Katharine H. Coward i^h 



possible through the points plotted from the results of the different doses and 

 then finding the dose corresponding to the result obtained from the one dose 

 of standard. If the relation between effect and dose is logarithmic, the line 

 relating effect to the logarithm of the given dose is a straight one and can be 

 calculated mathematically with very little difficulty. The whole process is 

 then independent of judging where exactly the line should go. 



Other workers like to give two groups of rats two different doses of the 

 standard and two other groups two different doses of the test substance. Then 

 two curves of response are obtained, their average slope determined and the 

 relative potencies of test and standard found by the distance between the tAvo 

 parallel lines. On the other hand, some -workers prefer to establish a general 

 curve of response for their test and compare the results from dose of standard 

 and dose of test substance through the doses corresponding to these results 

 on the general curve. Unless it has been shown that the slopes obtained in 

 various experim.ents are significantly different from each other, then the 

 method of using the general curve of response is probably tlie best one, as 

 having been determined by the use of a much larger number of rats than 

 would be used in any one assay alone. 



It is well worth while to realise, as R. A. Fisher has so clearly shown in his 

 Design of Experiments,^ that, by careful planning, an experiment can be 

 made to yield far more information than the one point primarily needed 

 and that by such planning the information on the chief point is also made 

 much more reliable. For example, the writer has lately compared two sam- 

 ples of irradiated ergosterol for vitamin D activity. It was also desired to 

 discover without independent experimentation whether the differences in 

 response to vitamin D were greater between different litters than between 

 different rats of the same litter. Several years ago the writer had sliown that 

 tliere was a fair correlation between the amount of liealing brought about 

 by vitamin D and the weight of the rat when it was first given the rachito- 

 genic diet. By careful arrangement of the rats within each litter given the 

 particular doses (three of one sample of irradiated ergosterol and three of the 

 other sample) it -was possible to arrange the results {a) as a "scjuare" in which 

 the row's were litters and the columns were doses, and {b) a second "square" 

 in which the rows were litters and the columns were rats in ascending order 

 of initial weight. Thus, information was gained on two points of impor- 

 tance in addition to the one for which the experiment was originally needed. 



Discarding the principle of standardising the conditions of experiments 

 by varying only one factor at a time, but varying several conditions instead, 

 makes it possible to gain not only more information on different points but 

 also information on the interaction of these conditions. The interaction 

 of the vitamins, already indicated almost by chance, is only one of many 

 fascinating lines which this method of procedure suggests in planning c\en 

 routine assays. 



