The accuracy of a quantitative polymerase chain reaction assay in quantifying the DNA of trichothecene-producing F. culmorum and F. graminearum within harvested wheat grains and head tissue was evaluated in comparison with incidences of infected kernels and deoxynivalenol levels. In a first experiment, six durum and bread wheat varieties were grown in randomized plots for a 2-year period, and inoculated with Fusarium macroconidia at six growth stages between heading and dough ripening, to obtain a wide range of Fusarium head blight incidences. There was a close relationship between fungal DNA and the amount of deoxynivalenol, and this relationship was consistent over Fusarium species, wheat species and varieties, and over a wide range of Fusarium head blight infection. In a second experiment potted wheat plants were grown under environmentally controlled conditions and inoculated with the two Fusarium species at full flowering; head samples were collected before inoculation and after 6 h to 12 days, and processed by the quantitative polymerase chain reaction assay. This assay made it possible to detect the dynamic of fungal invasion in planta after infection had occurred, and to single out the presence of infection before the onset of the disease symptoms: A robust detection of the infection occurred within 18-24 h for F. culmorum, and within 2-9 days for F. graminearum.