REPORT ON THE PRESSURE ERRORS OF THE THERMOMETERS. 41 



These numbers are not given as exact, though they are probably very near the truth. I have noticed 

 that the scale error, in a thermometer with two liquids in contact, varies within considerable limits at 

 auy one temperature, according as the thermometer has been raised to that temperature from a lower 

 one, or cooled down to it from a higher. I found an excellent illustration of this in some of my glass 

 pressure-gauges, where (for the purpose of allowing the interior plug to be seen) I at first employed a 

 transparent liquid in the bulb, with a short column of mercury in the stem to move the index. In some 

 of these instruments, after they had been several times exposed to high pressures, a film of the trans- 

 parent liquid entirely surrounded the column of mercury, which could then move pretty freely, even 

 in the narrow tube, under the action of gravity. Of course this mode of construction was at once 

 given up. 



There seems to be no necessity for the printing of the records of the very numerous experiments 

 which have been made on the various thermometers. In the text above, I have said enough to show 

 that the true pressure correction to be applied to the deep sea observations is exceedingly small, and 

 in column 11 of the annexed table it is calculated with all necessary accuracy. I have already said 

 that no fair comparison can be drawn between the numbers in columns 5 and 6. There is a general 

 resemblance between them, and that is all that could be expected where the modes of obtaining them 

 were so different. 



As regards the numbers in column 7 of the table, the remainder, when the corresponding number 

 for the maximum index in column 10 is subtracted, ought to be nearly the same for all the thermo- 

 meters if the vulcanite supports and cover were similarly applied to each. The differences among them 

 are mainly due to this cause, and it is somewhat surprising to find that they are so nearly alike. 



I have so often mentioned Amagat's determinations of the volume of air at different pressures, as 

 the basis of the whole of my measurements, that it is well to give, as I have done in fig. 3 of the plate, 

 a graphic representation of them. The horizontal axis gives pressure of air in atmospheres, the ver- 

 tical gives the corresponding densities, or (what comes to the same thing) the pressures calculated 

 from the densities by assuming the truth of Boyle's law. It will be seen that the straight line, which 

 would represent densities in terms of the actual pressures, if Boyle's law were true, lies Mow the 

 curve at first : i.e., air is more compressible at first than Boyle's law would make it At about a ton, 

 (or rather 140 atmospheres,) its volume is exactly that which Boyle's law would give ; and at higher 

 pressures its compressibility falls farther and farther short of that assigned by the law. But the error 

 caused by assuming Boyle's law to hold good up to one ton pressure is, at its greatest, only about 1 

 per cent. ; and this occurs considerably under 100 atmospheres. Practically, my gauge unit was 

 determined at pressures at which Boyle's law is almost exactly true. 



Finally, it may be interesting to mention that a fairly approximate determination of the com- 

 pressibility of water was made by counting the number of strokes of the pump required to produce 

 a measured pressure in the interior of the large apparatus. 



F 



