The Future of Citation Analysis

has grown dramatically in size and influence. The database has expanded from 1.4 million citations in 1964 to 550 million today.

Jeffrey Perkel(jperkel@the-scientist.com)
Oct 23, 2005
<p>THE HOUSE THAT GENE BUILT:</p>

ISI headquarters, in Philadelphia. The building's facade is supposed to evoke moving punch cards, which the company originally used to store citation data. The following page lists the top 10 cited papers from the past two years, 10 years, and of all time.

In the 50 years since Eugene Garfield first proposed it,1 the Science Citation Index has grown dramatically in size and influence. The database has expanded from 1.4 million citations in 1964 to 550 million today. Its list of source journals has grown from 613 to 15,721. And it has become a key tool for tenure, funding, and award committees.

The move to a Web interface that can analyze a century's worth of literature at the click of a mouse has made the Science Citation Index, now part of Thomson Scientific's Web of Science (WOS), more useful than ever. But the...

UNPUBLISHED BUT CRITICAL

Scientists can influence their peers beyond the published word, of course. Consider a scientist who develops a useful program, and posts it to a Web site from which it can be downloaded. Such contributions, says Blaise Cronin, the Rudy Professor of Information Science at Indiana University, are "subterranean, subcutaneous," and they are generally ignored in traditional citation analyses.

One place where they do sometimes appear, however, is in a paper's acknowledgments. "By analyzing acknowledgements, you can demonstrate just how much people rely on one another, even competitors, and especially in the life sciences, where you are required to share reagents after publication," says Cronin, who has spent 15 years mining acknowledgements in scientific literature for their citation value. His recently completed analysis of acknowledgements in four years of Cell issues found that "over the course of three decades, the intensity of acknowledgment behavior rose for each category, most notably in the cases of materials (from 17.6% to 65.1%) and conceptual contributions (30.1% to 84%)."

To date Cronin's analyses have been painstaking, manual processes. But he won't have to work manually for long: This past December Pennsylvania State University researchers C. Lee Giles and Isaac G. Councill reported a systematic effort to extract and parse acknowledgement text from 335,000 computer science papers.3 "Our work supports prior studies showing that acknowledgment trends for individuals do not correlate well with citation trends, perhaps indicating a need to reward highly acknowledged researchers with the deserved recognition of significant intellectual debt," the authors write.

Another metric that citation analysts are currently debating is the value of Web linkages (a link from one person's home page to another). Simply counting links isn't likely to be of much use, says Henry Small, chief scientist at Thomson Scientific and president of the International Society for Scientometrics and Informetrics. "Basically anything goes on the Web. You can have crackpots and charlatans linking to your stuff, [and] you can have Nobel Prize winners linking to your stuff."

Hypertext links reflect more informal, social contacts, says Small, while citations represent more formal expressions of intellectual debt. Nevertheless, he says ongoing efforts to map the Web, to visualize its connectivity and see who influences whom, are among the most sophisticated areas to evolve from traditional citation analysis. "People are attempting to use all the links to map the system of underlying communications or of ideas," he says.

Yale's Bauer suggests that with all the new options available, journals per se may lose their dominance, in favor of the papers within them. The playing field could be leveled: Authors may not choose particular journals based on impact factors, but choose publishing methods based on effectiveness.

For now, however, the traditional refereed paper, wherever it happens to be published, remains the coin of the realm. Says Cronin: "As more of scientific literature moves to the Web and becomes available, you're going to have a richer picture of the life and vitality of a scientific paper than you can have today. So citation analysis won't become passé, it will become one of a battery of indicators with which to measure the impact and influence of a publication."