With the imminent promise of constraints on the epoch of reionization from observations of microwave background anisotropies, the question of whether or not the standard cold dark matter (CDM) model permits early reionization has been subjected to detailed investigation by various authors, with the conclusion that reionization may occur at quite high redshift. However, it is widely accepted that this model is excluded, since when normalized to the COBE observations it possesses excessive galaxy clustering on scales below tens of megaparsecs. We examine the trends in observations, first in a fairly model-independent way, and secondly by considering variants on the standard CDM model introduced to resolve the observational conflicts. We conclude that the epoch of reionization favoured by the observational data is typically considerably later than the standard CDM model suggests, and amongst models that may fit the observational data only the introduction of a cosmological constant leads to a reionization redshift close to that of standard CDM.