HEIGHT

OF PLANT

generation and practically no mold in the fifth so that the coefficient for per cent. smut is based upon the second and fifth generations and for per cent. mold upon the first and fourth.

The figures show a fairly high association for height of plant and moldy ears. This means that by selecting the highest lines in the first generation the resulting inbred strains in the fourth generation would tend to be taller than the average. Similarly, by selecting lines at the start that were free from mold, inbred strains could

Figure 59. Graph showing the behavior of two lines with respect to height during four generations of self-fertilization, selected for vigor and productiveness but not for height. From one to three progenies are grown in each generation.

finally be attained that would on the average be freer from moldy ears than other strains which showed more mold at the start. This relation does not hold so well for the other characters; number of tillers and per cent. smut. For these the coefficients are low and in two of the varieties a negative correlation is shown. This means that lines without tillers and showing low smut infection may be obtained from plants at the start which have tillers and are susceptible to smut infection.

Another method of bringing out the relation between the several lines at the start and at the end of the selection period is to separate all the lines of each variety into the upper and lower halves, with respect to the characters studied, in the first generation and then compare the average of these two groups with the averages of the same lines after being inbred for four or five generations. This has been done in Table IX, making the separation within each variety into equal sized groups in the first generation. Thus the basis for separating the groups is the median instead of the mean. The results are summed up graphically in Figure 58. It will be seen that the relative position of the upper and lower halves remains nearly the same at the end of the period of selection as at the

TABLE IX.

The relative position of the same self-fertilized lines at the beginning and at the end of the period of selection.

84

81

44 100

100

Low

31

82 52 41 64 93

smut in the second and High fifth generations

Yield of grain in bu. per

Low

888

acre in the first and High 31 fourth generations beginning for such characters as height of plant, number of tillers, per cent. of moldy ears and smutted plants although the differences are generally less at the close than at the start. This tendency to change during the period of inbreeding is most marked for yield of grain. In this respect the high and low groups are very nearly alike at the end of the selection period in spite of the fact that all along attention has been given to productiveness. These results indicate that it is unwise to eliminate the unproductive strains in the first generations, as from them lines may be obtained that are as productive as those from high yielders at the start. Other characters can apparently be somewhat more surely selected for at the beginning of the inbreeding period. If such characters as freedom from mold and smut are of chief importance it might be advisable to eliminate those lines which show much mold and smut in the first inbred generations.

The general tendency for some of the lines to hold the same relative position throughout the process of selection is illustrated by the height of plant of two lines shown graphically in figure 59. In the first inbred generation the two lines averaged 69 and 77 inches in height. In the second generation two progenies overlapped but from then on they were clearly distinct, the difference in height increasing until the end of the selection period. The same result is shown in the average number of tillers per plant of two other lines as brought out in figure 60. Differing at the start the two lines remained distinctly different in all their progenies throughout the period of inbreeding. In marked contrast to this is the result shown graphically in figure 61. Two lines differing noticeably in their number of tillers changed positions so that in

Figure 60. Graph showing the behavior of two lines with respect to tillering during five generations of self-fertilization, selected for vigor and productiveness but not for the number of tillers. The relative position of these two lines remained the same.

the end the few tiller strain at the start averaged more tillers on all progenies than did the many tiller strain. Similarly two strains which were alike in this respect at the start became extremely different as uniformity and constancy was reached, as shown in figure 62.

LIMITING FACTORS.

In planning and carrying out a selection program the best procedure will depend upon the number of plants which can be grown and the number of hand pollinations which can be made in a season. Where the facilities available for artificial pollination is the limiting factor, and this is usually the case, the best procedure

AVE

ON

TILLERS

PER

PLANT

is to self-pollinate just enough plants to continue as many lines as possible until a reasonable degree of homozygosity is reached. If the amount of land available to grow the plants is the limiting factor it would be better to pollinate a larger number of plants within each line, although extensive selection within a progeny has been shown to have little value, as the better individuals are almost certain to be more heterozygous, making it difficult to arrive at their true value. More attention should be paid to increasing the number of progenies within the more desirable appearing lines, basing selection on their behavior throughout the season and their uniformity and productiveness at maturity.

The method now being used at this station is to grow three progenies in each line and to pollinate two plants in each progeny. On the basis of the general appearance of the plants in the field and

Figure 61. Graph showing two lines which differed in the amount of tillering and which changed positions during the five generations of selffertilization.

their productiveness at maturity the best and second best progenies are noted where there is an appreciable difference. Two ears from the best progeny and one from the second best are used for planting the following year. If no differences are shown, one ear from each of the three is planted. This procedure is based upon the results in the five-year selection experiment described above in which no reliable criterions of selection were found which could be used before the time of pollination. It is still provisional and will be modified as future experience justifies. It is possible that better results can be obtained by paying still less attention to selection during the reduction period than the method outlined. By expending the same amount of time and effort on more lines, growing only one progeny in each generation and pollinating only enough plants to insure the perpetuation of the strain until uniformity and constancy are reached, more diverse material would be available

from which to select the best inbred strains. In this procedure there would be the possibility, and even probability, of missing altogether valuable material which might exist in some lines. However, since it has been shown that many of the lines change greatly during the reduction process, selection during this period will always be somewhat ineffective. From a theoretical standpoint the best method is the one which will produce the largest number of fixed strains from which to choose the ones best suited to the purpose for which they are to be used.

In this connection one further point should be mentioned. Whenever any particularly outstandingly good strain has been obtained there is the possibility that still better material may exist in that strain in the earlier generations. This would indicate that it

Figure 62. Graph showing two lines which showed the same amount of tillering at the start but differed widely at the end.

might be well worth while to go back to the earlier generations and grow as much of this material as possible from the remaining seed in order to obtain the very best germplasm available in this strain. In fact, this procedure has already been followed with several of the more promising lines and it has been possible to isolate new strains which are distinctly superior in some respects to the old ones.

CONCLUSION.

The one fact that stands out from the results secured in this selection experiment is that there is no single criterion by which high-yielding strains can be obtained. During the process of inbreeding, with the resulting segregation and recombination and the automatic elimination of heterozygous combinations of factors, selection for particular characters is somewhat effective. By