The Drosophila eye is composed of about 800 ommatidia, each of which becomes dorsoventrally polarised in a process requiring signalling through the Notch, JAK/STAT and Wingless pathways. These three pathways are thought to act by setting up a gradient of a signalling molecule (or molecules) often referred to as the 'second signal'. Thus far, no candidate for a second signal has been identified.The four-jointed locus encodes a type II transmembrane protein that is expressed in a dorsoventral gradient in the developing eye disc. We have analysed the function and regulation of four-jointed during eye patterning. Loss-of-function clones or ectopic expression of four-jointed resulted in strong non-autonomous defects in ommatidial polarity on the dorsoventral axis. Ectopic expression experiments indicated that localised four-jointed expression was required at the time during development when ommatidial polarity was being determined. In contrast, complete removal of four-jointed function resulted in only a mild ommatidial polarity defect. Finally, we found that four-jointed expression was regulated by the Notch, JAK/STAT and Wingless pathways, consistent with it mediating their effects on ommatidial polarity.The clonal phenotypes, time of requirement and regulation of four-jointed are consistent with it acting in ommatidial polarity determination as a second signal downstream of Notch, JAK/STAT and Wingless. Interestingly, it appears to act redundantly with unknown factors in this process, providing an explanation for the previous failure to identify a second signal.