The Paradox of Tolerance

The interesting disagreements are not about whether the logical tension exists. They are about what follows from it.

The reciprocity reading. Toleration is a bargain, not a gift. Its protections extend to those who accept the bargain. This is the cleanest resolution and the most widely held, and it is essentially Popper's own. Its weakness is that "refuses to accept the bargain" is much harder to identify in practice than in a footnote.

Rawls's version. In A Theory of Justice (§35), Rawls argues from a different starting point and lands somewhere stricter than Popper. On his account an intolerant group has no title to complain of intolerance, but it does not follow that a just society should restrict them: it should do so only when it sincerely and with reason believes that its own constitutional security is genuinely in danger. Rawls sets a higher bar and states the trigger more precisely.

The free-speech objection. The most serious criticism is not that the paradox is unreal but that the criterion is unusable. Any authority empowered to decide who has "refused argument" can apply that label to ordinary opponents, and historically such powers have been used that way far more often than against the cases they were designed for. On this view the answer to bad argument is more argument, and building the exception into the rule is more dangerous than the problem it solves.

The scope objection. A related position holds that the paradox is real but that its remedy belongs to social rather than legal authority — that refusing to platform, associate with, or fund a position is different in kind from prohibiting it, and that conflating the two is what makes the debate go badly.

There is no consensus resolution here, which is why the status of this entry is "debated". What is not in dispute is the logical point that a completely unrestricted rule of toleration cannot guarantee its own survival.

1945. The Open Society and Its Enemies was published in two volumes, written between 1938 and 1943 while Popper taught at Canterbury University College in Christchurch, New Zealand. He described it as his "war effort" — a defence of liberal democracy written in response to the rise of fascism, aimed at what he saw as the authoritarian strands running from Plato through Hegel to Marx.

The paradox of tolerance is not in the main text. It is note 4 to chapter 7, which concerns Plato. It is a few sentences long, and Popper never developed it into a chapter, an essay, or a system.

A precedent. The structure was not new. Debates about whether a free society must tolerate those who would end it recur through liberal political thought, and Popper's contribution was the compressed formulation rather than the observation itself.

1971. John Rawls addressed the same question in A Theory of Justice, in a section titled "Toleration of the Intolerant". Rawls arrived independently at a qualified conclusion, and his formulation of the triggering condition is more precise than Popper's.

2010s onwards. The footnote acquired a second life on the internet, usually as an image or a short quotation. In that form the qualifying clause is routinely omitted. It is a striking case of a passage becoming famous in a version its author did not write — which is worth knowing before citing it in either direction.

The immediate lesson is narrow and worth having on its own: read the primary source. A famous position summarised in one line is often not the position. Popper's footnote is short enough to read in under a minute, and doing so changes what it supports.

The wider lesson is structural. A rule that applies without exception to everything, including itself, can license its own abolition. That is the same shape as the Liar sentence and as Russell's paradox, and the fixes available are also the same in outline: restrict the rule's scope, or accept that it must contain an exception clause. Constitutions with unamendable articles are the practical version of a type restriction.

The third lesson is the one this entry has tried hardest to make. A valid abstract argument does not carry its own application. The logical point here is solid, and it tells you almost nothing about any particular case. Most of the work — and all of the danger — is in the step from "unlimited toleration is unstable" to "therefore this specific thing should be restricted." Treating that step as automatic is a mistake regardless of which conclusion you reach.

Content moderation. Every platform with rules faces this directly, and every one of them has landed on a version of the reciprocity answer: participation is conditioned on accepting the rules of participation. The hard problems are all in the application — who decides, on what evidence, with what appeal — which is exactly what the abstract argument does not settle.

Constitutional design. Unamendable clauses, judicial review, and supermajority requirements are all mechanisms for stopping a system from being used to dismantle itself through its own procedures. They are engineering responses to a logical problem.

Institutional norms. Scientific and academic communities operate a version of the reciprocity reading: any claim may be argued for, provided it is argued for. Refusing to engage with evidence does not get you excluded by decree; it gets you excluded by irrelevance. That mechanism is closer to what Popper actually recommended than the version usually attributed to him.

Protocol design. Distributed systems face a structurally identical problem and solve it explicitly. A network that accepts every message including messages that shut it down does not survive; so real protocols specify which messages are in-protocol and reject the rest, and do it in advance rather than case by case. The lesson transfers: the time to decide the boundary is before you need it, in public, with the rule visible.