Quantum observers can communicate across multiverse branches Maria Violaris∗ (arXiv:2601.08102v2) Department of Physics, University of Oxford, Oxford, United Kingdom (Dated: January 2026) It is commonly thought that observers in distinct branches of an Everettian multiverse cannot communicate without violating the linearity of quantum theory. Here we show a counterexample, demonstrating that inter-branch communication is in fact possible, entirely within standard quantum theory. We do this by considering a Wigner’s-friend scenario, where an observer (Wigner) can have quantum control over another observer (the friend). We present a thought experiment where the friend in superposition can receive a message written by a distinct copy of themselves in the multiverse, with the aid of Wigner. To maintain the unitarity of quantum theory, the observers must have no memory of the message that they sent. Our thought experiment challenges conventional wisdom regarding the ultimate limits of what is possible in an Everettian multiverse. It has a surprising potential application which involves using knowledge-creation paradoxes for testing Everettian quantum theory against single-world theories. I. INTRODUCTION When observers are treated as quantum systems, many apparent paradoxes and ambiguities are resolved, leading to a counterintuitive but self-consistent Everettian multiverse account of reality [1–3]. In this account, while individual observers experience an approximately classical macroscopic reality locally, on a global scale they exist in superpositions of multiple such experiences occurring on different quantum “branches” [4]. A natural question to consider in such a setting is whether observers in a quantum multiverse can communicate across their branches, for instance to exchange a message with a copy of themselves in a different branch of the multiverse where they are having a different local experience. It is commonly thought that communication across branches of the multiverse is universally forbidden as it would violate the linearity of quantum theory [5, 6]. Here we challenge that notion with an explicit protocol that enables exchange of messages between observers across branches, within fully unitary, linear quantum theory. The impossibility of inter-branch communication appears to follow intuitively from decoherence theory [4, 7]. Once an observer measures a quantum system in superposition, and then interacts with a complex environment, then the global superposition of the system, observer and environment branches into two approximately independently evolving worlds [4]. The emergence of these approximately independent branches is now wellestablished, though there is continuing research on various aspects of the emergence of apparent classicality (e.g. [8]). To achieve inter-branch communication within standard linear, unitary quantum theory, we drop the assumption that the decohered observer branches cannot be coherently controlled. This assumption does not hold in so-called “Wigner’s friend” scenarios, a family of thought experiments which were originally used to investigate the measurement problem in quantum mechanics ∗ maria@violaris.com [9]. These have since become a popular tool in the quantum foundations community for formulating and testing core assumptions of quantum theory and potential successor theories (e.g. [10–12]). Our thought experiment to demonstrate the possibility of inter-branch communication works as follows. The protocol involves an observer inside an isolated laboratory subject to global control, that transports a classical n-bit message from one branch to another. The protocol works by (i) creating a superposition of the observer, (ii) letting the observer in one branch write a message, and (iii) applying a global unitary that exchanges the observer branches. Afterwards the observer who never wrote the message nevertheless possesses it, while the observer who wrote it no longer has the message. This observer retains only (at most) the fact of having written something. We present a unitary implementation of the protocol using a simple quantum circuit toy-model, and discuss how issues of knowledge-creation mean that this protocol provides a novel test of many-worlds against single-world accounts of quantum theory. II. PROTOCOL FOR INTER-BRANCH COMMUNICATION We consider an outside observer, Wigner, who has quantum control over a laboratory containing his friend. The global Hilbert space consists of five subsystems, ordered as follows: the measured qubit Q, a room-label register R, the friend F, the friend’s memory register M, and a piece of paper P. The register R functions as a classical record indicating which room the friend occupies. It serves as a branch label, analogous to a measurement pointer indicating the friend’s location. The protocol is visualised in Fig. 1, with a quantum circuit depiction in Fig. 2. The friend begins in the state |0⟩F , while the qubit to arXiv:2601.08102v2 [quant-ph] 26 Apr 2026 2 FIG. 1. Hello Worlds: Wigner cannot exchange the friends’ messages, but can instead switch the friends’ branches such that one obtains a message from the other. be measured is prepared in the state |+⟩Q = √ 1 2 (|0⟩Q + |1⟩Q). (1) After the friend measures the qubit, which can be modelled by a controlled-NOT (cnot) gate controlled on the qubit and targeted on the friend, the joint state of Q and F becomes √ 1 2 (|00⟩QF + |11⟩QF ). (2) The friend was told before the experiment that if they see the outcome 0, they enter a room labeled “0”, and if they see the outcome 1, they enter a different room labeled “1”. In our global statevector representation, we account for the friend’s location by including the roomlabel register R, and this room record interacts with the friend via a cnot gate controlled on F and targeted on R. Including this record explicitly, the global state becomes √ 1 2 (|000⟩QF R + |111⟩QF R). (3) We refer to the friend in the R = 0 branch as friend0 and in the R = 1 branch as friend-1. Each branch contains a blank piece of paper, initially in the state |0⟩P , and the friend’s memory register begins in the state |0⟩M. The friend was instructed in advance that if they observe outcome 1, they should write a message for friend0 on their paper, while if they observe outcome 0 they should leave the paper blank. We denote the classical message content by µ, and write |µ⟩M and |µ⟩P for its encoding in the memory and paper registers respectively. The friend encodes µ into their memory, evolving via a controlled-µ gate (controlled on F and targeted on M): √ 1 2

