The Possibilities of Science Policy

Irvine and Martin (IM) have been responsible for some serious questioning of British science policy. To date, the outcome of their work seems likely to be beneficial for those who seek some basic nourishment from the research budget, rather than the five-course dinners of Big Science. Nevertheless, their research is not without flaws, nor is it obviously the best way for science policy to make use of recent work in science studies. I will criticize it on two grounds. First, it misses the cognitive 'wood' for the institutional 'trees', because of the way in which it disaggregates science. Science policy ought not to be about maintaining efficient institutions, but about maintaining the sort of cognitive community that will produce the desired scientific products.' The second, interrelated, criticism is that their measures of scientific output are inappropriate, since they reflect only the internal reward system of science; optimum performance according to such measures is not necessarily optimum for science policy. My criticisms of their work on the particular topic of highenergy physics institutions2 are implicit in these more general remarks. IM have studied science by breaking it up into units of comparison defined by what I will call 'non-cognitive boundaries'.3 Usually the boundaries chosen have been those of institutions the laboratory, the university department, the discipline, the nation. The analysis has then turned on the comparison of output/input ratios of these units. The greater the output/input ratio the better that is, more cost-effective the unit. This is certainly an improvement on measures of output alone, and much better than the widespread practice of assessing ability by input alone (how much research money is attracted). However, to develop a policy with cognitive goals in view, it is essential to start by disaggregating science