366 MR T. C. BAILLIE ON THE 



to. The thermometer was observed through the telescope of a cathetometer, and the 

 time at which the top of the mercury column passed each degree division mark was 

 noted by looking at that instant at the dial of a watch. After some practice it was 

 found easy to note the time of such transits to within a couple of seconds from the 

 position of the seconds' hand, without paying much attention to the divisions round the 

 dial. The time of transit set down for each degree division was late by the time taken 

 to look from the thermometer to the watch, but as this is small and affects each reading, 

 it is of no consequence. When the cooling became comparatively slow, as it did below 

 100° C, it was possible to see the top of the mercury column disappear behind a degree 

 division mark, note the time, and have the eye in position at the telescope again in time 

 to see the top of the column reappear on the under edge of the division mark. 



§ 5. Reduction of the Cooling Readings. — The method employed for reducing 

 these readings was as follows : — On paper ruled in squares temperatures were plotte. 

 abscissae, and the times (in seconds) taken by the bar to cool through one degree were 

 plotted as ordinates corresponding to the mean temperatures for those degrees : thus, 

 for example, the ordinate corresponding to the temperature 102*5° C. was the observed 

 time taken by the bar to cool from 103° C. to 102° C. The advantage of this method 

 of reduction is the simplicity of correcting for errors of observation, &c. Suppose, for 

 example, that the time set down for the transit across the 175 degree division mark is 

 too late, the time noted for cooling from 176° to 175° is too great, and the amount by 

 which it is unduly increased is deducted from the time of cooling from 175° to 174°; 

 but the average time of cooling for a range including 176° to 174° is not affected by 

 the supposed error. An error in graduation, by which one of the division marks is 

 displaced, produces a similar effect. The ordinates would, if there were no errors of 

 any kind, increase in length continuously as the temperature diminishes. If the curve 

 formed by the ends of the ordinates is not continuous, all that is necessary is to make 

 a continuous curve by reducing the lengths of those ordinates which are obviously too 

 long and increasing the lengths of adjacent ordinates, and vice versa, so as to keep the 

 sum total of the lengths of all the ordinates constant. This treatment will get rid of 

 the effects of errors such as those considered above. The way in which this was carried 

 out was to form a new curve in which for each reading was substituted the average of 

 the five nearest readings. This gave a curve which was smooth but with small "waves" 

 along it. A mean curve was then drawn by means of a lath planed thinner towards one 

 end so as to produce the necessary variation of curvature along it. The ordinate at any 

 temperature of the curve so constructed is the reciprocal of the rate of cooling at that 

 temperature. 



A cooling experiment was done alongside of the statical experiment on several days, , 

 until the surface of the cooling bar was thought to be just perceptibly dimmer than that ' 

 of the long bar. It was confidently expected that the repeated heating of the short bar, j 

 especially as "sweating" was not entirely avoided, would affect the surface and increase 

 its emissivity. The readings taken show that each time the bar was heated its enns- 



