and subtract delta from all columns of J, - delta from all rows of I to obtain (9), which here holds for more general J except that c(r)= 0 for IJ is modified to c(r) > 0. Then (10) and (11) also hold. Step 3 is illustrated in the matrix C(2) of Table 3 where delta = c25(2) = 2. The contribution to the bounding set sum is S(2) = 2(19) = 38 units. There are corresponding results when column and row are interchanged. Thus the subtraction of 2 units from rows 1 and 2, -2 units from columns 1, 2, 4, lead to the same C(3). Application of step 3 is continued until the matrix is completely reduced with no subset of columns (rows) deficient.