The field experiment, as commonly executed, has sinned against both these requirements, but particularly the first, in that it has assumed uniformity of conditions instead of proving it. One of the most important of these conditions is the soil. It has too frequently been assumed that simple inspection is sufficient to assure one of the uniformity of this factor, but this is far from being the case.
Uniformity of soil over any considerable area is by no means an easy thing to attain. In our Northern States, or in any drift-region, one has only to examine the sides of the nearest ditch in order to convince himself that the character of the soil varies from rod to rod, and almost from foot to foot, and to cause serious doubts as to the value of comparative field trials to arise in his mind. But even in localities where such striking variations do not occur, sufficient differences between adjacent plots are frequently found to invalidate the results of such trials. These are not only differences in the amount of available plant-food present; the physical properties of the soil play a prominent rôle. Even slight differences in the depth or texture of soil or subsoil, a greater or less proportion of organic matter, a difference in surface color, a variation in the moisture of the soil, may have a decided effect on the crop. Repeated trials have shown that it is practically impossible to prepare a series of plots whose natural crop-producing power shall be uniform.
Under these circumstances trustworthy results in field trials can be expected only if the amount of probable variation between different plots is known, A preliminary cropping without manure naturally suggests itself as adapted to furnish this information, and such preliminary trials are of great value. At the same time they are not of themselves sufficient, A different season may cause the relative yield of two plots to vary considerably in different years. Moreover, two plots might show the same yield when unmanured, and yet be differently affected by the same manuring. As a control on the natural variation of the soil, Drechsler depends chiefly upon duplication of manurings, the same treatment being applied to a number of plots scattered over the field. By this he aims to accomplish two things: 1. The average yield of these several plots is more likely to correspond to the response which the field as a whole would make to the same manuring than is the yield of any single plot. 2. The variations of the several plots from the average furnish a measure of the uniformity of the soil, and serve to show whether a given difference in the final results of two kinds of manuring is significant, or is simply accidental and less than the errors of experiment. This is not the place to enter into a description of all the precautions required in the conduct of such trials. Those interested can consult Drechsler's original papers. One point may be noted, however, viz., that differences in the season, whether wet or dry, e. g., may have a deciding influence on the action of manures, and that only continuing the experiments for a number of
