(A) Differences in the reconstructed volumes between the current FIBSEM dataset and our previous ssEM reconstruction (Takemura et al., 2013). With ssEM, we densely reconstructed a single medulla column, but applied additional tracing of L1 and Tm3 cells in 18 surrounding columns. This was possible because we collected the original TEM images from thin sections that covered a much larger area than the cropped image series used to reconstruct the single column. The FIBSEM dataset used in the current study is superior to ssEM in the comprehensiveness of the dense reconstruction it supports, but was limited to seven medulla columns (Home and its six neighbors) contained wholly within the imaged volume. We thus lacked complete access to any columns beyond this FIB-imaged volume. (B) Calculation of the Tm3 to Mi1-offset, computed from the current FIBSEM data according the method used in Takemura et al., 2013. For each Mi1 and Tm3, we found a field centre of L1 connections, assumed to originate from within a regular hexagonal grid, weighted by their respective synapse counts as reported in Takemura et al., 2013. Then, for each T4, the Mi1 and Tm3 field centres were calculated, also weighted by their respective synapse counts. The Tm3 to Mi1 offset is plotted in x,y dimensions, with Mi1 at 0, 0, where 1.0 represents the distance between two adjacent ommatidia. A significant limitation is that most Tm3 cells extend beyond the FIB-imaged volume and hence are not fully reconstructed. The eight Tm3s that feed the 'fb' T4, for example, have an estimated completeness of 100, 100, 94, 85, 72, 61, 49, and 42%. Given this incompleteness we compute 'raw' displacements (from the partial cells) and 'adjusted' displacements, using estimates of completeness supplemented by the locations for input connections. While the direction and magnitude of these offsets are roughly within the previously reported range (Takemura et al., 2013), no consistent alignment with a T4’s directional sensitivity is observed. However, the errors are large, as shown by the difference between 'raw' and 'adjusted' values, and given only one fully reconstructed T4 for each direction no statistics are possible. Thus, we can neither confirm nor refute the previously reported displacements between Mi1 and Tm3 field centres.