Impact factors are a bit like television’s Nielsen ratings. You scrutinize them and take credit if you are a beneficiary, but they are a tad unsavory! Physicists ostensibly do not write to garner citations; they merely prefer to publish in journals with high impact factors.
If as an editor you were overcome by a desire to raise your journal’s impact factor, how should you go about it? (A journal’s impact factor is defined for a given year. PRL’s impact factor for 2006, for instance, is the number of citations in 2006 to Letters published in 2004–2005 divided by the number of those published Letters.) In essence, you should aim to publish at most a few dozen papers a month and focus on areas that are trendy and have adherents with good citing practices.
PRL’s mission is unique, with somewhat conflicting aims: publish in “all fields of physics” but only “important fundamental research.” During PRL’s existence physics has become broader and more international, and at the same time physics publishing has become more fragmented and more hurried. With the advent of better software and the Internet, generating and submitting (though not writing) a manuscript is easier. All of this means that PRL, in particular, has grown steadily over the years despite an increasing rejection rate, and now publishes 4,000 papers per year.
PRL’s impact factor hovers around 7. Journals with significantly higher impact factors typically publish no more than a twentieth of the number of physics papers that PRL does. Reviews of Modern Physics, which had an impact factor of 33.5 in 2006, published just over 30 papers last year.
In Fig. 1 we plot the “cumulative impact factor” ρc(n) for PRL, the sum of all citations in 2006 to each of the n most-cited 2004–2005 Letters, divided by n: ρc(n)=∑i=1ncitations(i) / nwhere citations(i) are the citations to the ith most-cited paper.
We plot the cumulative impact factor ρc(n) that the n most-cited Letters in PRL would have in 2006, if published separately. The 500 most-cited Letters published each year would have an impact factor of around 20; the 250 most-cited Letters would have an impact factor of over 25; and so on. The actual 2006 impact factors of PRL and of Reviews of Modern Physics are also shown. The data are from the Web of Science (WoS; collected September 2007) and Journal Citation Reports (JCR), Thomson Reuters, for 2006.
You can think of ρc(n) also as the “citation density” of the journal, which drops as we stack up more and more less-cited papers. The curve shows the hypothetical impact factor that PRL’s n most-cited Letters would have, if published separately. The rightmost point in the curve (corresponding to ρc(N) where N is the total number of Letters published in 2004–2005) shows the average citation density for the whole journal, which is PRL’s actual impact factor.
A consequence of PRL’s growth is that it now publishes a significant number of Letters that are cited many fewer times than its average impact factor would suggest. While one can think of many reasons to reduce this tail of least-cited Letters, raising the impact factor significantly is not one of them. As you can see from Fig. 1, rejecting the 1000 least-cited Letters per year would raise the impact factor by only 2 units (assuming that one could identify beforehand most papers that would get low citations). By contrast, the 500 most-cited Letters—if published separately, and again assuming one could spot them a priori—would have an impact factor of around 20. (No high-impact journal publishes more than 500 physics papers a year.)
And there’s the rub: even a journal that publishes very many highly cited papers cannot achieve an impact factor above 20 in physics unless it publishes at most only a few hundred papers per year. The averaging process that the impact factor entails means that attracting even dozens of stellar papers each year will not significantly increase the impact factor of a journal that publishes thousands of papers. Newer bibliometric measures, such as the h-index, the Eigenfactor, etc., provide different perspectives.
All journals—even scholarly physics journals—strive to increase readership, attract top-notch authors, and discourage unsuitable submissions. While we at PRL intend to tackle its long tail soon, we are not about to ramp up its impact factor in the short term to, say, twice its current value. To do so would entail drastically changing the scope of what we publish. However, as noted above, each year PRL contains several hundred Letters that as a set surpasses in citations any other physics journal in the world. We hope to attract even better submissions and more readers by highlighting this set. The new online publication Physics and PRL Suggestions are significant steps in that direction.