194 THE DEVELOPMENT OF RAW COTTON 



Taking next a small tuft picked at random from a good 

 sample of pure-strain ginned lint, and testing successive 

 bunches from it, we find a moderate amount of variation 

 due to the previous causes: Six fibres, 2-8, 3-3, 3-5, 3-7. 

 4-0; eight fibres, 3-5, 2-2; ten fibres, 3-2, 4-1 ; twelve fibres, 

 4- 2,5-9; fourteen fibres, 3- 6, 3- 4, 3-4, 3- 3. It will be noticed 

 with sufficient clearness for our present purpose how the 

 probable error is decreasing as larger bunches of fibres 

 are taken, and the fibre to fibre variation is eliminated 

 thereby. 



If now, instead of taking a small tuft from one part 

 of a good ginned sample, we take tufts from seed-cotton, 

 such as were used in one of the series hereafter to be 

 described, we find that it is impossible to sample effect- 

 ively. The strength of the lint in the 1913 series of 

 dated bolls was owing to circumstances necessarily 

 determined from unginned cotton. In order to obviate 

 the disadvantage of losing the mixing action of the gin, 

 the following routine was practised. Each sample was 

 broken into five lots, one for each impact test ; 

 each lot contained usually from fifty to 500 seeds, 

 and ten tufts dt lint were drawn from each lot, each tuft 

 taking some fibre from not less than three seeds. These 

 ten tufts were rolled up together, then pulled straight, 

 and a wide layer drawn out as in grading for regularity ; 

 this layer was then drawn down right and left until a 

 countable number of fibres remained, and thesk were 

 tested. The figures obtained with all these precautions 

 are given subsequently, ^and it will be seen that, although 

 they give five-day means with a probable error of 5 per 

 cent., which is good enough for our purpose, the bunches 



