Molecular clouds are characterized by large density contrasts and pronounced internal motions. Some information on their internal physics has been acquired through relatively simple data analysis and through theoretical modelling, mainly hydrodynamical calculations. More detailed radiative transfer models have been limited to spherical cloud models.

A related but less recognized problem is the classification of molecular clouds, or otherwise stated, can we compress the information in the observed two-dimensional maps for various line parameters in such a way that the physical differences are retained, while the random component has been removed? This is a more fundamental formulation of the special problem of comparing the appearance of two molecular clouds, or of the problems of quantifying the velocity correlation within the clouds.