The way a risk is framed influences people's judgment of the severity or importance of that risk. This study examined the effects of framing of earthquake risk on risk perception. Five frames described logically identical but semantically different statements describing the earthquake risk to the population of a hypothetical city. The frames varied frequencies versus probabilities, the time frame (1600 dead in 500 years, 10% chance of 1600 dead in 50 years, 2 deaths per year) and the sample frame (1.9 deaths per 100,000, 19 per million). Participants rated the five frames according to how risky they seem. The statement that describes the risk in relation to a 50-year time frame and uses a fatality frequency (10% chance of 1600 dead in 50 years) was rated more risky than the other four statements. This finding suggests that risk relating to low frequency hazards should be framed in terms of frequencies of deaths over a life-time. These results clarify which framing has the strongest effect on judgments of earthquake risk and can help agencies responsible for risk management to communicate risk information more effectively.