The first output of a millimeter interferometer is a set of calibrated visibilities in the plane. Scientific analysis of this kind of data is possible in two different ways

**Analysis in the image plane**- ,
*i.e.*analysis on images which are made from the visibilities. This is the most common kind of analysis but it implies (complex) mathematical transforms (*i.e.*imaging and deconvolution) which may bring some artifacts in the results. This is the subject of the next chapter. **Analysis in the plane**- ,
*i.e.*fitting of a source model through the visibilities. When possible, it is the best analysis mode because it avoids the imaging and deconvolution steps. For instance, this enables the precise determination of the properties of proto-planetary disks (geometry, Keplerian rotation, etc...). However, the fit process implies the use of an underlying model of the source. This limits the use of this analysis mode to the ``simplest'' objects. It must also be used with much critical sense as the use of a wrong fitting model may largely bias the result:*e.g.*trying to fit an circular Gaussian through an elliptical one will give a biased full width at half maximum. In practice, a really important use of this analysis mode is the fit of simple models (point, Gaussian, etc...) through unresolved (or slightly resolved) sources, in particular at low signal-to-noise ratio^{3.3}. MAPPING offers tools dedicated to this special goal.