A range of plant species typical of semi-natural grasslands were tested for their sensitivity to short-term ozone injury under normal and reduced irrigation, and in relationship to air vapour pressure deficit. Potted specimens of 24 herbs, legumes and grasses were exposed during two seasons to four O3 treatments in open-top chambers. The ozone treatments were: (a) charcoal-filtered air; (b) charcoal-filtered air plus ozone to match ambient levels; (c) charcoal-filtered air plus O3 to ambient levels 1.5 and (d) charcoal-filtered air with ozone added to twice ambient levels during selected episodes of 7–13 d. During these ozone episodes, half of the plants in each ozone treatment received reduced irrigation (dry treatment) while the rest was kept under full irrigation (wet treatment). Type and date of first occurrence of leaf injury were noted during individual growth periods. Plants were harvested three times per year, and the percentage of injured leaves was recorded. Depending on species, injury symptoms were expressed as flecking (O3-specific injury), leaf yellowing or anthocyanin formation. Carum carvi and most species of the Fabaceae family (Onobrychis sativa, Trifolium repens, Trifolium pratense) were found to be most responsive to O3, injury occurring after only a few days of exposure in treatment (b). An episodic reduction in irrigation tended to reduce the expression of O3-specific symptoms, but only in species for which a reduction in soil moisture potential and an associated reduction in stomatal conductance during the dry episodes were observed. In other species, the protection from O3 injury seemed to be of little importance. Using artificial neural networks the injury response of nine species was analysed in relation to Species, stomatal conductance, ozone as AOT40 (accumulated exposure above a threshold of 0.04 ppm for periods with global radiation [ges ]50 W m−2 (Fuhrer et al., 1997)), mean relative growth rate, air vapour pressure deficit and global radiation. In the model with all factors, Species was most important, and when Species was omitted, stomatal conductance was the most important determinant for leaf injury to occur, whereas mean relative growth rate was less important. With no plant-related factors included, air vapour pressure deficit and AOT40 were of highest importance. Only in eight species was a positive relationship found between these two factors during the five days before the onset of injury, indicating increasing protection from ozone with increasing air vapour pressure deficit in some but not all species. These data show that across a range of grassland species, leaf injury caused by elevated levels of ozone is most likely to occur in species with high stomatal conductance and that protection from ozone during dry periods is species-specific and depends on a reduction in stomatal conductance due to a decrease in soil moisture potential. Protection under increased vapour pressure deficit can occur in some but not all species, depending on the relationship between stomatal conductance and air vapour pressure deficit.