Another way of expressing these constant tendencies is to give in each case the number of people in the group of fifty observers who tend in each direction. These figures are given in the last two columns of the above table. It is clear at once that in the case of the first four traits the tendency is predominantly in the direction of overestimation; in the next three traits the two tendencies are evenly balanced, while in the last two the general tendency is strongly toward underestimation. It is of course difficult to say, in this connection, just how accurately the figures given portray the real self-estimation of the different individuals, and to what degree they indicate merely what the individual will do with her own name in the case of such an experiment. Natural modesty might easily lead one to place her own name lower in the scale for a given trait than she really believed herself to belong. If this were the case, we might then infer that the figures we have presented, although qualitatively suggestive, were not quantitatively reliable. They would, in other words, express smaller degrees of overestimation and underestimation than were really present in the consciousness of our observers. Here, as in all the results of this investigation, the figures are given only as indicating what individuals actually do when asked to rank themselves among their associates. Our conclusion on this point is that they tend to overestimate or to underestimate themselves, according to the "desirableness" or "undesirableness" of the trait in question. Individual differences in these tendencies are everywhere apparent. Thus, in neatness, individuals S and H stand about equally high (S being ninth and H being thirteenth), but S underestimates herself by thirteen places, while H overestimates herself by ten places.