292 PROCEEDINGS OF THE AMERICAN ACADEMY. 



In order to eliminate various sources of i^ossible error, including espe- 

 cially disagreement of the two thermometers, sets of observations were 

 made iu pairs, the stream entering at Ai being warmer than the other in 

 the first set of observations and colder than the other iu the following 

 set. The ditference of temperature was usually between 5° and 10° C. 



Some doubt was felt at first as to whether the ends of the bar would 

 have the same difference of temperature as the thermometers. It is evi- 

 dent that use of exits £Ji and j^o, whereby the streams were made to flow 

 a considerable distance along the bar, would be more effective than the 

 use of the dotted exits. The latter were used upon occasion, with the 

 idea that, if they gave about the same effect as £i and SJ2, the latter 

 could be regarded as satisfactory. The dotted exits gave a result some 

 four or five per cent less than that given by the other exits. It ap- 

 peared unlikely that any considerable error would be made in assuming 

 that, when J^i and iJ.. were used, the difference of temperature of the 

 ends of the bar was the same as the difference of temperature of the 

 thermometers. 



It will be observed, however, that the iron bar used in this test was 

 like a piece cut parallel to a certain diameter of tlie disk, not parallel to 

 the thickness of the disk. It was a matter of very grave doubt whether 

 the thermo-electric quality of the bar, taken lengthwise, could be re- 

 garded as identical with the thermo-electric rpuxlity of the disk taken 

 thickness-wise. It was the latter quality that came into play in the con- 

 ductivity experiment^ ; and some way of determining it was to be found. 

 The method which was finally adopted is described in Appendix I. of 

 this paper. It showed that scarcely an appreciable error would have 

 been made by using the results obtained from the first method, above 

 described. 



With the information thus obtained it was easy to calculate with con- 

 siderable accuracy the difference ii» temperature of the two faces of the 

 iron disk iu the conductivity experiments, the indications of the astatic 

 galvanometer being readily interpreted. The deflections of the galva- 

 nometer, upon reversal of the current, were usually about 9 cm. ; and 

 the difference of temperature of the faces of tlie iron was usually rather 

 more than 1° C. The difference of temperature of the streams, above 

 and below the disk, was usually about 8° C. 



