The Criterion for Reality

Definition

Deutsch generalizes Samuel Johnson's rebuttal of Bishop Berkeley's idealism — kicking a rock and declaring "I refute it thus" — into a working test for what belongs in your ontology: "if it can kick back, it exists." More precisely, if the simplest available explanation of some effect requires an entity that is complex and autonomous (its behavior isn't reducible to, or borrowed from, something already accounted for), then that entity is real. He sharpens this further with a computational version: if reproducing the illusion of an entity would require as much computation as the entity's own behavior does, positing that the entity doesn't really exist explains nothing — it just relocates the same complexity behind a curtain.

In the Book

Chapter 4 stages the criterion against the Roman Inquisition's defense against Galileo: the Church accepted that the heliocentric theory made correct predictions, but denied it described anything real, offering instead a geocentric theory elaborately rigged to mimic heliocentric predictions. Deutsch shows this "Inquisition's theory" fails on its own terms — to predict planetary motion under it, you must first understand and apply the heliocentric theory it claims to reject, so it is a disguised, more complicated version of the theory it competes with, not an alternative to it. The same move dispatches solipsism (a solipsist's mind turns out to need a rich, autonomous "unconscious" indistinguishable from an external world) and, crucially, feeds back into Chapter 2's multiverse argument: shadow photons pass the test because explaining the interference pattern without them (e.g., via "action at a distance") requires computing exactly the trajectories that positing shadow photons would — the complexity doesn't go away, it just loses its name. Physicist David Bohm's rival, ostensibly single-universe formulation of quantum theory is judged, by this criterion, to be a multiverse theory wearing a disguise, because its "guiding wave" must encode the same computation as the shadow particles it was invented to avoid.

Why It Matters

The criterion converts "is X real / does X really explain this" from a question about direct perception into a question about explanatory economy: does removing X from the picture actually simplify the explanation, or does it just rename the same irreducible complexity? That test travels well beyond physics — it is a way of catching redescriptions that pretend to eliminate a mechanism (a market force, a psychological drive, an organizational dynamic) while secretly still requiring all of its structure to make correct predictions. It gives you a principled reason to say "you haven't explained this away, you've just stopped naming it."