Detecting rows and columns of contingency table, which outline from a total positivity pattern

It is known that the procedure called Grade Correspondence Analysis (GCA) transforms any bivariate contingency table into an approximation of table with a very regular positive dependence, called total positivity of order two (TP2). This fact is reminded in the paper, with illustration by the GCA transformation of an artificial contingency table TP8×6. A search for rows and/or columns of table TP8×6, which most strongly outlay from the TP2 pattern, is then also described. The part of the paper, devoted to application, presents the outliers found in three large contingency tables, containing the occupational mobility data from Britain and Poland and the parliamentary election data from Poland.