. There is a factor that determines length of the feathers and a factor that determines their erectness. There is probably also an extension factor that controls the area that the crest occupies on the head. Thus flatness of position dominates over its absence (or erectness). This is seen even in the first generation. Figs. 5, 6, 8, and 17 of my report of 1906 show this very plainly. They also show that continued growth of feather is dominant over interrupted growth. Thus in the second hybrid generation I got birds with short and erect feathers and one of these is shown in fig. 11 of the 1906 report. That shortness is recessive is proved by various matings of extracted short × short crest. Of 29 offspring none have a higher grade than 1, grade 10 being of full length. On the other hand, two parents with long feathers in the crest (grades 6 to 8) give 5 offspring of grade 1, 12 of grades 5 to 10, thus approaching the 1:3 ratio expected from two DR parents. That erectness is recessive is proved by various matings of extracted erect × erect crest. Of 25 offspring none has a lower grade than 4 (1 case) or 5 (1 case). On the other hand, two parents with extracted non-erect feathers give in 46 offspring 13 with feathers whose grade of erectness is 6 or higher and 33 with a grade of 5 or below—of these half of grade 0—close to the expected 1:3. The evidence is conclusive that there are two factors in crest that behave in Mendelian fashion—a factor determining the prolonged growth of the feather and a factor causing the feathers to lie repent. The area of the head occupied by the crest is also variable. This was estimated in tenths for each of the parents and offspring. Two principal centers of variation appeared, at 3 and at 8, or roughly one-third and two-thirds the full area. The results, being based on estimates, are not wholly satisfactory, but so far as they go they indicate that when both parents have a crest that belongs to the lower center of variation their offspring belong chiefly if not exclusively to that center; but when they both belong to the upper center of variation a minority of the offspring belong to the lower center. Provisionally it may be concluded that extensive crest is dominant over the restricted crest or that there is an "extension factor."