The structure of the first significant transition state on the unfolding pathway of barnase has been analysed in detail by protein engineering methods. Over 50 mutations placed strategically over the whole protein have been used as probes to report on the local structure in the transition state. Several different probes for many regions of the protein give consistent results as do multiple probes at the same site. The overall consistency of phi values indicates that the mutations have not produced changes in the protein that significantly alter the transition state for unfolding. A fine-structure analysis of interactions has also been conducted by removing different parts of the same side-chains. Many of the results of simple mutations fall nicely into the two clear-cut cases of phi = 1 or 0, indicating that the local noncovalent bonds are either fully broken or fully made in the transition state. Much of the structure of barnase in the transition state for unfolding is very similar to that in the folded protein. Both major alpha-helices fray at the N terminus. The last two turns in helix1 are certainly intact, as is the C terminus of helix2. The general picture of the beta-sheet is that the three central beta-strands are completely intact while the two edge beta-strands are mainly present but certainly weakened. The first five residues of the protein unwind but the C terminus remains folded. Three of the five loops are unfolded. The edges of the main hydrophobic core (core1) are significantly weakened, however, and their breaking appears partly rate determining. The centre of the small hydrophobic core3 remains intact. Core2 is completely disrupted. The first events in unfolding are thus: the unfolding of several loops, the unwinding of the helices from the N termini, and the weakening and disruption of the hydrophobic cores. The values of phi are found to be substantially the same under conditions that favour folding as under conditions that are highly denaturing, and so the structure of the unfolding transition state is substantially the same in water as in the presence of denaturant. The structure of the final kinetically significant transition state for refolding is identical to that for unfolding. The final events in refolding are, accordingly, the consolidation of the hydrophobic cores, the closing of many loops and the capping of the N termini of the helices.