The matching behavior of honeybees in a patch of four artificial feeders was studied under two different environmental conditions in order to examine the involvement of different stimuli in the choice process. Matching fails if all nearby landmarks are removed but can, under certain conditions, be restored by subsequently introducing odors, colors, or landmarks showing that there is no unique stimulus modality that provides matching. We propose two fundamentally different memory processes, both of which affect feeding behavior and support matching. Wesuggest that in one case, the probability of choice is determined by the strength of direct associations between locally perceived odor stimuli and reward rates. In the second case, simultaneously perceived color stimuli predict the relative reward rates indirectly by the spatial representation of the four feeders. Both memory processes are likely to interact and lead to efficient feeding behavior during foraging under natural conditions. Foraging honeybees exhibit a great variety ofbehaviors, which have interested researchers for almost a century (Bitterman, 1988;Gould, 1984;Lindauer, 1963;Menzel, 1990;Opfinger, 1931;Seeley, 1989;von Frisch, 1967). Although the literature on foraging is strongly influenced by ultimate arguments and revolves around optimality criteria (Cheverton, Kacelnik, & Krebs, 1985;Kacelnik, Houston, & Schmid-Hempel, 1986), it is well known that honeybees have phenomenal learning capabilities with respect to the location, efficiency, and production cycles (timing) of different food sources (von Frisch, 1967;Gould, 1984;Heinrich, 1985;Menzel, 1990). Thus an analysis of the proximate mechanisms in the natural setting is called for. Supporting the idea that honeybees make specific use oftheir learning capabilities during foraging, Greggers and Menzel (1993) recently showed that honeybees foraging in a patch of four artificial feeding sites (feeders) matched their choice frequencies to the relative reward rates of the feeders-that is, they visited the higher rewarding feeders more frequently than the lower rewarding feeders. This applied to both "stay" flights (the bee revisits the feeder just visited) and "shift" flights (the bee chooses one ofthe three alternative feeders). Matching was first described by Herrnstein (1961) and refers to an appetitive choice behavior produced by concurrent reinforcement schedules of at least two alternatives. If the animal matches, the response rates to the We are thankful to the referees for providing many constructive criticisms. We thank T. 1. Carew, S. Fisher, M. Hammer, and R. Menzel for valuable commentary and criticism on earlier versions of the manuscript and A. Carney for essential help in preparing the English manuscript. Correspondence should be addressed to U. Greggers, Institut fur Neurobiologie, Freie Universitiit Berlin, Konigin-Luise-Str. 28-30, Berlin, Germany (e-mail: greggers@neuro.biologie.fu-berlin.de).alternatives reflect their relative reward rates (Herrnstein, 1961) or the relative amount o...