Extension of the proboscis was conditioned in restrained honeybees with odor as the conditioned stimulus (CS) and sucrose solution--delivered to the antenna (to elicit extension of the proboscis) and then to the proboscis itself--as the unconditioned stimulus (US). In a first series of experiments, acquisition was found to be very rapid, both in massed and in spaced trials; its associative basis was established by differential conditioning and by an explicitly unpaired control procedure (which produced marked resistance to acquisition in subsequent paired training); and both extinction and spontaneous recovery in massed trials were demonstrated. In a series of experiments on the nature of the US, eliminating the proboscis component was found to lower the asymptotic level of performance, whereas eliminating the antennal component was without effect; reducing the concentration of sucrose from 20% to 7% slowed acquisition but did not lower the asymptotic level of performance; and second-order conditioning was demonstrated. In a series of experiments on the role of the US, an omission contingency designed to eliminate adventitious response-reinforcer contiguity was found to have no adverse effect on acquisition. In a series of experiments designed to analyze the resistance to acquisition found after explicitly unpaired training in the first experiments, no significant effect was found of prior exposure either to the CS alone or to the US alone, although the unpaired procedure again produced substantial resistance that was shown to be due to inhibition rather than to inattention; extinction after paired training was found to be facilitated by unpaired presentations of the US. The relation between these results for honeybees and those of analogous experiments with vertebrates is considered.