Highly-Cited Papers are More Likely to Cite Highly-Cited Papers

An interesting paper in PLOS ONE performed an analysis of all papers published in 2003 that are included in the intersection of the Scopus and ISI databases. They find that the most cited papers in the following 5 years are more likely to cite other highly cited papers than lower ranked papers are to cite highly cited papers relative to how much each group cited less highly cited papers. This is the figure they give for the life sciences:



The black curve shows that more than 50% of the references in the 1% most cited papers were also to papers that were in the 1% most cited category. However, the green curve shows that less than 20% of the references in the papers in the bottom half of the citation distribution - were to papers in the 1% most cited group. The effects were less dramatic in physical and social sciences.

This isn't so surprising in retrospect but it's nice to see the data. The authors claim that this shows that innovative researchers "stand on the shoulders of giants" as Newton said.

At least a couple of other things could be going on:

1. Papers in small subfields or on speciality topics which won't get a huge amount of citations are citing other papers in their subfield or topic which also aren't highly cited. Think economic history for example in economics. No economic history journal has a high impact factor.

2. Top researchers are better at deciding which papers are important and are worthy of citation than are weaker researchers.

Also from PLOS ONE: referees suggested by authors rate papers better than referees suggested by editors and open access papers get cited more.