To compare two modes of administration (self-administered; by interviewer) and two response options format (using words; images of "facial-expressions") of the first question of SF-36 (Q1SF-36), and to test its validity. We included 825 participants (20-90 years). Q1SF-36, using words or images, was included in a global questionnaire interview and at the end participants filled the SF-36. The agreement was tested by weighted kappa coefficients (WKappa). Classification Trees were used in the calibration of Q1SF-36 responses, with the physical (PDSF36) and mental dimensions of SF-36. Polyserial correlation coefficients and areas under the ROC curves (AUC) were used for validation. After categorization, using PDSF36 classification trees, the WKappa were 0.770 (self-administered vs. interviewer), 0.569 (self-administered vs. facial-expressions) and 0.566 (interviewer vs. facial-expressions). The WKappa between the PDSF36 and the modes (self-administered, interviewer and facial-expressions) were 0.784, 0.713 and 0.579 and the corresponding polyserial correlation coefficients were 0.784, 0.713 and 0.579. A good discriminatory power was found comparing the modes with the PDSF36 (AUC = 0.907, 0.923 and 0.827), but not with mental dimension (AUC = 0.538, 0.501 and 0.629). The Q1SF-36, by self-administration or interviewer, may be a valid alternative for assessment of subjective physical health, but not mental health.