TABLE A. EXPERIMENTS OF SECTION I FOR 1880, 1881 and 1882. per acre 1880. 1881. 1882. Avr'ge 1880. 1881. 1882. Aver'ge P'ks lbs. lbs. lbs. lbe. bus. bus. bus. bushels 345678345678 12.30 13.42 16.68 17.23 14.13 1,880 1,856 1,600 1,778 8.10 17.00 17.33 14.16 17.73 15.18 15.73 18.48 21.15 21.33 Now, if we were only to consider the results of the six experiments with the Champion Amber in 1880, we would certainly decide in favor of thick seeding, for we have a marked increase (with a single exception, explained in the report of that year) in the yield of grain, for every additional peck of seed used. And in 1881 five out of the six experiments with this variety were equally as favorable to thick seeding. And again, both in 1880 and 1881, the three most thickly seeded plats of the Boughton gave the best yields of grain. If we had stopped this series of experiments at the end of the second year, we could have safely affirmed that the results of twenty-four careful experiments continued for two seasons, and these seasons, be it remembered, very dissimilar in character, were clearly in favor of thick seeding. If, on the other hand, we should have regard only to those of the present year, we would see, upon glancing down the column giving the yield of grain for 1882, first, that in the experiments with the Champion Amber the eight peck plat, which in 1880 gave the largest, and in 1881 the next largest yield, this year gave the next smallest, we might say the smallest, for there is practically no difference between its yield and that of the five-peck plat. Second, that in the Boughton the same plat which last year gave the next best, and the year before, the third best yield, this year gave the worst. Third, that this year the yields of the three heavily seeded plats of the Boughton were inferior to those of the other three; and fourth, that between the three heavily seeded and the three thinly seeded plats of the Champion Amber, there was this year practically no difference. Hence, I say, if this year had been the first of the series we might have concluded that the results of the twelve experiments were in favor, for the reasons above stated, of thin seeding. The longer I study the subject the more I am convinced that field experiments, as ordinarily conducted, are worse than useless; they are mischievous and misleading. In rash or untrained hands they are dangerous weapons, and even in the most skilled, twoedged swords which cut both ways. Other investigators in this field are reaching the same conclusion. Dr. Sturtevant affirms : "The system of plat experiment as usually understood, i. e., the planting of equal areas under presumably like conditions and then estimating the efficacy of treatment by the crop yielded is not capable of giving an answer which can be deemed trustworthy." His affirmation is based on the results of some field experiments in corn culture. Twenty plats were seeded in corn, ten of these "received ordinary farm cultivation," the other ten no cultivation whatever. The ten cultivated plats gave an average of 48.8 bushels per acre, the ten uncultivated an average of 51.4 bushels. In eight cases out of the ten the yield of the cultivated plats was inferior to that of the uncultivated. The Director goes on to observe : "Were these our only experiments we might well be justified in assuming that cultivation was a disadvantage to the crop. Fortunately, however, a second series of ten plats alongside received also the same treatment and served as a duplicate series." In this series the ten cultivated plats gave an average of 54.5 bushels per acre, the ten uncultivated an average of 44.1 bushels. "In this case the cultivated was superior to the uncultivated in eight cases out of ten, and in one case no difference. If we look at the average of these two series we find in Series I the uncultivated plats yielding 2.6 bushels per acre more corn than the cultivated; in the second series the cultivated yielding 10.4 bushels per acre more than the uncultivated. Attempt will not be made here to explain these discrepancies." He concludes: "Before progress in experiment can be * Bulletin No. XV. New York Agricultural Experiment Station. made it will be necessary to devise some more accurate and scientific method; one which is not only free from the objection to the plat system, but one which may be capable in itself of giving results which can be acted upon as if trustworthy." Now, while I am hardly prepared to admit that the plat system should be so summarily discarded, I am free to confess that it stands in great need of reformation and improvement. In the first place the experiments should be more carefully conducted. There is need of the greatest care and accuracy in the measurement and preparation of the land, in the selection and weighing and planting of the seed, in the harvesting of the crop, etc. The fact is too often overlooked that the slightest error made in estimating the yield of