An unpreserved artificial tear substitute containing 0.1% sodium hyaluronate was compared with a preparation containing 1.4% polyvinyl alcohol and 0.5% chlorobutanol in a controlled, double-masked, randomized study in patients with moderately severe keratoconjunctivitis sicca. Patients were evaluated initially, at 1, 4, and 8 weeks. The dry-eye status was evaluated by means of tear-film osmolality, tear breakup time, rose bengal staining, Schirmer's test (without anesthesia), and ocular surface-impression cytology. In general, neither preparation was found to be superior to the other. In both study groups, the mean tear-film osmolality and rose bengal staining score improved over the eight-week study, but the degree of squamous metaplasia of the bulbar conjunctival surface, as shown by impression cytology, did not change significantly.