|0000⟩QRFM + |111µ⟩QRFM . (4) Then they write this message to the paper, evolving via a cnot gate (controlled on M and targeted on P), giving the global state: √ 1 2

|0000⟩QRFM |0⟩P + |111µ⟩QRFM |µ⟩P  . (5) Next Wigner uncomputes the memory via a cnot from P to M. This uncomputation is independent of the value of µ and returns the memory register to |0⟩M in both branches. The resulting global state is √ 1 2

|0000⟩QRFM |0⟩P + |1110⟩QRFM |µ⟩P  . (6) Wigner then applies a global unitary that exchanges all degrees of freedom distinguishing the two branches except for the paper. In the abstraction we have used here for the degrees of freedom of the distinguishing features of the branches, this operation is U⇌ = XQ ⊗ XR ⊗ XF , (7) which swaps the friend, the measured qubit state, and the corresponding room-label record while acting trivially on the memory and paper. We refer to this as a partial branch-swap operation, depicted by a unitary with a subscript of opposing half-arrows ⇌ to represent the partial exchange of branches. After applying U⇌, the global state becomes √ 1 2

|1110⟩QRFM |0⟩P + |0000⟩QRFM |µ⟩P  . (8) At this point, the paper in the R = 0 branch contains the message µ, while the paper in the R = 1 branch is blank. We can summarise the conclusion of this protocol in the following theorem: 3 |+⟩Q |0⟩R |0⟩F |0⟩M |0⟩P X X X µ FIG. 2. Quantum circuit implementing the inter-branch message-transfer protocol. The room-label register R records the friend’s location via a cnot from F to R. The dashed box encloses the partial branch-swap operation XQ ⊗ XR ⊗ XF . All operations applied by Wigner are independent of the message value µ. Theorem 1 (Inter-branch message transfer). There exists a global quantum operation such that, starting from a superposition of two decohered branches labelled by R = 0 and R = 1, a classical message µ created locally by the observer in the R = 1 branch is transferred to a subsystem in the R = 0 branch which can be read by the observer in that branch. The global operation implementing this transfer is independent of the value of µ. Proof. The result follows by explicit construction. Starting from the initial state in Eq. (2), the friend’s location is recorded in the room-label register via a controlled operation from F to R, resulting in Eq. (3). The friend in the R = 1 branch then thinks of the message µ in their memory (Eq. (4)) and writes it on the paper, giving Eq. (5). Then the friend’s memory is uncomputed, yielding the state in Eq. (6). Applying the partial branchswap unitary U⇌ defined in Eq. (7) produces the state in Eq. (8), in which the paper in the R = 0 branch contains µ. No operation in the protocol depends on the value of µ. The protocol generalises straightforwardly to an n-bit classical message by promoting the memory register M and paper P to n-qubit systems, and the controlled-µ operation is modified so µ is an n-qubit message-encoding operation. The cnots between the memory and paper occur transversally between the n qubits in each state (meaning that there is a cnot between the first qubit in each set, one between the 2nd qubit, etc). The resulting generalised quantum circuit for transporting an n-bit classical message between multiverse branches is that in Fig. 3. III. FRIENDS CANNOT REMEMBER THEIR MESSAGES We now prove the necessity of the memoryuncomputation step for the success of the protocol in Sec. II. If the friend retains any memory of the message, then inter-branch message transfer is impossible under n n |+⟩Q |0⟩R |0⟩F |0⟩ ⊗n M |0⟩ ⊗n P X X X µ FIG. 3. Generalisation of the protocol to an n-qubit message. The room-label register R records the friend’s branch via a controlled operation from F to R. The dashed box again applies XQ ⊗XR ⊗XF , transporting the message across branches. the constraint that Wigner’s operations are independent of the message content. Corollary 1 (Memory erasure is necessary for inter-branch communication). There is no protocol in which a message µ is transferred from the R = 1 branch to the R = 0 branch while the friend who created the message retains a memory record of µ, if all operations applied by Wigner are independent of µ. Proof. We proceed by contradiction. Consider omitting the memory-uncomputation step from the protocol. After the friend in the R = 1 branch creates the message µ in their memory and writes it to the paper, the global state takes the form |ψ⟩ = √ 1 2

