i6o JOSEPH BARRELL 



The test by adjacent stations at different elevations. — 'There is, 

 however, another way of using the data given for stations situated 

 well above and below the mean elevation of mountainous regions. 

 If a pair of stations be taken close together, one far above the mean 

 elevation, the other far below, they will presumably, because of 

 their juxtaposition, be affected in much the same way by the errors 

 incident to the hypothesis of uniform compensation through a depth 

 of 114 km., with complete compensation at that depth. In order 

 that good results may be obtained, however, the specific gravity of 

 the local rocks should be carefully determined in order to have a 

 correction for the mass between the stations. The parts of the 

 anomalies due to the irregularities and incompleteness of compensa- 

 tion will ordinarily have the same sign and be of nearly the same 

 value at the adjacent stations. This is indicated by the contour 

 lines of Fig. 5, which show that in the same region the anomalies 

 are of sufficiently regular gradation in magnitude to make the 

 drawing of contour lines possible. The parts of the anomalies at 

 the high and low stations due to errors in the hypothesis of local or 

 regional compensation will, however, be of opposite sign. If, then, 

 the algebraic difference of the anomahes for such a pair of stations 

 be computed for successive hypotheses of broader regional com- 

 pensation, the part of the anomalies due to vertical imperfection of 

 the hypothesis will be largely eliminated. The algebraic difference 

 measures the horizontal imperfection of the hypothesis. That 

 hypothesis is favored whose assumed radius of regional compensa- 

 tion gives a minimum value to this algebraic difference. This test 

 may be made by combining data given on p. ico, Hayford and 

 Bowie, with p. 15, Bowie; although, because of incompleteness of 

 the tables, this combination gives the data for only a few of the 

 properly situated mountain stations. The best couple of stations 

 for the application of this test consist of 42, Colorado Springs, and 

 43 Pikes Peak. Somewhat more distant stations — ^44, Denver, and 

 45, Gunnison — -may also be added to the group. The tabulation is 

 shown on p. 161 (Table IV). 



It is seen that for three of the four Colorado stations the absolute 

 value of the anomaly is least with regional compensation to 166.7 

 km. For the fourth station it remains practically constant for 



