REPORT ON THE PRESSURE ERRORS OF THE THERMOMETERS. 



9 



is heated by compression if it is at a temperature higher, and cooled if it is at a 

 temperature lower, than that of the maximum density. One set of Captain Davis' 

 observations were made in water at temperatures near, but under, the maximum density 

 point : in which, therefore, very little effect can be produced, even by very great pressure 

 (and that little should be cooling, not heating), and he combined these with a number of 

 other observations made at temperatures approaching 55 F., in which a comparatively 

 large amount of heating is produced even by moderate pressures. The average of the 

 results of these determinations was taken, but, unfortunately, Captain Davis struck out 

 before taking the average all those observations which appeared to give much larger 

 effects than the others, taking them as being obviously erroneous. 



When we sift out from the observations all those made nearly at any one temperature 

 we find they agree fairly enough with the theoretical result of the compression. But 

 observations made at different temperatures were included in the group from which the 

 average effect was deduced. Such an average has no physical meaning. 



As this is a point of some importance, I shall give a graphic representation of one set 

 of the observations, those made with one of Sir William Thomson's thermometers ; 



J.'F 



O 



Fms 



500 



'00 



asoo 



showing which were rejected, the average thus obtained, what ought to have been 

 obtained, and also the strict theoretical result. In the diagram above, pressures are 

 measured in fathoms of sea water along the horizontal line, and the corresponding 

 changes of temperature, shown by one of the completely protected thermometers, are 

 represented by vertical lines. The centres of each of the series of white and black spots 

 inserted in the diagram represent the various observations made by Captain Davis and 

 Professor Miller at temperatures near 55 F. 



If we suppose that all these observations had been made in precisely similar circum- 

 stances, a fairly approximate way to get the average effect indicated would have been to 

 draw a line through the series of spots in such a way that the average distance from it of 



B 