a small plat becomes a very serious one when multiplied by 10, or 16, or 20. The chemist understands this thoroughly when making his analyses. He will use only balances of the utmost precision; he carefully tests all his reagents, and makes sure that his apparatus is clean and well adjusted. The agriculturist is only beginning to realize the necessity of similar precautions. Heretofore he has been content to step off his land, to roughly measure his seed, and to make use of scales that show no difference between an ounce and a pound. Again, it is not sufficient to duplicate an experiment. Where tests of fertilizers, of different varieties and quantities of seed, of methods, of culture, etc., are made, four or five plats, situated in different parts of the field should be devoted to each experiment. And again, these tests, thus made, must, to give trustworthy results, be continued for a number of years. For the experiments of any one year serve only to show the effects of a single season, and are valueless as far as any other is concerned. And it should be distinctly remembered that even the results of experiments thus made, and thus continued, are only applicable to the given soil and locality. For different soils and different localities other similar series must be made. General conclusions applicable to a given region can only be drawn from comparisons of the results of experiments made in every different soil and locality of that region. Agricultural experimentation is certainly in its infancy, and it is not capable as yet of giving very accurate results, but this is only the more reason why we should labor to perfect it. It is not more untrustworthy than chemical analysis was a generation ago, and it must be per fected by the same means which have given to analysis its accuracy and value, by intelligence, effort and experience. That well-devised and carefully conducted field tests continued for a period of years, are capable of giving trustworthy results is shown by the celebrated experiments of Lawes and Gilbert. The farmer in making his calculations has to deal for the most part with the doctrine of averages. Especially does he consider the effect of the average seasons in estimating the probable yield of his proposed crops. He makes no special provisions of extraordinary seasons, and yet, no one knows better how potential a factor the season is in determining the yield of any given crop. But he is well aware that such provision would be useless, even dangerous, and that he has to take the chances. He is therefore most concerned to know what processes and methods, what varieties and what amount of any given seed, etc., would give the best results, not in any one year, but in a given period of years. And experiments continued for a number of years and hence exhibiting in their average results the effects of average seasons are fairly capable of giving him the desired information. It is generally held, for example, that on ordinary wheat soils the hardier red varieties are more profitable than the smooth white varieties of wheat. For while the latter will give the larger crop in exceptionally favorable years, the former yield better in poor or average seasons, and experience shows that average results are safer to count on in the long run. From this point of view the average results of the experiments of section I are possessed of some value. For these experiments now cover a period of three years-years differing greatly in the character of their seasons. Now, while I have commented freely on the conflicting nature of the results of the experiments of the present year, and have pointed out that they are rather favorable than otherwise to thin seeding, I must draw attention to the fact that the difference is too slight to alter the verdict of the experiments of 1880 and 1881. Hence, the averages of the experiments for the three years are, as shown by Table A, still in favor of thick seeding. For in the Boughton plats 7 pecks gave the best average yield, 17.23 bushels per acre, and 6 pecks the next best ; 8 pecks and 3 pecks gave nearly the same average, while 4 pecks and 5 pecks were decidedly inferior to the others. In the Champion Amber, 8 pecks gave the best average, 7 pecks the next best, and 6 pecks the next. With the same amount of seed the average of the Champion Amber was in every case superior to the Boughton. These points are all well brought out by the following diagram, which explains itself. Scale 100 pounds to 1-10 inch. a b represents yield of grain of Boughton. cd represents yield of grain of Champion Amber. ef represents yield of straw, etc., of Boughton. gh represents yield of straw, etc., of Champion Amber. In this it is seen that the line representing the yield of grain of the Champion Amber keeps well above, and fairly parallel with, that representing the grain of the Boughton until the 7 peck vertical line is passed, when the latter suddenly falls; that both are |