432 DR W. PEDDIE ON 



the subjection of the wire to a comparatively small number of full oscillations (given 

 in brackets in Table V.) before an experiment was made, reduced n and b to values 

 like those which were obtained in the first series. This was the case even when y was 

 comparatively small — see experiment 12.11.97 (1). 



The most important object of the present series of experiments was to determine 

 whether or not, under different initial conditions, points representing simultaneous 

 values of log b and n still practically lay upon straight lines passing through the 

 point (2 "3, 1). This was found to be the case. At first the slope of the line was 

 found to be positive, as it was in the experiments described in the First Paper. The 

 slope of the line increased, under increased fatigue, until it became practically vertical. 

 The wire was very sensitive to variations of fatigue, whether due to magnitude of 

 initial range or to repeated oscillations. Increased fatigue causes an increase of n 

 and a diminution of b : see, for example, experiments 11.11.97 (1) and (2); experi- 

 ments 16.11.97 (1) and (2) ; and experiments 17.11.97 (1), (2) and (3). 



Fig. 13 represents a number of the results graphically. The group of three points 

 marked thus O corresponds to the first three experiments. The group marked x 

 corresponds to the next nine experiments ; those marked H correspond to the next 

 ten ; those marked v correspond to succeeding experiments in which fatigue was 

 large ; and those marked by single points correspond to some of the experiments in 

 which fatigue was small. It is evident that the various groups throughout each of 

 which fatigue was fairly constant are collected in the neighbourhood of straight lines 

 passing through the point (2 '3, 1). Variations may be due to slight differences of 

 condition as to fatigue or to the fact that a is always choseu as a whole number, while 

 the most suitable value may lie between two consecutive whole numbers. If, in any 

 case in which a is small, an error of unity were made in the value of a, the correspond- 

 ing value of n would change by 0'06 or 0'07, while the value of log b would only 

 change by about 0*015 or 0*02. As an error of unity, when a is small, is impossible, it 

 is evident that the grouping of the points round the lines cannot be regarded as 

 accidental. 



It therefore appears that the Oscillation Constant, A, is truly a constant throughout 

 all the treatment to which the wire has been subjected. 



Recovery from Fatigue. 



The data given, Table V., show that the wire recovers partially from the effect of 

 fatigue with considerable rapidity. Compare, for example, the data for the experiments 

 16.11.97 (2) and 25.11.97. This is most marked in the case of small oscillations — see 

 12.11.97 (1) and 17.11.97 (1), the former experiment being made immediately after 

 heavy fatigue, while the latter was made one day after heavy fatigue. 



There is another fact which may possibly bear on the question. In some of the 



