Massive gravity may be viewed as a suitable limit of bimetric gravity. The limiting procedure can lead to an interesting interplay between the "background" and "foreground" metrics in a cosmological context. The fact that in bimetric theories one always has two sets of metric equations of motion continues to have an effect even in the massive gravity limit. Thus, solutions of bimetric gravity in the limit of vanishing kinetic term are also solutions of massive gravity, but the contrary statement is not necessarily true.

We discuss the subtle relationship between massive gravity and bimetric gravity, focusing particularly on the manner in which massive gravity may be viewed as a suitable limit of bimetric gravity. The limiting procedure is more delicate than currently appreciated. Specifically, this limiting procedure should not unnecessarily constrain the background metric, which must be externally specified by the theory of massive gravity itself. The fact that in bimetric theories one always has two sets of metric equations of motion continues to have an effect even in the massive gravity limit, leading to additional constraints besides the one set of equations of motion naively expected. Thus, since solutions of bimetric gravity in the limit of vanishing kinetic term are also solutions of massive gravity, but the contrary statement is not necessarily true, there is not complete continuity in the parameter space of the theory. In particular, we study the massive cosmological solutions which are continuous in the parameter space, showing that many interesting cosmologies belong to this class.

We investigate the clustering properties of galaxies in the recently completed ELAIS-S1 redshift survey through their spatial two point autocorrelation function. We used a sub-sample of the ELAIS-S1 catalog covering approximately 4 deg^2 and consisting of 148 objects selected at 15 micron with a flux >0.5 mJy and redshift z<0.5. We detected a positive signal in the correlation function that, in the range of separations 1-10 h mpc is well approximated by a power law with a slope gamma = 1.4 and a correlation length s =5.4 h mpc. This result is in good agreement with the redshift-space correlation function measured in more local samples of mid infrared selected galaxies like the IRAS PSC$z$ redshift survey. This suggests a lack of significant clustering evolution of infrared selected objects out to $z=0.5$ that is further confirmed by the consistency found between the correlation functions measured in a local (z<0.2) and a distant (0.2<z<0.5) subsample of ELAIS-S1 galaxies. We also confirm that optically selected galaxies in the local redshift surveys, especially those of the SDSS sample, are significantly more clustered than infrared objects.