Two populations of the butterfly Euphydryas editha were repeatedly sampled for scientific study over a period of 22 yr. Bootstrap simulation models indicate that the effects of this perturbation on expected population sizes are not statistically detectable at any time during or after the period of sampling because of the large variance of yearly rates of population growth. However, the models indicate that sampling increased the chances of extinction for the two populations. The estimated amount of this increase depends on whether the populations are assumed to be density-independent or weakly density-dependent.