Following Fig. 1 in the following paper: Eberly (2002): Bell inequalities and quantum mechanics In an ideal case (this is far from easy to do): Send an incident beam from one side of entangled photon pairs into a beamsplitter, and then recombine the outputs back into a single stream. Let's call the outputs of the beamsplitter as X and Y, where X and Y are orthogonal. The theory is that the recombined stream X+Y will still evidence entanglement, as the polarization has been erased. In the next step, let's do the same thing for both Alice and Bob. You end up with 4 outputs from the 2 beamsplitters, which are before erasure: Alice X Alice Y Bob X Bob Y 1. So if you take the recombined streams (Alice X + Alice Y) and (Bob X + Bob Y), they are still entangled. As far as I know, this experiment (as described by Eberly in more complex versions) has not been performed. Does anyone know of a reference on this? 2. Here is a strange one: Alice X and Bob X are identical streams, as are Alice Y and Bob Y. I believe that theory would say that (Alice X + Bob Y) and (Bob X + Alice Y), if they could be combined, would be entangled! Were that true, it would indicate that the probability waves are "real" even though the observable properties apparently are not. Any thoughts? Has anyone seen anything on this?