Machine-readable data include not only raw experimental data but also such material as US Census files and bibliographic databases. The authors discuss both technological and sociological factors enabling and inhibiting the sharing of such scientific data.

The growth of computer networks and the availability of good data manipulation languages encourage such sharing. The prospects of detection of bias or fraud, possible disagreement, and motives of personal power and economic gain all tend to discourage it. In spite of the tradition that science and its data are public, the authors believe that the outlook for scientific data sharing is dim. They give many fascinating examples to illustrate their claim that the politics of data sharing discourage it, the biggest exception being data that legislation mandates be made available to others.