Odorants are primarily sensed by olfactory receptor neurons that send their axons to the antennal lobe, the first olfactory neuropil in the central brain of insects. There, each odorant or odorant mixture is encoded by a unique combination of co-activated neurons that represents its primary neural representation. In previous studies we showed that appetitive experience with an odorant changes the internal representation of a mixture containing that odorant. The pattern of activity elicited by the mixture becomes more similar to the representation of the rewarded odorant. This change was interpreted as a mechanism that improves the perception of the rewarded component, which otherwise might get occluded by the perception of the mixture. In the present study we used honey bees Apis mellifera as a model to analyze to what extent the degree of discrimination or generalization between two odors can be predicted based on the similarity between their respective neuronal representations in the antennal lobe. We evaluated whether the changes induced by learning at the level of the representation of mixtures in the antennal lobe, does indeed correlate with the ability of animals to detect the rewarded component immersed in the mixture. Here we confirm that the shift in the representation of the mixture toward the rewarded odor correlate with an increased ability to recognize that component and with a reduced perception of the novel odorant presented in the same mixture.