Abstract Detailed atomic models of a eubacterial 5 S rRNA (spinach chloroplast 5 S rRNA) and of a eukaryotic 5 S rRNA (somatic and oocyte 5 S rRNA from Xenopus laevis) were built using computer graphic. Both models integrate stereochemical constraints and experimental data on the accessibility of bases and phosphates towards several structure-specific probes. The base sequence was first inserted on to three-dimensional structural fragments picked up in a specially devised databank. The fragments were modified and assembled interactively on an Evans & Sutherland PS330. Modeling was finalized by stereochemical and energy refinement. In spite of some uncertainty in the relative spatial orientation of the substructures, the broad features of the models can be generalized and several conclusions can be reached: (1) both models adopt a distorted Y-shape structure, with helices B and D not far from colinearity; (2) no tertiary interactions exist between loop c and region d or loop e; (3) the internal loops, in particular region d, contain several non-canonical base-pairs of A · A, U · U and A · G types; (4) invariant residues appear to be more important for protein or RNA binding than for maintaining the tertiary structure. The models are corroborated by footprinting experiments with ribosomal proteins and by the analysis of various mutants. Such models help to clarify the structure—function relationship of 5 S rRNA and are useful for designing site-directed mutagenesis experiments.