OP ARTS AND SCIENCES. 105 



gauges would show its largest discrepancy at zero atmospheres follow- 

 ing a high pressure. Change of temperature of the water jacket (uu, 

 Fig. 4), or change of the water pressure, cuuld not give rise to persist- 

 ent cyclic curves, but would show itself in a displaced fiducial zero. 



I think the clue for the interpretation of the apparent hysteresis 

 is the observation that data of the '' on " series lie on straiglit lines, 

 whereas the locus of the data belonging to the " off" series is curvi- 

 linear. The cycles have the form of an archer's bow. Hence, bearing 

 in mind that the liquid in tt is being heated by compression of the oil 

 in the gauge during the " on " series, and that, even though gradually 

 parting with its heat increment, it is nevertheless continual!^ hotter 

 than the surrounding medium ; bearing in mind, moreover, that in the 

 "off" series the water in it is cooled by expansion of the oil in the 

 steel gauge, but that the water is hotter than, the surrounding medium 

 only during the high pressure stages, and colder than the surrounding 

 medium during the low pressure stages (heat having been continually 

 dissipated), — an explanation of the bow-shaped cycles is suggested. 

 Even my last and best Tait gauges, considered as thermometers, are 

 unfortunately very sensitive instruments, showing displacements of 1 

 or 2 centimeters per degree Centigrade. In Table IV. the cycles are 

 reduced in width, the distance apart being equivalent to only 17 atmos- 

 pheres ; but the error in question here occurs differentially. 



A bow-shaped cycle may also be conceived to result as a viscous 

 phenomenon, if a strain (set) were gradually impressed on the gauges 

 during the " on " series, which strain would then assert itself during 

 the " off " series. Without stopping to examine this consideration, 

 I believe the former explanation more probable. 



It follows from the present inquiry as a whole, that, to obtain thor- 

 oughly trustworthy data, pressure observations should be recorded 

 during the pressure increasing phase of the work. In most experi- 

 ments this condition is easily fulfilled, since the pressure datum sought 

 is usually reached from zero. Finally, the steel gauge tube must be 

 made with thinner walls than the above. 



20. Tait Gauge. — Volume Increase measured and computed. — 

 From the dimensions A^, A^, L, and p, given in Table I., it follows 

 that the increase of the external volume of the cylindrical steel tube 

 of the gauge is 47 / 10® per unit of external volume, per centimeter of 

 displacement of the thread of the capillary tube ss (Fig. 4). This 

 corresponds to a volume increment 



y/ F= .00000080 (1) 



per unit of external volume, per atmosphere. 



