Scholars are still debating the relative importance of science and technology in the innovation process. While most agree that the linear model - basic research followed by applied research then development is far from adequate, we have not yet arrived at an accepted theoryexplaining the relationships between science and technology. At best, we have surveys trying to identify the sources of innovations, or historical case studies limited in scope. When bibliometric analysis first appeared, it looked promising. Several bibliometric studies, mostly conducted in the last few years, have tried to correlate the output of papers with that of patents in order to see how science and technology interact. After surveying this literature, the author of the present study proceeds to a bibliometric analysis of the scientific production of the 199 firms which produce the most patents in the world. A database consisting of 11,814 papers published in 1989 and 84,658 products and processes patented between 1986 and 1989 has been developed. Papers have been classified into 9 scientific disciplines, and patents into 32 product groups. Statistical analyses have been performed on these data at the industrial level. The study has found that, while industries traditionally identified as science-based (chemicals/pharmaceuticals, electronics/computers) are obviously the one which publish the most, they are far from being the only ones. A second group of industries (instruments, aircraft, motor, metal, mining) is becoming more science-based. Their scientific publications increased by 20% between 1980 and 1989, and the total volume corresponds to 16% of our database. The author suggests that these trends reflect the evolutionary nature of industrial innovation. As one might have expected, a large share of papers has been found to be concerned with applied research. More interestingly, however, a greater share has been identified as science rather than technology. Finally, the study has developed a preliminary model of the incentives to publish within industry. The motivations identified in a survey of 113 industrial authors of papers have been shown to be the same as those of academics, irrespective of their research context. However, whereas academics' publications supply science, industrialists demand science. Consequences are drawn for science policy.