Temporal characteristics of retracted articles

There are three main reasons for retraction: (1) ethical misconduct (e.g. duplicate publication, plagiarism, missing credit, no IRB, ownership issues, authorship issues, interference in the review process, citation manipulation); (2) scientific distortion (e.g. data manipulation, fraudulent data, unsupported conclusions, questionable data validity, non-replicability, data errors-even if unintended); (3) administrative error (e.g. article published in wrong issue, not the final version published, publisher errors). The first category, although highly deplorable has almost no effect on the advancement of science, the third category is relatively minor. The papers belonging to the second category are most troublesome from the scientific point of view, as they are misleading and have serious negative implications not only on science but also on society. In this paper, we explore some temporal characteristics of retracted articles, including time of publication, years to retract, growth of post retraction citations over time and social media attention by the three major categories. The data set comprises 995 retracted articles retrieved in October 2014 from Elsevier's ScienceDirect. Citations and Mendeley reader counts were retrieved four times within 4 years, which allowed us to examine post-retraction longitudinal trends not only for citations, but also for Mendeley reader counts. The major findings are that both citation counts and Mendeley reader counts continue to grow after retraction.