(A) Simplified overview (based on e.g., Heisenberg, 2003; Perisse et al., 2013). Odors are coded combinatorially across the olfactory sensory neurons (OSN, blue). In the antennal lobe, these sensory neurons signal towards local interneurons (not shown) and projection neurons (PN, deep blue). Projection neurons have two target areas, the lateral horn (LH, orange) mediating innate approach, and the mushroom body (MB, yellow). Reinforcement signals (green and red for appetitive and aversive reinforcement, respectively) from the gustatory system reach the mushroom body, leading to associative memory traces in simultaneously activated mushroom body neurons. In the present analysis, this sketch focuses selectively on five broad classes of chemosensory behavior, namely innate odor approach, learned odor search and escape, as well as appetitive and aversive innate gustatory behavior. The boxed region is displayed in detail in (B–C). The break in the connection between mushroom body output and behavior is intended to acknowledge that mushroom body output is probably not in itself sufficient as a (pre-) motor signal but rather exerts a modulatory effect on weighting between behavioral options (Schleyer et al., 2013; Menzel, 2014; Aso et al., 2014). (B) Reinforcement processing by value (based on e.g., Heisenberg, 2003; Schleyer et al., 2011; Perisse et al., 2013): a reward neuron sums input from fructose and aspartic acid pathways and thus establishes a memory allowing for learned search for ‘good’. In a functionally separate compartment, a punishment neuron summing quinine and salt signals likewise establishes a memory trace for learned escape from ‘bad’. This scenario cannot account for quality-of-reinforcement memory. (C) Reinforcement processing by both value and quality: in addition to a common, value-specific appetitive memory, fructose and aspartic acid drive discrete reward signals leading to discrete memory traces in at least functionally distinct compartments of the Kenyon cells, which can be independently turned into learned search. For aversive memory, there may be only quality-specific punishment signals. This scenario is in accordance with the present data.