The long tail of papers

In an entry on pay per paper, I alluded to Chris Anderson's new book, The Long Tail. It should be straightforward to collect page views or down loads or citations of individual papers in a journal. You can plot the numbers of hits of individual papers against the rankings of the papers. Here is the curve for articles in Slate. (Not sure why data stopped at top 500 hits. Why not go further to see a really long tail?) Hope someone in Applied Mechanics will show the same data for JMPS, IJSS, MOM, etc. It will be fun.

Here is the gist of Anderson's observation: If you care about the total sale, as a publisher might, then what matters is the area under the curve; the contribution of the tail may rival that of the head. This much is objective, and should not be controversial.

Now allow me to play a variation of the theme, which is admittedly subjective and possibly controversial. Let's say the net contribution of a journal to new knowledge is proportional to the area under the curve (the subjective part). Then numerous less cited papers may make a significant contribution comparable to the contribution made by the best cited papers.

If you are interested in this argument, you might as well generalize the analysis from a single journal to all journals in a field, or to all journals in science, engineering and medicine. I'm not sure if such a curve has ever been plotted, but the job should not be too hard.

Now, if you are an individual author, surely you'd like to have a lot of hits for your own papers, just as Anderson is celebrating his book becoming a best seller. However, if your job is to increase the total knowledge, as the NSF is set up to do, then you might as well pay as much attention to the long tail as to the tall head.