METHODS OF INVESTIGATION 193 



sixteen fibres has only a quarter of the probable error of 

 a s ingle fibre. The whole difficulty lies in sampling, 

 and the results obtained with the impact tester will serve 

 to illustrate the points in general. 



Sampling. It might appear easy to obtain a uniform 

 lot of fibres by repeated drawing of the lint, overlaying 

 tufts on one another, and drawing again from these. In 

 point of fact the only reliable way is to pick the hairs 

 one by one at random from those projecting out of a 

 loose lump, of lint, as O'Neill did. Again, it should be 

 noticed, the amateur and the grader obtain the best 

 results from exactly opposite methods. 



Illustrating the difficulty of sampling from numerical 

 obtained with the impact tester, and leaving the numbers 

 in their arbitrary scale, taken direct from the notebook, 

 we obtain such results as the following : 



Small tuft of about 500 fibres taken from a square milli- 

 metre of the butt of a single seed, and tested in bunches of 

 4, 5, 7, 10, 12, 14, 16, and 18 fibres respectively, worked 

 out at the following strengths per fibre: 1-25, 1-60, 1-85, 

 2-00, 1-40, 1-85, 1-70, and 1-60. The variation is slight, 

 though the absolute strength is very low. 



Another seed of the same sample was tested from six 

 points round it, two on either side one from the short 

 hairs near the tip, and one from the butt; twelve hairs in 

 each sample ; strength per fibre was Tip 2- 5, butt 4- 2, left 

 side 5-4 and 6-4, right side 3-1 and 3^1. The variation 

 is increased, probably because the nutrition varies accord- 

 ing to the proximity of the particular hairs to the vascular 

 bundles which supply food-substances. 



13 



