146 



THE AGRICULTURAL NEWS. 



May 11, 1912 



necessary eventually to modify the whole scheme of 

 investigation. 



Even though such a scheme may have been drawn 

 up most carefully, it is not possible to expect an exact 

 answer to the questions that it is sought to elucidate — 

 a matter that is amply illustrated by the differences 

 obtained in various years or various places. The aim 

 of the investigation is to find the most probable 

 result that can be given by the differing answers. 

 In obtaining this most probable result, in the first 

 case, a selection must not be made of the results to 

 which the greatest importance is attached; all the 

 figures that are given by the trials must have con- 

 sideration. In the same way, it must be realized 

 that the average result of a series of experiments 

 cannot always receive dependence. This is illustrated 

 from the consideration that the effect of a manure on 

 different soils, to take an example, will not be truly 

 represented by the mean of the results, on account of 

 the fact that several among them may give no response 

 of any kind to the manure. In such a case the only 

 accurate method is the repetition of the trials on the 

 same soil. L■^stly, however great may be the care with 

 which the experiments are carried out, there is always 

 some error where measurements have to be made. It 

 is therefore necessary to recognize the existence of this 

 inevitable error, to reduce it as far as is practically 

 possible, and to measure its probable amount for the 

 purpose of determining what must be allowed for its 

 effect, in arriving at definite conclusions. Methods 

 have been devised for estimating this probable error, 

 and although they were invented for use in astronomy 

 and physics, they can be employed equally in agricul- 

 tural experimentation. 



In considering the causes of variation in the 

 results of field trials, it is evident that these can never 

 be repeated under exactly the same conditions. This 

 fact, together with what has just been stated, shows 

 that the differences observed arise in two ways: from 

 the true error of the experiment, and from the diver- 

 gence between the conditions of the various trials. 

 One of the most potent among such conditions is 

 lack of uniformity in the soil. Another is the effect 

 of hedges and trees, in subjecting plants in different 

 parts of the plot to varying circumstances. A slight 

 difference of level, even, in an experimental field may 

 suffice to render the undulating part of it useless 

 for the purpose of investigation. In temperate lati- 

 tudes, more than in the tropics, lasting effects of a dis- 

 turbing nature arise from old applications of pen 

 manure. Again, the conditions of plant growth may 



not be uniform, and this is why the outside row or 

 rows often does not receive consideration in arriving 

 at results. Further, there is the effect of the unequal 

 incidence of diseases and pests: it must be remem- 

 berod, however, that this may possess a useful 

 significance in regard to the subject at issue; thus 

 a lack of potash is liable to be shown by a greater 

 susceptibility to disease. These considerations would 

 not be complete without a reference to the effect of the 

 individuality of plants, which however is small because 

 of the hu-ge number that is usually employed in an 

 agricultural experiment. Lastly, there is the effect of 

 variation in season, which may be very large. All the 

 sources of variation that have been dealt with may be 

 divided into two groups: firstly, those that get smaller 

 with the decrease in the size of the plot, as for instance 

 variation in the soil and the conditions of growth; and 

 secondly, tho.se that become greater as the size of the 

 plot decreases, these being the variations in the indi- 

 viduality of the plant and in the incidence of disease, 

 and the effect of the outside rows. 



The facts that have been adduced are sufficient to 

 show that the chances of accuracy of experiments are 

 enhanced by increasing the number of plots. In this 

 way, the effect of unei[ual conditions in the soil tends 

 to become eliminated, and the errors made in experi- 

 mentation and observation go to balance one another. 

 As, however, the addition of plots increases the labor- 

 iousness of the work, the number of plots used must 

 be the result of compromise between accuracy and 

 convenience. 



At this stage, it is evident that the magnitude of 

 the experimental error in field trials will be dependent 

 on the character of the soil and sub-soil, the previous 

 history of the former, and the nature of the crop and of 

 the season, so that it is necessary for this error to be 

 calculated on each occasion that the trials are made. 

 It is, however, desirable to know what may be the 

 order of the error to expect, and the determination of 

 this and of other similar matters will receive attention 

 in the next number of the Agricultural News. 



Experiments in which the Bambarra ground nut ( I'oand- 

 zeia suhterranea) was fed, with hay, to wethers is described 

 in Ber Tropenpfan-.er, 1911, p. 41.3. The unshelled pods 

 were used, and the percentage co efficients of digestibility 

 proved to be: fat 100, proteids 84-2, nitrogen-free extract 84:3, 

 cellulose 25-6. Analysis of the pods thus employed gave the 

 following percentages: water 15, proteids 179, fat 3'9, nitro- 

 gen-free extract 491, fibre 107, ash 3-4. The Bambarra 

 ground nut was described in the Agriciiltuval News, Vol. IX, 

 p. 340. It is under trial at .<;everal of the Botanic Stations 

 in the West Indies. 



