USON et al.1 have detected the 21-cm hydrogen line at a redshift of z = 3.4, both in absorption against a radio galaxy, 0902 + 343, which is also at z = 3.4, and more interestingly in emission from a nearby region at the same redshift. They interpret the H I emission as arising from a Zeldovich 'pancake'2, a primordial supergalactic condensation of gas which has yet to fragment into protogalactic units. Pancake (or 'top-down') theories of galaxy formation have several well-known difficulties3,4, however, and 'bottom-up' theories, in which protogalactic units form first and only later aggregate into clusters, are currently more popular. We propose that the H I emission seen by Uson et al. may be emission from neutral hydrogen in a collection of protogalaxies, which are of the type needed to explain the damped Lyman-α absorption systems in quasar spectra5. For this explanation to work, more galaxies per unit mass must form in protocluster environments than in the field. The narrowness of the observed emission line implies that the protocluster is near gravitational turn-around, the moment at which it detaches itself from the general cosmic expansion and begins to collapse. The absorption line can arise from a gaseous halo rich in H I around the radio galaxy, or from other galaxies in the vicinity.