Olfaction appears to be a relatively shallow sense – it is likely to take rather fewer levels of neural processing to recognise a lemon by smell than by sight. We would like to understand the circuit basis of olfactory perception by combining both functional and neuroanatomical approaches. Our model system is Drosophila. The fruitfly olfactory system shares the organisation of vertebrates but is numerically much simpler. Furthermore, in contrast to vertebrates there are genetic methods to target specific classes of second and third order olfactory neuron down to the single neuron level. In Drosophila, as in mice, olfactory information is spatially organised in the first olfactory relay of the CNS, the antennal lobe. Olfactory receptor neurons expressing the same odorant receptor molecule send axons to the same subunit or glomerulus within the antennal lobe. Indeed the molecular identity of the odorant receptor input to 37 of the 50 anatomically identifiable glomeruli of the antennal lobe has recently been described [1,2]. At the next level, we and others have shown that the axons of second order projection neurons (equivalent to vertebrate mitral cells) generate a new spatial map in the lateral horn, one of two higher olfactory centres [3,4]. We are most interested in the transformation that occurs between the second and third order neurons that meet in the lateral horn. As a prelude to electrophysiological investigation of identified lateral horn neurons, we have applied new image processing techniques to generate a 3D atlas of this higher olfactory centre. We take 3D confocal images of individual brains each containing single fluorescently labelled neurons. Each brain is also stained with a synaptic marker; these images are registered with a reference brain using a high performance non-rigid intensity-based image registration algorithm, thus mapping every sample into a common coordinate system. 313/575 imaged brains registered sufficiently well to be included in our data set, which presently consists of 231 traced and identified second order projection neurons and 21 third order lateral horn neurons. 35 of an estimated 54 classes of olfactory projection neuron are present in our data set (median 5.5 samples per class, range 1-14). Four classes of projection neuron that originate from a pair of candidate pheromone-responsive glomeruli project to a distinctive region of the lateral horn, which is thus a candidate pheromone-processing centre. We have also used these data to investigate the spatial representations of different odours in the lateral horn. Finally we have applied the same technique to third order neurons of the lateral horn. We have investigated 21 individual neurons falling into three classes of lateral horn neuron and predicted which projection neurons are most likely to be their presynaptic partners. By comparing the input to these neurons with physiological data that we have recently begun to obtain, we hope to generate a detailed understanding of the transformation of odour representations that underlie olfactory perception in the fly.