OVERHANG AT BARBERS HILL, TEXAS 29 
A long series of combinations of these geologically probable as- 
sumptions are possible and have to be calculated directly or indirectly. 
For many such sets of assumptions, it was unnecessary to carry 
the calculations through to the point of finding the mean square. If 
the unsquared differences are plainly large, the mean square necessa- 
rily is large. The calculation of any set of assumptions was dropped as 
soon as it was very evident that the unsquared differences were 
large. 
The work of the construction of the charts and of the calculation 
was done by the writer’s then assistant, Maude Hickey. 
Approximately 1 month was spent in calculation in connection 
with the effect of the cap rock and approximately 3 months on the 
effect of the salt. The methods evolved considerably during the cal- 
culations. Much of the time was spent in refinement of the technique 
and in recalculating calculations which had been done earlier by less 
refined technique. It is probable that a similar calculation of a diame- 
tral profile could be made by an experienced calculator in 2 months, 
although if difficulties came up, a slightly longer time might be re- 
quired. The time required for satisfactory results would vary in- 
versely with the amount of geologic information that would limit 
the range of variations in our fundamental assumptions. 
RESULTS 
The calculations indicate that the overhang at Barbers Hill can 
be detected by calculations based on torsion-balance data, for the 
results of the calculations indicate that the series of assumed forms 
rank in the following order of decreasing probability. 
Relative 
Rating 
cd Maximum overhang 1.00 
ce Overhang both sides but on right not so much as in preceding 0.80 
of Overhang left; no overhang right 0.44 
be Overhang right; no overhang left 0.43 
bf No overhang 0.22 
Preliminary calculations showed clearly that all forms including 
a or g would produce calculated gradient profiles which fit the 
smoothed gradient profile very poorly. Those forms were dropped from 
the further calculations. 
A summary of the results of the calculations for the forms (0,c) 
and (d, e, f) are shown in Table I. Many trials which were started 
are not reported on that sheet, for trials of any set of assumptions 
were stopped as soon as the unsquared differences were seen to be 
689 
