, 2007, Gao et al., 2008 and Mank et al., 2008). Behavioral assays have been developed that are amenable to simultaneous

neuronal monitoring and a complete anatomical wiring diagram of the visual system appears within reach ( Seelig click here et al., 2010, Maimon et al., 2010 and Chklovskii et al., 2010). Taking advantage of these tools, two groups describe their first results concerning the mapping of the Reichardt model onto neuronal hardware. The minimal circuitry that is thought to be involved in motion detection consists of photoreceptors in the retina, which synapse onto two types of large monopolar cells called L1 and L2 in the next neuropil, the lamina. These cells project in turn onto neurons in the medulla called Mi1 and Tm1 that contact T4 and T5 cells before reaching large tangential cells in the lobula plate that are well characterized and known to represent the output of the Reichardt model ( Figure 1C). The starting point of the first article, by Eichner and colleagues (2011) (this issue of Neuron), is the recognition that multiplication see more over the

entire range of negative and positive brightness fluctuations, as required by the Reichardt model, is unlikely to be achieved by single neurons. This led to the proposal that brightness changes be initially half-wave rectified and then multiplied, which should be much easier to implement in single neurons. That is, multiplication would be carried out on signals that are clipped at zero, sON(t) = max(0, s(t)) and sOFF(t) = max(−s(t),0), resulting in four distinct subbranches of the Reichardt model: ON-ON, ADAMTS5 ON-OFF, OFF-ON, and OFF-OFF, respectively (Figure 1B of Eichner et al., 2011). Indeed, since this formulation is equivalent

to the original model, a wealth of experimental data supports it (e.g., Figure 2 of Eichner et al., 2011). Yet, the tangential cell recordings reported by Eichner and colleagues suggest that half-wave rectification of fast brightness fluctuations is not the only signal driving the Reichardt detector: quite remarkably, brightness changes occurring up to 10 s earlier in the first stimulated channel still impact changes in the second one (their Figure 3). Clark et al. (2011) (discussed below) essentially confirms this result at the behavioral level (their Figure 6D). This leads Eichner and colleagues (2011) to formulate a model that includes these much slower changes, or “DC” components (terminology borrowed from electrical engineering; their Figure 4A). As a byproduct, two of the four subbranches of the original implementation, the ON-OFF and the OFF-ON, can be entirely disposed of, while still reproducing a wide range of experimental data.