Radiation of Non-luminous Heat. 253 



'. Besides the necessity of making several experiments to obtain 

 a single result, it sometimes occurred that particular results re- 

 quired to be repeated for verification when apparent discrepan- 

 cies occurred ; this was done to ascertain if they were real or not. 



As it was obvious that the experiments must necessarily ex- 

 tend through a considerable time, during which the circumstan- 

 ces attending the cooling of the cylinders could not be expected 

 to remain uniform, a standard for comparison was provided, and 

 a cylinder of which the coating was not changed, and which was 

 observed in regular turn with the other cylinders. At first a 

 vessel without coating was used for this purpose, but as it was 

 found liable to tarnish, a cylinder was substituted having a coat- 

 ing of aurum musivura, which was one of the smoothest and 

 most uniform of the coloured substances used. The number 

 obtained on the different days from a mean of the trials made of 

 the cooling of the standard cylinder were applied to compare the 

 results of one day with those of another. This assumes that the 

 times of cooling of the different vessels would be affected propor- 

 tionally by a given change in the circumstances of the experi- 

 ment. This inability to preserve the circumstances constant is 

 the real objection to this method, and one which most affects the 

 certainty of the results.* 



The following example shews the application of this method. 

 The observed times of cooling of the standard cylinder, from 

 180° to 140° in two experiments on the 31st of October, were 

 969i and 968J seconds, mean 969. Three experiments on the 

 1st of November gave 898, 892, 8931 seconds, mean 894i. 

 Cylinder No. 4, coated with cochineal (crimson), gave for the 

 time of cooling from 180° to 140° on the 1st of November, 848 J. 

 To compare this with a result obtained with the same cylinder 

 on the 31st of October, we have 894^ : 969 : : 848^ : ^, the equi- 

 valent number for October 31st, 916.3 seconds. 



The results obtained with the same cylinder on different oc- 

 casions of experiment having been thus rendered comparable^ 

 the comparison of experiments with different cylinders was ef- 



• If the circumstances could be retained the same, three observations of 

 the temperature at equal known intervals would give a numerical expressioft 

 for the radiating power of the coating. 



VOL. XXI. NO. XLir.— OCTOBER 1836. S 



