Compositions of ∼2500 spinel–olivine pairs and 400 melt inclusion–spinel pairs have been analysed from 36 igneous suites from oceanic, arc and intraplate tectonic settings. Our data confirm that Cr-spinel mg -number is largely controlled by melt composition, but also influenced by octahedral site substitutions, and rate of cooling. Lavas quenched in submarine environments tend to have higher mg -number at a given cr -number than slowly cooled subaerial lavas and peridotites. Unlike mg -number, Cr-spinel Al 2 O 3 and TiO 2 contents show good correlations with melt composition, with only limited post-entrapment modifications. Our data suggest that increased activity of Al 2 O 3 decreases the partitioning of TiO 2 into spinels. The Al 2 O 3 content of Cr-spinel is a useful guide to the degree of partial melting of mantle peridotites; however, this same relationship is obscured in volcanic rocks. Al 2 O 3 contents of volcanic Cr-spinels are mostly determined by melt composition rather than mantle source composition. The data also suggest that most spinels from residual mantle peridotites can be readily differentiated from those hosted in volcanic rocks. Mantle peridotite spinel tend to have lower TiO 2 and higher Fe 2+ /Fe 3+ ratios than spinel from volcanic rocks. The spinel compositions in our database can be subdivided on the basis of tectonic setting and mode of occurrence using an Al 2 O 3 vs TiO 2 diagram. A total of seven fields can be distinguished with varying degrees of overlap. This diagram can then be used to determine the tectonic setting of spinel from altered mafic igneous rocks such as serpentinites or meta-basalts, or detrital spinel in sandstones.