In this paper, molecular structure generationÐ a central problem in analytical chemistryÐ is modelled as a constraint satisfaction problem (CSP). Because the data of molecular structure generation problems are often ambiguous, in the sense that many substructures may be plausible alternatives for a single item of structural information, techniques are proposed for handling structured and ambiguous data. Th is approach is described in the framework of dynamic constraint satisfaction problems (DCSPs).