1. Despite widespread recognition of the need for long-term monitoring of pollinator abundances and pollination service provision, such studies are exceedingly rare.2. In this study, we assess changes in bee visitation and net capture rates for 73 species visiting watermelon crop flowers at 19 farms in the mid-Atlantic region of the United States from 2005 to 2012.3. Over the 8 years, we found a 58% decline in wild bee visitation to crop flowers, but no significant change in honey bee visitation rate. Most types of wild bees showed similar declines in both the visitation and the net capture data; bumble bees, however, declined by 56% in the visitation data but showed no change in net capture rates. Trends in pollination services, that is, estimated pollen deposition, largely followed the trends in visitation and net capture rates.4. While we detected large and significant declines in wild bees when using generalised linear mixed models (GLMMs), permutation analyses that account for non-directional variation in abundance were non-significant, demonstrating the challenge of identifying and describing trends in highly variable populations. 5. As far as we are aware, this article represents one of fewer than 10 published time-series (defined as >5 years of data) studies of changes in bee abundance, and one of only two such studies conducted in an agricultural setting. More such studies are needed in order to understand the magnitude of bee decline and its ramifications for crop pollination.