Sir

In a super-egotistical response to the Commentary about duplication (Nature 451, 397–399; 2008), I decided to check the Déjà vu database (http://spore.swmed.edu/dejavu) to see whether anybody was copying my data. A search for “Brennan, P.” revealed 31 citations, but this came as no surprise, because my name is quite common. However, I was more surprised when I went through the list and found only two significant incidents of plagiarism. This indicates a 93% false-positive rate for genuine duplication on my small sample. The Commentary authors themselves report a false-positive rate of 27%, after manually checking a much larger sample of 2,600 entries.

Some false positives could be excluded with a little effort. Four of the 31 articles were a series — more of a rarity today, but quite common in the past; sometimes the word 'part' appears in the title, so these could be excluded. Seven of the 31 were clinical updates and four were updates of reviews or opinion pieces. However, reviews and clinical trials are indicated in Medline and could probably be excluded from analysis. Clinical updates are often three or more years apart but have similar authors; the Déjà vu database has a 'time lag' field that can help to identify this type of duplication.

Errami and Garner ask whether it is reasonable to publish the same information in different languages, enabling a piece of work to reach different audiences. In terms of promoting good practice in medicine or other fields, local languages are important. These could be excluded within Medline.

Of the remaining articles I uncovered, there were one-off explanations: research on similar and related topics that seemed like duplicates but were, to me, distinct; re-publication of an article with corrections; and one example of an administrative error causing the journal to publish the article twice. One reprint was an homage to classic citation with permission. The Déjà vu database's tagging system is too simple to differentiate all of these contributions.