This summary from a physics journal fills the gap. "Realism" is a way of saying that the "way of things" is not altered by the act of our observing them. The research seems to affirm an interpretation of quantum mechanics that gives special power to the act of perceiving, namely the power to collapse a wave function .... (emphases mine)
Quantum physics says goodbye to reality (April 2007) - News - PhysicsWeb
... Some 40 years ago the physicist John Bell predicted that many hidden-variables theories would be ruled out if a certain experimental inequality were violated – known as "Bell's inequality". In his thought experiment, a source fires entangled pairs of linearly-polarized photons in opposite directions towards two polarizers, which can be changed in orientation. Quantum mechanics says that there should be a high correlation between results at the polarizers because the photons instantaneously "decide" together which polarization to assume at the moment of measurement, even though they are separated in space. Hidden variables, however, says that such instantaneous decisions are not necessary, because the same strong correlation could be achieved if the photons were somehow informed of the orientation of the polarizers beforehand. [jf: In the transactional interpetation the "informing" can occur by meaning-free "messages" that travel back in time.]Bell's trick, therefore, was to decide how to orient the polarizers only after the photons have left the source. If hidden variables did exist, they would be unable to know the orientation, and so the results would only be correlated half of the time. On the other hand, if quantum mechanics was right, the results would be much more correlated – in other words, Bell's inequality would be violated.
Many realizations of the thought experiment have indeed verified the violation of Bell's inequality. These have ruled out all hidden-variables theories based on joint assumptions of realism ... [reality exists when we are not observing it].... and locality ... [separated events cannot influence one another instantaneously]. But a violation of Bell's inequality does not tell specifically which assumption – realism, locality or both – is discordant with quantum mechanics.
Markus Aspelmeyer, Anton Zeilinger and colleagues from the University of Vienna, however, have now shown that realism is more [?] of a problem than locality in the quantum world. They devised an experiment that violates a different inequality proposed by physicist Anthony Leggett in 2003 that relies only on realism, and relaxes the reliance on locality. To do this, rather than taking measurements along just one plane of polarization, the Austrian team took measurements in additional, perpendicular planes to check for elliptical polarization.
They found that, just as in the realizations of Bell's thought experiment, Leggett's inequality is violated – thus stressing the quantum-mechanical assertion that reality does not exist when we're not observing it. "Our study shows that 'just' giving up the concept of locality would not be enough to obtain a more complete description of quantum mechanics," Aspelmeyer told Physics Web. "You would also have to give up certain intuitive features of realism."
This article is the best so far, but there were a few awkward phrases. I've tried to edit it minimally to clarify.
My current (weak) understanding is that locality fell a while ago. There was still hope of preserving realism, but now realism is at least partly gone. We're well into the realm of Schrodinger's "cat" being both alive and dead until an "observer" inspects the cat. I think this result may favor the "decoherence" interpretation of QM, and goes against the "transactional" interpretation favored by (among others) Gribbin.
Alternatively, there's always the reassuring possibility that mathematics is a less trustworthy guide to reality than commonly thought...
Update 5/26/2007: It turns out that in 2004 I posted on a fascinating discussion about how reality can be emergently structured that seems to fit very well with this experiment. I need to find more along the lines of that 2004 post!