Recent studies have shown that distant red galaxies ( DRGs ) , which dominate the high-mass end of the galaxy population at z \sim 2.5 , are more strongly clustered than the population of blue star-forming galaxies at similar redshifts .
However these studies have been severely hampered by the small sizes of fields having deep near-infrared imaging .
Here we use the large UKIDSS Ultra Deep Survey to study the clustering of DRGs .
The size and depth of this survey allows for an unprecedented measurement of the angular clustering of DRGs at 2 < z _ { phot } < 3 and K < 21 .
The correlation function shows the expected power law behavior , but with an apparent upturn at \theta \lesssim 10 \arcsec .
We deproject the angular clustering to infer the spatial correlation length , finding 10.6 \pm 1.6 h ^ { -1 } \textrm { Mpc } .
We use the halo occupation distribution framework to demonstrate that the observed strong clustering of DRGs is not consistent with standard models of galaxy clustering , confirming previous suggestions that were based on smaller samples .
Inaccurate photometric redshifts could artificially enhance the observed clustering , however significant systematic redshift errors would be required to bring the measurements into agreement with the models .
Another possibility is that the underlying assumption that galaxies interact with their large-scale environment only through halo mass is not valid , and that other factors drive the evolution of the oldest , most massive galaxies at z \sim 2 .