16 Mr. J. P. Gassiot on Observations [June 20^ 



far as this cause of variation is concerned, bj tlie ordinates of the 

 straight line (dotted), fig. 2 ; and accordingly Mr. Stewart, assuming the 

 uniformity of instrumental change, takes the difference of ordinates of 

 the dotted and curve lines (fig. 2) as a quantity unexplained by known 

 causes. This quantity follows very w^ell the march of the latitude, 

 or of the temperature which marches ivith the latitude. It would be 

 removed by supposing the temperature-correction applied to be too 

 great by about 



0-30 or 0*33 for 20°. 



" Such an error in the temperature-correction can hardly be said to 

 be too great to be admitted. Still it is greater than we should expect 

 if the London and Kew determinations apply to the state of things on 

 hoard ship, 



" I do not think Mr. Browning's conjecture that the progressive 

 change was due to an alteration in the glass of the prisms probable. 

 I should think it more probably due to a slow release from a state of 

 constraint in which the instrument was left by the boltings, &c. If so, 

 I should expect that the instrument would shake itself down to a 

 permanent state — that the instrumental change would be more rapid 

 in the early than in the later part of the voyage. 



" In fig. 2 I have drawn a smooth curve by the eye following generally 

 the irregular curve, with a view to clearing in some degree the observa- 



Pig. 2. 



tions from casual errors. The curve ought perhaps to lie a little lower 

 on the right, but I was anxious to lean rather towards Mr. Stewart's view 

 than the reverse. The smooth curve of fig. 2, lifted so as to cut the 

 axis on the extreme right, is transferred to fig. 3. If we suppose the 

 instrumental change more rapid at first than afterwards, the correction 

 thence arising will be represented by the ordinate, not of a straight 

 line as in fig. 2, but of some such concave (upwards) curve as the dotted 

 curve of fig. 3, and the unexplained residue will be reduced to the ordi- 

 nates intercepted between the curves of fig. 3. If the curves are some- 

 what as I have drawn them, this residue will follow very well the march 



