³ÉÈËVRÊÓƵ

In the continuing debate on the appropriateness of metrics as a tool in the evaluation of research quality, particularly for the next research excellence framework, citation counts are frequently compared and contrasted with the ¡°gold standard¡± of peer review (¡°The weight of numbers¡±, Features, 9 July). It is considered axiomatic that any deviation between what is termed peer review and citation numbers indicates that metrics are an inadequate tool. What is generally ignored is that article citation counts themselves encompass two levels of peer review. The first is that a cited article will have been reviewed prior to its acceptance by a journal, which often involves several referees who will have been selected for being close to the specifics of the article. The second level of review is that citations themselves indicate that an article has been read and its content considered to be significant by the peers who then cite it.

The ¡°gold standard¡± peer review associated with the REF means in practice that several hundred articles are read over a period of a few months, and although they will be within the broad subject area of the panel, some, and perhaps many, will be at the periphery of (or beyond) the expertise of the individuals who are assessing them ¨C as those who have sat on panels, whether for the REF or the research councils, are aware. Indeed, in many cases, assessment would be by those sufficiently distant from the specific topic of the article that authors would regard them as inappropriate as referees were they to be asked to review the same material for a journal. It is therefore odd to argue that peer review by panel members is somehow more robust and reliable than citation numbers.

Citations do have their weaknesses for the REF, including being more relevant to some fields than to others, and clearly the most recently published articles will have had too little time to be cited. Perhaps the strongest argument against the widespread adoption of citation numbers for the REF is described by Howard Hotson (¡°A metrics-based REF: from sexennial pain to permanent headache¡±, Opinion, 9 July) ¨C that this may well lead to ever more frequent cycles of evaluation.

Paul Trayhurn
Research professor
University of Buckingham