438 BOTANICAL GAZETTE [December 



somewhat over 360°. In the meantime a reverse twist was produced 

 in the upper part, which continued to grow in length after the lower 

 part had stopped. In other words, homodromous torsion in the 

 lower part overpowered the younger and less rigid upper part 

 and produced antidromous torsion in it. 



If the frame was released before the internode had stopped grow- 

 ing in length, the upper part recovered to a certain extent from its 

 reverse twist. If this was not done until full growth was attained, 

 it recovered but little or not at all. It does not seem worth while 

 to give results in more detail at present, but it is planned to pursue 

 these points further by modifications of the method. 



A few experiments have been made with the first internode of 



the flowering bean when fastened in this apparatus. It is found to 



twist in the same direction as other internodes, but only to a limited 



extent (about 90° or less). Determinations of rigidity gave the 



same general results as for other internodes. Hop vines were 



studied to some extent. The indications are that the same general 



conclusions hold here also, but the experiments were not very 



complete. 



Discussion 



The question of rehability of measurements and of results is 

 first to be considered. The errors in measuring increases in length 

 are generally negligible compared with those in other measure- 

 ments. The first length, however, can be measured only to within 

 about one mm., and this error affects all other lengths, but to a 

 much less extent. Only the first values of rigidity and its coefficient 

 are very much affected by it. As already mentioned, when using 

 the mirror clamp, corresponding lengths can be measured with 

 only fair accuracy. A careful consideration of our readings of 

 the twist shows that these are generally accurate to within 2 or 3°. 

 Disturbing influences, such as air currents and imperfections in 

 apparatus, give at times greater errors than these, and account, in 

 part at least, for slight irregularities in the curves. Determinations 

 of the period of oscillation are generally accurate to about i or 2 per 

 cent. Since the square of this value is used in the calculations, this 

 introduces an error of about 2-4 per cent in the values for rigidity. 

 In the last part of each experiment care must be taken to have the 



