Does Probability Require a Single History?

The ingredients of the proposal are not new. Everett used squared-amplitude measure to characterize the distribution of records in repeated experiments. Although he sometimes described probability as deducible, his measure is naturally read as playing a typicality role rather than as an explicit postulate of objective chance (Everett 1957). Papineau argues that Everettian probability should not be held to an explanatory standard that other accounts of physical chance do not meet (Papineau 1996; Papineau 2010). McQueen and Vaidman explicitly postulate that squared amplitude gives the probability of self-location and compare their postulate with the probability postulate required by collapse (McQueen and Vaidman 2019). They locate that uncertainty after measurement and explain premeasurement behavior through concern for descendants rather than premeasurement uncertainty.

Wilson develops objective Everettian probability through indexicalism, a package combining divergent Everett worlds, individualism about possible worlds, and indexical actuality (Wilson 2013). The present account shares his use of indexical structure to make sense of nontrivial objective probability, but requires neither divergence nor the identification of Everett worlds with metaphysically possible worlds.

Wilhelm’s centered Everett interpretation likewise assigns objective Born chances to propositions about which record is present at an observer’s position, rather than whether that record exists globally (Wilhelm 2022; Wilhelm 2023). He develops this interpretation within a worm ontology and a best-system account of chance. The present paper retains his centered subject of probability without requiring that additional structure.

Tappenden proposes “pure wave mechanics plus a no-collapse probability postulate” and applies it prospectively through a one-to-many account of an observer’s future (Tappenden 2021). His broader account invokes stage theory, temporal counterparts, and a unitary interpretation of mind. The present paper retains the probability postulate and the one-to-many physical continuation relation while leaving those further metaphysical commitments open.

Other approaches address different parts of the problem. Decoherence identifies stable macroscopic records (Zurek 2003; Schlosshauer 2005). Postmeasurement self-location shows how an observer can be ignorant of their local record even while knowing the global state (Vaidman 1998; Tappenden 2011; Sebens and Carroll 2018). Saunders and Wallace use overlapping persons to recover ordinary premeasurement uncertainty (Saunders and Wallace 2008); Barrett argues that forward-looking probability requires suitable metaphysical scaffolding, especially an appropriate trans-temporal identity relation, and develops a many-threads account that supplies it (Barrett 2025). Typicality and confirmation arguments explain the distribution and evidential use of records (Greaves and Myrvold 2010). Decision-theoretic arguments use rational-choice constraints to justify Born-weighted action (Deutsch 1999; Greaves 2004; Wallace 2012). These are not all attempts to do the same job. On the account defended here, the explicit law supplies objective probability; the other approaches address complementary questions.

Existing Everettian accounts often combine a probability proposal with a substantive theory of personal persistence, using overlapping persons, temporal counterparts, worms, or threads to specify how subjects are related across time. A recurring strategy recovers a one-to-one relation by positing several physically indistinguishable premeasurement subjects, each with only one future. This paper rejects the need for that reconstruction while leaving broader questions of persistence open. It takes the one-to-many physical continuation relation at face value: each later observer is a continuation of the one present observer. The same local causal, physical, and psychological continuities that connect the pre- and postmeasurement observer under collapse connect the present observer to each later observer under Everett. Probability requires neither selecting one of them as that observer’s unique future nor positing a distinct present antecedent for each future observer.

Section 2 develops the positive account. It distinguishes global from local exclusivity; uses delayed revelation to clarify local-record uncertainty; constructs a prospective sample space over future continuations; states the Born law; and compares its predictions with those of collapse. Section 3 then considers objections from selection, emergence, personal identity, and empirical confirmation, including Adlam and Barandes’s recent argument that the axiomatic route cannot succeed (Adlam and Barandes 2026). Section 4 concludes.

Consider an idealized two-outcome measurement. Before measurement, write the state schematically as

Here $D$, $O$, and $E$ denote the apparatus, observer, and environment. Under a collapse theory, measurement produces one of the possible record states:

The outcome subscripts label the record encoded by the corresponding state. A stochastic Born law says that the $A$ transition occurs with probability $|c_{A}|^{2}$ and the $B$ transition with probability $|c_{B}|^{2}$.