|000⟩QRF |0⟩M |0⟩P + |111⟩QRF |µ⟩M |µ⟩P  . (9) Suppose Wigner now applies U⇌, from Eq. (7). The resulting state is U⇌ |ψ⟩ = √ 1 2

|111⟩QRF |0⟩M |0⟩P +|000⟩QRF |µ⟩M |µ⟩P  . (10) In this state, the friend occupying the R = 0 branch carries the memory |µ⟩M that was created in the R = 1 branch. The resulting state has friend-0 carrying the memory created by friend-1, and vice versa. As a result, the friends’ cognitive states are no longer independently well-defined within each branch. The global state no longer contains a branch that can be identified as “friend0 with an additional received message”. Instead, the friends’ minds have been mixed and they are no longer independently well-defined. Therefore without erasing the memory record, the protocol fails to realise inter-branch message transfer as defined in Theorem 1. The above argument shows that if the memory is left intact, swapping the branches cannot transfer a record between well-defined observers. One might attempt to repair this by modifying the branch-swap operation so as to exchange friend-0 and friend-1 while preserving their respective memories. We now show that this is impossible unless Wigner applies message-dependent operations. 4 Lemma 1 (No message-independent memory-preserving branch swap). There is no unitary operation, independent of µ, that exchanges the states |0⟩M and |µ⟩M while leaving the rest of the memory Hilbert space invariant. Proof. To preserve the friends’ memories while swapping branches, Wigner would require a unitary G acting on the memory register such that G |0⟩M = |µ⟩M , G |µ⟩M = |0⟩M . (11) When |0⟩M and |µ⟩M are orthogonal—which is necessarily the case when µ represents a classical message created by the friend—the desired action on their span uniquely determines the unitary G = |µ⟩⟨0| + |0⟩⟨µ| + X j |ej ⟩⟨ej | , (12) where {|ej ⟩} completes {|0⟩, |µ⟩} to an orthonormal basis. However, this operator depends explicitly on |µ⟩. Since G must be applied by Wigner, who by assumption has no access to the message content, such a unitary cannot be implemented. Any attempt by Wigner to learn µ would require measuring the friend, thereby entangling himself with the branches and destroying his ability to perform the required global operation. Hence no message-independent unitary can preserve the friend’s memory while swapping branches. Note that in the original protocol, the only operation that depends on µ is the controlled-µ gate describing the dynamics of the friend’s act of creating the message. All operations performed by Wigner must be independent of µ. This is precisely why swapping the observer between branches is possible, whereas swapping the message itself is not. Finally, we note a closely related limitation concerning branch amplitudes: Corollary 2 (Message-branch amplitudes cannot be modified). It is impossible to change the amplitude of the branch containing the message µ using only messageindependent unitaries. Proof. Consider an initial superposition with unequal amplitudes, q 1 3 |000⟩QRF + q 2 3 |111⟩QRF , (13) and follow the protocol in Sec. II up to the memoryuncomputation step. The resulting state is |ψ⟩ = q 1 3 |0000⟩QRFM |0⟩P + q 2 3 |1110⟩QRFM |µ⟩P . (14) Applying the partial branch-swap unitary exchanges the amplitudes of the branches that the friends each find themselves in, but leaves the amplitude of the branch containing the message µ unchanged. To instead exchange the amplitudes associated with |0⟩P and |µ⟩P would require a unitary that maps |0⟩P to |µ⟩P and vice versa. As shown in the proof of Lemma 1, any such operation necessarily depends on µ. Therefore, message-independent dynamics cannot modify the amplitude of the message branch. IV. DISCUSSION A. Testing many- vs single-world theories using a knowledge-paradox There are thought experiments involving modifications to quantum mechanics that enable the emergence of knowledge paradoxes, with a paradigmatic example being in the context of closed-timelike-curves (i.e. time-travel) [13–16]. By imposing a consistency condition, it is possible to resolve Grandfather paradoxes, which are logically inconsistent timelines such as where you travel back in time and kill your grandfather