Hering (1874) suggested that the memory colors of familiar objects may influence their immediate perceived color, as well as exert an indirect effect on the overall scene perception. Given the inherent ambiguity of the illumination color, it may be that familiar objects with diagnostic colors serve as references against which other object colors are calibrated. Method. We manipulated the colors of real, solid objects using our ‘illumination box’ (Ling & Hurlbert, J. Vis. 4(9), 2004). Observers (N = 40) viewed arrays containing a test object (dome or familiar fruit) and a ‘contextual’ object (fruit or rectangular patch). We varied the colors of all objects systematically between trials. After a masking interval, the observers selected the best match to the remembered test object’s color from 6 alternatives, under the same (memory) or changed illumination (constancy). Results. The ‘contextual’ object’s color significantly influenced the test object color matches. For example, when the contextual object was a yellow banana, the remembered test object colors were more bluish than when the contextual object was a yellow patch, for both memory and constancy tasks. We hypothesise that the memory color of the ‘contextual’ object biases the observer’s estimate of the illumination color.