Dark maps, which are also called dark modes or dark themes, have gained popularity, but their usability has not been experimentally evaluated. This article presents a user experiment that uses eye-tracking to assess the usability of dark and light maps. Here, two widely used web maps, Google Maps and Gaode Maps, are selected at the street and city scales.Eight map-use tasks are designed and cover four operations (identify, compare, rank, and associate) with space-alone and attributes-in-space distinctions. Four pairs of map-use tasks (light-during-the-day, dark-during-the-day, dark-at-night, and light-at-night) are examined from three aspects: effectiveness, efficiency, and cognitive load. The results provide preliminary evidence that the light-during-the-day performance is generally the best in most cases, followed by the dark-at-night performance; the dark-during-the-day performance is the worst in all cases, followed by the light-at-night performance, which suggests that aligning the map design with the environment (i.e., lighting environment) is critical for better communication. The light-during-the-day performance is the best for space-alone tasks, and the dark-at-night performance is the best for attributes-in-space tasks. Our investigation also indicates that dark maps are far less favored in practice, which suggests that users' preference for using the dark mode of public-facing web maps needs to be shaped. Since light and dark maps are associated with photopic and scotopic vision, respectively, the results