of the Heat-conducting Power' of Mercury. 489 



perceived; but, with the irregular differences of the numbers, 

 constancy also is not sufficiently apparent, although it seems more 

 probable. We even see that in this method the errors of observa- 

 tion are greater than are permitted for the decision of so difficult a 

 question. The reason for this is perhaps essentially the employ- 

 ment of an outflow-thermometer. Even the discordant fillings of 

 the tube indicate the uncertainty which is here presented. And 

 then, also, the outflow, notwithstanding the very narrow aper- 

 ture, did not always follow in the same manner ; and so the 

 determination of the principal decisive temperature m could not 

 be effected with the requisite exactness. 



§4. 



All these inconveniences would be avoided with a thermometer 

 of the usual construction, the production of which, however, is 

 more troublesome. I therefore now carried the experiment 

 further with such a thermometer, which I constructed so that a 

 capillary thread which issued at about a third of the length 

 of the tube from the bottom, after a bend, was carried up- 

 wards, and then at the height of the upper edge of the tube, 

 after another bend, ran horizontally past a long scale. When 

 the mercury stood in this horizontal part of the thread, there 

 was always present, reckoning the capillary force, a slight excess 

 of pressure towards the upper end of the tube and the thermo- 

 meter T, which made it impossible for air to penetrate inwards 

 through the fitting-in of the latter. Conversely, in order that 

 the thermometer might not retrograde through the overpressure, 

 and thereby a variable volume of the interior of the tube be ob- 

 tained, it was held firmly pressed into its fitting by a light stay 

 which could be tightened by means of screws and proceeded 

 from the closure of the tube. The aperture in this latter was 

 of course stopped. At the same time I had shortened the tube, 

 since a calculation of the previous experiments had shown that, 

 in each of them, a very near approximation to the surrounding 

 temperature prevailed in the tube long before reaching the 

 lower end. 



The tube, thus converted into a mercury-thermometer with a 

 large vessel, was now, in a bath, carefully calibrated as to its 

 temperature-indications; and it resulted that, on the average, 

 16 millims. on the scale was equal to one degree. Thus it was 

 made possible to read off the temperatures m rigorously to frac- 

 tions of a tenth of a degree ; and thereby the most decisive part 

 of the apparatus was present in very fine execution. That this 

 large thermometer did not undergo any gradual alteration du- 

 ring the experiments was shown by the circumstance that when, 

 after a day of experiments, the whole thermometer had during 



