Paul Krugman, "The Facebooking of Economics", 12/17/2013:
Economics journals stopped being a way to communicate ideas at least 25 years ago, replaced by working papers; publication was more about certification for the purposes of tenure than anything else. Partly this was because of the long lags — by the time my most successful (though by no means best) academic paper was actually published, in 1991, there were around 150 derivative papers that I knew of, and the target zone literature was running into diminishing returns. Partly, also, it was because in some fields rigid ideologies blocked new ideas. Don’t take my word for it: It was Ken Rogoff, not me, who wrote about the impossibility of publishing realistic macro in the face of “new neoclassical repression.”
In most of the fields where I've worked (computational linguistics, speech technology, computer science, …) real scientific and technical communication now takes place mainly via published conference papers or arXiv preprints, usually available on authors' websites. I've joked for a couple of decades that in those disciplines, publishing in old-fashioned journals has become a sort of cultural ritual, like wearing academic regalia at graduation exercises, which has no substantive role in the actual life of the field.
This is not yet true in most areas of linguistics, because the LSA (and similar organizations) still haven't gotten on board with the decades-old practice of requiring conference presentations to be accompanied by 4-8 page papers rather than 200-word abstracts. So one-to-two-year publication delays (and some blockages of new ideas) remain, alas, the norm.