of the Fishery Board for Scotland. 



195 



fore, that I am well within the mark in assuming 0*05 mgr. as the maxi- 

 mum error of any single weighing. 



The temperature of the balance-room and the height of the barometer 

 were regularly observed. I made no hygrometric observations, but read- 

 ings of the wet and dry bulb thermometer, taken within a mile of the 

 balance-room, by Mr Blackwood, were kindly supplied to me. In order 

 that the humidity of the air inside the room might be as nearly as possible 

 the same as that outside, the door and windows were kept open for some 

 time previous to weighing. 



I am well aware that the corrections for humidity thus arrived at are 

 but roughly approximate, but I thought it better to derive them in this 

 manner rather than to apply a constant correction. 



A very considerable error in the estimated amount of moisture in the 

 air does not affect the result in the fifth decimal place. 



From the data thus obtained the density of the air was calculated, and 

 the corresponding corrections applied to all the weighings. 



Discussion of Results. 



The degree of accuracy to which I have attained by this method will be 

 best seen by an inspection of Table II., in which I have collected all the 

 necessary data. After the above description of the mode of working the 

 table will speak for itself, but a short discussion of the results may be 

 useful. 



Sprengel, in a paper describing this form of pyknometer, gives the 

 results obtained with one of them in two sets of three weighings of water 

 at 15° C. and 16^ C. respectively, the maximum difference between any 

 two weighings at the same temperature being 0"1 m.gr. Though I have 

 often in individual cases obtained results fully equal to this, I have not 

 been able to attain to this degree of accuracy throughout. Sprengel's 

 weighings being, I suppose, all made within a short time, were not 

 reduced to vacuo ; and I at first thought that, as a considerable lapse of 

 time was unavoidable between my weighings, the neglect of this reduction 

 to vacuo might account for the discrepancy. This hope, however, was only 

 very partially fulfilled. No doubt in some cases the improvement was 

 very marked, yet in others the effect of the correction was to increase the 

 differences considerably. If the mean be taken of all the observed differ- 

 ences between the several duplicate weighings, both of distilled water and of 

 the samples of sea water, it will be found to be 0*22 m.gr. as against 0*18 

 m.gr. in the case of the weighings after reduction to vacuo. The gain 

 over the whole is therefore slight but still unmistakeable ; and in view of 

 the very considerable influence of this correction upon individual results, 

 it is evident that it would not be safe to neglect it. 



Sprengel points out in his paper a source of error, viz., the inconstancy, 

 even to the extent of 0*1° C, in the temperature of the bath. It is pos- 

 sible that, in spite of the care I took to insure perfect constancy of tem- 

 perature, that this may have had an appreciable effect, and that a more 

 complicated mechanical arrangement for maintaining constancy of tempera- 

 ture would result in a marked improvement, but I am inclined to doubt 

 this. 



A comparison of the results (corrected to vacuo) obtained with the 

 different tubes throws a good deal of light on the question. 



