One of the questions most often asked about the feedforward model of orientation selectivity is how the underlying circuitry develops. The question becomes particularly puzzling in light of the observation that orientation selectivity develops in the absence of patterned visual input, before a kitten's eyes open (Movshon and Van Sluyters, 1981, Fregnac and Imbert, 1984, Crair et al., 1998, Hubel and Wiesel, 1970). Miller (1994) showed that the arrangement of ON and OFF inputs to a simple cell predicted by the feedforward model can develop from activity-dependent rules of synaptic modification, provided that spontaneous LGN activity patterns in the developing animal have certain simple structures. The development of intracortical connections can be explained by these same rules (Miller et al., 1999, Kayser and Miller, 1998). Excitatory cells would develop connections to correlated cells, whereas inhibitory cells would develop connections with anticorrelated cells. These rules yields the required push-pull inhibition. In addition, they yield excitation among cells that have roughly superimposed ON regions and superimposed OFF regions (same-phase excitation). The inhibition yields contrast-invariant tuning as we have seen, while the excitation amplifies the tuned responses (Chung and Ferster, 1998, Troyer et al., 1998, Ferster et al., 1996).