Colocalization measurements aim to characterize the relative distribution of two molecules within a biologically relevant area. It is efficient to measure two distinct features, co-occurrence, the extent to which the molecules appear together, and correlation, how well variations in concentration of the two molecules match. The Manders overlap coefficient (MOC) appears in most colocalization software but the literature contains three interpretations of its measurements: (a) co-occurrence, (b) correlation, or (c) a combination of both. This is surprising given the simplicity of the underlying equation.Testing shows that the MOC responds both to changes in co-occurrence and to changes in correlation. Further testing reveals that different distributions of intensity (Gaussian, gamma, uniform, exponential) dramatically alter the balance between the contribution from co-occurrence and correlation. It follows that the MOC's ability to differentiate between different patterns of colocalization is very limited, since any value is compatible with widely differing combinations of co-occurrence, correlation, and intensity distribution. To characterize colocalization, we recommend reporting both co-occurrence and correlation, using coefficients specific for each attribute. Since the MOC has no clear role in the measurement of colocalization and causes considerable confusion, we conclude that it should be discarded.