264 



VARIATIONS OF SPECIFIC GRAVITY. 



peated experiments with each of the three samples at each stage of 

 its treatment. These measurements were made without exception by 

 my colleague, Mr. F. Sturm, and the mean error of the 69 deter- 

 minations has the extremely satisfactory value of only 0.0004, 



Table I. — Tartied cylinders (/) and rolled and luwimered cylinders {II). 



The.se numbers confirm my earlier statement, and show that even 

 with such a favorable material as platinum the results are strongly 

 affected by the individuality of the sample examined. Samples cut 

 from immediate proximity otF the same bar show a difference of 

 density of 0.1734. 



By rolling and hammering the density is increased l)y, roughly, 1 

 per cent, and the difference of density between the three samples is 

 at the same time reduced to about one-tenth its former magnitude. 

 There is then nothing new in Table I, wdiich merely proves the pres- 

 ence of pores or faults, which are diminished by the application of 

 mechanical force. 



Table II is, however, more interesting: 



Table II. — Rolled and hammered samples {II) and 1 mm. cold-drawn ivire {III). 



Sample 1, which had tlic highest density as hammered platinum, 

 .showed a diminution in s})eciHc gravity of 0.0178, which is a differ- 

 ence far beyond the mean error of measurement (0.0004) already 

 given. Sam])k's '2 and ->, wliich had been somewhat lower in density 

 than No. 1, advanced somewliat, but, as we shall see, it is probable 

 that behind this apparent increase a decrease is really hidden, and 

 that sampk's 2 and )> also would have showed a falling off had they 



