The Las Campanas Infrared (LCIR) Survey, using the Cambridge Infra-Red Survey Instrument (CIRSI), reaches H∼21 over nearly 1 deg2. In this paper we present results from 744 arcmin2 centred on the Hubble Deep Field South for which UBVRI optical data are publicly available. Making conservative magnitude cuts to ensure spatial uniformity, we detect 3177 galaxies to H=20.0 in 744 arcmin2 and a further 842 to H=20.5 in a deeper subregion of 407 arcmin2. We compare the observed optical–infrared (IR) colour distributions with the predictions of semi-analytic hierarchical models and find reasonable agreement. We also determine photometric redshifts, finding a median redshift of ∼0.55. We compare the redshift distributions N(z) of E, Sbc, Scd and Im spectral types with models, showing that the observations are inconsistent with simple passive-evolution models while semi-analytic models provide a reasonable fit to the total N(z) but underestimate the number of z∼1 red spectral types relative to bluer spectral types. We also present N(z) for samples of extremely red objects (EROs) defined by optical–IR colours. We find that EROs with R-H>4 and H<20.5 have a median redshift zm∼1 while redder colour cuts have slightly higher zm. In the magnitude range 19<H<20 we find that EROs with R-H>4 comprise ∼18 per cent of the observed galaxy population, while in semi-analytic models they contribute only ∼4 per cent.We also determine the angular correlation function w(θ) for magnitude, colour, spectral type and photometric redshift-selected subsamples of the data and use the photometric redshift distributions to derive the spatial clustering statistic ξ(r) as a function of spectral type and redshift out to z∼1.2. Parametrizing ξ(r) by ξ(rc,z)=[rc/r(z)]-1.8, where rc is in comoving coordinates, we find that r(z) increases by a factor of 1.5–2 from z=0 to z∼1.2. We interpret this as a selection effect – the galaxies selected at z∼1.2 are intrinsically very luminous, about 1–1.5 mag brighter than L. When galaxies are selected by absolute magnitude, we find no evidence for evolution in r over this redshift range. Extrapolated to z=0, we find r(z=0)∼6.5 h-1 Mpc for red galaxies and r(z=0)∼2–4 h-1 Mpc for blue galaxies. We also find that, while the angular clustering amplitude of EROs with R-H>4 or I-H>3 is up to four times that of the whole galaxy population, the spatial clustering length r(z=1) is ∼7.5–10.5 h-1 Mpc, which is only a factor of ∼1.7 times r(z=1) for R-H<4 and I-H<3 galaxies lying in a similar redshift and luminosity range. This difference is similar to that observed between red and blue galaxies at low redshifts.