Quantitative experimental results on the antiherbivorous effect of cyanogenesis are rare. In our analyses, we distinguished between the total amount of cyanide-containing compounds stored in a given tissue [cyanogenic potential (HCNp)] and the capacity for release of HCN per unit time (HCNc) from these cyanogenic precursors as a reaction to herbivory. We analyzed the impact of these cyanogenic features on herbivorous insects using different accessions of lima beans (Phaseolus lunatus L.) with different cyanogenic characteristics in their leaves and fourth instars of the generalist herbivore Schistocerca gregaria Forskål (Orthoptera, Acrididae). Young leaves exhibit a higher HCNp and HCNc than mature leaves. This ontogenetic variability of cyanogenesis was valid for all accessions studied. In no-choice bioassays, feeding of S. gregaria was reduced on high cyanogenic lima beans compared with low cyanogenic beans. A HCNp of about 15 micromol cyanide/g leaf (fresh weight) with a corresponding HCNc of about 1 micromol HCN released from leaf material within the first 10 min after complete tissue disintegration appears to be a threshold at which the first repellent effects on S. gregaria were observed. The repellent effect of cyanogenesis increased above these thresholds of HCNp and HCNc. No repellent action of cyanogenesis was observed on plants with lower HCNp and HCNc. These low cyanogenic accessions of P. lunatus were consumed extensively--with dramatic consequences for the herbivore. After consumption, locusts showed severe symptoms of intoxication. Choice assays confirmed the feeding preference of locusts for low over high cyanogenic leaf material of P. lunatus. The bioassays revealed total losses of HCN between 90 and 99% related to the estimated amount of ingested cyanide-containing compounds by the locusts. This general finding was independent of the cyanogenic status (high or low) of the leaf material.