Four gray crows were trained to choose the larger of two sets presented to them within the “1–12” range. The color, form, and size of the stimulus elements on cards, as well as the patterns of their disposition, were varied constantly during training, while at the same time prohibiting the use of extraneous features not relating to the number of elements. In order to exclude choice on the basis of the total area of the elements, series of presentations with variation of the ratio of the number of elements and their total area were carried out: in half of the presentations the larger set consisted of small elements, and the smaller set of large elements; in the second half, the reverse. In this range all of the birds demonstrated a high accuracy of evaluations (75.3±2.4%), including when there were minimal differences between the sets to be compared. Thus, in the “1–12” range, the crows were found to be capable of accurately comparing sets precisely on the basis of the number of elements in them. When a transition was made to a new “10–20” range, all of the crows demonstrated a successful transfer of the developed reaction (71.5±2.3%). However, when the ratio of the number of elements and their total area was varied, only two birds were able to cope with this task (71.9±3.6%). Overall, the results obtained indicate a capacity of crows for the formation of the concept “larger set,” and are regarded as a manifestation of a high level of development of reflective activity.