Abstract: Under American regulatory law, the dominant contemporary test involves cost-benefit analysis. The benefits of regulation must justify the costs; if they do, regulation is permissible and even mandatory. Under American free speech law, in sharp contrast, the dominant contemporary test involves clear and present danger. Regulators cannot act on the ground that the benefits justify the costs. They may proceed only if the speech is likely to produce imminent lawless action. In principle, it is not simple to explain why the free speech test does not involve cost-benefit analysis, as indeed both Judge Learned Hand and the Supreme Court insisted that it should in the early 1950s. An initial explanation points to the difficulty of quantifying both costs and benefits in the context of speech. That is indeed a serious challenge, but it does not justify the clear and present danger test, because some form of cost-benefit balancing is possible on a more informal, intuitive basis. A second and more plausible explanation points to the serious risk of institutional bias in any assessment of both costs and benefits of speech. This explanation has considerable force, but it depends on questionable assumptions, because institutional safeguards could be introduced to increase accuracy and to reduce any such bias. The third and best justification of the clear and present danger test is that in practice, it does not impose high costs, because the speech that ends up being immunized from regulation has not, in practice, turned out to be harmful. On this view, the benefits of the clear and present danger test turn out to justify its costs. From 1960 or until 2001, this assessment was probably correct for the United States, and it may continue to be correct; but the problem of terrorism, and of recruitment to commit terrorist acts, raises legitimate questions about whether the assumptions on which it rests are correct today.