before you were born [13]. However, the condition does not by itself straightforwardly resolve knowledge paradoxes. These are timetravel paradoxes in which someone copies a mathematical proof from the future, and that copied proof then becomes the very proof that exists in the future [13]. Such scenarios are logically consistent, and compatible with the consistency condition, but there is no explanation for where the knowledge of the proof came from; hence a knowledge paradox. We can construct a similar form of paradox within the inter-branch communication protocol, such that the generation of knowledge provides a test of the multiverse account of quantum theory against single-world accounts. Consider participating in the experiment, from the perspective of friend-0. The friend has a blank sheet of paper, and then after Wigner performs the operation that swaps the friends’ branches, friend-0 sees a message from friend-1 appear on the paper. In principle, this message could contain new knowledge, such as a novel mathematical proof. Friend-0 requires an explanation for how that knowledge was created, and if the only known mechanism for generating knowledge is another agent, then it provides evidence that friend-1 physically existed in a form that is just as real as friend-0. This test has a similar spirit to Deutsch’s 1985 thought experiment for testing unitary quantum theory (including the Everettian interpretation) against theories where measurement causes an irreversible collapse (such as objective collapse theories) [17]. However the inter-branch communication test may distinguish between a different class of rival theories, because explaining the generation of knowledge requires the real existence of friend1 in another branch, not just unitary dynamics where measurements are reversible. Thus, the former thought experiment distinguishes collapse and no-collapse theories, while the latter distinguishes single-world and manyworld theories (more precisely: worlds where all branches of the quantum state are physically real and can thus 5 sustain knowledge-creating observers, and worlds where multiple observers in superposition do not all physically exist). An example would be to distinguish many-worlds and Bohmian interpretations of quantum mechanics [18], which are indistinguishable by many conceivable empirical tests due to both having unitary dynamics and being no-collapse theories. B. Does inter-branch communication violate linearity? Deterministic nonlinear modifications of quantum mechanics provide a useful point of comparison, such as those introduced by Weinberg [19, 20]. Nonlinear dynamics can make the evolution of a subsystem depend on the global state decomposition rather than solely on its reduced density matrix, and hence local operations can influence observables associated with other decohered branches. Polchinski showed that if Weinberg-type nonlinear dynamics are constrained to forbid superluminal signalling, they generically permit communication between decohered Everettian branches, a phenomenon he termed an “Everett phone” [6]. In that protocol, the preparation and reception of a message occur in different branches of the observer’s wavefunction. Polchinski emphasised that such inter-branch communication may be conceptually even more unsettling than faster-than-light signalling, remarking that “communication between branches of the wave function seems even more bizarre than faster-than-light communication . . . but it is not clear that it represents an actual inconsistency.” Polchinski’s result suggested that inter-branch communication was a possible pathology of nonlinear dynamics; our protocol shows that this bizarre form of information transfer can be accessed even within strictly linear quantum mechanics, given sufficient global control. C. Implementation and complexity One may consider what the difficulty is of actually implementing such an experiment. A simple qubit simulation of the protocol can likely already be done on today’s quantum computers, given it only requires five qubits and a shallow circuit consisting of a small number of standard gates. To become more realistic, the size of the message can