Three groups of 6 rats each were trained for 10 days in a straight runway on a reinforcement schedule that provided 14, 7, 3, 1, and 0.045-g food pellets over five successive runs each day. For Group FN, the first four food quantities (elements) of the series were unsignaled and the terminal 0-pellet element was signaled by flashing lights suspended over the runway, a feature-negative discrimination problem. For Group FP, the first four elements were signaled by flashing lights and the 0-pellet element was unsignaled, a feature-positive discrimination. Group C experienced the flashing lights as a signal for Elements 1-4 or for Element 5, on alternate days. Under the limited training conditions used here, only Groups FN and FP learned to anticipate, and run slowly to, the 0-pellet element, and that behavior developed faster and was more pronounced in Group FN than in Group FP. Implications of the results for analyses of feature discrimination learning are discussed.