By contrast, Everettian quantum mechanics retains both states:

Environmental decoherence identifies those two states as stable macroscopic record structures and makes interference between them negligible in the relevant regime. It removes neither from the resulting state and gives neither a probabilistic interpretation.

Collapse makes three claims coincide:

1. (1)

   an $A$-record exists globally;

2. (2)

   the record at this observer position is $A$;

3. (3)

   the observer’s future record will be $A$.

Within the $A$-collapse history shown in Equation (2), all three are true; within the $B$-collapse history shown there, all three are false. The phrase “the outcome is $A$” can therefore stand for any of the three without changing its truth value. Collapse thus supplies global exclusivity: within any resulting history, only one of the possible outcome records for that measurement exists.

Everettian quantum mechanics does not supply global exclusivity: both $O_{A}$ and $O_{B}$, bearing records $A$ and $B$, are present in the universal state. It nevertheless supplies local exclusivity at each later observer-continuation. $O_{A}$ bears $A$ and not $B$, while $O_{B}$ bears $B$ and not $A$. Here “local” means relative to one observer-record, not spatially local.

A delayed-revelation experiment makes the distinction vivid. Suppose the observer is put to sleep before the measurement. The result controls a transport mechanism, so the $A$-relative observer wakes in one room and the $B$-relative observer in another. The rooms are internally identical and each contains a sealed local record. The two observer states are physically distinct and correlated with different records, but have the same cognitively available information. The record is revealed only later (Vaidman 1998; McQueen and Vaidman 2019). The state during the delay has the form

Writing $i$ for either $A$ or $B$, the superscript in $O^{i}_{?}$ identifies the definite record physically local to the observer; the question mark indicates that this value is not cognitively available. The prime marks the environment state during the delay.

For each record value $i$, compare the prerevelation observer in the possible collapse history containing $i$ with the actual Everettian observer associated with $i$. In both cases, record $i$ is physically local to the observer, but its value is not cognitively available. Knowing whether the experiment involved collapse or unitary evolution changes what they know about the global state, but does not reveal the local record. Delayed revelation therefore establishes parity in the target of uncertainty: the hidden local record is a coherent subject of uncertainty under either theory. Collapse actualizes one such observer, whereas Everett instantiates one for every record value. The existence of the other Everettian observers does not reveal the local record to any one of them.

Once the result is revealed, each observer learns which record is in their environment:

The record value $i$ is unchanged. The subscript in $O_{i}$ indicates that the record is now encoded in the observer’s cognitive state as well as being physically local.

Delayed revelation establishes the uncertainty parity just described, but is not itself part of the probability model. The prospective construction starts from $O_{\mathrm{pre}}$. Its future-record variable is defined over the later states $O_{i}$, rather than the intermediate states $O^{i}_{?}$, so no actual prerevelation interval is required. We now specify that future-record domain.

Let $I$ be the finite set of record values with nonzero Born weight. As above, $O_{i}$ denotes the future observer state in which record $i$ is cognitively available. The $O_{i}$ are schematic, coarse-grained observer-record states, distinguished here only by their records. Under unitary evolution, each $O_{i}$ is a future continuation of $O_{\mathrm{pre}}$:

For each $i$, the future local observer state $O_{i}$ is the same, at the record level, as the observer state in the collapse history with result $i$. Thus, from $O_{\mathrm{pre}}$’s prospective standpoint, the two theories present the same record-indexed family of future local experiences. They differ globally: collapse actualizes one member of that family, whereas Everett actualizes them all.

