From researchers at 15 institutions collaborating on the effects of earthquakes to anthropologists accessing primate DNA samples with detailed accompanying information, "virtual organizations" on the Internet promise new scientific discoveries that would not have been possible without modern high-speed communications and data sharing.

But not all scientists are eager to log on. Some are held back by their "research culture."

The National Science Foundation (NSF) has created a program to invest in technology to support collaboration. But it must make sure the scientists involved are prone to collaborate, say Carl Lagoze and Theresa Velden, researchers in the Information Science department at Cornell. With a new two-year, $400,000 NSF grant, they plan to study a variety of research cultures to find out how best to form effective virtual organizations.

"A brute force approach that assumes a uniform inclination to adopt new technology and the collaborative models it supports will fail to gain the necessary wide acceptance," they argue. "By studying fields that have so far not embraced cyberinfrastructure ... we expect to provide guidance to research investment decisions."

Research cultures vary surprisingly from one discipline to another, note the researchers. They will begin by comparing groups of scientists in chemistry and physics. Physicists who depend on large, shared resources like the Large Hadron Collider in Switzerland eagerly embrace such online tools as the pre-print arXiv at Cornell, where researchers freely share their latest results. Chemists, on the other hand, tend to work with their own apparatus in small, independent research groups and in general have shown less intense collaboration between groups. Earlier research by Velden also has shown that chemists are often constrained from data sharing by their associations with industry, particularly in drug development. The team will expand the study later to examine other sciences, starting with the life sciences and astronomy.

Velden and Lagoze will use computer analysis of journal publications to map collaborations between scientists, and Velden will work in the field, observing and interviewing scientists whose collaborations -- or non-collaborations -- have been mapped, to understand how their culture affects their interconnectivity.

The researchers have identified three characteristics of a research field that will affect its propensity for electronic collaboration: the kind and intensity of collaborations research groups already maintain, the data intensity of the research, and the ease with which research information can be digitized and communicated electronically. A specific goal of their study, they say, will be to determine how these factors determine what type of collaboration will fit a given field.

They offer an early off-the-cuff guess: Among small, independent research groups, informal social networking offers the most promise.