As Isaac Newton famously put it, today's scientists are "standing on the shoulders of giants" by relying on the work of their predecessors. Scientists give a nod to those predecessors by citing their papers. But bibliometric researchers have debated whether older work is becoming obsolete more quickly, with scientists increasingly citing the recent work of their contemporaries. Now, the team behind Google Scholar has weighed in with a study of their own massive index of papers—and it appears that Newton’s aphorism is truer than ever.
There’s no doubt that scientific papers become obsolete. Although some papers are continually cited and become immortal, the vast majority end up in the dustbin of scientific history. The question is whether the rate of obsolescence has been increasing or decreasing over time.
A 2007 study published in the Journal of the Association for Information Science and Technology concluded that obsolescence has been slowing down since the 1960s as authors cite ever older work. But a 2008 study published in Science reached the opposite conclusion: Obsolescence has accelerated over the past 2 decades as journals have gone online, with authors tending to ignore older papers.
For a study to mark Google Scholar’s 10th anniversary celebration, its researchers analyzed scientific papers published between 1990 and 2013. They divided the papers into nine broad research areas and 261 subject categories. Then they compared the publication dates of the papers cited in all those papers. (Google Scholar is universally acknowledged to index more scientific documents than anyone else, but as usual, the researchers are keeping the size of their data set secret.)
The broadest trend in the Google Scholar data is clear: The fraction of cited papers that are at least 10 years older than the paper citing them has increased steadily, from about 28% in 1990 to 36% in 2013, the team reports today in a paper posted to the arXiv preprint server. The trend held in seven out of the nine broad areas and 231 out of the 261 subject categories. So in spite of the huge growth in the number of journals and the number of papers published each year, scientists' collective memory is deepening. And as for the 30 categories where obsolescence held steady? "These are usually categories that have become major fields in the last 2 decades [such as] nanotechnology and AIDS/HIV," says team leader Anurag Acharya. These fields don't have enough history yet to cite.
"I am obviously happy to see our results confirmed by Google using another data source," says Vincent Larivière, a library scientist at the University of Montreal in Canada who co-authored the 2007 study that found the same trend. "However, I disagree on … the main cause." The Google Scholar team suggests that obsolescence is slowing because digitization and online search engines make it easier to find older papers. But Larivière says that "the age of cited literature has been increasing since the [1960s and 1970s], way before the digitization of older papers." He has a much simpler explanation. “In periods of exponential growth of science,” he argues, “the [average] age of existing literature, and thus of cited—or citable—literature, is always younger and younger." The rate of scholarly publication exploded after World War II, growing 10-fold between 1945 and 1965, Larivière says, and pushing down the age of cited papers. These days, however, "science is growing at a lower rate than it used to, and this affects the age of what is cited."
Of course, there could also be a darker explanation: The average quality of published research may simply be decreasing, driving scientists to look further and further back in time to find papers worth citing. "But if this were to be the case," Acharya wrote in an e-mail to ScienceInsider, "the quality of newer papers would need to suddenly start dropping starting in 2002 to 2004," when the obsolescence rate turned down significantly and across nearly all fields of research. That would be truly depressing.