The arrow in Equation (5) marks the direction of stable record formation, not a fundamental irreversibility in the unitary dynamics.¹¹ 1 Explaining this macroscopic arrow therefore requires more than unitary dynamics alone. Low initial entanglement has been proposed as a special cosmological boundary condition that could supply the asymmetry, perhaps in connection with the low gravitational entropy of the early universe and the thermodynamic arrow (Al-Khalili and Chen 2024). Whatever explains this macroscopic asymmetry, I assume here, as Everettian accounts standardly do, that sufficiently decoherent record structures support distinct observer perspectives. Defending that interpretive step lies outside the scope of this paper.

Decoherence identifies the stable local records that distinguish the $O_{i}$. The probability model is formulated entirely at this record level.²² 2 The construction uses the coarse-grained observer-record states $O_{i}$, not an inventory of branches. In contrast with branch-counting proposals, refining a record-bearing component without changing the observer’s record has no effect on the present construction. It creates no new value of the future-record variable and leaves the Born weights assigned to the records unchanged. Branch counting instead treats the number of branches bearing each record as relevant to rational credence, but then faces the difficulty that the branches to be counted are not canonically determined by the underlying quantum dynamics. For a recent treatment of this problem, see Khawaja 2026’s (Khawaja 2026) observation-indexed proposal. Assessing such proposals lies beyond the present argument because nothing in the probability model developed here depends on a branch count.

The Everettian sample space is therefore

Here “sample space” means the mathematical domain over which the probability law is defined. The sample space is constructed relative to $O_{\mathrm{pre}}$: its elements are that observer’s future observer states. Probabilities defined on $\Omega_{E}$ are therefore already prospective.

Define the Everettian future-record random variable by

A random variable assigns a value to each sample point. Here $X_{E}$ returns the record borne by each future observer state. Because $X_{E}$ has exactly one value at each such state, the events $X_{E}=i$ and $X_{E}=j$ are mutually exclusive whenever $i\neq j$. This is local rather than global exclusivity: every $O_{i}\in\Omega_{E}$ is instantiated in the universal state, but each future subject bears only one record. The probability law applies to these mutually exclusive local-record events without selecting one observer state for actualization.

The normalized Born measure $\mu_{\Psi}$, introduced below, supplies the measure on $\Omega_{E}$, completing the mathematical ingredients needed to treat $X_{E}$ as a random variable. The definition of $X_{E}$ itself fixes only which record corresponds to each future observer state; it introduces no new physical structure and assigns no probabilities.

The argument for describing $O_{\mathrm{pre}}$ as uncertain before measurement, developed more fully in Section 3.3, rests on the collapse–Everett parity just described. Each $O_{i}$ has the same relevant causal, physical, and psychological continuity with $O_{\mathrm{pre}}$ that makes the corresponding observer a future version of $O_{\mathrm{pre}}$ under collapse. Each is therefore a future subject continuing $O_{\mathrm{pre}}$, and none is privileged as its one genuine future. In each continuation, the observer experiences exactly one such record as local—“the record here is $i$”—an experience that $O_{\mathrm{pre}}$ does not yet have. This is the sense in which $O_{\mathrm{pre}}$’s uncertainty is qualitative and prospective. It is not uncertainty about which continuation $O_{\mathrm{pre}}$ will uniquely become; there is no such continuation. Formally, the continuation relation supplies the future observer states over which $X_{E}$ is defined, and the Born law assigns probabilities to their records.

The quantum state already gives a squared-amplitude measure to the record alternatives. In the two-outcome example, their normalized Born measures are $|c_{A}|^{2}$ and $|c_{B}|^{2}$. More generally, if $c_{i}$ is the amplitude of the component bearing record $i$, write

Here $\mu_{\Psi}$ denotes the Born measure fixed by the quantum state. At this stage, it has not yet been interpreted as probability. For coarse-grained macroscopic records, $|c_{i}|^{2}$ is the total squared amplitude of the components bearing that record. Equivalently, one may represent a record value by a projector $\Pi_{i}$ and take its weight to be $\langle\Psi\rvert\Pi_{i}\lvert\Psi\rangle$. The simpler amplitude notation is sufficient here.

