Previous headache studies have been unable to verify the presumed presence of headache. Attempting to correct this design fault, the present study assessed four psychophysiological measures (frontalis EMG, temporal BVP, temporal and finger skin temperature) and salient subjective measures in 13 migraineurs, eight muscle-contraction headache sufferers, and 13 age-matched normals. All subjects submitted to two 30-minute sessions of quiet monitoring, and for the headache subjects, one of the sessions was headache active. A bogus, but convincing preliminary "assessment" revealed insufficient headache activity in the headache active session, forcing subjects to reschedule that session in the future when a strong headache was present. A parallel manipulation was employed with the normal subjects. A MANOVA failed to discriminate within- or between-group differences on the psychophysiological measures. Self-reported pain was uncorrelated with the psychophysiological indices. These results cast further doubt on the validity of the psychophysiological measures employed in this study, the same ones routinely endorsed by headache researchers and therapists. We discuss problems of recruitment, compliance, and attrition in basic headache research.