Drosophila melanogaster detects odors using olfactory sensory neurons (OSNs), the primary neurons of the olfactory system, located on two types of head appendages, the antennae and the maxillary palps. These OSNs differ in their ligand specificities. Some of them are specific for signature chemicals of ecological interest to the fly. Others can be activated by a broad spectrum of odorants. In this dissertation we aimed at a deeper understanding of the contribution of individual OSN classes to the evaluation of odor signals. Using – among other methods – a high-resolution bioassay for assessing odor-guided behavior in flies, we were able to identify and behaviorally characterize hitherto unidentified Drosophila pheromones. Methyl laurate, one of the pheromones, activates two OSN classes, one of which exclusively mediates attraction towards this compound, while the other one contributes to male copulation success without affecting attraction. In this intra-specific communication system, the non-redundant behavioral relevance of two different OSN classes could be demonstrated by examining flies mutant for either of the two pheromone receptors. To allow for the examination of other OSN classes expressing less specific ORs, we aimed at establishing genetic tools to silence different OSN populations. Unfortunately, none of the tools we tested proved dependable enough for a large-scale investigation to assess valence weights of less specific OSN classes. In an alternative approach, we examined whether the hedonic valence of odor mixtures can be predicted on the basis of mixture constituent valences. Physiologically, binary odor mixtures often retain component information and my results show that the valence of a binary odor mixture can be even quantitatively predicted on the basis of component valences. Hence, our results support the idea that hedonic valence of a complex odor results from the integration of the valence weights of individual processing channels.