186 liijlmiici' of S(((l Wdijht, clc. on the rofdfu Crop 



paper, in the case of tlie smaller lots in the lowest row, the return per 

 tuber set is calculated after correcting for the lesser number of plants. 



As there was no disease in the tubers at all, there being but very 

 little in this part in 1921, and the variety itself exhibiting considerable 

 resistance to I'hytophthora, no allowance was needed for this possible 

 source of error. If the diagram lie studied it will appear that: 



(rt) Tliere is much less variation in tlie crop return on the guard rows 

 both as regards plots alongside and plots above or below eacli (jther, 

 than there is in the case of the A-K series. 



{b) That whilst yields in the outermost guard row on the north side 

 are uniformly less in all four of its divisions than those of its inner 

 neighbour, those of the latter do not materially differ from those of the 

 two guard rows on the south side, which, in their turn, are almost com- 

 pletely ahke. 



(c) The variation between the uj)i)er and lower divisions in the guard 

 rows, more especially the inner northern and the two southern rows, is 

 such that the crop is gradtially increased from east to west so that that 

 of the latter is twice the value of that of the former rows. On the other 

 hand, the returns of any of the four rows sliows no appreciable variation 

 from north to south. 



Turning to the A-K series, it will be seen that B, C, D and (f are 

 represented at least once in every row, so that in their case the variation 

 in the soil condition which nuinifested itself is, in its effect on crop, 

 neutralised. 



In respect to A. E and F we are on less sure ground, but as will be 

 seen later, it is ])ossil)le to arrive at results in respect to them wliich would 

 appear to be almost equally satisfactor}'. 



The first problem was to find the standard deviation of differences. 

 As B, C, D and G occur in all rows, tl.e pairs BC, BD, Bb, CD, Cb, CG 

 were taken first. Whenever B, C or D was duplicated in a given row, 

 both possible pairs were taken, but throughout, only pairs in the same 

 row were considered. To tliis group of thirty-five pairs was added 

 fifteen composed of the pairs EB, EC, and ED, each occurring four 

 times, and E(! which occurs three times. Thus ten pairs of varieties gave 

 fifty pairs of plots, each pair in the same row. 



The standard deviation thus obtained is I-IO lbs., which allows of a 

 probable error for one 2>air of 2-77, or for four jmirs of l-;58. 



There is next the difficulty of obtaining comparable mean yields for 

 each variety. The data for B and C are : 



