The winner-take-all (WTA) circuit is a useful computational circuit for signal processing and learning tasks. By adding spatial coupling between pixels, local regions of competition can be delineated. Recently, we described a normalising circuit which enhances Lazzaro's WTA circuit (1989) only with the addition of a transistor and a global bias to each pixel. This new circuit allows the network to transition between a soft-max and a WTA function through the global bias. The WTA network of Lazzaro together with spatial coupling forms the current-mode silicon retina of Boahen and Andreou (1992). This retina models the center surround processing performed in biological retinas to enhance responses to spatial contrasts. Here, we show how our normalising network together with spatial coupling performs as a silicon retina. Results are presented from a fabricated circuit in a 2 /spl mu/m CMOS process.