The petrochemistry of the Yellowknife volcanic belt at Yellowknife (Kam, Duck and Banting formations) shows the suite to be more chemically complex than previously thought. The Kam Formation (65% basalt, 27% basaltic andesite, 17% andesite, and 2% dactite) has a strong tholeiitic trend, involving pyroxene fractionation, which shows iron enrichment to FeO/(FeO + MgO) = 0.75. After altered samples are removed from consideration, it can be shown that the Kam is chemically analogous to modern ocean-floor or ocean-rise tholeiites. The andesitic members of the Kam were partially melted from a hornblendic region, presumably former oceanic crust, possibly with a slight sialic component. The Duck Formation (11% basalt, 21% basaltic andesite, 59% andesite, and 9% dacite) shows considerable evidence of alkali mobility, with varied carbonate content. It is thought to have been largely derived as partial melts from former oceanic crust via hornblende fractionation. The Banting Formation (14% basalt, 7% basaltic andesite, 38% andesite, and 41% dacite) was also derived from an ocean-crust parent, with hornblende fractionation forming a calc-alkaline suite. There is also consistent major- and trace-element evidence that the relationship between Kam parental magmas and Duck–Banting parental magmas involved magnetite as well as hornblende. A petrogenetic model, starting from garnet or spinel lherzolite, with extensive fractionation and (or) partial melting of a preexisting oceanic crust, which may be either true oceanic crust or the lower part of a thin continental crust, is proposed. The whole suite appears to have been formed in one series of related events, developing a rifted ensialic environment at a late stage in the evolution of the Slave craton. It is shown that major- and trace-element discrimination diagrams yield highly varied tectonic conclusions about this sequence, and it is concluded that in general the diagrams are inapplicable to Archean volcanic sequences.