A suitable quantity for evaluating the image quality in mammography is the smallest visible size of an object. This quantity, called the image quality index (IQI), can be derived from the basic image parameters: contrast, MTF, Wiener spectrum. Several evaluation methods of the IQI, all based on statistical decision theory, have been considered. An experimental visibility test using simulated microcalcifications has been performed in order to compare the results obtained with different IQI models. A previous approach, based on simplifying assumptions, yields a good correlation with the visibility test but fails to predict the actual size of the visible objects. Improved models have been derived for an ideal observer and for a 'quasi-ideal' one with perfect or with realistic visual characteristics. The experimental visual results are well modelled by the IQI method, provided that a suitable threshold signal-to-noise ratio is used for each of these models.