246 Journal of Agricultural Research voi. xn, No. $ 



measure its probable amount so as to make sure it is not likely to vitiate 

 our conclusions. 



One of the chief difficulties in obtaining reliable results in field trials 

 is the natural variability of the material with which we are deaUng. 

 Crops are living organisms with inherent tendencies to vary, even though 

 it were possible to make their environmental conditions identical. 



In agronomic experiments the number of plants taken is usually so large 

 that inherent variability ceases to be a factor of importance. In horti- 

 cultural experiments, however, where fruit trees are under observation, 

 the limited number of trees possible to include in a plot may make the 

 factor of inherent variability an important consideration. 



Further variation is induced as a result of the many factors of the 

 environment which are beyond the control and possibly the recognition 

 of the experimenter. Some of these factors are independent; others 

 react upon one another. In designing a set of field trials, we try to 

 avoid, as far as possible, all secondary factors which may exert a 

 disturbing influence. 



Lack of uniformity in both the physical and chemical characteristics 

 of the soil is one of the foremost factors causing variation in productivity 

 of plants. Apparently uniform surface soils may be underlain with a 

 heterogeneous subsoil. Differences also occur which are not evident on 

 a careful inspection of both the soil and the crops, but which are easily 

 measured by weighing the yields. In other words, the weighing machine 

 is more sensitive than the eye and reveals differences that mere inspection 

 can not detect. 



The past treatment of the soil brings in variables the significance of 

 which may not be comprehended at the time a field trial is begun. The 

 persistent effects left by the application of stable manure on some of the 

 Rothamsted plots show how large a part is played by the past history of 

 the field. Plots which for 40 years have had identical treatment still 

 give different crop yields because of the effect of dressings of barnyard 

 manure applied at an earlier period. 



Unequal prevalence of diseases and insects may bring about further 

 error in the results. 



Besides the above sources of variation and possibly outweighing them 

 at times is the effect of season. No season is entirely normal; and it is 

 only when the experiment has been repeated for several years, or, in other 

 words, until "a fair sample of seasons" has been made, that any sort of 

 allowance can be made for seasonal effects. 



As is shown by observations at Rot hamsted, it is not possible to estab- 

 lish a schedule of relative jdelds for a series of plots, even after several 

 years' comparison. In this case two grass plots were treated alike for 

 50 years; by taking the yield of one plot as the standard, the yield on 

 the other in the same season has been as low as 90 per cent and as high as 

 196 per cent. 



