We study unique data from a dynamic natural experiment involving more than 7,000 American women to understand how a woman's propensity to perform an annual mammography changes over time after a co-worker is diagnosed with breast cancer. We find that in the year this event occurs the probability that a woman performs a mammography drops by about 8 percentage points, off a base level of about 70%. This impact effect is persistent during at least the following 2 years, is driven by cases of breast cancer diagnosed at non-early stages, and by the behavior of individuals who are less knowledgeable about health issues. This negative effect is confirm ed when we allow for serial correlation in screening behavior and when we estimate the effect of the treatment on the hazard of not screening, at the daily frequency. However, the effect vanishes in placebo experiments.