Despite decades of research with laboratory rodents, the mechanisms underlying learning and memory, and their impairment, are still not fully understood. The zebrafish is a newcomer in this research area, but it has shown great promise. Food is often employed as a reinforcer in learning tasks with rodents. However, for zebrafish, food has been a problematic reinforcer. Controlling timing and localization of its delivery is difficult. What food types zebrafish prefer is also rarely studied? Here, we describe a novel food delivery hardware and procedure. The apparatus is simple, cheap to manufacture, and easy to employ. Using this new method, we compare how zebrafish respond to three food types, artemia nauplii, crushed tropical fish flakes, and small zebrafish pellets. In binary choice tasks, we show that zebrafish spend significantly more time near the artemia delivery cylinder, swim closer to, and visit this cylinder more frequently compared to food cylinders delivering flakes or pellets, while responses to these latter two cylinders do not differ from each other. We conclude that the newly developed method allows the quantification of food preference in zebrafish, and that it will lead to the identification of highly rewarding food types for learning studies in this species.