This paper aims to present our experience in analysing multiple entries during the term harmonization process in the Croatian terminology database Struna. The first part will demonstrate that a great number of multiple entries before harmonization were the result of: a) Its organizational structure, in which there is no direct cooperation between field experts from individual projects, and b) Inexperience of the members of individual projects in distinguishing between language for special purpose and general-purpose language. The need for harmonization became evident at the moment when merging of all terms merged into a single search engine. The harmonization was based on a list of multiple entries and began at term level. The central part of the paper will present the terminographical methods and theoretical background used in the process of distinguishing homonyms from synonyms, and in the process of synonym normalization.