The possible origin of the immature hair cells that appear in the utricular maculae of guinea pigs following gentamicin-induced hair cell death was investigated. Guinea pigs were continuously infused with bromodeoxyuridine, to label proliferating cells and their progeny, for 2 weeks after inducing damage to the inner ear on one side with gentamicin. The opposite ear in each animal served as control. Serial sections were cut through the entire utricular maculae of both ears of each animal and the number of labelled cells in the epithelium and underlying connective tissue was counted. Label was present in cells in the sensory epithelium in the utricles from the drug exposed ears but not in the controls. The nuclei of cells in the underlying connective tissue were also labelled in both ears. Some of the labelled nuclei in the epithelium were at the level normally occupied by hair cells, but most were at the level of supporting cell nuclei. However, the total number of labelled nuclei in the sensory epithelium was small; the maximum was 12 in one animal. The number of labelled nuclei in the connective tissue of the treated ears was significantly greater than the number in the untreated ear. This confirms that cell proliferation is stimulated in the mature mammalian utricular macula after hair cell loss, but the extent to which it occurs appears to be insufficient to explain the recovery in hair cell numbers which is observed. Detailed thin section studies of the utricular maculae of gentamicin-treated animals over a prolonged post-treatment period were also performed. In utricles which had suffered damage, there were cells which, like supporting cells but unlike hair cells, were resting on basement membrane, but which possessed at their apical ends organized bundles of microvilli similar to immature hair cell stereocilia. Other cells with more obvious stereocilia remained in contact with the basement membrane via and a small feet process. In still other cells, where a stereociliary bundle was obvious and almost mature in appearance, there was a foot process extending towards the basement membrane but not quite in contact, suggesting it had just detached. All these cells were contacted by nerve endings and specialization of the membranes were apparent at the site of cell-neurone contact. The morphological characteristics of these cells are consistent with phenotypic conversion of supporting cells into hair cells and this may account for some of the hair cell production in the mature mammalian vestibular sensory epithelia after hair cell death.