The present study provides a first attempt to estimate the validity of the Guilty Knowledge Test (GKT) in real-life criminal investigations. Skin resistance responses in GKT records of 50 innocent and 48 guilty Ss, for whom actual truth was established by confession, were analyzed by two methods. On the basis of an a priori decision rule, 98% of the innocent and 42% of the guilty Ss were correctly classified. Signal detection measures based on various possible decision rules indicated that the distribution of guilty Ss was highly differentiated from the distribution of innocent Ss. An optimal decision rule yielded correct classifications for 94% of the innocent Ss and 65% of the guilty Ss. These results suggest that the GKT could be a useful investigative tool that protects innocent suspects from being falsely classified as guilty. The false-negative identifications were accounted for in terms of the realistic conditions. Two main techniques of psychophysiological detection are used in field practice. The Control Question Technique (CQT; Reid & Inbau, 1977) is the most commonly used and the more controversial mainly because it uses questionable controls (e.g, Furedy & Heslegrave, in press; Lykken, 1974,1981). Other investigators (e.g, Raskin, 1978; Raskin & Kircher, in press; Raskin & Podlesny, 1979) have argued that the questions in the CQT were never intended to function as true experimental controls for relevant questions but rather to set up a situation wherein the innocent examinee, who is truthful concerning the relevant questions, will be more concerned about the controls. Therefore, the term control in the CQT is a misnomer. The Guilty Knowledge Test (GKT) is less frequently used in field practice, but it is based on genuine questions that differ from the relevant questions only with respect to the event under investigation. Therefore, its use has been recommended by many investigators (e.g., Ben-Shakhar, in press; Lykken, 1974,1981). The GKT has been the focus of research on psychophysiological detection since the early 1960s (Davidson, 1968; Giesen & Rollison, 1980; Lykken, 1959, I960; Podlesny & Raskin, 1978), and in some cases, excellent detection rates have been reported. In these studies, the electrodermal measure (which is relatively easy to quantify) was usually used for several multiple-item questions, and detection scores were based on differential responsivity to the relevant items across questions (e.g, Davidson, 1968; Lykken, 1959). All this research was conducted in laboratory settings, however, and used either cardtest or mock-crime procedures. The present study constitutes a first attempt to gather and analyze data based on the GKT as applied in field practice. The purpose of the GKT is to detect the presence of guilty knowledge by means of physiological measurements. During