Data on thunder days, thunder events (discrete periods of nearly continuous thunder), and cloud-to-ground lightning flashes for portions of the United States were compared to discern how well the frequencies of thunder days, events, and lightning flashes are related. Across the United States, thunder day frequencies related well to events with correlation coefficients of +0.97 to +0.84, being lowest in areas where multiple storm periods in a day occur most frequently. The number or duration of thunder events does not relate well to lightning flash frequencies; annual thunder events and durations explain less than 30% of the variance in lightning flashes. This poor relationship is caused by four factors: more than 50% of all thunder events had no cloud-to-ground lightning; many flashes (22--26% in the east and 31--40% in the west) occurred at times not in thunder periods, the frequency of lightning in thunder periods was extremely skewed, with 50% of all flashes occurring in 10% of the period; and locational errors inherent in the detection systems affect the preciseness of the relationships. The poor event-lightning relationships indicate that historical values of thunder activity cannot be used to reconstruct historical lightning frequencies. The results also raise serious questions about the accuracy of the thunder event and thunder day records; clearly, they underestimate the actual storm activity at a weather station. ¿ American Geophysical Union 1988