be increased, requiring a more complex encoding unitary and uncomputation of the friend’s memory. Another significant consideration for performing such a protocol is how different friend-0 and friend-1 become in their branches by the time Wigner needs to swap them. In the protocol presented, flipping between friend-0 and friend-1 is achieved simply by an X-gate. Imagine that friend-0 and friend-1 can be represented by bit-strings, and the bit-strings become more unique as friend-0 and friend-1 evolve in their respective branches. For example, the friends could both begin in the state |0111100001⟩ before branching, and after branching are in the states |0101110101⟩ and |1101100100⟩. In this case, the partial branch-swap operation applied by Wigner to the friend would be X1X6X10. The more distinct friend-0 and friend-1 become, the more complex the string of X operations required by Wigner to flip between their states; loosely speaking, “twins” can communicate more easily across branches than “cousins” can, due to reduced complexity of the swapping operation for the former. This reinforces why it is vital that Wigner has detailed knowledge of the evolution of the friends in the branches, outside of the content of the message. It is useful to situate our protocol within the broader landscape of implementation strategies for quantum observer-based experiments. In the near term, the interbranch communication protocol can be explored further using the approach developed in related pedagogical work by the author, which demonstrates how quantum thought experiments involving observers can be encoded as explicit quantum circuits, and varied to probe conceptual features of quantum paradoxes on existing quantum hardware [21, 22]. There is also growing interest in the scientific community to extend beyond such toy models and implement increasingly realistic Wigner’s-friend–type experiments on quantum information processing platforms, including explicit proposals towards encoding an observer within a quantum computer [12, 23, 24]. These efforts suggest a pathway by which inter-branch communication protocols of the type proposed here may ultimately be explored experimentally. D. Predictability and identity Considering the level of detail Wigner requires to know about the branches leads to an interesting subtlety: if Wigner knows friend-1’s state in exact detail, including full details of friend-1’s experiences in her branch, then could Wigner deterministically predict the contents of the message which friend-1 comes up with? On the assumption that knowledge-creation is fundamentally unpredictable, it is possible that friend-1 comes up with a message that Wigner cannot predict, despite knowing the full details of friend-1’s state. We could similarly consider friend-0 attempting to predict the contents of friend-1’s message, thus directly bypassing the need for communicating between branches in order to retrieve the message. Again, if we assume that friend-1 has had a different experience in her branch to friend-0, then friend1’s knowledge creation becomes unpredictable to friend0. Even if it was somehow the case that Wigner and friend-0 could deterministically predict the contents of friend-1’s message, this would not falsify the inter-branch communication protocol. Instead it would render its existence trivial, considering all three agents involved can deduce the contents of the message with no inter-branch communication required. The protocol also raises philosophical questions regard- 6 ing the continuity of identity. Here we identified friend0 and friend-1 with their abstract information content represented here by the |0⟩ and |1⟩ states. However an alternative interpretation of the protocol could be that the swapping operation is actually transforming friend0 into friend-1, and friend-1 into friend-0, rather than moving the friends (and by proxy, the messages) between branches. These interpretations are indistinguishable at this level of abstraction, creating an interesting problem to analyse in terms of the philosophy of identity across multiverse branches, and how the precise implementation of the swapping operation may play a role (if at all).

