During a COLIPA multicentre study carried out on ‘Alternatives to Eye Irritation’ two laboratories undertook the evaluation of the samples with the fluorescein leakage test (FLT). The lead laboratory (L'Oréal) proposed a prediction model (PM), which converts the in vitro data from this assay to a prediction of eye irritation. All the surfactant-containing substances were tested using the FLT if they were soluble in Hanks’ balanced salt solution (HBSS). Briefly, confluent MDCK cells were exposed for 15 minutes to five fixed concentrations of the test substance. The amount of damage caused to the cellular monolayer was determined by the amount of fluorescein that leaked through the cell layer after a 4-hour period. The amount of test substance that produced 20% leakage compared with a cell-free control was determined (FL20 H4). The FLT results transformed using the PM, generally predicted the in vivo classification (linear Kappa: 0.87±0.17 and 0.75±0.25). Where there were misclassifications, the category assigned was only one different from the observed in vivo score. There was generally good agreement between the two laboratories in this study. The model seems to be more appropriate for evaluation of non-irritants or moderate irritants than for severe irritants. This is relevant for the selection of cosmetic ingredients. Overall, the results of the FLT assay appeared to be encouraging.