Equation (8) supplies a measure, not its interpretation. The Everettian account defended here interprets that measure as probability:

This is the Born probability postulate. It is not a theorem of unitary quantum mechanics. The future-record variable $X_{E}$ makes explicit what these probabilities are probabilities of: the record values borne by the present observer’s future continuations.

A critic may still deny that objective probability can apply when every observer state in $\Omega_{E}$ is actual. The remaining disagreement is therefore whether objective probability requires unique global actualization.

Ordinary first-person grammar makes this simple probability difficult to express. We can say, precisely but awkwardly, “The future-local-record variable defined over the continuations of the present speaker has value $A$ with probability $p$.” The natural alternative—“I will record $A$ with probability $p$”—presupposes that the present speaker has one future bearer and invites the question, “Which future continuation will be me?” Some approaches preserve that one-to-one structure by positing a distinct premeasurement antecedent for each future observer. No such multiplication is needed here. Every $O_{i}$ is a future continuation of the one $O_{\mathrm{pre}}$.

Expressing this one-to-many relation while keeping the present speaker as the grammatical subject would require a future-plural first-person pronoun. For expository purposes, I will use we-I. “We-I will record $A$ with probability $p$” is shorthand for

The same relation can be described without new vocabulary: “My future continuations will each record one outcome.” But the possessive formulation puts the continuations at a distance. It can make them sound like later observers related to me rather than future versions of me. We-I keeps the claim in the first person: “We-I will each record one outcome.”

The expression refers separately to each future observer position; it does not describe a single experience shared by all the continuations. It posits neither a group mind nor several hidden observers already present before measurement.

We-I will therefore record $i$ with probability $|c_{i}|^{2}$.

A collapse theory assigns the same Born probabilities to the possible recorded results, but over a different sample space. In the two-outcome model of Equation (2), write $G_{i}:=\lvert i,D_{i},O_{i},E_{i}\rangle$, $i\in\{A,B\}$, for the possible global states after collapse. For a general record set $I$, define

Thus $X_{C}$ is the random variable that returns the unique record in each possible postcollapse global state. Its stochastic law is

Both laws give record value $i$ probability $\lvert c_{i}\rvert^{2}$. However, their sample spaces differ. Under collapse, a sample point $G_{i}$ is a possible postcollapse global state, and $X_{C}$ returns the record it contains. Under Everett, a sample point $O_{i}$ is one of the present observer’s actual future observer states, and $X_{E}$ returns the record local to that observer. The variables $X_{E}$ and $X_{C}$ therefore have different domains but the same possible values and probability distribution. Thus

This is a record-level probabilistic correspondence, not an isomorphism between the theories as a whole.

The single-trial law extends by independence. For $n$ repetitions, write $\mathbf{i}=(i_{1},\ldots,i_{n})$ for a record sequence, and let $X_{E,m}$ return the record of trial $m$. The joint distribution is

The collapse law assigns the same product distribution.

Let $K$ be the number of $A$-records in the sequence. Either law gives

Under collapse, this distribution ranges over $2^{n}$ possible global histories, only one of which becomes actual. Under Everett, the same $2^{n}$ record sequences distinguish $2^{n}$ actual coarse-grained future observer states. In this context, when $O_{\mathrm{pre}}$ uses we-I, the pronoun refers to these future continuations, each bearing its own record sequence. Equation (13) gives the prospective probability distribution over those sequences. Accordingly, as $n$ increases, we-I will record an $n$-trial sequence whose relative frequency of $A$ is close to $|c_{A}|^{2}$, with probability approaching one. The existence of an anomalous sequence does not make it probable as a future local record.

The same point applies to evidence. An observed record sequence $\mathbf{i}$ has its usual Born likelihood, not probability one merely because that sequence is instantiated somewhere globally. Past frequencies can therefore support the Everettian Born law in the same ordinary way that they support a collapse law with the same distribution. This use of objective probability to guide credence involves whatever general connection between chance and rational belief is required in a collapse theory; it requires no further bridge unique to Everettian quantum mechanics.

