A survey is given of principles for obtaining an estimate of the numerical density of profiles not biased by the edge effect in planar samples. Various practical problems of implementation of the principles are discussed. None of the methods represent the ideal combination of independence of assumptions concerning both profiles and frame as well as independence of information external to the frame. The counting rules usually employed in the determination of numerical density are biased and also unnecessarily complicated. For manual work only two unbiased principles apply. The choice between them is mainly determined by the ease by which they are implemented.