OBJECTIVE To evaluate the accuracy of confrontation visual field testing with regard to the density of the visual field defect and its location in the peripheral visual field. METHODS A prospective comparison of confrontation visual field testing with full-threshold Humphrey automated static perimetry C24-2 or C30-2 was conducted at a university eye center over a 3-month period. Seventy-two patients with a variety of neurologic and ophthalmologic conditions underwent confrontation visual field testing and automated perimetry as a part of their evaluation. One visual field from each patient was analyzed for this study. RESULTS Confrontation visual field testing yielded an overall sensitivity for detecting an abnormal visual field (full-field analysis) of 63%, when sensitivity of confrontation testing rested on the detection of just one abnormal quadrant. The sensitivity of confrontation testing varied depending on the type of visual field loss present: 51% for arcuate scotomas, 67% for visual field constriction, 78% for altitudinal scotomas, and 90% for hemianopias. The sensitivity of detecting abnormal visual field quadrants, rather than the full-field analysis, was, however, poor at 38%. The sensitivity of confrontation testing was lower for superior quadrant defects and higher for inferior quadrant defects. The estimated probability of detecting an abnormal visual field quadrant occurring at a -26-decibel sensitivity loss from age-matched healthy patients for superior quadrant defects and a -19-decibel sensitivity loss for inferior quadrant defects was 50%. The increased sensitivity noted for visual field defects and for inferior quadrant defects appears to be related, in part, to the density of the visual field loss present. CONCLUSIONS Confrontation visual field testing is relatively insensitive unless a moderate to dense defect is present, and as such is a poor screening test. However, when visual field defects are identified with confrontation visual field testing, the defects often are real as per the high specificity (97%) and high positive predictive value (96%).