25 years of Science Citation Index--some experiences.

Science Citation Index (SCI) depends for intellectual content entirely on citations by authors, who are sometimes prodded by editors and referees. Its patchiness is therefore not surprising, but frequently it gives access to relevant and up-to-date documents not easily accessible by other means. Two contrasting "citation families" are described. The first family, dealing with the various ascorbic acid derivatives having C substitution at C-2, actually retrieved very nearly all the relevant documents (other than patent specifications) that were retrieved by a CAS ONLINE substructure search. Organic chemists are clearly careful authors. The second family, dealing with amino acid residues covalently bound in soil organic matter, yielded documents having surprisingly little overlap with those retrieved by using a carefully devised Boolean "profile" on the general subject index of Chemical Abstracts. This was only partly because SCI is beset by language-barrier problems to which Chemical Abstracts is immune. The SCI management might extend its journal coverage, but otherwise improvement can only come from a more serious attitude to placing references in primary publications. SCI remains a complement to, not a substitute for, other data-bases.