52 
PUTNAM. 
above the sea level is exactly compensated by the deficiency 
in density of the mass below as compared with sea water, but 
there is a lack of local compensation, and if the berg were 
not sufficiently rigid and strong it would bend up at the 
sides or split in the middle. Let gravity be measured at A 
on the surface of the water ; at B at the same level, but over 
the ice ; at C the average level of the surface of the ice, and 
at D higher than the average surface. Neglecting the fact 
that the different surfaces of the iceberg are not indefinitely 
extended, it is evident that if we reduce to sea level by allow¬ 
ing for the attraction of the ice lying above that level we will 
obtain at the three stations B, C, and D equal values for 
gravity, but values which are less than that at A, and that 
this difference will be a measure of the deficiency of density 
in the mass below sea level. If, on the other hand, we 
neglect the attraction term and reduce for elevation only, we 
will find as compared w T ith A that at B gravity is in defect, 
at C normal, and at D in excess, and the differences at B 
and D will be a measure of the lack of local compensation 
and will be equal in amount to the attraction of a horizontal 
plain whose thickness is the difference in elevation between 
the station and the average surface. B is overcompensated 
and D is undercompensated—a condition maintained by the 
rigidity of the ice. 
If we consider that these conditions apply to continental 
elevations, and that local irregularities in surface are not 
compensated by the general lack of density or other cause 
below sea level, we may then, following the idea of M. Faye, 
apply a further correction to the observed force of gravity 
at any point to reduce to the normal condition. This cor¬ 
rection may be expressed dg — 2g ~ which represents the 
attraction of an indefinitely extended horizontal plain of 
thickness h and density <5, and the correction is evidently posi¬ 
tive for stations below the average level and negative for 
those above. h = H x — H where H x is the average elevation 
and H the station elevation. This method has been tested by 
applying it to the stations most affected in the present series, 
