A recent publication from a standard information science journal actually tested the relationship between impact factor and actual number of cites that an article receives. They found that before 1990, the link was very tight. After 1990, when papers became more freely available online, and were no longer closely tied to the journal in which they were published, the link has become weaker. From the abstract “Furthermore, since 1990 the overall proportion of highly cited papers coming from highly cited journals has been decreasing and, of these highly cited papers, the proportion not coming from highly cited journals has been increasing.”
This seems a fairly reasonable & obvious outcome. In the past libraries used IF to determine what journals to buy. But probably more importantly, everyone read Science & Nature, and not the 2nd tier journals for something that might be interesting outside of the immediate circle of the subdiscipline. Papers get highly cited when more than the Olde Boye’s Clubbe reads & cites. In fact, last time I looked within my subdicipline, lots of the “Golden” papers were methods that everyone cited because they had to, because it gave credibility to their work. And frequently, these new methods papers are not going to be in the glamour journals.
Then there is the “we all hate the glamour journals (except when we can get published in them)” effect. There are going to be a number of Bright Young Things who can’t get their somewhere between slightly and greatly revolutionary schtuff published in the G. J.’s. They turn to the next tier, and (sometimes) eventually get recognized. If G.J.’s are a limited resource (i.e. they can’t publish a lot), but non-G.J.’s are a more abundent resource (cause there are more of them), even if every paper in GJ turns out to be important, eventually, just the size of the other journals will ensure that some highly cited papers are published elsewhere. Online publishing only accelerates this.
The weakening relationship between the impact factor and papers’ citations in the digital age. George A. Lozano, Vincent Larivière, Yves Gingras. Journal of the American Society for Information Science and Technology. Volume 63, Issue 11, pages 2140–2145, November 2012