Two recent decisions, one by the Supreme Court and one by the Ninth Circuit, have occasioned an ink spill of Exxon Valdez proportions and no little contention. The question, broadly stated, is when the First Amendment should protect speech alleged to constitute a threat by the speaker to kill or seriously injure someone. Given the level of discord, a notable feature of the debate is the acceptance, by judges and commentators alike, of the general proposition that a threat is not protected by the First Amendment, as the Supreme Court told us as early as 1969. In Watts v. United States, where the Court held that no "true threat" had been issued by the speaker, it also took the occasion to announce the "threats exception."

Because the Supreme Court offered little direction for more than two decades, the state supreme courts and federal circuit courts were left to their own devices in fashioning mediating principles to define the contours of the category. On their own, these courts have achieved a considerable consensus around a general formula, even though claims about threats are made in widely diverse factual settings. As this Article shows, the prevailing formula is a set of abstractions offering minimal predictability of results from one case to the next. Remarkably, however, judges typically recite one version of the formula or another as if it were determining the outcome. The result is a collection of opinions that are long on assertion and short on evaluation of anything that matters. In the discussion that follows, we shall see the doctrinal weakness of such an approach. The threats exception, as a First Amendment category, has largely been shaped to fit the very facts it is supposed to govern.