The exact nature of the olfactory signals that arrive in the brain from the periphery, and their reproducibility, remain essentially unknown. In most organisms, the sheer number of olfactory sensory neurons (OSNs) makes it impossible to measure the individual responses of the entire population. We measured the individual in situ electrophysiological activity of OSNs in Drosophila larvae, in response to stimulation with 10 aliphatic odors (alcohols and esters). We studied control larvae (a total of 296 OSNs) and larvae with a single functional OSN, created using the Gal4-upstream activator sequence system. Most OSNs showed consistent, precise responses (either excitation or inhibition) in response to a given odor. Some OSNs also showed qualitatively variable responses ("fuzzy coding"). This robust variability was an intrinsic property of these neurons: it was not attributable to odor type, concentration, stimulus duration, genotype, or interindividual differences, and was seen in control larvae and in larvae with one and two functional OSNs. We conclude that in Drosophila larvae the peripheral code combines precise coding with fuzzy, stochastic responses in which neurons show qualitative variability in their responses to a given odor. We hypothesize that fuzzy coding occurs in other organisms, is translated into differing degrees of activation of the glomeruli, and forms a key component of response variability in the first stages of olfactory processing. Copyright © 2008 Society for Neuroscience.