There are rare accounts of ganglion cells being confined to retinal segments. For example, somatostatin-positive alpha cells in cat and rabbit retina are largely restricted to inferior retina (Sagar, 1987; White and Chalupa, 1991) and in the mouse, a subset of RGCs selective for upward motion occurs CH5424802 in dorsal retina (Kay et al., 2011). Because our approach is a functional characterization at the voxel level, retinotopic variations need not necessarily reflect regional variations in the distribution of different morphological classes of RGC. For example, there may be regional differences in the circuits driving RGC activity within the retina. Such differences, which

have been reported in the distributions of neurotransmitter and neuromodulator systems with various classes of retinal neurons (Wilson et al., 2011), could influence the function of a cell in a given morphological class. Presynaptic calcium levels may also be modulated by inputs onto RGC terminals within the tectum itself (Edwards and Cline, 1999). An obvious question is whether find more the functional retinotopic biases in RGC input is reflected in the postsynaptic tectal neurons. The rostro-caudal dendritic

extent for at least three tectal cell types is well below the size of the RGC functional domains that we have identified (Robles et al., 2011). Such spatially restricted sampling of RGCs by tectal neurons suggests that the regional biases in direction and orientation selectivity could L-NAME HCl be preserved in the population of postsynaptic tectal neurons. Indeed, a previous functional imaging study has suggested that retinotopic biases in direction-selective responses do indeed exist in the population of tectal cells

in the zebrafish tectum (Niell and Smith, 2005). While the underlying reasons for it are unknown, the functional architecture we have described may reflect an evolutionary solution that minimizes wiring costs associated with integrating and processing visual stimuli that are perhaps ethologically related (Chklovskii and Koulakov, 2004). Our description of the diversity and organization of inputs to the tectum will also provide a platform for studying emergence in tectal circuits. The property of emergence in neural networks, whereby a neuron produces an output that is not explicitly present in any of its individual inputs, is not well understood. The principal reason is that while the output of individual cells is simple to quantify, determining information about the input that may arise from tens to hundreds of cells is incredibly difficult. Indeed, recent attempts in the retina (Briggman et al., 2011) and visual cortex (Bock et al., 2011) using serial reconstruction at the nanoscale resolution are revealing the size of the challenge. Both of these studies are searching for rules of connectivity that explain the emergent functional properties of neurons.