348 PROCEEDINGS OF THE AMERICAN ACADEMY. 



served to admit the thermometer (®) and the stirrer. The tempera- 

 ture in the outer very large tube fluctuated but very slightly, and that 

 in the inner tube containing the substance was almost exactly con- 

 stant. There was no difficulty in regulating the heat of the water in 

 the thermostat to within less than one tenth of a degree by an ordinary 

 gas regulator. For reading the thermometer (0), a telescope with a 

 very exact micrometer was used, by means of which the smallest scale 

 divisions could easily be divided into hundredths. The danger of irreg- 

 ular readings of the thermometer through the various media, which 

 might cause errors due to parallax, was wholly overcome, in that on the 

 one hand all the glass walls were arranged as vertically as possible, 

 and the telescope was made exactly horizontal, and on the other hand 

 every reading of the thermometer was made both from before and from 

 behind. Obviously, the mean of these two readings must represent the 

 true value, even if a slight displacement due to refraction had been 

 present. The thermometer was so arranged that it could easily be 

 turned on a vertical axis, so that there was no difficulty in making 

 these readings. As a matter of fact, the readings before and behind 

 never differed more than four thousandths of a degree, and usually 

 differed much less than that. The true value was always taken as the 

 mean of these readings. In the case of the Beckmann thermometer, 

 the telescope was so placed that the scale division lines appeared 

 perfectly straight through the tube, without a trace of bending. 



The concordance of the results furnishes yet another proof that these 

 methods of reading were entirely satisfactory and thoroughly trust- 

 worthy. The great advantage of this apparatus is that the tempera- 

 ture of the scale can be kept indefinitely at a temperature as nearly as 

 possible to the true value, and this is no small advantage, because with 

 such a length of thread a single tenth of a degree difference of tempera- 

 ture causes a thread -correction of tuo o°- We conclusively proved that 

 it was not possible to attain the necessary constancy if even a milli- 

 meter of the mercury thread projected beyond the thermostat into the 

 temperature of the room. 



As has been said, in Table I the accurate results with the first ther- 

 mometer and the Beckmann are given, and also the corrections, in so far 

 as these could be determined. The final determinations with the large 

 new thermometer are given in Table II. On the basis of these results, 

 we think it is safe to say that the transition temperature of manganous 

 chloride for the transition from the crystal form with 4 molecules of 

 water into that with 2 of water, has a value 58.089° (±0.005) referred 

 to the international hydrogen scale. 



In conclusion, it is a great pleasure to express our thanks to the 



