Effectiveness of testing criteria is the ability to detect failures in a software program. We consider not only effectiveness of some testing criterion in itself but a variance of effectiveness of different test sets satisfied the same testing criterion. We name this property 'tolerance' of a testing criterion and show that, for practical using a criterion, a high tolerance is as well important as high effectiveness. The results of empirical evaluation of tolerance for different criteria, types of faults and decisions are presented. As well as quite simple and well-known control-flow criteria, we study more complicated criteria: Full Predicate Coverage, Modified Condition/Decision Coverage and Reinforced Condition/Decision Coverage criteria.