Assuming perfect detection efficiency, we present an (indeterministic) model for an EPR-Bohm experiment which reproduces the singlet correlations, without contradicting Bell's original locality condition. In this model we allow the probability distribution $\rho_{ \lambda}$ of the state $\lambda$ at the source to depend parametrically on the orientation $\xi$ of one of the measuring devices: $\rho_{ \lambda}(\lambda,\xi)$. In a Bell experiment, no-signaling between the source and each one of the devices would seem clearly sufficient to rule such an influence; however, not even schemes where the choice of observables takes place during the on-flight time of the particles can prevent, in some situations, a model of this type from violating the local bounds. In particular, a random shift $\rho_{ \lambda}(\lambda,\xi_1) \rightarrow \rho_{ \lambda}(\lambda,\xi_2) \rightarrow...\rightarrow \rho_{ \lambda}(\lambda,\xi_n)$ allows the model to perform a "subensemble selection" for each of the terms involved in the inequality (analogous to what goes on with the efficiency loophole), whenever some correlation of those $\rho_{ \lambda}$-shifts with the sequence of measurement choices is allowed. That correlation does not necessarily imply signaling during the photon on-flight time.
Categories: quant-ph