Maximum credible accident analysis is one of the most widely used concepts in risk assessment of chemical process industries. Central to this concept is the aspect of 'credibility' of envisaged accident scenarios. However, thus far the term credibility is mostly treated qualitatively, based on the subjective judgement of the concerned analysts. This causes wide variation in the results of the studies conducted on the same industrial unit by different analysts. This paper presents an attempt to develop a criterion using which credible accident scenarios may be identified from among a large number of possibilities. The credible scenarios thus identified may then be processed for detailed consequence analysis. This would help in reducing the cost of the analysis and prevent undue emphasis on less credible scenarios at the expense of more credible ones. (C) 2002 Elsevier Science Ltd. All rights reserved.