Oxford and Pittsburgh, contemporary

Simon Saunders and David Wallace

If every outcome of a quantum measurement happens, what is left to be uncertain about? Simon Saunders and David Wallace answer: which of the many persons now sharing your present you are.

Two men in work clothes sit facing each other across a small table, each intent on his cards, one smoking a pipe, with a bottle between them
Paul Cézanne, The Card Players, early 1890s. One of five canvases on the subject, in its two-player form. Each player has a stake, and neither knows how the hand will fall.Public domain, via Wikimedia Commons.

The view

On Everett's theory, a quantum measurement does not have one outcome. Every outcome is realised, each in its own branch, with its own version of the observer. This raises the problem of probability in its sharpest form. Quantum mechanics predicts, with great precision, that some outcomes are more likely than others. But if you know that every outcome will happen, and that there will be a version of you seeing each, what could you be uncertain about, and what do the probabilities measure?

Simon Saunders and David Wallace, both then at Oxford, answered in "Branching and Uncertainty" (2008). They took David Lewis's account of fission: persons are extended through time, and in a fission case two persons share their stages up to the point of division. Applied to Everett, this means that before a measurement there are already many persons present, sharing your current stage, whose futures will differ. Each of them can sensibly wonder which outcome it will see. The uncertainty is real, it is about something, and the quantum weights are what it should be proportioned to.

The argument

The view answers a worry of Hilary Greaves and others: that Everettians must give up uncertainty altogether, and replace it with a "caring measure", a rule for how much a branching agent should care about each future branch. Saunders and Wallace argued that the Lewisian account restores ordinary talk about what will happen and what is likely, and that it matches the way physicists actually speak. Wallace developed the wider case, including a derivation of the Born rule from the theory of rational decision, in The Emergent Multiverse (2012).

On the Lewisian reading the persons are already distinct before the measurement, sharing one stage. Each is uncertain which outcome is its own.

Where they meet Parfit

Like Lewis, Saunders and Wallace accept that what matters in survival is psychological continuity, and that fission is no worse than ordinary survival. In the multiverse that is just as well: if branching were death, everyone would die countless times a second. The Many of you page sets their view beside Parfit's.

Where they part from Parfit

For Parfit, a person about to divide has nothing to be uncertain about: he knows exactly what will happen, and the question "which will be me?" is empty. Saunders and Wallace need the question to have content, since it is what gives probability its meaning in a branching world. So they take Lewis's side on identity, and with it the view that identity and what matters never come apart. Wallace's account of emergent worlds has a page on the reality site.

Saunders and Wallace and Parfit side by side
Saunders and WallaceParfit
What is a person?A person extended through time in the multiverse, a branching history of stages, on Lewis's model.Nothing over and above a brain, a body and a series of interrelated physical and mental events.
What makes you the same person over time?Psychological and physical continuity along one branch; persons on different branches share their pre-branching stages.Relation R, psychological connectedness and continuity, holding in one line without branching.
Is identity what matters?Yes. Identity is kept, as on Lewis's view, by counting persons rather than stages.No. What matters is Relation R, which can hold without identity, as in division.
One of you, or many?Many. Before a measurement, many persons share your present stage; each is uncertain which outcome it will see.Normally one, but the concept fails when R branches; then the question may be empty.

See every view compared →

Further reading