Gravity is attractive. Mass curves spacetime toward itself, and every test particle follows geodesics that converge on the source. There is no negative mass, no gravitational charge reversal, no configuration of matter that pushes rather than pulls. This is so fundamental that the attractive nature of gravity is often taken as definitional — the force between masses is always toward, never away.
Bose and Vaidman (arXiv 2602.22715, February 2026) describe a scenario in which a probe mass moves away from a source mass under gravitational interaction, conditioned on a specific quantum measurement outcome.
The setup places the source mass in a spatial superposition — a quantum state where the mass is simultaneously in two distinct locations. The probe mass starts at rest between these locations and interacts gravitationally with the superposed source. Without measurement, the probe feels the weighted average of the gravitational pulls from both positions, which is attractive in the usual sense. But if the source mass is subsequently measured and found in a specific position — one of the two superposition components, selected by postselection — the probe mass is found to have moved away from that position.
The mechanism operates through weak values. A weak value is the expectation value of an observable conditioned on both a prepared initial state and a postselected final state. Weak values can fall outside the eigenvalue spectrum of the observable — they can be negative, complex, or larger than any eigenvalue. When the gravitational force operator has a negative weak value, the effective force acting on the probe during the conditioned evolution is repulsive. The probe moves as if gravity pushed it.
This is not antigravity. The gravitational interaction is purely attractive at every instant and for every branch of the superposition. The repulsion appears only in the conditioned subensemble — the subset of experimental runs where the source mass is found in the specific postselected state. Runs where the source is found in the other state show enhanced attraction. Averaged over all outcomes, the probe moves toward the source, as gravity requires. The repulsion is a property of the conditional statistics, not the fundamental force.
What makes this more than a curiosity is what it requires. For the probe to exhibit repulsive conditioned motion, the gravitational interaction must create entanglement between the source and probe positions. This entanglement requires gravity to act as a quantum channel — coupling the superposition of the source to the motion of the probe coherently, not just classically. If gravity were a classical stochastic force, it would create classical correlations but not the quantum correlations needed for anomalous weak values. The observation of conditioned repulsion, if achieved experimentally, would constitute evidence that gravity operates quantum-mechanically — that spacetime itself participates in superposition.
The proposed experiment uses spin-bearing nanocrystals, exploiting the Stern-Gerlach effect to create spatial superpositions of mesoscopic masses. The masses are small enough for quantum coherence, large enough for measurable gravitational interaction, and the spin degree of freedom provides the measurement basis for postselection. The experimental challenges are severe — isolating nanocrystals from decoherence while maintaining gravitational coupling — but the protocol is concrete.
The result inverts the usual relationship between gravity and quantum mechanics. Normally, quantum gravitational effects are sought at the Planck scale — energies and distances far beyond experimental reach. Here the quantum nature of gravity manifests at the mesoscopic scale, detectable not through extreme conditions but through conditioned measurement statistics. The evidence would be statistical, not dynamical: gravity remains attractive, but its quantum superposition creates conditional outcomes that no classical gravitational theory can reproduce.