1. Photoreceptor terminals in the flies Musca domestica and Drosophila melanogaster have been reconstructed in three dimensions from serial EM to reveal the surface distributions of afferent tetrad synapses. 2. The terminals are cylindrical and surround two target cells; they have synaptic sites distributed along their length and around their circumference, except for a strip along the face that lies furthest away from the target cells. 3. Over their inner faces, the terminals have presynaptic sites that are distributed evenly. 4. The distribution of sites in maps plotted from reconstructed membrane surfaces was examined by quadrat analyses. The frequency of sites per quadrat division was not Poissonian, i.e. was non-random. Thus, some form of site selection must exist during synaptogenesis. 5. The sites were shown by variance ratio analysis to be regular (evenly dispersed, not clustered). This suggests that some form of interaction exists, so as to reduce the probability that a synapse will form close to an already existing synaptic site. 6. Distances between nearest-neighbour pairs of synapses had a closest minimum spacing of about 0.8 micron in Musca that was violated by about 5% of pairs, whereas the corresponding distances were about 0.2 micron shorter in Drosophila, which had 13% of pairs situated closer together than 0.8 micron. 7. During synaptogenesis, either initially in the pupa or later in the adult, the probability that a synapse will form is therefore effectively zero within these distances from an existing synaptic site, perhaps through an inhibitory influence exerted by the latter. The nearest-neighbour distances are normally distributed. 8. Unlike the distribution of presynaptic sites, the distribution of postsynaptic sites over the surfaces of the dendrites of the target cells is not even. Although not studied in detail, the corresponding nearest-neighbour distances are much smaller, as little as 0.1 micron. Thus the wider spacing seen between sites over the receptor terminals is a function of the presynaptic cells, and not of their postsynaptic partners, and implies the existence of interactions between synaptic sites.