Adlam and Barandes compare branch weights with the proportions of differently colored marbles in a jar. If 43 of 100 marbles are red, the fraction of red marbles is $0.43$, but it is not yet the probability of anything. A random drawing or some other selection process must be specified before $0.43$ can be interpreted as the probability of drawing red. Likewise, they argue, the mere coexistence of branches with different weights supplies no probability unless something selects one branch.

They are right that a measure over coexisting objects is not by itself a probability, so Equation (8) does not entail Equation (9). Equation (9) supplies the missing step axiomatically: it interprets Born measure as probability. Their marble analogy does not, however, show that such a law must describe a global selection. Here the law applies to $X_{E}$, which returns the record borne by each future observer state continuing $O_{\mathrm{pre}}$, not to the global existence of those records.

Collapse and Everett provide exclusivity differently. Collapse leaves a single record in the global state. In the Everettian sample space, every future observer state is actual, but $X_{E}$ has exactly one value at each. Adding a process that selected one record as the unique global result would not show that $X_{E}$ is ill-defined. It would replace $X_{E}$ with the collapse variable $X_{C}$.

A collapse advocate might reply that this is precisely the problem: collapse supplies genuine premeasurement possibilities, whereas every Everettian future observer state is actual. But treating this difference as disqualifying assumes the point at issue: that probability must concern global possibilities of which only one becomes actual. Everettian uncertainty instead concerns the local record borne at a future observer state.

Selection also does not determine the Born distribution. A process that produces one global result could follow any probability law. Collapse assigns that result probability $\lvert c_{i}\rvert^{2}$ only through its stochastic Born postulate. The account developed here makes the parallel move for $X_{E}$: its law has a different domain but postulates the same distribution. The two questions remain distinct: what probability concerns and what probabilities the law assigns.

Adlam and Barandes also argue that a fundamental axiom cannot assign a new primitive property to an emergent object. Their analogy is an axiom declaring that tables and chairs derived from an underlying physical theory must be yellow. Because branches and observers emerge through decoherence, they conclude that probabilities attaching to them must be derived from the fundamental physics rather than postulated.

However, the Born postulate proposed here does not add an arbitrary property to an emergent record. Unitary measurement dynamics correlates each microscopic alternative with a macroscopically distinct apparatus and observer record while preserving its amplitude. Environmental decoherence makes the resulting record structures effectively orthogonal. The Born weight of a macroscopic record is the squared norm of the corresponding part of the quantum state; if that part is divided into finer orthogonal components, their weights add to the same total.

Thus, although the interpretation of Born measure as probability remains a postulate, the measure itself is defined by the fundamental quantum state rather than assigned arbitrarily to emergent records.

A chair provides a useful analogy only for the issue of precision. There is no perfectly sharp fact about which particles throughout its surface region belong to it, so its exact mass is correspondingly vague. But that does not make mass inapplicable at the macroscopic scale. Likewise, uncertainty at the boundaries of a record structure may make its precise Born weight approximate without making Born weight inapplicable.

There are structural reasons to privilege squared amplitude over other ways of assigning weights to the record alternatives. Squared amplitude is additive across orthogonal record components and does not depend on the basis used to describe the same quantum state. Gleason-style results make the uniqueness claim precise under suitable assumptions (Gleason 1957). Such results explain why Born measure is the natural measure supplied by the quantum formalism. They do not turn that measure into objective probability. That final interpretation is exactly what Equation (9) states.

Collapse is in the same position. A collapse theory may build the Born values into its stochastic dynamics, but their probabilistic status is still part of the theory’s law. Unique actualization supplies neither those values nor a reductive analysis of objective chance. Everettian quantum mechanics need not derive what collapse also postulates.

