Water-Conductivity of the Wood in Trees and Shrubs. 235 



during this investigation were mainly statistical. This appeared, in the 

 present lack of accurate knowledge on the subject, to be the most promising 

 way to attack the subject, and to offer the best chance of clearly formulating 

 further problems. But the conclusions drawn from the data themselves 

 must be cautiously and critically drawn. The nature of the material, 

 especially in the case of deciduous trees, often shows rather wide fluctuations. 

 When there is reason to believe that these fluctuations arise owing to the 

 lack of real homogeneity in the material, it is obvious that nothing is gained 

 by determining the arithmetic mean and the probable error. The curves 

 drawn from a large series of observations sufficiently indicate the interaction 

 of different factors, and increase of the observations does not lead to a 

 smoothing of a single curve for the total results. A frequency -grouping of 

 values at different intervals is what is obtained, and in this way one arrives 

 at a clearer recognition of the existence of the several influences which affect 

 the general result. 



The observations made of the Birch afford an illustration of the foregoing 

 remarks. They show that a smooth curve cannot be obtained, nor is it likely 

 that a very large increase in the experiments would bring this about. Even 

 if it did the results would be misleading, because in collecting data from an 

 indiscriminately selected lot of Birch twigs one is dealing with really hetero- 

 geneous material. The leading shoots are worse water-conductors (per unit 

 area of wood) than the laterals, and the presence of the two maxima leads to 

 a flattening of the curve in the intervening values. At the same time, as the 

 subjoined Tables show, a few unusually high or low figures may seriously 

 disturb the position of the arithmetic mean of the result if taken from such 

 a relatively small number of specimens as 28 sets of laterals and terminals 

 here presented. 



The shoots were obtained from a lot of trees all about five to six years old, 

 and as much care as possible was taken to secure as fairly average a set as 

 possible both of terminals and laterals. The results show that by taking the 

 mean of each series, it would be concluded that there was but little difference 

 (at most only 2'8) between them — too small to be regarded as significant. 

 But the numerical average for the terminals is really too high, and this is due 

 to a small detached group standing at about 67. The density average for the 

 terminals lies about 40, whilst that of the laterals is at about 52. The Tables 

 give the areas as well as the absolute specific volumes, and the close agree- 

 ment between the average specific volume, as recorded at the bottom of the 

 last column, with that obtained from the sum of the areas and absolute 

 volumes at the foot of the third column, affords a ready check on the calcula- 

 tions as a whole. 



u 2 



