Abstract This article presents a detailed analysis and partial reconstruction in POP-2 of the control structure of MYCIN, arguably the best-known and most significant Expert System currently in existence. The aim is to aid the development of theory in the field and to assist those who wish to build their own working systems. Attention is focused, inter alia, on the production rules, the goal-directed backward chaining of rules that comprises the control structure and the parameters and context types employed, together with the data structures created during a consultation, and the system's use of “certainty factors” to handle uncertain information. A detailed account is given of how a typical consultation proceeds and some variants that can occur are considered. Developments to MYCIN since its original implementation in 1976 and its generalized version, EMYCIN, are also briefly described. The article identifies a number of gaps in the original reporting of MYCIN and presents a critique of a number of its features and an appraisal of the value of a MYCIN-like approach as a starting point for further Expert Systems development.