Within 40 years, J. D. Bernal's radical commentary on `the social function of science' has become the conventional wisdom of public science policy. It is now commonplace that science should be organized and financed on a large scale, and directed towards societal goals. This policy also includes basic research, which needs financial support for expensive apparatus. Science is thus being transformed from an individualistic community into a homogenous collective enterprise, which now covers all types of research from the academic to the technological. The modern `R & D system' is undoubtedly beneficial to society, and to the advancement of knowledge. But collectivization has not only changed the societal function of science; it has also changed its internal sociology. Personal discretion in the choice of research problems is now severely limited, even in the university sector, because most projects are now funded by outside agencies. Tension between the individualist norms of the academic tradition and managerial principles derived from the industrial tradition has made research an ambivalent profession. Should scientists be regarded as members of a transnational community devoted to `the search for truth', or are they simply typical employees of governmental and commercial organizations with very worldly aims? This ambivalence is evident in controversies over scientific freedom and responsibility, and in the ethical problems of military research. Collectivization generates pressure for efficiency and public accountability. The R & D system is thus driven from the top towards more urgent and utilitarian programmes. Is highly innovative research fostered adequately by the method of awarding project grants on the basis of peer review, which is often accused of being ponderous and unadventurous? It is suggested that this method of funding basic research should give way to the award of block grants to `institutes', e.g. whole university departments, on the basis of periodic evaluation of their past work. This reform would locate the risks and responsibilities of innovative research within individual institutes, and would reduce the irrationalities and administrative rigidities of the present system; but it would need to be done rigorously, and with sincere concern for the dangers of institutional elitism and internal autocracy.