This paper argues that the lexicographic mediostructure is a network structure that deals with a set or sets of relations that exist between different parts of data by way of cross-referencing, dictionary-internal as well as dictionary, external. The abstract mediostructure consists of all the possible sets of cross-referential relations, whether realised by concrete sets or not in the dictionary. The actual realisation of these referential networks may be function-related and the primary function of the dictionary may then be given priority. The actual cross-references at this level are then the concrete sets of relations depending on the function of the dictionary, the distribution structure and the search path involved in retrieving the information. The paper introduces a distinction between use-related and funtion-related corss-references and focuses on cross-references supporting the comprehension of texts. It is possible to show hierachical relationships (genus/species relation) between terms as well as sequential relations by way of cross-references. Examples taken from bilingual law dictionaries also show that cross-references can facilitate the use of synopsis articles and the placing of factual information in the field introduction in stead of in the articles. It is further argued that the compilers of a dictionary need to take a broad approach to the lexicographic mediostructure as it is directly linked to several other structures. Finally, the concept of lexicographic information costs is introduced.