A population of the tortoise Testudo hermanni near Olympia in southern Greece was studied by mark-recapture from 1975 to 1984. Part of the site was sprayed with the herbicides 2,4-dichlorophenoxyacetic acid (2,4-D) and 2,4,5-trichlorophenoxyacetic acid (2,4,5-T) each year from 1980, producing symptoms of poisoning (swollen eyes, fluid discharge from the nose and immobility) in tortoises. Survival rates of tortoises 10 cm or larger were significantly lower in the affected areas, with extra mortality of about 34% year-1, against an annual survival rate of 0.85-0.90 in unaffected areas. Changing population structures showed that juveniles were even more strongly affected, with the proportion of juveniles in samples decreased by half. The population in the sprayed area declined to near zero by 1984, due to mortality rather than to emigration, since more movements were recorded into than out of the affected area. There was no difference in body mass condition between sprayed and unsprayed areas, showing that effects were acute; mortality was not due to starvation from loss of food plants. The scale and pattern of mortality was similar to that from a severe scrub fire; spraying is potentially more catastrophic since often repeated at shorter intervals than burning. Possible physiological mechanisms of death are discussed. The susceptibility of tortoises to 2,4-D and 2,4,5-T (or to associated dioxin impurities) presents a warning for conservation of these late-maturing animals.