Viscosity of Solids and its Physical Verification. 205 



twists, were reckoned alternately positive and negative, con- 

 formably with the sign of the deformation ((/> — $')/ T - In this 

 oscillatory march (time as abscissa), since each deformation 

 (ordinate) now begins where the preceding deformation 

 ceased, a continuous series of open cycles is necessarily 

 generated. The positions of these cycles shift at a gradually 

 retarded rate towards a final, very flat cycle, which for the 

 constant values of time and stress is fixed in position and 

 closed. 



Cycles here, closed or not, are expressions of the fact that 

 the " past histories " (in Maxwell's words) of the molecular 

 configurations in the " stress positive " and " stress negative " 

 phase of ench cycle are not the same. Shifting is brought 

 about by permanent molecular break-up, the amount of which 

 gradually vanishes. In the ultimate and fixed cycle as many 

 configurations are broken during the " stress positive " 

 as are reconstructed in the " stress negative " phase, though 

 they need not be the same configurations. 



These considerations suggest a comparison between " accom- 

 modation " and Prof. E wing's " hysteresis," * for the purpose 

 of detecting the extent to which like causes are discernible 

 in each phenomenon. Both exhibit a static character. But 

 such a comparison, to be fruitful, calls for direct experiments ; 

 for instantaneous values of stress and viscosity must be co- 

 ordinated. 



9. Having thus discussed one phase of the results in 

 Table L, I will pass to Table II., which is a digest of the mean 

 values of Table L, in so far as such a digest can be made. 

 Following the scheme at the end of the preceding paragraph, 

 this comparison should be made after an infinite number of 

 twists have been imparted to each wire. In such a case, 

 however, the original number of unstable configurations has 

 been seriously reduced, so that, apart from the inconvenience 

 of a time-consuming method like this, the original properties 

 of the wire are not clearly present in the results. In wires 

 perfectly free from strain at the outset, the first twist leads to 

 better indications of the viscous quality. As this condition 

 was not always guaranteed for the wires of this paper, I 

 have accepted the mean viscous behaviour during the first 

 and second twists as the best available index for comparison. 

 It is sufficient, at least, for the present purposes. Again, 

 raking the mean for rods of the same nominal temper, I 

 obtain the data from which figure 1 is constructed. Mean 

 viscous deformation ((/> — </>')/ T varying with time is here 



* Cf. Phil. Trans, ii. 1885, p. 523 ; ibid. ii. 1886, p. 301. Prof. Ewing-'s 

 earlier pap ers are there mentioned. 



