410 



PROCEEDINGS OF THE AMERICAN ACADEMY 



perlments in Table III. the pillars were cut out of the fragments 

 formed in the preceding experiments in the same table. The ends of 

 all of them were turned flat and pei'pendicular to the axis. It will be 

 seen that the breaking weights by experiment vary from about thirty- 

 three per cent greater to about twenty-three per cent less than the 

 breaking weights computed by formulas (1) and (2), the mean being 

 about five per cent greater. 



Table III. 



In his second paper, Hodgkinson also describes a few experiments 

 on square pillars and on pillars with the section of an equilateral tri- 

 angle ; these were compared with cylindrical pillars ; two of each pat- 

 tern were tested, all being cast from the same kind of iron, and were 

 of the same length and nearly the same weights. The relative 

 strengths, reduced to the same weights, were as follows : — 



Cylindrical pillar, 100. 

 Square " i)3. 



Triangular " 110. 



