Ox THE GEISSLER THERMOMETEKS 483 



and the column so short that the capillary action is sufficient to prevent 

 the fall. Now I think that a thermometer in which there is this layer 

 of air around the mercury in the bulb must be uncertain in its action; 

 hence my opinion is unaltered that all thermometers in which we can- 

 not remove this layer or at least make certain of its absence should be 

 rejected. 



Furthermore, with respect to calibration, the reservoir is not essen- 

 tial to the calibration of thermometers whose range is and 100 C. 

 But my remarks apply better to those whose range is between and 

 30 C. or 40 C. Here calibration is impossible with a short column 

 at ordinary temperatures unless some of the mercury can be stored up 

 in the reservoir so as to allow the column to move over the whole scale. 

 And it is within this limit that thermometers are of the greatest value 

 in the physical laboratory. 



The other defects of the Geissler thermometer, the scale which was 

 always coming loose, the metal cap which was never tight and always 

 allowe'd water to enter, the small capillary tube which wandered with 

 perfect irregularity from side to side over the scale, all these were so 

 obvious that I confined my remarks to the more obscure errors. 



Furthermore, I believe there is some error in most Geissler ther- 

 mometers from the small size of the bulb and the capillary tube, and 

 this I have mentioned on p. 124 ' of the paper referred to. Pfaundler 

 and Platter, in a paper on the specific heat of water, in Poggendorff's 

 Annalen for 1870, found an immense variation within small limits. In 

 a subsequent paper 2 the authors traced this'error to the lagging of the 

 thermometer behind its true reading. 



The authors used Geissler thermometers graduated to ^j- C. ! in a 

 series of experiments made by plunging the thermometer into water 

 after slightly heating or cooling the thermometer so that in one case 

 the mercury fell and the other rose to the required point. When the 

 thermometer fell about 6 or 8 C. it lagged behind 0-0654 and when 

 it rose 3 or 4 it lagged 0-022, making a difference of 0-087 C.! Now 

 my thermometers made by Baudin show no effect of this kind. They 

 indicate accurately the temperature whether they rise or fall to the 

 given point, provided the interval is not too great. The fact then 

 remains that a Geissler thermometer graduated to 7 V C. may be uncer- 

 tain to 0-087C., while a Baudin graduated to mm., one mm. being 

 from T V to T V C. is not uncertain to 0-01 or 0-02 C. May not the 



1 [p. 393 this volume.! * Poggendorff's Annalen, cxli, p. 537. 



