1 Introduction

Extremely red objects (R-K>5, EROs hereafter) are providing increasingly stringent constraints on our understanding of the formation of galaxies in general, via their spectral evolution and clustering properties. The very red colors of EROs are well known to be both consistent with old passively evolving distant (z>0.8) elliptical galaxies (e.g. Cohen et al. 1999; Spinrad et al. 1997) or dust-reddened starburst galaxies (e.g. Cimatti et al. 1998; Smail et al. 1999). Purely passive evolution of the present day population of elliptical galaxies is consistent with the measured surface density of faint EROs with $K\sim17$-22, while current renditions of the semianalytical hierarchical merging models fail to reproduce the surface density of EROs by a large factor (Daddi et al. 2000a; Smith et al. 2001; Firth et al. 2001).

Recently, we have completed a relatively large deep survey of very red galaxies covering 700 arcmin² (Daddi et al. 2000b, D00 hereafter), concluding that EROs are strongly clustered in projection, by an order of magnitude more than all galaxies at the same limits of $K\leq 18$-19.2. With careful attention to the measurement uncertainty inherent in narrow field data, Daddi et al. (2001, D01 hereafter) showed that the angular clustering of EROs implies a spatial correlation length of $r_0=12\pm3$ h^-1 comoving Mpc, consistent with the assumption that the ERO population is dominated by elliptical galaxies. This large clustering amplitude is not at odds with recent hierarchical merging models, which require that the most massive galaxies are clustered more strongly than the general galaxy population at high z (e.g. Mo & White 1996). Our results on the angular and spatial clustering of EROs have been substantially confirmed by the Las Campanas Redshift Survey data (McCarthy et al. 2001; Firth et al. 2001; Moustakas & Somerville 2001).

In our recent large K20 redshift survey of a flux limited sample of $\sim$500 galaxies with $K\leq20$ (Cimatti et al. 2002, C02 hereafter), we obtained redshifts for a sub-sample of 35 EROs. For red objects with R-K>5 and $K\leq 19.2$, about 1/3 were identified as old systems (consistent with being passively evolving elliptical galaxies), 1/3 were found to be dusty starburst galaxies and 1/3 remain unidentified. While the derived fraction of early-type galaxies, $50\pm20$%, is consistent with previous estimates based on morphology (Moriondo et al. 2000; Stiavelli & Treu 2000), C02 showed that the dusty star-forming (SF) objects do contribute significantly to the ERO population at faint magnitudes, thus complicating the interpretation of both ERO surface density and clustering, as measured in earlier analyses. In particular, given the strong interest in the clustering amplitude of early-type galaxies, it is important to estimate separately the clustering properties of the old and of the dusty-SF EROs, hence their relative contribution to the clustering of the whole ERO population. This is attempted in this letter, where we adopt a cosmology with $\Omega_\Lambda = 0.7$, $\Omega_{\rm m} = 0.3$ and H₀ = 100 h kms^-1/Mpc.