ON THR DRTEUMTNATTON OF Gr.AVTTY AT SEA. 551 



raent it is not impossible. If siiccessivp readings cannot be taken, an equal 

 number of maxima and minima should be measured. It would be prc^fer- 

 ablo to employ an assistant to read and record the dial indications, but if 

 the experiment is conducted in a refrigerator the presence of a second 

 observer must be avoided. Telephonic communication with an assistant 

 outside should be arranged. 



If the instrument records photographically the film may be studied nt 

 leisure, and only those portions chosen for measurement in which the 

 ]nimping is small and fairly regular. It is found simplest not to measure 

 each crest and hollow separately, but to set the wires in the eyepiece first 

 along the mean line of crests and then along the mean line of hollows ; 

 these can be judged with considerable accuracy. The mean of these two 

 readings is then taken as the level of the undisturbed mercury surface. 

 Unfortunately the photographic record involves difficulties, arising from 

 the density of the photographic image and from parallax. 



If the ship's motion is regular, and the barometer free from unsymme- 

 trical errors, it is impossible to improve upon the height of the barometer 

 as given by the mean of the lines of crests and of hollows. Hecke/, 

 however, finds that the Kew pattern bai'ometer gives unsymmetrical 

 pumping, but his experimental evidence is open to criticism. (See 0. E.) 

 When the ship's motion is irregular, the pumping is necessarily imsymme- 

 trical. It is doubtful if it is practicable to deal usefully with observations 

 made when the photographic trace shows the dissymmetry to be marked. 

 A prolonged comparison between the readings of the marine barometer 

 carried on board a ship straining at anchor in seas of all kinds, and a 

 standard barometer on a neighbouring pier or headland, might settle this 

 ])oint ; or it might be feasible to imitate the motion of a ship with the 

 aid of a lift oscillating about the floor containing a standard, barometer. 



Rolling and pitching produce pumping both liy adding to the vertical 

 acceleration of the point of support of the apparatus and by throwing 

 the barometer slightly out of the vertical position, an inevitable accom- 

 paniment of the motion, since friction at the suspension cannot be com- 

 pletely avoided. The former, which is the more important, is dependent 

 upon the positio'ii of the apparatus in the ship, and could be eliminated by 

 taking the mean of simultaneous readings of similar barometers placed 

 equidistant from the two axes, and on opposite sides of them. The 

 usefulness or otherwise of introducing linear terms to correct for the 

 various types of pumping is briefly discussed in the original Report. 



4. Temperature Correction. — Since an error of 0'1° introduces an un- 

 certainty in the value of gravity of 02 cms./sec.^, accurate measurement 

 of the stem temperature is essential. Inequalities of lagging ma}^ occasion 

 a temperature gi'adient which is difficult to allow for, and it would be 

 preferable to immerse the barometer in a well-stirred water bath. This 

 would obviate difficulties such as are occasioned by the approach of the 

 observer, which is especially troiiblesome if the observations are carried 

 out in a refrigerator, since there is usually a difference in the temperatun' 

 of the attached thermometer and that of the mercury in the stem of the 

 l)arometer. 



.'5. Other matters to be considered are the cajiacity correction, capillary 

 depression, pressure of mercury vapour, and tlie loading of the shii). 

 The consumption of fuel and food lightens the ship and may tilt the 

 apparatus if it is not suspended. 



