Sean Carroll has an excellent account of a new paper by Rachel Bean. I don't always share Carroll's attitudes or appreciate his reporting, but here's a case where I thought his post was top notch. You may have already read his posting and if not I hope you will. Meanwhile lets take a cue from Sean and look at the original paper: http://arxiv.org/abs/0909.3853 ==quote from Rachel Bean== The COSMOS data gives the ﬁrst indication that dark energy might be a modiﬁcation to GR, rather than Λ. It will be extremely interesting to see if this signature is seen in other lensing datasets... ==endquote== She has two functions psi and phi: "ψ and φ are the two Newtonian potentials respectively describing temporal and spatial perturbations to the metric." As time goes on (or equivalently as the scalefactor a increases) the geometry of the universe gets more wrinkly/bumpy and warty. This is called "structure formation". Things curdle, condense, gather together in structure. So the geometry gets locally less smooth. So there are these local perturbations---both to psi (curvature in timelike directions) and phi (curvature in spacelike). Slow things like planets are more affected by temporal curvature (because their worldlines move a lot more in time than they do in space). But fast stuff like light is equally affected by both curvatures, and both kinds of bumpiness, both psi and phi. Rachel looks at the ratio ETA = spatial/temporal = phi/psi If General Relativity is obeyed the two should be equal and ETA = 1. She plots 1/ETA = temporal/spatial = psi/phi, as derived from her data, and it has a peak at around 3 or 4. Somehow temporal curvature, the kind of thing that orbiting planets and other slow bodies are more responsive to, has more structure formation than does spatial curvature. Look at her figure 1, something of a shocker. Sean points out that anomalies like this usually go away----some flaw in the analysis shows up, or more data comes in and contradicts the finding. He wisely cautions against jumping to conclusions. At this stage we still bet on GR being right. But figure 1 says that this analysis of this batch of data "disfavors GR at 98% significance level." You can read more of what Sean Carroll has to say about it here: http://blogs.discovermagazine.com/cosmicvariance/2009/10/12/a-new-challenge-to-einstein/ His headline is "A new challenge to Einstein?"