The alternative process to make use of the chi-squared statistic for assessing the independence of the two variables forming a two-by-two contingency table particularly when expected frequencies are quite small. The method comprises of evaluating the sum of the probabilities connected with the observed table and all possible two-by-two tables which have the same row and column totals as the observed data but exhibit much more extreme departure from the independence. The probability of each table is computed from the hyper- geometric distribution.