Advanced quantitative chemical and image analysis techniques based on scanning and scanning-transmission electron miCroscopy have been used to study hot cracking and micro-segregation in an 18%Cr-10%Ni weld metal. It is found that the main cause of hot cracking is sulphur segregation. In fully austenitic welds, manganese remains in solution and when the last of the S-rich 'melt solidifies Mn is simply not available to break up the sulphide films. The present practice of substantially increasing the amount of Mn in austenitic welds, or adding cerium, would thus seem to be justified. Phosphorus also aggravates the problem of hot cracking, but fortunately it does not appear to segregate so markedly as sulphur, being much more soluble in austenite. If the weld metal is so alloyed that solidification to the two-phase austenite plus ferrite occurs, hot cracking does not occur. The two main reasons are concluded to be: (1) manganese is less soluble in ferrite than austenite and a sufficient quantity of Mn is evidently released to combine with S and form inclusions; (2) the structural refinement of two-phase structures, compared to the single-phase austenite, results in a more even distribution of Sand P throughout the weld metal, and marked segregation is thus avoided.