Typically, cognitive control is assessed using tasks triggering a conflict between two responses. In these tasks, the measure is the congruency effect. This is computed as the performance difference between trials triggering a response conflict (thereafter called incongruent trials) and trials triggering no or low response conflict (congruent trials). Previous research has often reported low reliability for these measures. Furthermore, the correlations between the congruency effects of different tasks have been shown to be low. Moretti et al. (2023) have suggested solving these measurement problems by selecting the trials with the largest congruency effects, such as the trials following congruent trials. Using this approach, Moretti et al. (2023) reported higher reliability estimates and correlations. We addressed two limitations of this approach. First, one may argue that congruency effects may be more valid following incongruent trials because these trials have been reported to trigger an upregulation of cognitive control. A second limitation may be the reliability metric used by Moretti et al. (2023). This was sensitive to the number of trials and their correction accounting for this may have artificially increased the estimates. We addressed both limitations by analyzing the congruency effect following incongruent trials and by using the signal-to-noise ratio, a reliability metric independent of the number of trials. In a first preliminary analysis including one dataset with eight tasks, selecting trials did not consistently result in higher reliability. Thus, these results seem to suggest that this trial selection does not solve the reliability problem recognized in the field.