Tuesday, July 17, 2012

The incredible growing Transactions on Information Theory

My complimentary copy of the Transactions on Information Theory came in the mail today, my reward for publishing this paper. I haven't received one of these copies in a few years, so I was pretty surprised by the size of it. You're looking at 891 pages of journal right there.
It's not just me. The Transactions on Information Theory are getting much, much bigger.

 Here's what I found going back to 1992:
In 1992, the Transactions published under 2000 pages per year in 6 issues, and around 300 pages per issue; today, it publishes over 4 times as many pages per year in 12 issues, and well over twice as many pages per issue.

This increase has gone largely unnoticed, probably because nobody gets the paper version of the journal any more. But if you did, you would notice that 2011 takes up as much shelf space as everything from 1992-1995. The July 2012 issue pictured above -- longer than average even by this year's standards -- would take up about half as much as 1992. The transactions have doubled in size since 2005.

The society has faced this problem before: you'll note that pages per issue increased dramatically up to 2001, then dropped significantly in 2002; that was the year when the Transactions went from 6-7 issues to 12 issues per year.

But now, with the sheer volume of material being published in the Transactions, maybe it's time to think about splitting the transactions into two or more series. Here are three off the top of my head: Series A, Fundamentals; Series B, Multiuser Information Theory; Series C, Applications. Each series could adopt its own editorial board and standards, which might also solve the persistent problem of time-to-first-decision.


Adam Williamson said...

Interesting info, Andrew. Do you have any data on the lag from submission to publication in this time frame?

Also, I didn't see your ISIT tweets until I'd returned from Cambridge. I'll jump on the hashtag bandwagon in Istanbul next year.


Andrew Eckford said...

I don't have any statistics -- they sometimes print them in the Information Theory Newsletter. But from what I've heard, delays of 18 months are typical.