Structures of large RNAs are not easily solved by X-ray crystallography or by NMR spectroscopy. This paper reviews the alternate methodology based on enzymatic and chemical mapping data collected on RNAs combined with graphical modeling for the construction of three-dimensional models. The different steps that lead to the establishment of the models are critically discussed. It is shown how the correctness of an RNA model can be strengthened by establishing correlations between the structure and the functionality of the molecule and its variants. Finally, the predictive potential of a model is discussed The approach is illustrated by results obtained on plant viral tRNA-like structures, and particularly on that of brome mosaic virus (BMV) RNA.