Mitochondrial inheritance is generally assumed to be maternal. However, there is increasing evidence of exceptions to this rule, especially in hybrid crosses. In these cases, mitochondria are also inherited paternally, so "paternal leakage" of mitochondria occurs. It is important to understand these exceptions better, since they potentially complicate or invalidate studies that make use of mitochondrial markers. We surveyed F1 offspring of experimental hybrid crosses of the 17-year periodical cicadas Magicicada septendecim, M. septendecula, and M. cassini for the presence of paternal mitochondrial markers at various times during development (1-day eggs; 3-, 6-, 9-week eggs; 16-month old 1st and 2nd instar nymphs). We found evidence of paternal leakage in both reciprocal hybrid crosses in all of these samples. The relative difficulty of detecting paternal mtDNA in the youngest eggs and ease of detecting leakage in older eggs and in nymphs suggests that paternal mitochondria proliferate as the eggs develop. Our data support recent theoretical predictions that paternal leakage may be more common than previously estimated.