ANOMALOUS BEHAVIOUR OF DELICATE BALANCES, ETC. 411 



The small undulations which occur in the first part of this curve cannot be due to 

 flexure of the beam, for it was found that the sensibility of this particular balance 

 was precisely the same for all loads. Neither could they have been caused by 

 fluctuations in the temperature, for those would have been detected by the bolometer. 



The observed variations therefore probably owe their origin, as we have already 

 suggested, to some slight relative movement of the agate knife-edges consequent upon 

 the loading of the beam. The correctness of this view may also be inferred from the 

 fact that when the above experiment is repeated with the balance duly fatigued, the 

 obtained R.P. curve is a right line wholly free from undulations. 



We may therefore conclude that the slight periodic variations which the bolometer 

 has shown to exist in the temperature within the usual balance case, are responsible 

 for the more or less harmonic motions recorded in the R.P. curve ; and it appears very 

 probable that these effects must always be detectable in a sensitive balance, the beam 

 of which is not guarded in some such manner as we have indicated. 



Of the Major Undulations to be Found in a Resting-Point Curve. 



It may be observed that when an experienced worker is using a balance, he places 

 the required weights upon the pan at approximately equal intervals of time. As each 

 weight is introduced the currents of warm air rising from the observer's hand will 

 envelope the end of the beam above ; these currents will produce heat-waves, roughly 

 harmonic, within the beam. 



The magnitude of these temporary heat-waves will be considerable when compared 

 with the magnitude of those which are usually flowing along the beam. As soon as 

 the loading is ended and the shutter dropped, the larger waves will begin to die away, 

 and after a short time disappear. But whilst they remain their existence is proved 

 and their influence indicated, by the exceptionally large undulations which appear in 

 the majority of the R.P. curves drawn from data obtained immediately after the 

 loading of the beam has been completed. 



It may be noted in passing that the first portions of such R.P. curves which we 

 have often obtained, forcibly remind one of the graphs representing the results of the 

 experiments of DESPRETZ, FORBES, and others, upon the heat conductivity of uniform 

 bars, and it is not improbable that further investigation would show the curves to be 

 logarithmic in form. 



Experience has taught us that the particular source of error in weighing which we 

 are now discussing, is sufficiently guarded against by allowing a period of about 

 10 minutes to elapse between the completion of the loading and the actual weighing. 

 And. if the balance is fatigued during this same time, we also avoid the error which 

 appears to be due to the slight initial movements in the several groups of adjusting 

 screws. These precautions are now almost invariably taken by observers who are 

 here engaged in researches for which very accurate weighing is essential. 



3 G 2 



