We have constructed a model of the complex between tyrosyl-tRNA synthetase (TyrRS) from Bacillus stearothermophilus and tRNA(Tyr) by successive cycles of predictions, mutagenesis of TyrRS and molecular modeling. We confront this model with data obtained independently, compare it to the crystal structures of other complexes and review recent data on the discrimination between tRNAs by TyrRS. Comparison of the crystal structures of TyrRS and GlnRS, both of which are class I synthetases, and comparison of the identity elements of tRNA(Tyr) and tRNA(Gln) indicate that the two synthetases bind their cognate tRNAs differently. The mutagenesis data on tRNA(Tyr) confirm the model of the TyrRS:tRNA(Tyr) complex on the following points. TyrRS approaches tRNA(Tyr) on the side of the variable loop. The bases of the first three pairs of the acceptor stem are not recognized. The presence of the NH2 group in position C6 and the absence of a bulky group in position C2 are important for the recognition of the discriminator base A73 by TyrRS, which is fully realized only in the transition state for the acyl transfer. The anticodon is the major identity element of tRNA(Tyr). We have set up an in vivo approach to study the effects of synthetase mutations on the discrimination between tRNAs. Using this approach, we have shown that residue Glu152 of TyrRS acts as a purely negative discriminant towards non-cognate tRNAs, by electrostatic and steric repulsions. The overproductions of the wild type TyrRSs from E coli and B stearothermophilus are toxic to E coli, due to the mischarging or the non-productive binding of tRNAs. The construction of a family of hybrids between the TyrRSs from E coli and B stearothermophilus has shown that their sequences and structures have remained locally compatible through evolution, for folding and function, in particular for the specific recognition and charging of tRNA(Tyr).