Individuals are happy to make estimates of the probabilities of unique events. Such estimates have no right or wrong answers, but when they suffice to determine the joint probability distribution, they should at least be consistent, yielding one that sums to unity. Mental model theory predicts two main sources of inconsistency: the need to estimate the probabilities that events do not happen, and the need to estimate conditional probabilities as opposed, say, to conjunctive probabilities. Experiments 1 and 2 corroborated the first prediction: when the number of estimates of non-events increased for a problem, so did the degree of overall inconsistency. Experiment 3 corroborated the second prediction: when the number of estimates of conditional probabilities increased, the degree of overall inconsistency was larger as well.