The conversion of sypHy signals to rates of vesicle release is shown for light steps of three different intensities in Figure 4A (ON terminals) and Figure 4B (OFF). These records were obtained by averaging over the two populations, irrespective of sensitivity. The variation within each population is illustrated by the individual examples in Figures 2E and 2F and by averaging responses from the 20% of terminals at the two extremes of the sensitivity distribution, as shown in Figures 4C and 4D. For both ON and OFF cells, we only analyzed the initial response at light onset, measured from a dark-adapted state. The intensity-response relations of each of these four subsets of synapses

is shown in Figures 4E and 4F. A good description was obtained using the NVP-BGJ398 molecular weight Hill equation: equation(Equation 2) R=Rmax(IhIh+I1/2h)where sensitivity is quantified as the intensity producing the half-maximal response (I1/2), and the Hill coefficient (h) is the power law describing how the response grows at low intensities. For cones, h is ∼1 and I1/2 is constant across the whole population when measured at the optimal wavelength (Baylor et al., check details 1987 and Normann and Perlman, 1979). The synaptic output of cones and voltage responses in the soma

of bipolar cells also display a Hill coefficient around 1 (Choi et al., 2005 and Euler and Masland, 2000). But in synapses of bipolar cells, both h and I1/2 varied widely.

The distribution of h is shown by the histograms in Figure 5A. Two components can be seen: a sharp peak at h below about 1.5, and a much more widely distributed component at h greater than about 2.0. Supralinearity, which we defined as h > 2, was observed in 66% of OFF and 62% of ON terminals. In other words, some terminals signaled luminance almost in an all-or-none manner. Individual examples of this behavior are shown in Figures 2E (ON) and F (OFF) and Figures S5A and S5B. Thresholding in the synaptic output of bipolar cells is not easily explained by the idea that these are graded neurons that simply respond to linear synaptic inputs and is more likely to reflect active conductances within the synaptic terminals (Burrone either and Lagnado, 1997 and Baden et al., 2011). The value of I1/2 across the population of bipolar cells varied over 4 log units and the distribution had a characteristic shape for both ON and OFF channels—normal on a log scale (Figures 5B and 5C). Strikingly, a number of studies have found that the distribution of luminance in natural scenes is also log normal (Richards, 1982, Brady and Field, 2000 and Geisler, 2008). Although the shape of these distributions appears relatively constant, the width varies: a scene in bright sunlight containing deep shadows might contain luminances varying across 4–5 log units (Pouli et al., 2010 and Rieke and Rudd, 2009).