184 



SCIENTIFIC AGRICULTURE. 



Februai-y, l9"-2. 



there is considerable variation in differ- 

 ent parts of a field, averages may fur- 

 nish a doubtful basis for measuring the 

 effect of treatments. 



The number of questions "put to the 

 soil and the plant" in a given plat ex- 

 periment has usually been far too large. 

 For example, the customary rotation-fer- 

 tilizer experiment has often covered prac- 

 tically the whole range of soil fertility 

 and plant nutrition. This wide range has 

 limited the amount of replication prac- 

 ticable, and it has failed to reflect the 

 discrimination in gathering data and the 

 simplification of the problem dictated 

 by the method of science. 



Such experiments have relied quite 

 largely on what the field results them- 

 selves were interpreted to show, prim- 

 arily the crop returns. True, most of the 

 later experiments have embodied plans 

 for chemical, bacteriological and other 

 laboratory studies, but only to a limited 

 extent have these been developed with the 

 progress of the work so as to shed new 

 light. The chemical studies have often 

 become of a routine nature — analyses of 

 the crops and of the soils at stated inter- 

 vals, and the bacteriological studies by 

 the technique developed have largely 

 failed to meet expectations in establish- 

 ing correlations between soil treatment 

 and bacterial flora. Such observations 

 have now almost ceased in connection 

 with these experiments. 



Reduced to such a simple collection of 

 experimental data, the conduct of these 

 extensive field experiments has often be- 

 come largely a matter of routine. The 

 niceties of plat work are observed, but 

 the element of actual inquiry is deferred 

 until many years have supplied their 

 data. When that time is reached the pub- 

 lication is more often a summary of field 

 and laboratory records than a critical 

 analysis of the data -and their a-ctual 

 meaning. At best the product is quite apt 

 to consist of empirical observations rath- 

 er than definite contributions to funda- 

 mental principles. "We have not yet learned 

 how to interpret except superficially the 

 answer which the soil and tHie plant give 

 as to just what has happened or to what 

 the apparent effects are due. Wo have not 

 yet learned how to examine a plat of soil 

 so as to determine the changes occurring 



from time to time or brought about by a 

 long continued system of treatment, or how 

 to connect these changes with the res- 

 ponse of the crop in a given season or 

 period. Indeed, relatively little study is 

 now given in such experiments to the soil 

 itself, and only to a limited extejit are 

 underlying questions suggested by such 

 experiments being given intensive study. 



In a word, the indications are that in 

 the majoritj' of cases the use is not be- 

 ing made of such long-time field exper- 

 iments that ought to be made at this 

 stage. They are rarely being simplified 

 as time goes on, with a narrowing down 

 to specific problems for intensive re- 

 search, and they are not being increas- 

 ingly supplemented by definitely direct- 

 ed laboratory study. They ought them- 

 selves to be progressive both in metliod 

 and outlook. They ought to be used as 

 the source of problems and material with 

 which to make further and more profound 

 inquiries. 



We can hardly fail to recognize the 

 T'lianged status at the present time, both 

 as to practical requirements and the stage 

 wliich has been reached in research and 

 its problems. What is especially needed at 

 this stage is the study of factors ^nd 

 their relationships, rather tlian gross com- 

 parisons of one complex of conditions 

 with other complexes. This will call for 

 the kind of team work which has been 

 applied to the Rothamsted expei'iments, — 

 the association of the cliemist and the bac- 

 teriologist with the agronomist and soil 

 expert, and the guidance of the statis- 

 tician in both planning and interpreta- 

 tion. 



In iiiaiiy of the feeding exn<M'iments, 

 also, llie unchecked sources of possible 

 error are too great for safety. The small 

 number of animals in the lots gives large 

 chance for tllie influence of individual 

 variation. The conditions and infrequency 

 of weigliine: may also give misleading in- 

 dications. Some of tlie results of such ex- 

 periments can be measured quite accur- 

 ately, while others can only be describ- 

 ed. Some are not .strictly experimental 

 because they embody so many factors not 

 under experimental control, whose prob- 

 able variation can not be estimated. Tliis 

 is true of tlie cost or fiiumcial returns 

 in feeding, as Dr. IT. IT. Mitchell has re- 



