September, 1911.] 



221 



Edible Products. 



EXPERIMENTS BEARING ON THE 

 CULTIVATION OF PADDY. 



By R. H. Lock, m.a., sc.d. 



(Paper read before the Board of Agri- 

 culture at its August Meeting.) 



h On the Nature of Agricultural 

 Experiments. 

 The notes which I have the honour to 

 submit to you this afternoon deal with 

 the method of experiment as applied to 

 a particular branch of agriculture. In 

 order that there may be no mistake 

 about my meaning, I should like, with 

 your permission, to explain the sense in 

 which I use these terms. Agriculture, 

 I take it, is the cultivation of the soil 

 for profit, and the best agriculture is 

 that which results in the largest profit 

 without exhausting the soil and so re- 

 ducing its capacity for yielding further 

 profit. An experiment is a test planned 

 scientifically for the purpose of obtaining 

 definite knowledge, and the best experi- 

 ment is that which leads to the most 

 accurate information. What I have 

 said so far may seem at first sight to be 

 in agreement with the opinion univer- 

 sally prevalent in Ceylon among all 

 classes — the opinion that the objects of 

 agriculture and those of science are 

 totally opposed to one another. It is 

 not so very long since we heard the 

 opinion expressed in this room that a 

 commercial experiment is a different 

 thing from a scientific experiment. That 

 view is not held in all agricultural 

 countries, and it is not my view. One 

 of my main objects to-day is to express 

 as forcibly as I can the opinion that an 

 experiment which is not a scientific ex- 

 periment is not an experiment at all. 



Two things only are essential in a 

 scientific experiment, namely, common 

 sense and accuracy, and any man who 

 possesses these most uncommon qualifi- 

 cations has the right to call himself 

 scientific. Without them no amount of 

 labour and information can produce 

 science. Let us see what science has to 

 say on the subject of agricultural experi- 

 ments. 



We will take a simple case of an ex- 

 periment designed to afford accurate 

 information on some point important to 

 agriculture, that is to say, affecting the 

 question of profit. For example, we 

 may wish to know whether a certain 

 quantity of a particular manure applied 

 to a particular crop will produce an 

 increase in the yield, the value of which 

 will be greater than the cost of the 

 manure, the cost of transport to the 



field, and the cost of application, all 

 added together. For this purpose we 

 may mark out two plots of equal area; 

 sow each with the same quantity of seed 

 of the crop in question, and treat them 

 in the same way in all other respects, 

 except that we apply the manure to one 

 of the plots and not to the other. If the 

 plots or their treatment differ in any 

 other way, it will be quite impossible to 

 tell whether any difference which may 

 be found between the yields from the 

 two plots should be ascribed to the 

 effect of the manure or to some other 

 cause. 



We will suppose that our two plots 

 are each one-hundredth of an acre in 

 extent ; and further, that the crop from 

 the unmanured plot weighs 100 lb., and 

 that from the manured plot 110 lb; Are 

 we therefore justified in assuming that 

 the same amount of manure applied in 

 the same way always cause an increased 

 yield of 10 per cent.? 



The answer to this question is " no." 

 To come to such a conclusion would be 

 to suppose that our work is perfectly 

 accurate, and that natural conditions 

 can be made perfectly uniform. In 

 practice we can only make an approach 

 to accuracy. Two plots in a field can 

 never be made exactly alike, there will 

 be slight differences in soil, aspect, 

 drainage, and the like, and all these will 

 affect the crop. It is therefore most 

 important to know how close an 

 approach to accuracy may be expected 

 in an experiment like the present. In 

 order to find this out it is necessary to 

 know how much difference is to be 

 expected between two plots which have 

 been made as much alike as possible, 

 and which have not been manured 

 differently or otherwise differently 

 treated. 



How is this most important point to 

 be ascertained ? The method is to grow 

 a large number of pairs of like plots and 

 to observe to what extent the crops do 

 actually differ from one another. We 

 may take the average crop of all the 

 plots as the amount which each plot 

 ought to yield theoretically, and we 

 shall find that the actual yield of each 

 plot differs to some extent from this 

 amount. From the data thus obtained 

 it is possible to work out the odds that 

 the crop of a single plot will differ from 

 the average by more or less than a given 

 quantity. 



This has been done in a very interest- 

 ing paper published by Professor Wood 

 and Mr. Stratton in the last number of 

 the " Journal of Agricultural Science." 

 A brief summary of their conclusions 

 may not be without interest. 



