The Conservative Exception

Peacock’s Principle of Permanence says: when you extend a mathematical system, preserve the laws that governed the original system. If addition is commutative for natural numbers, it should be commutative for the integers, the rationals, the reals. Each extension inherits the algebra of its predecessor. The principle guided nineteenth-century algebra and seemed vindicated by every extension that worked.

Then Hamilton discovered quaternions. Multiplication of quaternions is not commutative: ij ≠ ji. The standard narrative says Peacock’s Principle was thereby refuted — a casualty of algebra’s expansion into non-commutative territory.

Toader (arXiv:2603.07592) argues the standard narrative is wrong. Hamilton himself endorsed Peacock’s Principle. He did not see quaternions as violating it. Toader reconstructs why: the principle, correctly understood, is not a rigid rule (“preserve all laws”) but a conservative strategy (“preserve laws unless there is sufficient reason not to”). The framework is Humean — maintain your reasoning habits broadly, but permit exceptions when the evidence demands them.

Under this reading, non-commutative multiplication is permitted by the principle, not prohibited by it. Hamilton weighed the cost of losing commutativity against the gain of a consistent algebraic system in four dimensions. Commutativity was sacrificed because the alternative — no quaternions at all — was worse. The principle told him to hold the line on as much algebraic structure as possible, which he did: quaternions preserve associativity, distributivity, the existence of inverses. They drop only commutativity, and only because no four-dimensional division algebra can have it.

The through-claim: conservatism in mathematics is a strategy, not a dogma. “Preserve the laws” means “preserve as many laws as the new domain permits, and drop only those whose cost of preservation exceeds the benefit.” The principle was never about rigidity. It was about minimizing the distance between the old system and the new one — a least-action principle for algebraic extension.