A critic might object that the previous experiments do not show that ICET is superior to the other algorithms due to its sensitivity to both test costs and classification error costs. Perhaps ICET is superior simply because it can handle delayed tests, while the other algorithms treat all tests as immediate. That is, the method of estimating the average classification cost (Section 2.2) is biased in favor of ICET (since ICET uses the method in its fitness function) and against the other algorithms. In this experiment, we labelled all tests as immediate. Otherwise, nothing changed from the baseline experiments. Table 7 summarizes the results of the experiment. ICET still performs well, although its advantage over the other algorithms has decreased slightly. Sensitivity to delayed tests is part of the explanation of ICET's performance, but it is not the whole story.