A further worry is that $X_{E}$ cannot represent what the present observer will record unless each future observer has a distinct premeasurement antecedent with only that future. Some Everettian responses accept this demand, using overlapping or divergent persons, spacetime worms, or nonbranching threads to recover a one-to-one structure (Saunders and Wallace 2008; Wilhelm 2022; Barrett 2025). Each future observer then has its own physically indistinguishable premeasurement antecedent, which may be uncertain which future is theirs. This restores a familiar model of uncertainty and probability, but accounts constructed this way must defend both the probability proposal and the accompanying theory of personal persistence. Lewis considers the same one-to-one strategy as one of several accounts of branching selves, but argues that it cannot provide uncertainty capable of grounding probability (Lewis 2007).

In this limited respect, the multiplication of physically indistinguishable premeasurement selves plays a role analogous to the luminiferous ether in pre-relativistic physics. The ether preserved a familiar mechanical picture: if there is a wave, something must be waving. The additional premeasurement selves preserve a familiar picture of personal persistence: if there are several future observers, each must already have its own present antecedent.

Not every theory of personal persistence imposes this one-to-one structure. Tappenden’s account, for example, retains a one-to-many relation within a theory of temporal counterparts (Tappenden 2021). The account developed here likewise takes personal continuation to be one-to-many. This is a substantive but limited claim about personal persistence. The $O_{i}$ are later subjects, not merely later physical objects, and each is genuinely a future continuation of $O_{\mathrm{pre}}$. This relation is not numerical identity: if $O_{\mathrm{pre}}$ were numerically identical to several $O_{i}$, those later subjects would be numerically identical to one another. It is instead the causal, physical, and psychological continuation relation by which one observer state develops into several later observer states.

For each record $i$, the corresponding postmeasurement observer state has the same relevant causal, physical, and psychological continuity with $O_{\mathrm{pre}}$ under collapse and unitary evolution. Unless membership in a uniquely actual history is itself constitutive of continuation, whatever makes the collapse observer a future version of $O_{\mathrm{pre}}$ also makes the corresponding $O_{i}$ a future version under Everett.

A collapse theorist might reject this conclusion by making global uniqueness a condition of personal continuation. But that condition does not follow from the relations between the earlier and later observer states. It therefore requires independent justification and cannot be assumed in an argument intended to show that probability requires unique actualization. For the Everettian probability model, what matters is that each $O_{i}$ is a later continuation of $O_{\mathrm{pre}}$ and bears record $i$. Any further theory of personal persistence that preserves these facts leaves the sample space $\Omega_{E}$, the future-record variable $X_{E}$, and the Born probability distribution unchanged. Its additional claims may matter to broader questions of personal identity, but they do not affect this probability model.

If every sufficiently long sequence of results occurs, it can seem that no observed sequence favors the Everettian Born law over a rival. Adlam has developed this objection to Everettian confirmation (Adlam 2014), and Adlam and Barandes present it as an obstacle to inductive approaches. The objection is decisive against an account that appeals only to global existence: on such an account, every nonzero-amplitude record exists, so existence alone gives every such record the same trivial likelihood of one.

The objective-probability postulate changes the evidential question. Under the Everettian Born law, a record sequence $\mathbf{i}=(i_{1},\ldots,i_{n})$ borne by one of the observer’s future continuations has its usual Born probability. A rival probability law generally assigns it a different probability. When these probabilities are used as likelihoods, observed frequencies can discriminate between the laws even though copies of every allowed sequence exist globally. The distinction is the same one with which the paper began: the probability that a record exists somewhere is not the probability that it is the record at the observer’s position.

This is not an attempt to derive the Born law inductively from records while presupposing that law. The law is a hypothesis, and the records are evidence for or against it through the likelihoods it assigns. A collapse theory uses the same order of explanation: stochastic law, probability distribution over records, and empirical testing. An Everettian law can be unlucky at some observer positions just as a collapse law can be unlucky in some possible histories. Low-probability records do not change the law that made them unlikely.

Questions remain about the general foundations of objective chance and its connection to rational credence, but they arise for collapse as well. So whatever principle licenses an observer to use known collapse chances in prediction can also license the use of the objective probabilities in Equation (9). Decision theory, self-location, and typicality may illuminate the connection between objective probability and credence; on the present account, they do not supply the objective probability itself. That is supplied by the probability law.