18 Mr. D. Thoday. Experimental Researches on [June 11, 





Avoiding veins. 



Including veins. 



(1) Average difference per square decimetre . . . 

 Difference per square decimetre taking all 

 together 



Percentage difference taking all together . . . 



milligrammes. 

 ±9 

 + 5 



±1-4 

 + 0-8 



milligrammes. 

 -37 

 -37 



±5-9 



- 5 -9 



Table XIV. — Catalpa bignmioides. 



Except in the case of Leaf 1, the midrib was removed before cutting pieces from the 

 half-leaves with the templet : area in each case 50 sq. cm. 



Leaf. 



(1) 



(2) 





milligrammes. 



per cent. 



1 



+ 19 -0 



+ 4-0 



2 



-1-6 



-0-3 



3 



+ 10 -6 



+ 1-7 



4 



-10-6 



-1 -8 



5 



+ 38 -8 



+ 8-1 



•6 



--17 -8 



-3-1 



The left half of Leaf 5 was seen to be not perfectly flat ; the results 



for this leaf have therefore been omitted from the following averages. 



Excluding Leaf 5. 





Average 



±11 -9 



±2-2 



Taking all together 



- 0-1 



Nil 



The results as a whole prove the necessity for determining this error for 

 each plant used in an investigation. Asymmetry, to the extent of an average 

 of 1^ to 3 per cent., is shown by all the leaves examined, and in extreme 

 cases it varies from zero to 4 per cent, or more. Catalpa hignonioides shows a 

 higher degree of asymmetry than the average. 



The seriousness of this asymmetry in assimilation experiments is greater 

 in thick than in thin leaves. Tims in Cherry Laurel, in an experiment 

 lasting seven hours, asymmetry to the extent of 2 per cent, will mean an 

 error of about +.':! milligrammes per square decimetre (+ 0'3 gramme per 

 square metre) per hour, while in Helianthus tuhcroma it will be less than 

 + 1 milligramme per square decimetre pef hour. For a rate of increase of 

 10 milligrammes per square decimetre per hour, such as Brown and Morris 

 found for detached leaves of JfelianlJms annuus, tlw average percentage error 

 would be 30 per cent, in Ciierry Laurel, 10 per cent, in JfrlianfJinfi. 



J'.y using a number of leaves this error may be reduced roughly in the 



