Both selfcal and gpscal can be given multiple input model images. The input models can be cubes for multi-channel data. These different models represent some different aspect of the data. For example, the data could contain multiple pointing centres from a mosaiced observation, multiple spectral lines, or have significant polarised flux density. The self-calibration tasks assume that the gains are not dependent on these different aspects of the data ( e.g. the gains are independent of frequency and pointing centre). All the information will be used in a simultaneous solution of the gains. The self-calibration tasks will generally determine the appropriate correspondence between data and model ( e.g. it will associate the data from a particular pointing with a model with the same pointing). There are, however, some caveats described below.