The purpose of this study was to assess the ability of in vivo confocal microscopy (CM) to provide noninvasively derived histopathologic correlates of surfactant-induced eye irritation from which specific pathologic mechanisms can be identified and later evaluated in alternative in vitro models. Rats and rabbits, divided into groups of 5, received 10 μl of an anionic or cationic surfactant in one eye with the other eye used as a control. At specified times, eyes were examined and scored for ocular irritancy using a penlight and slit-lamp. Subsequently, corneas were evaluated by in vivo CM to evaluate epithelial layer thickness and surface epithelial cell area, corneal thickness, depth of necrosis, inflammation, fibrosis, and endothelial injury. At 3 hr, the anionic surfactant produced slight irritation with peak scores of 12.4 and 8.0 out of a possible 110 in the rats and rabbits, respectively. In vivo CM revealed changes limited to the corneal epithelium that decreased in thickness to 78% in rats and 81% in rabbits at 3 hr. This decrease in the thickness correlated with a significant decrease in surface epithelial cell area from 2,061 & p l u s m n ; 395 μm 2 to 567 & p l u s m n ; 330 μm 2 in the rats and 1,523 ± 185 μm 2 to 934 ± 71 μm 2 in the rabbits (p < 0.005 and 0.005, respectively). The cationic surfactant produced severe irritation in both the rats and rabbits with peak scores of 85.4 and 80.2 occurring at day 2, respectively. In vivo CM in the rats showed complete loss of corneal epithelium, lysis of keratocytes, and loss of corneal endothelium. In the rabbits, injury appeared limited to the anterior cornea with complete loss of epithelium and loss of keratocytes extending to 52% of the corneal thickness. These findings establish the application of noninvasive, in vivo CM to qualitatively and quantitatively characterize the pathobiology of ocular irritation in situ. This information will be important in the development and evaluation of mechanistically based in vitro alternatives for ocular irritancy testing.