I have followed with interest a growing body of opinion among physicists that gravity (and space itself) is best thought of as an emergent phenomenon (most recently here). Erik Verlinde has a paper, called On the Origin of Gravity and the Laws of Newton, which presents a heuristic case for gravity as emergent.
Unlike most of the other research papers I've blogged about, this is not a quantum gravity theory, but rather uses a number of concepts in mainstream physics (thermodynamics, the holographic principle) to derive emergent gravity. He says that if one coarse grains a microscopic theory (whose precise dynamics need not be known), and applies the holographic principle to measure information on partition screens between particles, the information on the screens will give rise to an entropic force - this is gravity.
The paper has engendered discussion (I first saw it mentioned by Peter Woit here; there is some appreciation here, and criticism here -- Verlinde responds here). The main criticisms are that Verlinde's points are either not new, or that they embody circular reasoning (since concepts from Newtonian and post-Newtonian physics are used to derive Newtonian gravity). Verlinde responds that he is bringing out a new insight which should help convince people that gravity is not a fundamental force, but is emergent.
I can't adjudicate the disagreements, but I think it's very suggestive that the argument for emergence continues to gain adherents.
I also think it is interesting to note that in Verlinde's model the microscopic theory, while not defined in any detail, must have a well-defined asymmeterical time dimension, as in the emergent quantum gravity theories I've reviewed. "Time is fundamental, while space is not".
[UPDATE 22 Jan.2010: A couple of more related links (HT). A New Scientist article, and an illuminating preprint from Lee Smolin, who works through a Verlinde-type derivation in a different way, utilizing ideas from Loop Quantum Gravity research (altho note the specifics of LQG are actually not very important to the analysis). He does a very good job placing the Verlinde work in context of other research and shows where it seems to add new value.]