This paper attempts to broaden our understanding of retail pricing dynamics by providing some systematic evidence about U.S. grocery prices. Using a large data set containing prices on twenty categories of goods from thirty U.S. metro areas for the period 1988-1997, we find a number of empirical regularities. Sales are common phenomenon in that retailers seem to have a "regular" price, and most deviations from that price are downward. There is also considerable heterogeneity in sale behavior across goods within a category, such as cereal. Within each category of goods, retailers regularly put some items on sale, while other items are rarely, if ever, put on sale. Finally, the probability of a sale on an item appears to be greater when demand for that item is higher. These results suggest that retailers use complicated strategies in pricing the items they sell that differ across items and over time. Studies that use retail prices and do not account for the process determining retail prices are likely to yield misleading results.