V. CONCLUSION

Using a Wigner’s friend scenario, we have presented a protocol that exchanges a classical message between observers in distinct Everett branches by swapping observer states. The protocol is fully consistent with standard unitary quantum mechanics, and therefore challenges intuitive notions of information locality in the multiverse and invites refined theoretical, philosophical, and experimental investigation.

ACKNOWLEDGEMENTS I thank David Deutsch, Sam Kuypers, Ryan Mann and Tony Short for helpful discussions and comments, including suggestions regarding the interpretation of the protocol. I also thank Chiara Marletto, Vlatko Vedral, and the other participants of the Cocconato Constructor Theory Workshop 2024 for valuable discussions. Finally I thank Rob Sullivan for discussions that included posing the question that motivated this work.

[1] H. Everett, Relative state formulation of quantum mechanics, Reviews of Modern Physics 29, 454–462 (1957).

[2] D. Deutsch, The structure of the multiverse, Proceedings of the Royal Society A 458, 2911–2923 (2002).

[3] D. Wallace, The Emergent Multiverse: Quantum Theory according to the Everett Interpretation (Oxford University Press, Oxford, 2012).

[4] W. H. Zurek, Decoherence, einselection, and the quantum origins of the classical, Reviews of Modern Physics 75, 715–775 (2003).

[5] D. Deutsch, The Fabric of Reality (Allen Lane, London, 1997).

[6] J. Polchinski, Weinberg’s Nonlinear Quantum Mechanics and the Einstein–Podolsky–Rosen Paradox, Physical Review Letters 66, 397–400 (1991).

[7] M. Schlosshauer, Decoherence, the measurement problem, and interpretations of quantum mechanics, Reviews of Modern Physics 76, 1267–1305 (2005).

[8] C. J. Riedel, W. H. Zurek, and M. Zwolak, The rise and fall of redundancy in decoherence and quantum darwinism, New Journal of Physics 14, 083010 (2012).

[9] E. P. Wigner, Remarks on the mind-body question, Philosophical reflections and syntheses , 247–260 (1995).

[10] D. Frauchiger and R. Renner, Quantum theory cannot consistently describe the use of itself, Nature communications 9, 3711 (2018).

[11] Č. Brukner, A no-go theorem for observer-independent facts, Entropy 20, 350 (2018).

[12] K.-W. Bong, A. Utreras-Alarcón, F. Ghafari, Y.-C. Liang, N. Tischler, E. G. Cavalcanti, G. J. Pryde, and H. M. Wiseman, A strong no-go theorem on the Wigner’s friend paradox, Nature Physics 16, 1199–1205 (2020).

[13] D. Deutsch, Quantum mechanics near closed timelike lines, Physical Review D 44, 3197 (1991).

[14] C. H. Bennett, D. Leung, G. Smith, and J. A. Smolin, Can closed timelike curves or nonlinear quantum mechanics improve quantum state discrimination or help solve hard problems?, Physical Review Letters 103, 170502 (2009).

[15] S. Aaronson and J. Watrous, Closed timelike curves make quantum and classical computing equivalent, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 465, 631–647 (2009).

[16] S. Lloyd, L. Maccone, R. Garcia-Patron, V. Giovannetti, and Y. Shikano, Quantum mechanics of time travel through post-selected teleportation, Physical Review D—Particles, Fields, Gravitation, and Cosmology 84, 025007 (2011).

[17] D. Deutsch, Quantum theory as a universal physical theory, International Journal of Theoretical Physics 24, 1– 41 (1985).

[18] S. Goldstein, Bohmian mechanics, Stanford Encyclopedia of Philosophy (2017), https://plato.stanford.edu/entries/qm-bohm/.

[19] S. Weinberg, Precision tests of quantum mechanics, Physical Review Letters 62, 485 (1989).

[20] S. Weinberg, Testing quantum mechanics, Annals of Physics 194, 336–386 (1989).

[21] M. Violaris, A physics lab inside your head: Quantum thought experiments as an educational tool, 2023 IEEE International Conference on Quantum Computing and Engineering (QCE) , 203–213 (2023).

[22] M. Violaris, Entangling disciplines: Causality, entropy and time-travel paradoxes on a quantum computer, 2024 IEEE Quantum Science and Engineering Education Conference (QSEEC) , 71–81 (2024).

[23] W. J. Zeng, F. Labib, and V. Russo, Towards violations of local friendliness with quantum computers, Quantum 9, 1851 (2025).

[24] H. M. Wiseman, E. G. Cavalcanti, and E. G. Rieffel, A" thoughtful" local friendliness no-go theorem: a prospective experiment with new assumptions to suit, Quantum 7, 1112 (2023).