A derivation is given for the statistics of breakdown, including the effect of defects. The particular case where breakdowns are random in time and defects are randomly distributed is considered in detail. Results are obtained that are applicable to any kind of breakdown test. It is shown how relationships between the breakdown distributions for various breakdown tests and for different conditions in a particular test are generated. These relations are given for life tests (constant field) and for ramp tests (field proportional to time). It is found that an ordering of defect types according to their susceptibility to breakdown is not unique, but generally depends on the test conditions. This implies that the fields and temperatures used in screening procedures must be chosen with care.