The error I discovered in the Thompson Scientific ISI database ("Metrics marred by doubt", September 21) could not have arisen because of "the way a citation had been written", as alleged by the Thompson spokesperson.
From cross-checking all the references of the citing biology articles, it is clear there was one in common. True, this is by an author with the same first four letters, "Schn", as the author of the industrial relations article. But there are thousands of such authors.
The error obviously came from the manual process of entering citations into the database. This raises some questions:
- Who does this work and where, and how much are they paid?
- What are the quality control procedures?
- Does Thompson Scientific have a service level agreement with its customers, guaranteeing a particular level of accuracy, with financial penalties for any shortfall?
- If not, are those customers now going to demand this?
- Should we not treat citation evidence as probabilistic rather than absolute? For example, "There is a 99 per cent chance that this article has been cited 12 times, and a 1 per cent chance it has never been cited at all."
Newcastle University Business School