Counting the number of galaxies as a function of their apparent brightness is one of the fundamental cosmological tests, providing an important probe of both the geometry and evolutionary history of the Universe. CCD detectors have in recent years enabled astronomers to explore magnitude limits undreamed of a decade or so ago, and where important constraints can be placed on the allowable combinations of q0 and evolution. Recent work has shown that the B-band counts keep rising with a power-law distribution, with a fivefold excess in the number of galaxies at B = 26.5 over that expected from simple non-evolving models. Indeed, it has been suggested that the total numbers of galaxies already seen may be too high for a q0 = 0.5 universe, assuming there is a redshift cut-off in the galaxy distribution caused either by galaxies having strong Lyman limit systems or a low redshift of formation. As q0 = 0.5 is favoured by theoretical arguments, it is important to see if the behaviour of the counts at even fainter magnitudes can be reconciled with a high density universe. Most published counts are unreliable faintward of B ≈ 26, as the incompleteness corrections required become comparable in size to the data. We have now extended the counts to B ∼ 28, using a ∼ 24 hour CCD exposure taken on the 2.5 m Isaac Newton telescope (INT) on La Palma, together with a ∼ 10 hour exposure on a small part of this field taken using the 4.2 m William Herschel Telescope (WHT).