This paper presents the scanpaths analysis technique in pursuit of the data distribution variability in the electrocardiogram. Similar methods are already in use in many studies on active displays, man-machine interaction and human factors in interpretation of images. In all applications, the principle is a relationship between the local features of the scene and the parameters of eyeglobe trajectory captured in controlled circumstances. Experimental study of such relationship for the electrocardiogram was originally inspired by the research on adaptive discrete ECG representation using non-uniform sampling, but the area of possible application includes also physiological distortion assessment, perception-based interpretation learning and others. Our experiment reveals important variability of signal conspicuity in context of cardiac events and significant differences between novices and experienced observers.