Populations of Drosophila were trained by alternately exposing them to two odorants, one coupled with electric shock. On testing, the flies avoided the shocka~sociated odor. Pseudoconditioning, excitatory states, odor preference, sensitization, habituation, and subjective bias have been eliminated as explanations. The selective avoidance can be extinguished by retraining. All flies in the population have equal probability of expressing this behavior. Memory persists for 24 hr. Another paradigm has been developed in which flies learn to discriminate between light sources of different color.Because the hereditary mechanics of Drosophila melanogaster are understood in detail, the behavioral repertoire of this organism and the neural system that specifies it are amenable to genetic analysis. Many flies of identical genotype are readily produced, so that behavioral measurements can be made on populations rather than individuals, yielding instant statistics.If a mutation is found in a gene affecting behavior, methods using genetic mosaics exist for localizing the site of the gene's action to a specific region ("focus") in the fly (1). Anatomical or biochemical changes at the foci of various mutants mav then be correlated with alterations in behavior.One aspect of behavior that so far has been inaccessible to this form of analysis is learning. Conditioning experiments in Drosophila and other dipterans are fraught with complications, and most such studies have been inconclusive. A major problem is pseudoconditioning, in which the training schedule nonspecifically alters the state of the organism, producing changes in behavior that can be misinterpreted as associative learning. An example is the "central excitatory state" (2) in the blowfly Phormia regina; exposure of a hungry fly to sucrose solution arouses it so that afterward it extends its proboscis in response to a variety of unrelated stimuli. This probably accounts (3) for the results of Frings (4). The proper control for pseudoconditioning is to disassociate the reinforcement in time from the stimulus; if the response results from true learning it should depend on simultaneous or near-simultaneous presentation of stimulus and reinforcement.Another pitfall is the possibility of odor cues laid down by the flies. Our early experiments indicated that a stimulus, presumably an odor, was left in the apparatus by flies when shocked and later used by them as a cue for avoidance. The presence of odor trails may have affected the results of Murphey (5) on T-maze learning by Drosophila; these have recently been contradicted by Yeatman and Hirsch (6).Habituation is the decrease in a response on repeated presentation of the same stimulus. Although it can be considered a rudimentary form of learning, in some cases it occurs at the sensory receptors (7), so it is not necessarily related to higher learning in the central nervous system. Exposure of Drosophila larvae to odor altered their behavior as adults (8). This was interpreted as associative learning (9), but has si...