A method is described that allows calibration and assessment of the linearity of response of an array of photomultiplier tubes. The method does not require knowledge of the photomultiplier single photoelectron response model and uses science data directly, thus eliminating the need for dedicated data sets. In this manner all photomultiplier working conditions (e.g. temperature, external fields, etc.) are exactly matched between calibration and science acquisitions. This is of particular importance in low background experiments such as ZEPLIN-III, where methods involving the use of external light sources for calibration are severely constrained.