No two roses smell exactly alike, yet we still perceive their scents as being the same. Most natural odors are made up of odorant mixtures that evoke complex patterns of neural activity, and it is rare for an odor to have the exact same components in the exact same proportions. Encoding these odorant mixtures therefore requires both the identification of individual odorants (pattern separation) and perceptual stability despite the presence of different odorant mixtures (pattern completion). In this issue (p 1378), Barnes and colleagues investigated the neural signals underlying these processes.

The authors recorded responses from neurons in both the anterior piriform cortex and the olfactory bulb in anesthetized rats to complex odorant mixtures that were variants of a core mixture. Some presentations were missing one or more odorants and other presentations involved a replacement of one or more of the odors. Repeated presentations of the same odor mixture elicited highly correlated responses from ensembles of olfactory bulb neurons, and small changes to the mixture (for example, removing or replacing even a single odorant) resulted in a significant decorrelation in the ensemble response. In contrast, cortical ensemble responses were not decorrelated by removing a single odorant. However, removing additional components or replacing even a single component did result in a significant decorrelation.

The authors then tested the ability of rats to discriminate between odor mixtures. As predicted by their electrophysiological results, they found that animals had difficulty discriminating between the full mixture and the mixture with one component missing, but not between the full mixture and a mixture with one component replaced (even though the two mixtures were nearly entirely overlapping). These results suggest that ensembles of piriform cortical neurons perform pattern completion, possibly providing a substrate for perceptual stability.