: The number of times a scientific paper is downloaded is a reasonable measure of the paper's popularity. But when do those downloads occur? A new study sponsored by the Association of American Publishers in Washington, DC, has the answer. Philip Davis, a publishing industry consultant in Ithaca, New York, was granted access to the download statistics of 2812 journals in 10 science, engineering, and social science disciplines. Averaged across all disciplines, the median number of years it took for a paper to reach half of its total, eventual downloads was 2–4 years. For physics, the half-life was 4–5 years. Davis's findings could influence the ongoing debate about how to grant the public free access to papers that describe government-funded research.