Strange attractors in the Web of Science database

Accurate computation of h indices or other indicators of research impact requires access to databases supplying complete and accurate citation information. The Web of Science (WoS) database is widely used for this purpose and it is generally deemed error-free. This note describes an inaccuracy that seems to affect differentially non-English sources and targets in WoS, namely, “phantom citations” (i.e., papers reported by WoS to cite some article when they actually did not) and their concentration around particular articles that are thus dubbed “strange attractors”. The analysis of references in (and citations to) papers in two English sources and two non-English sources reveals that phantom citations and other errors of indexing occur about twice as often with non-English items. These and other errors of commission affect about 1% of the cited references in theWoS database, and they may reveal large-scale problems in the reference matching algorithm in WoS. © 2010 Elsevier Ltd. All rights reserved.