The ratio bias refers to decision makers' propensity to overestimate probabilities that are specified as ratios of large numbers in comparison to a ratio of smaller numbers (e.g., 9:100 vs. 1:10). To reconcile conflicting findings about this bias, we study such deviations in a more general setting allowing for deviations in both directions, irrational indifference, and inability to decide. We find that the predominant direction of deviations depends on the probabilities involved and that there is no uniform bias in favor of high numbers across all settings. Our findings indicate a strong negative impact of risk literacy on all such bias phenomena. We also test the relationship of the ratio bias to rational/experiential system theory and find that, contrary to the theoretical predictions, biased responses tend to take longer time than logically correct answers.