Abstract
A multistep model called the User Approach Model is proposed as an aid for determining terms to be used in an indexing thesaurus. Information problems and questions are used to derive words which are directly related (i) to a user's potential need for information and (ii) to the spontaneous language in which a user expresses his request. Search words identified by a User Approach could provide specifications for human indexers as well as for electronic equipment fed by character readers. A User Approach is designed to maximize the probability that words in storage will be matched by the words of system users. The contention is that information system designers should more explicitly build on the actual language inclinations of system users. Empirical comparisons of effectiveness and cost should be made for the indexing products of variants of the User Approach and for the current Expert and Statistical approaches to indexing.