374 Prof. Harrison and Mr. S. K. Chakravarti on the 



were similar to that recently described in detail (loe. cit. 

 p. 13) except that additional precautions were taken to keep 

 the temperature constant, and some slight modifications were 

 adopted in treating the results. The temperature was varied 

 between 27° and 140° G. 



With regard to tbe accuracy of the experiments, it may be 

 useful to give a summary of the estimated errors, though for 

 a detailed discussion reference may be made to the previous 

 paper already referred to. 



As described in that paper, the wire was mounted hori- 

 zontally, and was stretched by means of a calibrated spring- 

 in compression. The elongation under stress of the central 

 portion of the wire was measured by microscopes fitted with 

 micrometer eyepieces. 



1. The error on a single observation of the stretching force 

 is about 20 grams weight. 



2. Allowing for the estimated error in setting the cross- 

 wires on the specimen, the error of a single observation of 

 the stretch for a given load is about 0*00019 cm. 



3. Tbe length of the portion of wire whose stretch is 

 determined is 20 cm. The error on the measurement of this 

 length is negligible. 



4. The mean of 50 readings of the diameter of the wire 

 gives a probable error of 0' 00002 5 cm, on the radius. 



The error on the radius due to expansion of the wire when 

 heated through 100° is about 0*00003 cm. 



5. The temperature of the electrically heated wire was 

 measured by observing the resistance of its central portion r, 

 the apparatus was so adjusted that a length of about 1 cm. 

 on the potentiometer bridge wire was equivalent to 1° C. 

 change of temperature in the wire. 



During any one determination of the modulus the tempe- 

 rature of the wire was kept constant to o, 3 C; that is to say, 

 the balance point did not vary by more than 3 mm. This 

 limit of temperature steadiness is necessary in order that the 

 effect of expansion by heat may not vitiate the observed 

 stretches by more than half the " setting " error, that is to 

 say, by more than 0*0001 cm. When the temperature 

 changed by more than 0°*3 C. during any series of obser- 

 vations, that particular series was rejected. 



The temperature coefficient of resistance* of the specimen of 

 iron used was 0*0039 ohm/ deg. over the range of temperature 

 employed, and was subject to a temperature error of about 1°. 

 The resistance at 0° of the length of wire between the 

 potential leads was 0*1 ohm ; consequently the change in 



* Detevmir.t'd by a separate series of measurements. 



