The structure of yeast transfer RNA aspartic acid has been refined in one crystal form to 3 A resolution using the restrained least-squares method of Hendrickson and Konnert and real-space fitting using the FRODO program of Jones. The final crystallographic discrepancy index R is 23.5% for 4585 reflections with magnitudes twice their standard deviations between 10 and 3 A. With lower occupancies for some residues of the D-loop, the phosphate U1, and the base U33, the R-factor is 22.3%. The adaptation of the restrained least-squares program for nucleic acids and the progress of the refinement are described. The conformations are analysed with respect to stereochemistry and folding of the backbone. The contacts and hydrogen bonds of the secondary structure are compared with those of yeast tRNAPhe. The presence of only four bases in the variable loop, instead of five as in yeast tRNAPhe, leads to a rotation of residue 48 and a lateral movement of residue 46. These two rearrangements induce different environments for [U8 . . . A14] . . . A21 as well as for A9 and G45. Otherwise, all tertiary contacts observed in yeast tRNAPhe are present in yeast tRNAAsp, except for the absence of hydrogen-bonding between G18 of the D-loop and C56 of the T-loop. The presence of anticodon triplet pairing leads to a distribution of temperature factors different from that observed in yeast tRNAPhe with a stabilization of the AC stem-and-loop and a destabilization of the T and D-loops. We are inclined to suggest that the labilization of the interactions between the T and D-loops is a consequence of the interaction of the anticodon triplets of symmetry-related molecules through hydrogen bonding, which mimics the interaction between the anticodon and its cognate codon on the messenger RNA.