Reality just got a one-two punch. A new experiment has tried to suss out which of two counterintuitive ingredients is more basic to quantum theory, only to find that they go hand in hand.

Einstein was famously bugged by what are now well-established facts of quantum theory: the randomness of a particle's choices and the possibility of instantaneous linkages between far-flung light or matter. Experimenters now conclude that Einstein cannot even pick his poison, because allowing for instant links kills any simple notion of reality, too.

The team updated a classic 1982 experiment in which researchers measured the polarizations, or spatial orientations, of twin pairs of photons. In quantum theory, photons and other particles do not have definite values for properties such as location or polarization but rather acquire a specific property randomly when measured in an experiment.

"The big question always was … whether one can go beyond this probabilistic description," says physicist Markus Aspelmeyer of the Institute for Quantum Optics and Quantum Information in Vienna. Perhaps Einstein was right that "God does not play dice," and a photon has a true state that is somehow hidden from experiments.

Researchers learned that they could test a related question using photons that are entangled, meaning they are instantaneously connected over any distance in such a way that the measured property of one depends on the other—like a pair of dice that always comes up doubles.

In the 1982 experiment, if the photons "rolled doubles" more than a certain fraction of the time, it meant that particles violated something called local realism: the idea that influences between particles ripple through spacetime like waves (locality) and that particles have hidden nonrandom properties (realism).

But which assumption might be wrong? "It could still be possible," Aspelmeyer says, "that you maintain realism … and that you just relax this locality condition." So he, along with team leader Anton Zeilinger and colleagues, tested a proposed antiquantum model in which influences travel instantaneously but particles have real properties (no locality but realism).

They split red laser photons into entangled pairs and sent the twinned light particles along separate paths. They then measured the polarizations of the photon at different angles to see how often they scored "doubles," called correlations.

Aspelmeyer says the group's hunch was that "if you allow for nonlocal interactions, anything goes, [so] you can recover quantum physics completely" without losing a grip on reality. But, as in the older experiment, they once again saw more correlations than nonlocal realism allowed.

In other words, Aspelmeyer says, nonlocality is not enough to save realism from quantum theory. In effect, quantum naysayers like Einstein would have to swallow the spider of nonrealism to catch the fly of nonlocality. "You have to pay a price," Aspelmeyer says. "I'm still amazed [at the experiment's outcome], I have to say."

There are still other models of nonlocal realism that the experiment does not address, including some that are indistinguishable in principle from quantum theory, writes Alain Aspect of the Institute of Optics in Palaiseau, France, the leader of the 1982 experiment, in a comment published along with the findings in this week's Nature.

"The conclusion one draws is more a question of taste than logic," Aspect says. "But I rather share the view that such … experiments … allow us to look deeper into the great mysteries of quantum mechanics."