Synopsis Experience in a large manufacturing unit showed that splashes of undiluted shampoo only produced transient effects in the eyes of production and packaging staff; over the same period, very few complaints of eye irritation were received from shampoo users, none of which were of a serious nature. Predictive testing is needed to make sure that new formulations are not severe eye irritants though rabbit eye testing may tend to exaggerate the likely human response. Detergent concentration is the principal factor influencing findings in the rabbit eye and an active matter content of 2.5-3.0% (usually representing a 20% aqueous dilution of shampoo) gives the best discrimination between formulations whilst producing no signs of distress in the animals; for typical anionic shampoos, testing at this concentration is recommended, without subsequent rinsing. Rabbit eye tests should never be conducted at higher concentrations until their innocuousness at 20% dilution has been shown and, even then, stronger preparations should only be instilled if they contain new ingredients for which the potential risk in human use are unknown. Conduct of all rabbit eye testing as a comparison with a 'control' formulation of equivalent concentration, known to be well tolerated under market conditions, is recommended. For preliminary screening, an in vitro test using freshly-isolated buccal mucosa cells from human volunteers may be useful, irritancy being assessed by the proportion of cells showing loss of visible nuclei when examined by phase-contrast microscopy.