The Schön case: in-text distribution of references to papers before and after retraction

Over the last decades scientific integrity became a hot issue in science policy. The physicist J.H. Schön who co-authored more than 90 papers, was at the beginning of the century at the center of a famous fraud case. In September 2002 an investigation commissioned by Bell Labs, his employer concluded that 17 papers contained manipulation and misrepresentation of data. They were retracted along with an additional 14 papers based on them. As is frequently the case, some of these papers remain cited even several years after retraction.

In this study using natural language processing tools, the full text of the subset of the citing articles published in Elsevier journals is used to analyze the in-text distribution of the citations to Schön’s retracted oeuvre, and their precise mention. Between the publication year and 2016 these papers were cited in Elsevier journals 211 times (self-citations were excluded) and mentioned 998 in the text of the citing articles.

This in-text distribution evolves over time with 2004 as caesura: after that year most references were located in the introduction and before they were more evenly distributed throughout the text. Each mention in the text was classified: 87% were neutral with no judgment on the cited work’s validity, 7% were negative or mentioned fraud, and 6% were positive. However the latter category contains mostly theoretical work on models.

In ongoing work the evolution of the distribution of in-text references of retracted papers by Schön is benchmarked using a stratified random sample from the same Elsevier journals.