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



criticism, for he soaked the experimental half-leaves in water at the end of the 

 experiment to make them turgid. Soaking in water does in most cases 

 roughly counteract shrinkage errors, but more accurate means must be adopted 

 for general use. 



As an alternative, changes of area may be measured by the planimeter 

 method, or estimated from measurements of the changes in longitudinal 

 dimensions, and a corresponding correction then applied to the increase of dry 

 weight. Errors from shrinkage can also be entirely eliminated by the simpler 

 alternative of marking out a given area on both control and experimental 

 half-leaves at the beginning of the experiment. The stamping method has been 

 devised for this purpose : in this method an ink impression of a rectangle is 

 made on each half-leaf with a specially constructed rubber stamp. 



3. The other important error to which the dry weight method is liable 

 arises from asijmmetrij, from the fact that equal areas, taken at the same time 

 from the two halves of a leaf, have not accurately the same dry weight. This 

 source of error is inherent in the method and cannot be eliminated : it 

 determines the ultimate limit of the accuracy obtainable. 



The error can be diminished : (1) by avoiding outstanding veins, and so 

 reducing the degree of asymmetry shown by individual leaves, and (2) by 

 using a number of leaves for each experiment. 



In testing leaves of Paulownia iniperialis, using only parts without 

 prominent veins, differences of the order of 1| per cent, were found, whereas 

 portions with the minor veins slightly projecting gaye differences four timi's (ts 

 great. 



The reduction of the asymmetry error by the use of a large number of 

 leaves is limited in practice. An improved templet, or the rotating punch, 

 would make possible the simultaneous manipulation of a number of leaves of 

 smooth, firm texture ; but the result would have to be corrected for slirinkage, 

 and the estimation of this correction would prove laborious unless it could be 

 made from measurements of a few representative leaves. To deal singly with 

 a large number of leaves would take too long, whatever the method employed, 



4. Included in measurements of asymmetry are the experimental errors 

 involved in measuring the area and finding the dri/ weight of individual 

 portions of leaf material. Altliough these are small compared with tlie error 

 from actual asymmetry, it is nevertlieless important to reduce them as much 

 as possible, as they are all cumulative in their ellect upon the result. 



4a. The important methods of area detc^rmination are four in number : 

 Sachs' templet method ; Jhown and Escombe's planimeter method ; and the 

 rotating puncli and stamping methods devised in the course of this research. 



A test applied to Sachs' tc7)iplet methoil showed an error which was consider- 



