High-pressure synthesis is increasingly being used in the search for new materials. This is particularly the case for superconductors(1), but the synthesis products are difficult analyse because they are small in size (similar to 50 mg) and often consist of a mixture of unknolvn phases exhibiting a low superconducting volume fraction, X-ray or electron diffraction cannot identify a superconductor unambiguously if it is a minority constituent. Here we report a methodology-’scanning SQUID petrology’-that combines the use of a scanning SQUID microscope(2) with petrological techniques to image and identify low concentrations of superconducting phases in complex phase assemblages. We demonstrate the power of this methodology by investigating the poorly understood origin of superconductivity in the high-pressure Sr-Cu-O system(1). A Sr2CuO3 + KClO3 diffusion couple(3) processed at 60 kbar and 950 degrees C yielded the superconductor Sr3Cu2O5Cl at the similar to 3% level adjacent to the oxidizer. In addition to the unexpected participation of chlorine from an ostensibly ’inert’ oxidizer that is commonly used in high-pressure synthesis work, the sample was highly zoned owing to limited oxygen diffusion kinetics, and contained non-superconducting Sr2CuO3.2. These contamination and diffusion problems probably affected all previous high-pressure copper oxide diffusion-couple experiments. Scanning SQUID petrology has general applicability to heterogeneous samples and is capable of detecting magnetic or superconducting phases at concentrations of less than 1 p.p.m.